diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,11896 @@
+{
+  "best_metric": 0.7569891957933067,
+  "best_model_checkpoint": "wav2vec2-base-mirst500-ac/checkpoint-19560",
+  "epoch": 14.999616711383672,
+  "global_step": 19560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5337423312883438e-07,
+      "loss": 2.0826,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0674846625766876e-07,
+      "loss": 2.0811,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.6012269938650306e-07,
+      "loss": 2.0803,
+      "step": 30
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6.134969325153375e-07,
+      "loss": 2.0809,
+      "step": 40
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.668711656441718e-07,
+      "loss": 2.0797,
+      "step": 50
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.202453987730061e-07,
+      "loss": 2.0802,
+      "step": 60
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.0736196319018406e-06,
+      "loss": 2.0782,
+      "step": 70
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.226993865030675e-06,
+      "loss": 2.0776,
+      "step": 80
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.3803680981595093e-06,
+      "loss": 2.0757,
+      "step": 90
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.5337423312883435e-06,
+      "loss": 2.0767,
+      "step": 100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.6871165644171778e-06,
+      "loss": 2.0727,
+      "step": 110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.8404907975460122e-06,
+      "loss": 2.0713,
+      "step": 120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9938650306748465e-06,
+      "loss": 2.0673,
+      "step": 130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.147239263803681e-06,
+      "loss": 2.0645,
+      "step": 140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.3006134969325154e-06,
+      "loss": 2.0599,
+      "step": 150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.45398773006135e-06,
+      "loss": 2.0556,
+      "step": 160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.607361963190184e-06,
+      "loss": 2.0405,
+      "step": 170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.7607361963190186e-06,
+      "loss": 2.0452,
+      "step": 180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.914110429447853e-06,
+      "loss": 2.0183,
+      "step": 190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.067484662576687e-06,
+      "loss": 2.022,
+      "step": 200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.2208588957055217e-06,
+      "loss": 2.0177,
+      "step": 210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.3742331288343555e-06,
+      "loss": 2.0092,
+      "step": 220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.52760736196319e-06,
+      "loss": 1.9919,
+      "step": 230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.6809815950920245e-06,
+      "loss": 1.9732,
+      "step": 240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.834355828220859e-06,
+      "loss": 1.9817,
+      "step": 250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.987730061349693e-06,
+      "loss": 1.9678,
+      "step": 260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.141104294478528e-06,
+      "loss": 1.9719,
+      "step": 270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.294478527607362e-06,
+      "loss": 1.9792,
+      "step": 280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.447852760736197e-06,
+      "loss": 1.9729,
+      "step": 290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.601226993865031e-06,
+      "loss": 1.9424,
+      "step": 300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.754601226993865e-06,
+      "loss": 1.9597,
+      "step": 310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.9079754601227e-06,
+      "loss": 1.9569,
+      "step": 320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 5.061349693251534e-06,
+      "loss": 1.923,
+      "step": 330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 5.214723926380368e-06,
+      "loss": 1.9158,
+      "step": 340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 5.3680981595092024e-06,
+      "loss": 1.9466,
+      "step": 350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 5.521472392638037e-06,
+      "loss": 1.9243,
+      "step": 360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 5.674846625766872e-06,
+      "loss": 1.9048,
+      "step": 370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 5.828220858895706e-06,
+      "loss": 1.9481,
+      "step": 380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 5.981595092024539e-06,
+      "loss": 1.9287,
+      "step": 390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.119631901840491e-06,
+      "loss": 1.9206,
+      "step": 400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.273006134969325e-06,
+      "loss": 1.8873,
+      "step": 410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.426380368098159e-06,
+      "loss": 1.9009,
+      "step": 420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.579754601226995e-06,
+      "loss": 1.9162,
+      "step": 430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.7331288343558285e-06,
+      "loss": 1.9053,
+      "step": 440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.886503067484662e-06,
+      "loss": 1.895,
+      "step": 450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.039877300613497e-06,
+      "loss": 1.9202,
+      "step": 460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.193251533742332e-06,
+      "loss": 1.8949,
+      "step": 470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.346625766871166e-06,
+      "loss": 1.8945,
+      "step": 480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.5e-06,
+      "loss": 1.8978,
+      "step": 490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.653374233128835e-06,
+      "loss": 1.8694,
+      "step": 500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.806748466257669e-06,
+      "loss": 1.851,
+      "step": 510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.960122699386504e-06,
+      "loss": 1.8775,
+      "step": 520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.113496932515338e-06,
+      "loss": 1.8612,
+      "step": 530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.266871165644172e-06,
+      "loss": 1.8656,
+      "step": 540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.420245398773006e-06,
+      "loss": 1.8852,
+      "step": 550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57361963190184e-06,
+      "loss": 1.8791,
+      "step": 560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.726993865030675e-06,
+      "loss": 1.8619,
+      "step": 570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.88036809815951e-06,
+      "loss": 1.87,
+      "step": 580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.033742331288344e-06,
+      "loss": 1.837,
+      "step": 590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.187116564417178e-06,
+      "loss": 1.8714,
+      "step": 600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.340490797546012e-06,
+      "loss": 1.8338,
+      "step": 610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.493865030674848e-06,
+      "loss": 1.8605,
+      "step": 620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.647239263803681e-06,
+      "loss": 1.8143,
+      "step": 630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.800613496932515e-06,
+      "loss": 1.8239,
+      "step": 640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.953987730061349e-06,
+      "loss": 1.7989,
+      "step": 650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107361963190184e-05,
+      "loss": 1.7968,
+      "step": 660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026073619631902e-05,
+      "loss": 1.8004,
+      "step": 670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.0414110429447854e-05,
+      "loss": 1.8003,
+      "step": 680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.0567484662576688e-05,
+      "loss": 1.7964,
+      "step": 690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.0720858895705521e-05,
+      "loss": 1.7865,
+      "step": 700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.0874233128834355e-05,
+      "loss": 1.7934,
+      "step": 710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.102760736196319e-05,
+      "loss": 1.769,
+      "step": 720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.1180981595092025e-05,
+      "loss": 1.7856,
+      "step": 730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1334355828220858e-05,
+      "loss": 1.7329,
+      "step": 740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1487730061349694e-05,
+      "loss": 1.7759,
+      "step": 750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1641104294478528e-05,
+      "loss": 1.7598,
+      "step": 760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1794478527607363e-05,
+      "loss": 1.768,
+      "step": 770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1947852760736197e-05,
+      "loss": 1.7258,
+      "step": 780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2101226993865031e-05,
+      "loss": 1.6965,
+      "step": 790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2254601226993865e-05,
+      "loss": 1.7436,
+      "step": 800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2407975460122699e-05,
+      "loss": 1.739,
+      "step": 810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.2561349693251534e-05,
+      "loss": 1.7256,
+      "step": 820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.271472392638037e-05,
+      "loss": 1.7222,
+      "step": 830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.2868098159509203e-05,
+      "loss": 1.7297,
+      "step": 840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3021472392638037e-05,
+      "loss": 1.7278,
+      "step": 850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3174846625766871e-05,
+      "loss": 1.6908,
+      "step": 860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3328220858895707e-05,
+      "loss": 1.6969,
+      "step": 870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.348159509202454e-05,
+      "loss": 1.6793,
+      "step": 880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3634969325153374e-05,
+      "loss": 1.6909,
+      "step": 890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3788343558282208e-05,
+      "loss": 1.6867,
+      "step": 900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3941717791411042e-05,
+      "loss": 1.6758,
+      "step": 910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4095092024539879e-05,
+      "loss": 1.6615,
+      "step": 920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.4248466257668713e-05,
+      "loss": 1.6315,
+      "step": 930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.4401840490797547e-05,
+      "loss": 1.6183,
+      "step": 940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.455521472392638e-05,
+      "loss": 1.6285,
+      "step": 950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.4708588957055214e-05,
+      "loss": 1.6369,
+      "step": 960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.486196319018405e-05,
+      "loss": 1.5819,
+      "step": 970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015337423312882e-05,
+      "loss": 1.6646,
+      "step": 980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5168711656441718e-05,
+      "loss": 1.6153,
+      "step": 990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.5322085889570555e-05,
+      "loss": 1.5623,
+      "step": 1000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.5475460122699385e-05,
+      "loss": 1.6165,
+      "step": 1010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.5628834355828222e-05,
+      "loss": 1.5586,
+      "step": 1020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.5782208588957053e-05,
+      "loss": 1.5359,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.593558282208589e-05,
+      "loss": 1.5645,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6088957055214727e-05,
+      "loss": 1.5308,
+      "step": 1050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6242331288343558e-05,
+      "loss": 1.5733,
+      "step": 1060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6395705521472395e-05,
+      "loss": 1.5346,
+      "step": 1070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6549079754601225e-05,
+      "loss": 1.5405,
+      "step": 1080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6702453987730063e-05,
+      "loss": 1.5256,
+      "step": 1090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6855828220858896e-05,
+      "loss": 1.4989,
+      "step": 1100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.700920245398773e-05,
+      "loss": 1.531,
+      "step": 1110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.7162576687116564e-05,
+      "loss": 1.5069,
+      "step": 1120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7315950920245398e-05,
+      "loss": 1.4504,
+      "step": 1130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7469325153374235e-05,
+      "loss": 1.4477,
+      "step": 1140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.762269938650307e-05,
+      "loss": 1.398,
+      "step": 1150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7776073619631903e-05,
+      "loss": 1.4772,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.7929447852760737e-05,
+      "loss": 1.4016,
+      "step": 1170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.808282208588957e-05,
+      "loss": 1.4365,
+      "step": 1180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8236196319018404e-05,
+      "loss": 1.4039,
+      "step": 1190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.838957055214724e-05,
+      "loss": 1.4422,
+      "step": 1200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8542944785276075e-05,
+      "loss": 1.3787,
+      "step": 1210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.869631901840491e-05,
+      "loss": 1.3919,
+      "step": 1220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8849693251533743e-05,
+      "loss": 1.4059,
+      "step": 1230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9003067484662577e-05,
+      "loss": 1.3692,
+      "step": 1240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9156441717791414e-05,
+      "loss": 1.3892,
+      "step": 1250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9309815950920244e-05,
+      "loss": 1.406,
+      "step": 1260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.946319018404908e-05,
+      "loss": 1.3683,
+      "step": 1270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9616564417177912e-05,
+      "loss": 1.3689,
+      "step": 1280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.976993865030675e-05,
+      "loss": 1.382,
+      "step": 1290
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9923312883435586e-05,
+      "loss": 1.3718,
+      "step": 1300
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.42550846848573415,
+      "eval_loss": 1.442152500152588,
+      "eval_runtime": 309.6901,
+      "eval_samples_per_second": 134.79,
+      "eval_steps_per_second": 67.396,
+      "step": 1304
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0076687116564417e-05,
+      "loss": 1.4846,
+      "step": 1310
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.0230061349693254e-05,
+      "loss": 1.3437,
+      "step": 1320
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.0383435582822084e-05,
+      "loss": 1.3393,
+      "step": 1330
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.0536809815950922e-05,
+      "loss": 1.3249,
+      "step": 1340
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.0690184049079756e-05,
+      "loss": 1.3778,
+      "step": 1350
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.084355828220859e-05,
+      "loss": 1.3059,
+      "step": 1360
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.0996932515337423e-05,
+      "loss": 1.3326,
+      "step": 1370
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.1150306748466257e-05,
+      "loss": 1.3107,
+      "step": 1380
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.1303680981595094e-05,
+      "loss": 1.273,
+      "step": 1390
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.1457055214723928e-05,
+      "loss": 1.2811,
+      "step": 1400
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.1610429447852762e-05,
+      "loss": 1.2907,
+      "step": 1410
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.1763803680981596e-05,
+      "loss": 1.2771,
+      "step": 1420
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.191717791411043e-05,
+      "loss": 1.3234,
+      "step": 1430
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.2070552147239263e-05,
+      "loss": 1.2977,
+      "step": 1440
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.22239263803681e-05,
+      "loss": 1.3684,
+      "step": 1450
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.2377300613496934e-05,
+      "loss": 1.33,
+      "step": 1460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.2530674846625768e-05,
+      "loss": 1.2584,
+      "step": 1470
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.2684049079754602e-05,
+      "loss": 1.3037,
+      "step": 1480
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.2837423312883436e-05,
+      "loss": 1.2465,
+      "step": 1490
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.2990797546012273e-05,
+      "loss": 1.3108,
+      "step": 1500
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.3144171779141103e-05,
+      "loss": 1.2868,
+      "step": 1510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.329754601226994e-05,
+      "loss": 1.2254,
+      "step": 1520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.345092024539877e-05,
+      "loss": 1.2612,
+      "step": 1530
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.360429447852761e-05,
+      "loss": 1.2729,
+      "step": 1540
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.3757668711656446e-05,
+      "loss": 1.2281,
+      "step": 1550
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.3911042944785276e-05,
+      "loss": 1.2368,
+      "step": 1560
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.4064417177914113e-05,
+      "loss": 1.258,
+      "step": 1570
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.4217791411042944e-05,
+      "loss": 1.2843,
+      "step": 1580
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.437116564417178e-05,
+      "loss": 1.2428,
+      "step": 1590
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.4524539877300615e-05,
+      "loss": 1.2301,
+      "step": 1600
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.467791411042945e-05,
+      "loss": 1.2689,
+      "step": 1610
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.4831288343558282e-05,
+      "loss": 1.2719,
+      "step": 1620
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.4984662576687116e-05,
+      "loss": 1.252,
+      "step": 1630
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.5138036809815953e-05,
+      "loss": 1.232,
+      "step": 1640
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.5291411042944787e-05,
+      "loss": 1.2372,
+      "step": 1650
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.544478527607362e-05,
+      "loss": 1.216,
+      "step": 1660
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.5598159509202455e-05,
+      "loss": 1.2221,
+      "step": 1670
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.575153374233129e-05,
+      "loss": 1.2125,
+      "step": 1680
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.5904907975460122e-05,
+      "loss": 1.2319,
+      "step": 1690
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.605828220858896e-05,
+      "loss": 1.2291,
+      "step": 1700
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.621165644171779e-05,
+      "loss": 1.2521,
+      "step": 1710
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6365030674846627e-05,
+      "loss": 1.2216,
+      "step": 1720
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.651840490797546e-05,
+      "loss": 1.2214,
+      "step": 1730
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6671779141104295e-05,
+      "loss": 1.2131,
+      "step": 1740
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.682515337423313e-05,
+      "loss": 1.2014,
+      "step": 1750
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.6978527607361963e-05,
+      "loss": 1.2065,
+      "step": 1760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.71319018404908e-05,
+      "loss": 1.2146,
+      "step": 1770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.728527607361963e-05,
+      "loss": 1.2159,
+      "step": 1780
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.7438650306748467e-05,
+      "loss": 1.2331,
+      "step": 1790
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.75920245398773e-05,
+      "loss": 1.1944,
+      "step": 1800
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.7745398773006135e-05,
+      "loss": 1.1684,
+      "step": 1810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.7898773006134972e-05,
+      "loss": 1.1836,
+      "step": 1820
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.8052147239263803e-05,
+      "loss": 1.2277,
+      "step": 1830
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.820552147239264e-05,
+      "loss": 1.2311,
+      "step": 1840
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.835889570552147e-05,
+      "loss": 1.1541,
+      "step": 1850
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8512269938650308e-05,
+      "loss": 1.1939,
+      "step": 1860
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.866564417177914e-05,
+      "loss": 1.215,
+      "step": 1870
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8819018404907975e-05,
+      "loss": 1.2073,
+      "step": 1880
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.8972392638036812e-05,
+      "loss": 1.2172,
+      "step": 1890
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.9125766871165643e-05,
+      "loss": 1.1671,
+      "step": 1900
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.927914110429448e-05,
+      "loss": 1.1746,
+      "step": 1910
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.9432515337423314e-05,
+      "loss": 1.2045,
+      "step": 1920
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.9585889570552148e-05,
+      "loss": 1.2052,
+      "step": 1930
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.973926380368098e-05,
+      "loss": 1.2098,
+      "step": 1940
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.9892638036809815e-05,
+      "loss": 1.1804,
+      "step": 1950
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.9994887525562376e-05,
+      "loss": 1.1601,
+      "step": 1960
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9977845944103613e-05,
+      "loss": 1.1998,
+      "step": 1970
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9960804362644855e-05,
+      "loss": 1.2058,
+      "step": 1980
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.9943762781186096e-05,
+      "loss": 1.2042,
+      "step": 1990
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.9926721199727337e-05,
+      "loss": 1.156,
+      "step": 2000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.9909679618268578e-05,
+      "loss": 1.1389,
+      "step": 2010
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.9892638036809815e-05,
+      "loss": 1.2077,
+      "step": 2020
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9875596455351056e-05,
+      "loss": 1.1721,
+      "step": 2030
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9858554873892298e-05,
+      "loss": 1.1552,
+      "step": 2040
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.984151329243354e-05,
+      "loss": 1.1389,
+      "step": 2050
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.9824471710974776e-05,
+      "loss": 1.1423,
+      "step": 2060
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.980743012951602e-05,
+      "loss": 1.1087,
+      "step": 2070
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.979038854805726e-05,
+      "loss": 1.1672,
+      "step": 2080
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.9773346966598503e-05,
+      "loss": 1.1412,
+      "step": 2090
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.975630538513974e-05,
+      "loss": 1.2531,
+      "step": 2100
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.973926380368098e-05,
+      "loss": 1.2117,
+      "step": 2110
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.9722222222222223e-05,
+      "loss": 1.1797,
+      "step": 2120
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.9705180640763464e-05,
+      "loss": 1.1103,
+      "step": 2130
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.9688139059304705e-05,
+      "loss": 1.1961,
+      "step": 2140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.9671097477845946e-05,
+      "loss": 1.1998,
+      "step": 2150
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.9654055896387183e-05,
+      "loss": 1.1118,
+      "step": 2160
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.9637014314928428e-05,
+      "loss": 1.0952,
+      "step": 2170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.9619972733469666e-05,
+      "loss": 1.1666,
+      "step": 2180
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.9602931152010907e-05,
+      "loss": 1.1119,
+      "step": 2190
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.9585889570552148e-05,
+      "loss": 1.1351,
+      "step": 2200
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.956884798909339e-05,
+      "loss": 1.0964,
+      "step": 2210
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.955180640763463e-05,
+      "loss": 1.1115,
+      "step": 2220
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.953476482617587e-05,
+      "loss": 1.1286,
+      "step": 2230
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.951772324471711e-05,
+      "loss": 1.0792,
+      "step": 2240
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.9500681663258353e-05,
+      "loss": 1.2333,
+      "step": 2250
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.948364008179959e-05,
+      "loss": 1.1588,
+      "step": 2260
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.9466598500340832e-05,
+      "loss": 1.1164,
+      "step": 2270
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.9449556918882073e-05,
+      "loss": 1.1471,
+      "step": 2280
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9432515337423314e-05,
+      "loss": 1.1687,
+      "step": 2290
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9415473755964555e-05,
+      "loss": 1.1317,
+      "step": 2300
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.9398432174505796e-05,
+      "loss": 1.1144,
+      "step": 2310
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.9381390593047034e-05,
+      "loss": 1.1691,
+      "step": 2320
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.9364349011588278e-05,
+      "loss": 1.1186,
+      "step": 2330
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.9347307430129516e-05,
+      "loss": 1.1265,
+      "step": 2340
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.933026584867076e-05,
+      "loss": 1.139,
+      "step": 2350
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9313224267211998e-05,
+      "loss": 1.0644,
+      "step": 2360
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.929618268575324e-05,
+      "loss": 1.1752,
+      "step": 2370
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.927914110429448e-05,
+      "loss": 1.0796,
+      "step": 2380
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.926209952283572e-05,
+      "loss": 1.1441,
+      "step": 2390
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.924505794137696e-05,
+      "loss": 1.0841,
+      "step": 2400
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.9228016359918203e-05,
+      "loss": 1.1328,
+      "step": 2410
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.921097477845944e-05,
+      "loss": 1.1451,
+      "step": 2420
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.9193933197000685e-05,
+      "loss": 1.0597,
+      "step": 2430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.9176891615541923e-05,
+      "loss": 1.1627,
+      "step": 2440
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.9159850034083164e-05,
+      "loss": 1.1068,
+      "step": 2450
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.9142808452624405e-05,
+      "loss": 1.1241,
+      "step": 2460
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.9125766871165643e-05,
+      "loss": 1.0758,
+      "step": 2470
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.9108725289706884e-05,
+      "loss": 1.0849,
+      "step": 2480
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.9091683708248125e-05,
+      "loss": 1.1497,
+      "step": 2490
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.9074642126789366e-05,
+      "loss": 1.1167,
+      "step": 2500
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.9057600545330607e-05,
+      "loss": 1.1317,
+      "step": 2510
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.9040558963871848e-05,
+      "loss": 1.1173,
+      "step": 2520
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.9023517382413086e-05,
+      "loss": 1.0971,
+      "step": 2530
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.900647580095433e-05,
+      "loss": 1.0416,
+      "step": 2540
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.8989434219495568e-05,
+      "loss": 1.0694,
+      "step": 2550
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.8972392638036812e-05,
+      "loss": 1.112,
+      "step": 2560
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.895535105657805e-05,
+      "loss": 1.15,
+      "step": 2570
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.893830947511929e-05,
+      "loss": 1.1234,
+      "step": 2580
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8921267893660532e-05,
+      "loss": 1.1142,
+      "step": 2590
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8904226312201773e-05,
+      "loss": 1.1285,
+      "step": 2600
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.5869247538509451,
+      "eval_loss": 1.1061099767684937,
+      "eval_runtime": 307.6801,
+      "eval_samples_per_second": 135.67,
+      "eval_steps_per_second": 67.837,
+      "step": 2608
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.888718473074301e-05,
+      "loss": 1.1984,
+      "step": 2610
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.8870143149284255e-05,
+      "loss": 1.0601,
+      "step": 2620
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.8853101567825493e-05,
+      "loss": 1.058,
+      "step": 2630
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.8836059986366738e-05,
+      "loss": 1.0598,
+      "step": 2640
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.8819018404907975e-05,
+      "loss": 1.1398,
+      "step": 2650
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.8801976823449216e-05,
+      "loss": 1.114,
+      "step": 2660
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.8784935241990457e-05,
+      "loss": 1.1505,
+      "step": 2670
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.87678936605317e-05,
+      "loss": 1.0187,
+      "step": 2680
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.875085207907294e-05,
+      "loss": 1.0999,
+      "step": 2690
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.873381049761418e-05,
+      "loss": 1.0493,
+      "step": 2700
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.8716768916155418e-05,
+      "loss": 1.0676,
+      "step": 2710
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.8699727334696663e-05,
+      "loss": 1.0944,
+      "step": 2720
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.86826857532379e-05,
+      "loss": 1.094,
+      "step": 2730
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.866564417177914e-05,
+      "loss": 1.0826,
+      "step": 2740
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.8648602590320382e-05,
+      "loss": 1.0625,
+      "step": 2750
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.8631561008861624e-05,
+      "loss": 1.0935,
+      "step": 2760
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.8614519427402865e-05,
+      "loss": 1.0389,
+      "step": 2770
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.8597477845944106e-05,
+      "loss": 0.9919,
+      "step": 2780
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.8580436264485343e-05,
+      "loss": 1.0992,
+      "step": 2790
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.8563394683026588e-05,
+      "loss": 1.076,
+      "step": 2800
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.8546353101567825e-05,
+      "loss": 1.0939,
+      "step": 2810
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.8529311520109067e-05,
+      "loss": 1.0875,
+      "step": 2820
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.8512269938650308e-05,
+      "loss": 1.052,
+      "step": 2830
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.849522835719155e-05,
+      "loss": 1.059,
+      "step": 2840
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.847818677573279e-05,
+      "loss": 1.0816,
+      "step": 2850
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.846114519427403e-05,
+      "loss": 1.0246,
+      "step": 2860
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.844410361281527e-05,
+      "loss": 1.0586,
+      "step": 2870
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8427062031356513e-05,
+      "loss": 1.075,
+      "step": 2880
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.841002044989775e-05,
+      "loss": 1.0355,
+      "step": 2890
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.8392978868438995e-05,
+      "loss": 1.0758,
+      "step": 2900
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.8375937286980233e-05,
+      "loss": 1.0573,
+      "step": 2910
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.835889570552147e-05,
+      "loss": 1.0119,
+      "step": 2920
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8341854124062715e-05,
+      "loss": 1.0956,
+      "step": 2930
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8324812542603953e-05,
+      "loss": 1.0599,
+      "step": 2940
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8307770961145194e-05,
+      "loss": 1.075,
+      "step": 2950
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8290729379686435e-05,
+      "loss": 1.0769,
+      "step": 2960
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.8273687798227676e-05,
+      "loss": 0.9885,
+      "step": 2970
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8256646216768917e-05,
+      "loss": 1.0223,
+      "step": 2980
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8239604635310158e-05,
+      "loss": 1.0363,
+      "step": 2990
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.8222563053851396e-05,
+      "loss": 1.0527,
+      "step": 3000
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.820552147239264e-05,
+      "loss": 1.111,
+      "step": 3010
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.8188479890933878e-05,
+      "loss": 1.1154,
+      "step": 3020
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.8171438309475122e-05,
+      "loss": 0.9981,
+      "step": 3030
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.815439672801636e-05,
+      "loss": 1.0888,
+      "step": 3040
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.81373551465576e-05,
+      "loss": 0.9907,
+      "step": 3050
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.8120313565098842e-05,
+      "loss": 1.0645,
+      "step": 3060
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.8103271983640083e-05,
+      "loss": 1.0524,
+      "step": 3070
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.808623040218132e-05,
+      "loss": 1.049,
+      "step": 3080
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.8069188820722565e-05,
+      "loss": 1.0694,
+      "step": 3090
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8052147239263803e-05,
+      "loss": 1.0486,
+      "step": 3100
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8035105657805047e-05,
+      "loss": 1.0324,
+      "step": 3110
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8018064076346285e-05,
+      "loss": 1.038,
+      "step": 3120
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8001022494887526e-05,
+      "loss": 1.0528,
+      "step": 3130
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.7983980913428767e-05,
+      "loss": 1.073,
+      "step": 3140
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7966939331970008e-05,
+      "loss": 1.0368,
+      "step": 3150
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.7949897750511246e-05,
+      "loss": 0.9978,
+      "step": 3160
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.793285616905249e-05,
+      "loss": 1.0586,
+      "step": 3170
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.7915814587593728e-05,
+      "loss": 0.9979,
+      "step": 3180
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.7898773006134972e-05,
+      "loss": 1.0324,
+      "step": 3190
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.788173142467621e-05,
+      "loss": 1.1098,
+      "step": 3200
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.786468984321745e-05,
+      "loss": 1.042,
+      "step": 3210
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.7847648261758692e-05,
+      "loss": 0.9994,
+      "step": 3220
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.7830606680299933e-05,
+      "loss": 1.0479,
+      "step": 3230
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.7813565098841174e-05,
+      "loss": 1.0468,
+      "step": 3240
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.7796523517382415e-05,
+      "loss": 1.0405,
+      "step": 3250
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.7779481935923653e-05,
+      "loss": 1.0647,
+      "step": 3260
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.7762440354464897e-05,
+      "loss": 1.0701,
+      "step": 3270
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.7745398773006135e-05,
+      "loss": 1.0446,
+      "step": 3280
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.7728357191547376e-05,
+      "loss": 1.0427,
+      "step": 3290
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.7711315610088617e-05,
+      "loss": 0.972,
+      "step": 3300
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.7694274028629858e-05,
+      "loss": 1.0934,
+      "step": 3310
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.76772324471711e-05,
+      "loss": 1.03,
+      "step": 3320
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.766019086571234e-05,
+      "loss": 1.0371,
+      "step": 3330
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.7643149284253578e-05,
+      "loss": 1.0878,
+      "step": 3340
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.7626107702794823e-05,
+      "loss": 1.0889,
+      "step": 3350
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.760906612133606e-05,
+      "loss": 1.0513,
+      "step": 3360
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.75920245398773e-05,
+      "loss": 1.0305,
+      "step": 3370
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7574982958418542e-05,
+      "loss": 1.0411,
+      "step": 3380
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.755794137695978e-05,
+      "loss": 1.0554,
+      "step": 3390
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.7540899795501024e-05,
+      "loss": 1.0393,
+      "step": 3400
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.7523858214042262e-05,
+      "loss": 1.0301,
+      "step": 3410
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.7506816632583503e-05,
+      "loss": 1.0652,
+      "step": 3420
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.7489775051124744e-05,
+      "loss": 1.0494,
+      "step": 3430
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.7472733469665985e-05,
+      "loss": 0.9602,
+      "step": 3440
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.7455691888207226e-05,
+      "loss": 1.001,
+      "step": 3450
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.7438650306748467e-05,
+      "loss": 1.0052,
+      "step": 3460
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.7421608725289705e-05,
+      "loss": 1.0637,
+      "step": 3470
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.740456714383095e-05,
+      "loss": 1.0437,
+      "step": 3480
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.7387525562372187e-05,
+      "loss": 1.0599,
+      "step": 3490
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.737048398091343e-05,
+      "loss": 1.0933,
+      "step": 3500
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.735344239945467e-05,
+      "loss": 0.9721,
+      "step": 3510
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.733640081799591e-05,
+      "loss": 1.0125,
+      "step": 3520
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.731935923653715e-05,
+      "loss": 1.0616,
+      "step": 3530
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.7302317655078393e-05,
+      "loss": 1.0571,
+      "step": 3540
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.728527607361963e-05,
+      "loss": 0.9663,
+      "step": 3550
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.7268234492160875e-05,
+      "loss": 0.9366,
+      "step": 3560
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.7251192910702112e-05,
+      "loss": 1.0649,
+      "step": 3570
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.7234151329243357e-05,
+      "loss": 1.028,
+      "step": 3580
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.7217109747784594e-05,
+      "loss": 1.0266,
+      "step": 3590
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.7200068166325836e-05,
+      "loss": 0.9777,
+      "step": 3600
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7183026584867077e-05,
+      "loss": 0.9919,
+      "step": 3610
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7165985003408318e-05,
+      "loss": 1.0263,
+      "step": 3620
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7148943421949555e-05,
+      "loss": 1.0144,
+      "step": 3630
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.71319018404908e-05,
+      "loss": 1.0182,
+      "step": 3640
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.7114860259032037e-05,
+      "loss": 1.057,
+      "step": 3650
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.7097818677573282e-05,
+      "loss": 1.0242,
+      "step": 3660
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.708077709611452e-05,
+      "loss": 1.0005,
+      "step": 3670
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.706373551465576e-05,
+      "loss": 0.9629,
+      "step": 3680
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.7046693933197002e-05,
+      "loss": 0.952,
+      "step": 3690
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.7029652351738243e-05,
+      "loss": 0.9913,
+      "step": 3700
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.7012610770279484e-05,
+      "loss": 1.0045,
+      "step": 3710
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.6995569188820725e-05,
+      "loss": 1.0066,
+      "step": 3720
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.6978527607361963e-05,
+      "loss": 1.0154,
+      "step": 3730
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.6961486025903207e-05,
+      "loss": 1.0153,
+      "step": 3740
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.6944444444444445e-05,
+      "loss": 1.0212,
+      "step": 3750
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.6927402862985686e-05,
+      "loss": 1.0058,
+      "step": 3760
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.6910361281526927e-05,
+      "loss": 1.0003,
+      "step": 3770
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.6893319700068168e-05,
+      "loss": 1.0151,
+      "step": 3780
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.687627811860941e-05,
+      "loss": 1.0239,
+      "step": 3790
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.685923653715065e-05,
+      "loss": 1.0336,
+      "step": 3800
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.6842194955691888e-05,
+      "loss": 0.9447,
+      "step": 3810
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.682515337423313e-05,
+      "loss": 0.9884,
+      "step": 3820
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.680811179277437e-05,
+      "loss": 1.0046,
+      "step": 3830
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.6791070211315608e-05,
+      "loss": 0.957,
+      "step": 3840
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.6774028629856852e-05,
+      "loss": 0.9851,
+      "step": 3850
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.675698704839809e-05,
+      "loss": 0.9733,
+      "step": 3860
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.6739945466939334e-05,
+      "loss": 0.9672,
+      "step": 3870
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6722903885480572e-05,
+      "loss": 1.0147,
+      "step": 3880
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6705862304021813e-05,
+      "loss": 1.0262,
+      "step": 3890
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.6688820722563054e-05,
+      "loss": 1.0566,
+      "step": 3900
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.6671779141104295e-05,
+      "loss": 1.0275,
+      "step": 3910
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.6723522506767602,
+      "eval_loss": 0.8824578523635864,
+      "eval_runtime": 309.7994,
+      "eval_samples_per_second": 134.742,
+      "eval_steps_per_second": 67.373,
+      "step": 3912
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.6654737559645536e-05,
+      "loss": 1.0538,
+      "step": 3920
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.6637695978186777e-05,
+      "loss": 0.9989,
+      "step": 3930
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.6620654396728015e-05,
+      "loss": 0.9771,
+      "step": 3940
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.660361281526926e-05,
+      "loss": 0.9699,
+      "step": 3950
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.6586571233810497e-05,
+      "loss": 1.0131,
+      "step": 3960
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.6569529652351738e-05,
+      "loss": 1.0406,
+      "step": 3970
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.655248807089298e-05,
+      "loss": 1.0048,
+      "step": 3980
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.653544648943422e-05,
+      "loss": 1.0277,
+      "step": 3990
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.651840490797546e-05,
+      "loss": 1.0329,
+      "step": 4000
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.6501363326516702e-05,
+      "loss": 0.9593,
+      "step": 4010
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.648432174505794e-05,
+      "loss": 1.0028,
+      "step": 4020
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.6467280163599184e-05,
+      "loss": 0.9851,
+      "step": 4030
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.6450238582140422e-05,
+      "loss": 0.9802,
+      "step": 4040
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.6433197000681666e-05,
+      "loss": 1.0274,
+      "step": 4050
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.6416155419222904e-05,
+      "loss": 0.9856,
+      "step": 4060
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.6399113837764145e-05,
+      "loss": 0.9629,
+      "step": 4070
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.6382072256305386e-05,
+      "loss": 0.9939,
+      "step": 4080
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.6365030674846627e-05,
+      "loss": 0.963,
+      "step": 4090
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.6347989093387865e-05,
+      "loss": 0.9996,
+      "step": 4100
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.6332651670074983e-05,
+      "loss": 0.9749,
+      "step": 4110
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.6315610088616224e-05,
+      "loss": 1.024,
+      "step": 4120
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.6298568507157465e-05,
+      "loss": 0.991,
+      "step": 4130
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.6281526925698703e-05,
+      "loss": 0.9887,
+      "step": 4140
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.6264485344239948e-05,
+      "loss": 0.9711,
+      "step": 4150
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.6247443762781185e-05,
+      "loss": 0.9961,
+      "step": 4160
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.623040218132243e-05,
+      "loss": 1.0036,
+      "step": 4170
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.6213360599863667e-05,
+      "loss": 1.0061,
+      "step": 4180
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.619631901840491e-05,
+      "loss": 0.9465,
+      "step": 4190
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.617927743694615e-05,
+      "loss": 0.9747,
+      "step": 4200
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.616223585548739e-05,
+      "loss": 0.9397,
+      "step": 4210
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.6145194274028628e-05,
+      "loss": 0.9874,
+      "step": 4220
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.6128152692569873e-05,
+      "loss": 0.9868,
+      "step": 4230
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.611111111111111e-05,
+      "loss": 0.9647,
+      "step": 4240
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.6094069529652355e-05,
+      "loss": 0.9621,
+      "step": 4250
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6077027948193592e-05,
+      "loss": 1.0602,
+      "step": 4260
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6059986366734834e-05,
+      "loss": 0.994,
+      "step": 4270
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.6042944785276075e-05,
+      "loss": 1.0052,
+      "step": 4280
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.6025903203817316e-05,
+      "loss": 0.9876,
+      "step": 4290
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.6008861622358557e-05,
+      "loss": 0.9701,
+      "step": 4300
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5991820040899798e-05,
+      "loss": 1.012,
+      "step": 4310
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5974778459441035e-05,
+      "loss": 0.9664,
+      "step": 4320
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.595773687798228e-05,
+      "loss": 1.0146,
+      "step": 4330
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.5940695296523518e-05,
+      "loss": 0.9747,
+      "step": 4340
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.592365371506476e-05,
+      "loss": 0.9283,
+      "step": 4350
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.5906612133606e-05,
+      "loss": 1.0066,
+      "step": 4360
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.588957055214724e-05,
+      "loss": 1.0028,
+      "step": 4370
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.5872528970688482e-05,
+      "loss": 0.9699,
+      "step": 4380
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.5855487389229723e-05,
+      "loss": 0.9756,
+      "step": 4390
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.583844580777096e-05,
+      "loss": 0.9688,
+      "step": 4400
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.58214042263122e-05,
+      "loss": 0.9974,
+      "step": 4410
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.5804362644853443e-05,
+      "loss": 0.9451,
+      "step": 4420
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.5787321063394684e-05,
+      "loss": 0.9801,
+      "step": 4430
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.5770279481935925e-05,
+      "loss": 0.9746,
+      "step": 4440
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.5753237900477162e-05,
+      "loss": 0.9941,
+      "step": 4450
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.5736196319018407e-05,
+      "loss": 0.995,
+      "step": 4460
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.5719154737559645e-05,
+      "loss": 0.9153,
+      "step": 4470
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.5702113156100886e-05,
+      "loss": 1.0165,
+      "step": 4480
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.5685071574642127e-05,
+      "loss": 0.9787,
+      "step": 4490
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.5668029993183368e-05,
+      "loss": 0.98,
+      "step": 4500
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.565098841172461e-05,
+      "loss": 0.9741,
+      "step": 4510
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.563394683026585e-05,
+      "loss": 0.9948,
+      "step": 4520
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.5616905248807088e-05,
+      "loss": 0.9794,
+      "step": 4530
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.5599863667348332e-05,
+      "loss": 0.9691,
+      "step": 4540
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.558282208588957e-05,
+      "loss": 0.959,
+      "step": 4550
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.556578050443081e-05,
+      "loss": 1.0421,
+      "step": 4560
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.5548738922972052e-05,
+      "loss": 0.9879,
+      "step": 4570
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.5531697341513293e-05,
+      "loss": 0.9786,
+      "step": 4580
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.5514655760054534e-05,
+      "loss": 0.9374,
+      "step": 4590
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.5497614178595775e-05,
+      "loss": 1.01,
+      "step": 4600
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.5480572597137013e-05,
+      "loss": 0.9834,
+      "step": 4610
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.5463531015678257e-05,
+      "loss": 0.9414,
+      "step": 4620
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.5446489434219495e-05,
+      "loss": 0.9533,
+      "step": 4630
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.542944785276074e-05,
+      "loss": 0.9458,
+      "step": 4640
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.5412406271301977e-05,
+      "loss": 0.9218,
+      "step": 4650
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.5395364689843218e-05,
+      "loss": 0.9742,
+      "step": 4660
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.537832310838446e-05,
+      "loss": 0.9645,
+      "step": 4670
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.53612815269257e-05,
+      "loss": 0.9541,
+      "step": 4680
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.5344239945466938e-05,
+      "loss": 0.9492,
+      "step": 4690
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.5327198364008182e-05,
+      "loss": 0.8779,
+      "step": 4700
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.531015678254942e-05,
+      "loss": 0.9436,
+      "step": 4710
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.5293115201090664e-05,
+      "loss": 0.9214,
+      "step": 4720
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.5276073619631902e-05,
+      "loss": 0.9998,
+      "step": 4730
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.5259032038173143e-05,
+      "loss": 1.0091,
+      "step": 4740
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.5241990456714384e-05,
+      "loss": 0.9827,
+      "step": 4750
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.5224948875255625e-05,
+      "loss": 0.9845,
+      "step": 4760
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.5207907293796866e-05,
+      "loss": 1.0062,
+      "step": 4770
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.5190865712338107e-05,
+      "loss": 0.9933,
+      "step": 4780
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.5173824130879345e-05,
+      "loss": 0.9468,
+      "step": 4790
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.515678254942059e-05,
+      "loss": 0.9004,
+      "step": 4800
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.5139740967961827e-05,
+      "loss": 0.9518,
+      "step": 4810
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.5122699386503068e-05,
+      "loss": 1.0116,
+      "step": 4820
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.510565780504431e-05,
+      "loss": 0.9716,
+      "step": 4830
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.508861622358555e-05,
+      "loss": 0.9479,
+      "step": 4840
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.507157464212679e-05,
+      "loss": 0.974,
+      "step": 4850
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.5054533060668033e-05,
+      "loss": 0.9598,
+      "step": 4860
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.503749147920927e-05,
+      "loss": 0.9747,
+      "step": 4870
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.502044989775051e-05,
+      "loss": 0.9206,
+      "step": 4880
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.5003408316291752e-05,
+      "loss": 0.9525,
+      "step": 4890
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.498636673483299e-05,
+      "loss": 0.9175,
+      "step": 4900
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.4969325153374234e-05,
+      "loss": 0.9487,
+      "step": 4910
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.4952283571915472e-05,
+      "loss": 0.943,
+      "step": 4920
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.4935241990456717e-05,
+      "loss": 0.9856,
+      "step": 4930
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.4918200408997954e-05,
+      "loss": 0.9749,
+      "step": 4940
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.4901158827539195e-05,
+      "loss": 0.9868,
+      "step": 4950
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.4884117246080436e-05,
+      "loss": 0.9264,
+      "step": 4960
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.4867075664621677e-05,
+      "loss": 0.9407,
+      "step": 4970
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.485003408316292e-05,
+      "loss": 0.926,
+      "step": 4980
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.483299250170416e-05,
+      "loss": 0.9907,
+      "step": 4990
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.4815950920245397e-05,
+      "loss": 0.9418,
+      "step": 5000
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.479890933878664e-05,
+      "loss": 1.0033,
+      "step": 5010
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.478186775732788e-05,
+      "loss": 1.0064,
+      "step": 5020
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.476482617586912e-05,
+      "loss": 1.0107,
+      "step": 5030
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.474778459441036e-05,
+      "loss": 0.9168,
+      "step": 5040
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.4730743012951603e-05,
+      "loss": 0.9635,
+      "step": 5050
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.4713701431492844e-05,
+      "loss": 1.0064,
+      "step": 5060
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.4696659850034085e-05,
+      "loss": 1.048,
+      "step": 5070
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.4679618268575322e-05,
+      "loss": 0.9586,
+      "step": 5080
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.4662576687116567e-05,
+      "loss": 0.9419,
+      "step": 5090
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.4645535105657804e-05,
+      "loss": 0.9623,
+      "step": 5100
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.4628493524199046e-05,
+      "loss": 0.981,
+      "step": 5110
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.4611451942740287e-05,
+      "loss": 0.9509,
+      "step": 5120
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.4594410361281528e-05,
+      "loss": 0.9246,
+      "step": 5130
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.457736877982277e-05,
+      "loss": 0.9549,
+      "step": 5140
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.456032719836401e-05,
+      "loss": 0.966,
+      "step": 5150
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.4543285616905247e-05,
+      "loss": 0.9389,
+      "step": 5160
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.4526244035446492e-05,
+      "loss": 0.9596,
+      "step": 5170
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.450920245398773e-05,
+      "loss": 0.9833,
+      "step": 5180
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.4492160872528974e-05,
+      "loss": 0.9875,
+      "step": 5190
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.4475119291070212e-05,
+      "loss": 0.9285,
+      "step": 5200
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.4458077709611453e-05,
+      "loss": 0.9982,
+      "step": 5210
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.6712742256186666,
+      "eval_loss": 0.9181127548217773,
+      "eval_runtime": 307.6067,
+      "eval_samples_per_second": 135.703,
+      "eval_steps_per_second": 67.853,
+      "step": 5216
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.4441036128152694e-05,
+      "loss": 1.0134,
+      "step": 5220
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.4423994546693935e-05,
+      "loss": 0.9497,
+      "step": 5230
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.4406952965235173e-05,
+      "loss": 0.9867,
+      "step": 5240
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.4389911383776417e-05,
+      "loss": 0.9043,
+      "step": 5250
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.4372869802317655e-05,
+      "loss": 0.9055,
+      "step": 5260
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.43558282208589e-05,
+      "loss": 0.9698,
+      "step": 5270
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.4338786639400137e-05,
+      "loss": 0.8946,
+      "step": 5280
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.4321745057941378e-05,
+      "loss": 0.9638,
+      "step": 5290
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.430470347648262e-05,
+      "loss": 0.8915,
+      "step": 5300
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.428766189502386e-05,
+      "loss": 0.9603,
+      "step": 5310
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.42706203135651e-05,
+      "loss": 0.9579,
+      "step": 5320
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.425357873210634e-05,
+      "loss": 0.9358,
+      "step": 5330
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.423653715064758e-05,
+      "loss": 0.9576,
+      "step": 5340
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.421949556918882e-05,
+      "loss": 0.9395,
+      "step": 5350
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.4202453987730062e-05,
+      "loss": 0.9557,
+      "step": 5360
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.41854124062713e-05,
+      "loss": 0.9278,
+      "step": 5370
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.4168370824812544e-05,
+      "loss": 0.971,
+      "step": 5380
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.4151329243353782e-05,
+      "loss": 0.9275,
+      "step": 5390
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.4134287661895026e-05,
+      "loss": 0.9013,
+      "step": 5400
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.4117246080436264e-05,
+      "loss": 0.8966,
+      "step": 5410
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.4100204498977505e-05,
+      "loss": 0.8906,
+      "step": 5420
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.4083162917518746e-05,
+      "loss": 0.9914,
+      "step": 5430
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.4066121336059987e-05,
+      "loss": 0.8998,
+      "step": 5440
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.4049079754601228e-05,
+      "loss": 0.9504,
+      "step": 5450
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.403203817314247e-05,
+      "loss": 0.9486,
+      "step": 5460
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.4014996591683707e-05,
+      "loss": 0.9767,
+      "step": 5470
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.399795501022495e-05,
+      "loss": 0.9343,
+      "step": 5480
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.398091342876619e-05,
+      "loss": 0.9008,
+      "step": 5490
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.396387184730743e-05,
+      "loss": 0.9998,
+      "step": 5500
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.394683026584867e-05,
+      "loss": 0.9439,
+      "step": 5510
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.3929788684389912e-05,
+      "loss": 0.9725,
+      "step": 5520
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.3912747102931153e-05,
+      "loss": 0.9589,
+      "step": 5530
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.3895705521472394e-05,
+      "loss": 0.9439,
+      "step": 5540
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.3878663940013632e-05,
+      "loss": 0.9006,
+      "step": 5550
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.3861622358554876e-05,
+      "loss": 0.9801,
+      "step": 5560
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.3844580777096114e-05,
+      "loss": 0.936,
+      "step": 5570
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.3827539195637355e-05,
+      "loss": 0.9216,
+      "step": 5580
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.3810497614178596e-05,
+      "loss": 0.9283,
+      "step": 5590
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.3793456032719837e-05,
+      "loss": 0.9001,
+      "step": 5600
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.377641445126108e-05,
+      "loss": 0.9054,
+      "step": 5610
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.375937286980232e-05,
+      "loss": 0.9597,
+      "step": 5620
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.3742331288343557e-05,
+      "loss": 0.8701,
+      "step": 5630
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.37252897068848e-05,
+      "loss": 0.9429,
+      "step": 5640
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.370824812542604e-05,
+      "loss": 0.958,
+      "step": 5650
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.3691206543967284e-05,
+      "loss": 0.9559,
+      "step": 5660
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.367416496250852e-05,
+      "loss": 0.8959,
+      "step": 5670
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.3657123381049762e-05,
+      "loss": 0.977,
+      "step": 5680
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.3640081799591003e-05,
+      "loss": 0.973,
+      "step": 5690
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.3623040218132245e-05,
+      "loss": 1.0055,
+      "step": 5700
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.3605998636673482e-05,
+      "loss": 0.8854,
+      "step": 5710
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.3588957055214727e-05,
+      "loss": 0.9261,
+      "step": 5720
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.3571915473755964e-05,
+      "loss": 0.9147,
+      "step": 5730
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.355487389229721e-05,
+      "loss": 0.9169,
+      "step": 5740
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.3537832310838446e-05,
+      "loss": 0.8848,
+      "step": 5750
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.3520790729379688e-05,
+      "loss": 0.871,
+      "step": 5760
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.350374914792093e-05,
+      "loss": 0.9723,
+      "step": 5770
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.3486707566462166e-05,
+      "loss": 0.951,
+      "step": 5780
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.3469665985003407e-05,
+      "loss": 0.9554,
+      "step": 5790
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.345262440354465e-05,
+      "loss": 0.9429,
+      "step": 5800
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.343558282208589e-05,
+      "loss": 0.9401,
+      "step": 5810
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.341854124062713e-05,
+      "loss": 0.9516,
+      "step": 5820
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.340149965916837e-05,
+      "loss": 0.9745,
+      "step": 5830
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.338445807770961e-05,
+      "loss": 0.9055,
+      "step": 5840
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.3367416496250854e-05,
+      "loss": 0.9409,
+      "step": 5850
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.335037491479209e-05,
+      "loss": 0.9141,
+      "step": 5860
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 0.9605,
+      "step": 5870
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.3316291751874573e-05,
+      "loss": 0.912,
+      "step": 5880
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.3299250170415815e-05,
+      "loss": 0.9628,
+      "step": 5890
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.3282208588957056e-05,
+      "loss": 0.9333,
+      "step": 5900
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.3265167007498297e-05,
+      "loss": 0.917,
+      "step": 5910
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.3248125426039534e-05,
+      "loss": 0.9308,
+      "step": 5920
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.323108384458078e-05,
+      "loss": 0.9495,
+      "step": 5930
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.3214042263122016e-05,
+      "loss": 0.9661,
+      "step": 5940
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.319700068166326e-05,
+      "loss": 0.9299,
+      "step": 5950
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.31799591002045e-05,
+      "loss": 0.8665,
+      "step": 5960
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.316291751874574e-05,
+      "loss": 0.955,
+      "step": 5970
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.314587593728698e-05,
+      "loss": 0.9194,
+      "step": 5980
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.3128834355828222e-05,
+      "loss": 0.9222,
+      "step": 5990
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.3111792774369463e-05,
+      "loss": 0.8949,
+      "step": 6000
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.3094751192910704e-05,
+      "loss": 0.8949,
+      "step": 6010
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.307770961145194e-05,
+      "loss": 0.9611,
+      "step": 6020
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.3060668029993186e-05,
+      "loss": 0.9869,
+      "step": 6030
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.3043626448534424e-05,
+      "loss": 0.9109,
+      "step": 6040
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.3026584867075665e-05,
+      "loss": 0.9174,
+      "step": 6050
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.3009543285616906e-05,
+      "loss": 0.9297,
+      "step": 6060
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.2992501704158147e-05,
+      "loss": 0.9299,
+      "step": 6070
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.2975460122699388e-05,
+      "loss": 0.9322,
+      "step": 6080
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.295841854124063e-05,
+      "loss": 0.9049,
+      "step": 6090
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.2941376959781867e-05,
+      "loss": 0.8777,
+      "step": 6100
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.292433537832311e-05,
+      "loss": 0.9345,
+      "step": 6110
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.290729379686435e-05,
+      "loss": 0.8861,
+      "step": 6120
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.289025221540559e-05,
+      "loss": 0.8881,
+      "step": 6130
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.287321063394683e-05,
+      "loss": 0.8607,
+      "step": 6140
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.2856169052488072e-05,
+      "loss": 0.9473,
+      "step": 6150
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.2839127471029313e-05,
+      "loss": 0.8965,
+      "step": 6160
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.2822085889570554e-05,
+      "loss": 0.9397,
+      "step": 6170
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.2805044308111792e-05,
+      "loss": 0.913,
+      "step": 6180
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.2788002726653036e-05,
+      "loss": 0.9166,
+      "step": 6190
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.2770961145194274e-05,
+      "loss": 0.9946,
+      "step": 6200
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.275391956373552e-05,
+      "loss": 0.9883,
+      "step": 6210
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.2736877982276756e-05,
+      "loss": 0.9176,
+      "step": 6220
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.2719836400817994e-05,
+      "loss": 0.8989,
+      "step": 6230
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.2702794819359238e-05,
+      "loss": 0.9483,
+      "step": 6240
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.2685753237900476e-05,
+      "loss": 0.884,
+      "step": 6250
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.2668711656441717e-05,
+      "loss": 0.8576,
+      "step": 6260
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.2651670074982958e-05,
+      "loss": 0.9186,
+      "step": 6270
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.26346284935242e-05,
+      "loss": 0.9656,
+      "step": 6280
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.261758691206544e-05,
+      "loss": 0.9269,
+      "step": 6290
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.260054533060668e-05,
+      "loss": 0.9007,
+      "step": 6300
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.258350374914792e-05,
+      "loss": 0.8641,
+      "step": 6310
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.2566462167689163e-05,
+      "loss": 0.9741,
+      "step": 6320
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.25494205862304e-05,
+      "loss": 0.8326,
+      "step": 6330
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.2532379004771645e-05,
+      "loss": 0.9173,
+      "step": 6340
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.2515337423312883e-05,
+      "loss": 0.9842,
+      "step": 6350
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.2498295841854124e-05,
+      "loss": 0.9539,
+      "step": 6360
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.2481254260395365e-05,
+      "loss": 0.9222,
+      "step": 6370
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.2464212678936606e-05,
+      "loss": 0.9458,
+      "step": 6380
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.2447171097477844e-05,
+      "loss": 0.9298,
+      "step": 6390
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.243012951601909e-05,
+      "loss": 0.938,
+      "step": 6400
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.2413087934560326e-05,
+      "loss": 0.9401,
+      "step": 6410
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.239604635310157e-05,
+      "loss": 0.9286,
+      "step": 6420
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.2379004771642808e-05,
+      "loss": 0.9209,
+      "step": 6430
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.236196319018405e-05,
+      "loss": 0.8586,
+      "step": 6440
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.234492160872529e-05,
+      "loss": 0.9236,
+      "step": 6450
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.232788002726653e-05,
+      "loss": 0.9389,
+      "step": 6460
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.231083844580777e-05,
+      "loss": 0.8837,
+      "step": 6470
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.2293796864349014e-05,
+      "loss": 0.8943,
+      "step": 6480
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.227675528289025e-05,
+      "loss": 0.9505,
+      "step": 6490
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.2259713701431496e-05,
+      "loss": 0.9061,
+      "step": 6500
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.2242672119972733e-05,
+      "loss": 0.9036,
+      "step": 6510
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.2225630538513974e-05,
+      "loss": 0.9482,
+      "step": 6520
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.6971228709005103,
+      "eval_loss": 0.8717342019081116,
+      "eval_runtime": 308.249,
+      "eval_samples_per_second": 135.42,
+      "eval_steps_per_second": 67.711,
+      "step": 6520
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.2208588957055215e-05,
+      "loss": 1.0116,
+      "step": 6530
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.2193251533742334e-05,
+      "loss": 0.9246,
+      "step": 6540
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.217620995228357e-05,
+      "loss": 0.9363,
+      "step": 6550
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.2159168370824813e-05,
+      "loss": 0.9852,
+      "step": 6560
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.2142126789366054e-05,
+      "loss": 0.9244,
+      "step": 6570
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.2125085207907295e-05,
+      "loss": 0.9283,
+      "step": 6580
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.2108043626448536e-05,
+      "loss": 0.8593,
+      "step": 6590
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.2091002044989777e-05,
+      "loss": 0.8738,
+      "step": 6600
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.2073960463531014e-05,
+      "loss": 0.9051,
+      "step": 6610
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.205691888207226e-05,
+      "loss": 0.9083,
+      "step": 6620
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.2039877300613497e-05,
+      "loss": 0.8397,
+      "step": 6630
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.2022835719154738e-05,
+      "loss": 0.9232,
+      "step": 6640
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.200579413769598e-05,
+      "loss": 0.8942,
+      "step": 6650
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.198875255623722e-05,
+      "loss": 0.9277,
+      "step": 6660
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.197171097477846e-05,
+      "loss": 0.935,
+      "step": 6670
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.1954669393319702e-05,
+      "loss": 0.932,
+      "step": 6680
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.193762781186094e-05,
+      "loss": 0.8927,
+      "step": 6690
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.1920586230402184e-05,
+      "loss": 0.9473,
+      "step": 6700
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.190354464894342e-05,
+      "loss": 0.9632,
+      "step": 6710
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.1886503067484666e-05,
+      "loss": 0.8969,
+      "step": 6720
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.1869461486025904e-05,
+      "loss": 0.8751,
+      "step": 6730
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.1852419904567145e-05,
+      "loss": 0.83,
+      "step": 6740
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.1835378323108386e-05,
+      "loss": 0.9196,
+      "step": 6750
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.1818336741649627e-05,
+      "loss": 0.9417,
+      "step": 6760
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.1801295160190865e-05,
+      "loss": 0.8637,
+      "step": 6770
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.178425357873211e-05,
+      "loss": 0.9238,
+      "step": 6780
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.1767211997273347e-05,
+      "loss": 0.868,
+      "step": 6790
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.175017041581459e-05,
+      "loss": 0.8751,
+      "step": 6800
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.173312883435583e-05,
+      "loss": 0.8452,
+      "step": 6810
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.1716087252897067e-05,
+      "loss": 0.9365,
+      "step": 6820
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.169904567143831e-05,
+      "loss": 0.9037,
+      "step": 6830
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.168200408997955e-05,
+      "loss": 0.8345,
+      "step": 6840
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.166496250852079e-05,
+      "loss": 0.9125,
+      "step": 6850
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.164792092706203e-05,
+      "loss": 0.868,
+      "step": 6860
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.1630879345603272e-05,
+      "loss": 0.8535,
+      "step": 6870
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.1613837764144513e-05,
+      "loss": 0.9009,
+      "step": 6880
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.1596796182685754e-05,
+      "loss": 0.9118,
+      "step": 6890
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.1579754601226992e-05,
+      "loss": 0.9099,
+      "step": 6900
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.1562713019768236e-05,
+      "loss": 0.892,
+      "step": 6910
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.1545671438309474e-05,
+      "loss": 0.9109,
+      "step": 6920
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.1528629856850718e-05,
+      "loss": 0.8643,
+      "step": 6930
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.1511588275391956e-05,
+      "loss": 0.8763,
+      "step": 6940
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.1494546693933197e-05,
+      "loss": 0.8659,
+      "step": 6950
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.1477505112474438e-05,
+      "loss": 0.9253,
+      "step": 6960
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.146046353101568e-05,
+      "loss": 0.8241,
+      "step": 6970
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.1443421949556917e-05,
+      "loss": 0.8764,
+      "step": 6980
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.142638036809816e-05,
+      "loss": 0.8899,
+      "step": 6990
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.14093387866394e-05,
+      "loss": 0.8461,
+      "step": 7000
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.1392297205180643e-05,
+      "loss": 0.8903,
+      "step": 7010
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.137525562372188e-05,
+      "loss": 0.9119,
+      "step": 7020
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.1358214042263122e-05,
+      "loss": 0.9059,
+      "step": 7030
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.1341172460804363e-05,
+      "loss": 0.9097,
+      "step": 7040
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.1324130879345604e-05,
+      "loss": 0.9419,
+      "step": 7050
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.1307089297886845e-05,
+      "loss": 0.8518,
+      "step": 7060
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.1290047716428086e-05,
+      "loss": 0.8507,
+      "step": 7070
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.1273006134969324e-05,
+      "loss": 0.8547,
+      "step": 7080
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.125596455351057e-05,
+      "loss": 0.9453,
+      "step": 7090
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.1238922972051806e-05,
+      "loss": 0.8818,
+      "step": 7100
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.1221881390593047e-05,
+      "loss": 0.8539,
+      "step": 7110
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.120483980913429e-05,
+      "loss": 0.8854,
+      "step": 7120
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.118779822767553e-05,
+      "loss": 0.8575,
+      "step": 7130
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.117075664621677e-05,
+      "loss": 0.9199,
+      "step": 7140
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.115371506475801e-05,
+      "loss": 0.9484,
+      "step": 7150
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.113667348329925e-05,
+      "loss": 0.8678,
+      "step": 7160
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.1119631901840494e-05,
+      "loss": 0.9069,
+      "step": 7170
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.110259032038173e-05,
+      "loss": 0.9627,
+      "step": 7180
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.1085548738922972e-05,
+      "loss": 0.9177,
+      "step": 7190
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.1068507157464213e-05,
+      "loss": 0.9266,
+      "step": 7200
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.1051465576005454e-05,
+      "loss": 0.8668,
+      "step": 7210
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.1034423994546696e-05,
+      "loss": 0.8637,
+      "step": 7220
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.1017382413087937e-05,
+      "loss": 0.8736,
+      "step": 7230
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.1000340831629174e-05,
+      "loss": 0.8764,
+      "step": 7240
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.098329925017042e-05,
+      "loss": 0.9237,
+      "step": 7250
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.0966257668711656e-05,
+      "loss": 0.8943,
+      "step": 7260
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.0949216087252897e-05,
+      "loss": 0.9121,
+      "step": 7270
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.093217450579414e-05,
+      "loss": 0.8388,
+      "step": 7280
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.0915132924335376e-05,
+      "loss": 0.8678,
+      "step": 7290
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.089809134287662e-05,
+      "loss": 0.8809,
+      "step": 7300
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.088104976141786e-05,
+      "loss": 0.9125,
+      "step": 7310
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.08640081799591e-05,
+      "loss": 0.9039,
+      "step": 7320
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.084696659850034e-05,
+      "loss": 0.913,
+      "step": 7330
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.082992501704158e-05,
+      "loss": 0.8653,
+      "step": 7340
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.0812883435582823e-05,
+      "loss": 0.8841,
+      "step": 7350
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.0795841854124064e-05,
+      "loss": 0.8931,
+      "step": 7360
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.07788002726653e-05,
+      "loss": 0.923,
+      "step": 7370
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.0761758691206546e-05,
+      "loss": 0.9512,
+      "step": 7380
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.0744717109747783e-05,
+      "loss": 0.9233,
+      "step": 7390
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.0727675528289028e-05,
+      "loss": 0.8609,
+      "step": 7400
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.0710633946830266e-05,
+      "loss": 0.9083,
+      "step": 7410
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.0693592365371507e-05,
+      "loss": 0.9157,
+      "step": 7420
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.0676550783912748e-05,
+      "loss": 0.8578,
+      "step": 7430
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.065950920245399e-05,
+      "loss": 0.9037,
+      "step": 7440
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.0642467620995226e-05,
+      "loss": 0.8742,
+      "step": 7450
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.062542603953647e-05,
+      "loss": 0.8667,
+      "step": 7460
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.060838445807771e-05,
+      "loss": 0.957,
+      "step": 7470
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.0591342876618953e-05,
+      "loss": 0.8937,
+      "step": 7480
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.057430129516019e-05,
+      "loss": 0.9001,
+      "step": 7490
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.0557259713701432e-05,
+      "loss": 0.8399,
+      "step": 7500
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.0540218132242673e-05,
+      "loss": 0.8656,
+      "step": 7510
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.0523176550783914e-05,
+      "loss": 0.8681,
+      "step": 7520
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.050613496932515e-05,
+      "loss": 0.8925,
+      "step": 7530
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.0489093387866396e-05,
+      "loss": 0.9304,
+      "step": 7540
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.0472051806407634e-05,
+      "loss": 0.9501,
+      "step": 7550
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.0455010224948878e-05,
+      "loss": 0.8909,
+      "step": 7560
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.0437968643490116e-05,
+      "loss": 0.9035,
+      "step": 7570
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.0420927062031357e-05,
+      "loss": 0.9192,
+      "step": 7580
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.0403885480572598e-05,
+      "loss": 0.8084,
+      "step": 7590
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.038684389911384e-05,
+      "loss": 0.8796,
+      "step": 7600
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.036980231765508e-05,
+      "loss": 0.9236,
+      "step": 7610
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.035276073619632e-05,
+      "loss": 0.8263,
+      "step": 7620
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.033571915473756e-05,
+      "loss": 0.9037,
+      "step": 7630
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.0318677573278803e-05,
+      "loss": 0.8947,
+      "step": 7640
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.030163599182004e-05,
+      "loss": 0.8799,
+      "step": 7650
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.0284594410361282e-05,
+      "loss": 0.8887,
+      "step": 7660
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.0267552828902523e-05,
+      "loss": 0.8198,
+      "step": 7670
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.0250511247443764e-05,
+      "loss": 0.9119,
+      "step": 7680
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.0233469665985005e-05,
+      "loss": 0.8808,
+      "step": 7690
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.0216428084526246e-05,
+      "loss": 0.9084,
+      "step": 7700
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.0199386503067484e-05,
+      "loss": 0.8418,
+      "step": 7710
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.018234492160873e-05,
+      "loss": 0.8738,
+      "step": 7720
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.0165303340149966e-05,
+      "loss": 0.8637,
+      "step": 7730
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.0148261758691207e-05,
+      "loss": 0.9228,
+      "step": 7740
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.0131220177232448e-05,
+      "loss": 0.8914,
+      "step": 7750
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.0114178595773686e-05,
+      "loss": 0.8815,
+      "step": 7760
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.009713701431493e-05,
+      "loss": 0.9084,
+      "step": 7770
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.0080095432856168e-05,
+      "loss": 0.8887,
+      "step": 7780
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.006305385139741e-05,
+      "loss": 0.9019,
+      "step": 7790
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.004601226993865e-05,
+      "loss": 0.8782,
+      "step": 7800
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.002897068847989e-05,
+      "loss": 0.8873,
+      "step": 7810
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.0011929107021132e-05,
+      "loss": 0.8687,
+      "step": 7820
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.7163835852717821,
+      "eval_loss": 0.8041061162948608,
+      "eval_runtime": 304.7077,
+      "eval_samples_per_second": 136.994,
+      "eval_steps_per_second": 68.498,
+      "step": 7824
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.9994887525562373e-05,
+      "loss": 0.9292,
+      "step": 7830
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.997784594410361e-05,
+      "loss": 0.8409,
+      "step": 7840
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.9960804362644855e-05,
+      "loss": 0.8915,
+      "step": 7850
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.9943762781186093e-05,
+      "loss": 0.9016,
+      "step": 7860
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.9926721199727334e-05,
+      "loss": 0.9252,
+      "step": 7870
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.9909679618268575e-05,
+      "loss": 0.8744,
+      "step": 7880
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.9892638036809816e-05,
+      "loss": 0.8928,
+      "step": 7890
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.9875596455351057e-05,
+      "loss": 0.9114,
+      "step": 7900
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.98585548738923e-05,
+      "loss": 0.8299,
+      "step": 7910
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.9841513292433536e-05,
+      "loss": 0.9184,
+      "step": 7920
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.982447171097478e-05,
+      "loss": 0.9793,
+      "step": 7930
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.9807430129516018e-05,
+      "loss": 0.8285,
+      "step": 7940
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.9790388548057263e-05,
+      "loss": 0.8312,
+      "step": 7950
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.97733469665985e-05,
+      "loss": 0.825,
+      "step": 7960
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.975630538513974e-05,
+      "loss": 0.8562,
+      "step": 7970
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.9739263803680982e-05,
+      "loss": 0.8231,
+      "step": 7980
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.9722222222222224e-05,
+      "loss": 0.8703,
+      "step": 7990
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.970518064076346e-05,
+      "loss": 0.894,
+      "step": 8000
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.9688139059304706e-05,
+      "loss": 0.8525,
+      "step": 8010
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.9671097477845943e-05,
+      "loss": 0.8741,
+      "step": 8020
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.9654055896387188e-05,
+      "loss": 0.9095,
+      "step": 8030
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.9637014314928425e-05,
+      "loss": 0.9502,
+      "step": 8040
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.9619972733469666e-05,
+      "loss": 0.8367,
+      "step": 8050
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.9602931152010908e-05,
+      "loss": 0.8978,
+      "step": 8060
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.958588957055215e-05,
+      "loss": 0.9311,
+      "step": 8070
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.956884798909339e-05,
+      "loss": 0.8711,
+      "step": 8080
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.955180640763463e-05,
+      "loss": 0.8664,
+      "step": 8090
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.953476482617587e-05,
+      "loss": 0.888,
+      "step": 8100
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.9517723244717113e-05,
+      "loss": 0.8415,
+      "step": 8110
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.950068166325835e-05,
+      "loss": 0.9107,
+      "step": 8120
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.948364008179959e-05,
+      "loss": 0.8921,
+      "step": 8130
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.9466598500340833e-05,
+      "loss": 0.8661,
+      "step": 8140
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.9449556918882074e-05,
+      "loss": 0.8543,
+      "step": 8150
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.9432515337423315e-05,
+      "loss": 0.8428,
+      "step": 8160
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.9415473755964556e-05,
+      "loss": 0.8853,
+      "step": 8170
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.9398432174505794e-05,
+      "loss": 0.8344,
+      "step": 8180
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.9381390593047035e-05,
+      "loss": 0.8504,
+      "step": 8190
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.9364349011588276e-05,
+      "loss": 0.8673,
+      "step": 8200
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.9347307430129513e-05,
+      "loss": 0.9013,
+      "step": 8210
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.9330265848670758e-05,
+      "loss": 0.8323,
+      "step": 8220
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.9313224267211995e-05,
+      "loss": 0.9234,
+      "step": 8230
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.929618268575324e-05,
+      "loss": 0.8636,
+      "step": 8240
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.9279141104294478e-05,
+      "loss": 0.8638,
+      "step": 8250
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.926209952283572e-05,
+      "loss": 0.8078,
+      "step": 8260
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.924505794137696e-05,
+      "loss": 0.9014,
+      "step": 8270
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.92280163599182e-05,
+      "loss": 0.8429,
+      "step": 8280
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.9210974778459442e-05,
+      "loss": 0.808,
+      "step": 8290
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.9193933197000683e-05,
+      "loss": 0.8893,
+      "step": 8300
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.917689161554192e-05,
+      "loss": 0.8875,
+      "step": 8310
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.9159850034083165e-05,
+      "loss": 0.8907,
+      "step": 8320
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.9142808452624403e-05,
+      "loss": 0.871,
+      "step": 8330
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.9125766871165644e-05,
+      "loss": 0.8514,
+      "step": 8340
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.9108725289706885e-05,
+      "loss": 0.8627,
+      "step": 8350
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.9091683708248126e-05,
+      "loss": 0.8872,
+      "step": 8360
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.9074642126789367e-05,
+      "loss": 0.8402,
+      "step": 8370
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.9057600545330608e-05,
+      "loss": 0.7844,
+      "step": 8380
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.9040558963871846e-05,
+      "loss": 0.9087,
+      "step": 8390
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.902351738241309e-05,
+      "loss": 0.8536,
+      "step": 8400
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.9006475800954328e-05,
+      "loss": 0.906,
+      "step": 8410
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.898943421949557e-05,
+      "loss": 0.8363,
+      "step": 8420
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.897239263803681e-05,
+      "loss": 0.8959,
+      "step": 8430
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.895535105657805e-05,
+      "loss": 0.8767,
+      "step": 8440
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.8938309475119292e-05,
+      "loss": 0.8679,
+      "step": 8450
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.8921267893660533e-05,
+      "loss": 0.8454,
+      "step": 8460
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.890422631220177e-05,
+      "loss": 0.9341,
+      "step": 8470
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.8887184730743015e-05,
+      "loss": 0.847,
+      "step": 8480
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.8870143149284253e-05,
+      "loss": 0.8328,
+      "step": 8490
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.8853101567825497e-05,
+      "loss": 0.848,
+      "step": 8500
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.8836059986366735e-05,
+      "loss": 0.7977,
+      "step": 8510
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.8819018404907976e-05,
+      "loss": 0.8293,
+      "step": 8520
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.8801976823449217e-05,
+      "loss": 0.8847,
+      "step": 8530
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.8784935241990458e-05,
+      "loss": 0.8483,
+      "step": 8540
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.8767893660531696e-05,
+      "loss": 0.8853,
+      "step": 8550
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.875085207907294e-05,
+      "loss": 0.8406,
+      "step": 8560
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.8733810497614178e-05,
+      "loss": 0.8828,
+      "step": 8570
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.8716768916155422e-05,
+      "loss": 0.9019,
+      "step": 8580
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.869972733469666e-05,
+      "loss": 0.8747,
+      "step": 8590
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.86826857532379e-05,
+      "loss": 0.9091,
+      "step": 8600
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.8665644171779142e-05,
+      "loss": 0.9015,
+      "step": 8610
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.8648602590320383e-05,
+      "loss": 0.8597,
+      "step": 8620
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.8631561008861624e-05,
+      "loss": 0.8499,
+      "step": 8630
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.8614519427402862e-05,
+      "loss": 0.8363,
+      "step": 8640
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.8597477845944103e-05,
+      "loss": 0.817,
+      "step": 8650
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.8580436264485344e-05,
+      "loss": 0.8623,
+      "step": 8660
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.8563394683026585e-05,
+      "loss": 0.8424,
+      "step": 8670
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.8546353101567823e-05,
+      "loss": 0.8686,
+      "step": 8680
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.8529311520109067e-05,
+      "loss": 0.8855,
+      "step": 8690
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.8512269938650305e-05,
+      "loss": 0.8077,
+      "step": 8700
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 1.849522835719155e-05,
+      "loss": 0.8936,
+      "step": 8710
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 1.8478186775732787e-05,
+      "loss": 0.8479,
+      "step": 8720
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 1.8461145194274028e-05,
+      "loss": 0.8263,
+      "step": 8730
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 1.844410361281527e-05,
+      "loss": 0.9022,
+      "step": 8740
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 1.842706203135651e-05,
+      "loss": 0.8775,
+      "step": 8750
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 1.841002044989775e-05,
+      "loss": 0.888,
+      "step": 8760
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 1.8392978868438993e-05,
+      "loss": 0.8565,
+      "step": 8770
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 1.837593728698023e-05,
+      "loss": 0.8894,
+      "step": 8780
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 1.8358895705521475e-05,
+      "loss": 0.8077,
+      "step": 8790
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 1.8341854124062712e-05,
+      "loss": 0.8819,
+      "step": 8800
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 1.8324812542603953e-05,
+      "loss": 0.852,
+      "step": 8810
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 1.8307770961145194e-05,
+      "loss": 0.8538,
+      "step": 8820
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 1.8290729379686436e-05,
+      "loss": 0.8332,
+      "step": 8830
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 1.8273687798227677e-05,
+      "loss": 0.9331,
+      "step": 8840
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 1.8256646216768918e-05,
+      "loss": 0.8948,
+      "step": 8850
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 1.8239604635310155e-05,
+      "loss": 0.8291,
+      "step": 8860
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 1.82225630538514e-05,
+      "loss": 0.8053,
+      "step": 8870
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 1.8205521472392637e-05,
+      "loss": 0.922,
+      "step": 8880
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 1.818847989093388e-05,
+      "loss": 0.8618,
+      "step": 8890
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 1.817143830947512e-05,
+      "loss": 0.8584,
+      "step": 8900
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 1.815439672801636e-05,
+      "loss": 0.8537,
+      "step": 8910
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 1.81373551465576e-05,
+      "loss": 0.863,
+      "step": 8920
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 1.8120313565098843e-05,
+      "loss": 0.8412,
+      "step": 8930
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 1.810327198364008e-05,
+      "loss": 0.8383,
+      "step": 8940
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 1.8086230402181325e-05,
+      "loss": 0.8601,
+      "step": 8950
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 1.8069188820722563e-05,
+      "loss": 0.8295,
+      "step": 8960
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 1.8052147239263807e-05,
+      "loss": 0.8313,
+      "step": 8970
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 1.8035105657805045e-05,
+      "loss": 0.8541,
+      "step": 8980
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 1.8018064076346286e-05,
+      "loss": 0.8496,
+      "step": 8990
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 1.8001022494887527e-05,
+      "loss": 0.8761,
+      "step": 9000
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 1.7983980913428768e-05,
+      "loss": 0.8615,
+      "step": 9010
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 1.7966939331970006e-05,
+      "loss": 0.8919,
+      "step": 9020
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 1.794989775051125e-05,
+      "loss": 0.8309,
+      "step": 9030
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 1.7932856169052488e-05,
+      "loss": 0.7937,
+      "step": 9040
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 1.7915814587593732e-05,
+      "loss": 0.8417,
+      "step": 9050
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 1.789877300613497e-05,
+      "loss": 0.8701,
+      "step": 9060
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 1.788173142467621e-05,
+      "loss": 0.874,
+      "step": 9070
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 1.7864689843217452e-05,
+      "loss": 0.8842,
+      "step": 9080
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 1.784764826175869e-05,
+      "loss": 0.8768,
+      "step": 9090
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 1.783060668029993e-05,
+      "loss": 0.8806,
+      "step": 9100
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 1.7813565098841172e-05,
+      "loss": 0.8069,
+      "step": 9110
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 1.7796523517382413e-05,
+      "loss": 0.8841,
+      "step": 9120
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7034472845746592,
+      "eval_loss": 0.8869236707687378,
+      "eval_runtime": 310.0381,
+      "eval_samples_per_second": 134.638,
+      "eval_steps_per_second": 67.321,
+      "step": 9128
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 1.7779481935923654e-05,
+      "loss": 0.8652,
+      "step": 9130
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 1.7762440354464895e-05,
+      "loss": 0.8653,
+      "step": 9140
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 1.7745398773006133e-05,
+      "loss": 0.8367,
+      "step": 9150
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 1.7728357191547377e-05,
+      "loss": 0.8573,
+      "step": 9160
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 1.7711315610088615e-05,
+      "loss": 0.8888,
+      "step": 9170
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 1.769427402862986e-05,
+      "loss": 0.8236,
+      "step": 9180
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 1.7677232447171097e-05,
+      "loss": 0.8333,
+      "step": 9190
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 1.7660190865712338e-05,
+      "loss": 0.8048,
+      "step": 9200
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 1.764314928425358e-05,
+      "loss": 0.8088,
+      "step": 9210
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 1.762610770279482e-05,
+      "loss": 0.8372,
+      "step": 9220
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 1.7609066121336058e-05,
+      "loss": 0.8931,
+      "step": 9230
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 1.7592024539877302e-05,
+      "loss": 0.8734,
+      "step": 9240
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 1.757498295841854e-05,
+      "loss": 0.8339,
+      "step": 9250
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 1.7557941376959784e-05,
+      "loss": 0.8066,
+      "step": 9260
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 1.7540899795501022e-05,
+      "loss": 0.8337,
+      "step": 9270
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 1.7523858214042263e-05,
+      "loss": 0.8617,
+      "step": 9280
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 1.7506816632583504e-05,
+      "loss": 0.814,
+      "step": 9290
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 1.7489775051124745e-05,
+      "loss": 0.8013,
+      "step": 9300
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 1.7472733469665986e-05,
+      "loss": 0.8822,
+      "step": 9310
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 1.7455691888207227e-05,
+      "loss": 0.8639,
+      "step": 9320
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 1.7438650306748465e-05,
+      "loss": 0.836,
+      "step": 9330
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 1.742160872528971e-05,
+      "loss": 0.825,
+      "step": 9340
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 1.7404567143830947e-05,
+      "loss": 0.8309,
+      "step": 9350
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 1.7387525562372188e-05,
+      "loss": 0.891,
+      "step": 9360
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 1.737048398091343e-05,
+      "loss": 0.8418,
+      "step": 9370
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 1.735344239945467e-05,
+      "loss": 0.7954,
+      "step": 9380
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 1.733640081799591e-05,
+      "loss": 0.8647,
+      "step": 9390
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 1.7319359236537152e-05,
+      "loss": 0.8526,
+      "step": 9400
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 1.730231765507839e-05,
+      "loss": 0.7903,
+      "step": 9410
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 1.7285276073619635e-05,
+      "loss": 0.8643,
+      "step": 9420
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 1.7268234492160872e-05,
+      "loss": 0.8957,
+      "step": 9430
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 1.7251192910702113e-05,
+      "loss": 0.8942,
+      "step": 9440
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 1.7234151329243354e-05,
+      "loss": 0.8732,
+      "step": 9450
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 1.7217109747784595e-05,
+      "loss": 0.8563,
+      "step": 9460
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 1.7200068166325836e-05,
+      "loss": 0.778,
+      "step": 9470
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 1.7183026584867077e-05,
+      "loss": 0.7973,
+      "step": 9480
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 1.7165985003408315e-05,
+      "loss": 0.9155,
+      "step": 9490
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 1.714894342194956e-05,
+      "loss": 0.8396,
+      "step": 9500
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 1.7131901840490797e-05,
+      "loss": 0.8294,
+      "step": 9510
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 1.7114860259032042e-05,
+      "loss": 0.8643,
+      "step": 9520
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 1.709781867757328e-05,
+      "loss": 0.8353,
+      "step": 9530
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 1.7080777096114517e-05,
+      "loss": 0.9013,
+      "step": 9540
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 1.706373551465576e-05,
+      "loss": 0.8018,
+      "step": 9550
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 1.7046693933197e-05,
+      "loss": 0.8817,
+      "step": 9560
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 1.702965235173824e-05,
+      "loss": 0.7889,
+      "step": 9570
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 1.701261077027948e-05,
+      "loss": 0.8447,
+      "step": 9580
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 1.6995569188820722e-05,
+      "loss": 0.8404,
+      "step": 9590
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 1.6978527607361963e-05,
+      "loss": 0.8013,
+      "step": 9600
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 1.6961486025903205e-05,
+      "loss": 0.8675,
+      "step": 9610
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 1.6944444444444442e-05,
+      "loss": 0.832,
+      "step": 9620
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 1.6927402862985687e-05,
+      "loss": 0.9056,
+      "step": 9630
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 1.6910361281526924e-05,
+      "loss": 0.8293,
+      "step": 9640
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 1.689331970006817e-05,
+      "loss": 0.8508,
+      "step": 9650
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 1.6876278118609406e-05,
+      "loss": 0.8022,
+      "step": 9660
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 1.6859236537150648e-05,
+      "loss": 0.8738,
+      "step": 9670
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 1.684219495569189e-05,
+      "loss": 0.8502,
+      "step": 9680
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 1.682515337423313e-05,
+      "loss": 0.8043,
+      "step": 9690
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 1.6808111792774367e-05,
+      "loss": 0.8405,
+      "step": 9700
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 1.6791070211315612e-05,
+      "loss": 0.8143,
+      "step": 9710
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 1.677402862985685e-05,
+      "loss": 0.8702,
+      "step": 9720
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 1.6756987048398094e-05,
+      "loss": 0.8336,
+      "step": 9730
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 1.673994546693933e-05,
+      "loss": 0.8358,
+      "step": 9740
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 1.6722903885480573e-05,
+      "loss": 0.885,
+      "step": 9750
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 1.6705862304021814e-05,
+      "loss": 0.8908,
+      "step": 9760
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 1.6688820722563055e-05,
+      "loss": 0.8493,
+      "step": 9770
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 1.6671779141104296e-05,
+      "loss": 0.8316,
+      "step": 9780
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 1.6654737559645537e-05,
+      "loss": 0.8101,
+      "step": 9790
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 1.6637695978186775e-05,
+      "loss": 0.865,
+      "step": 9800
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 1.662065439672802e-05,
+      "loss": 0.8503,
+      "step": 9810
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 1.6603612815269257e-05,
+      "loss": 0.8982,
+      "step": 9820
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 1.6586571233810498e-05,
+      "loss": 0.8388,
+      "step": 9830
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 1.656952965235174e-05,
+      "loss": 0.8443,
+      "step": 9840
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 1.655248807089298e-05,
+      "loss": 0.7976,
+      "step": 9850
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 1.653544648943422e-05,
+      "loss": 0.9361,
+      "step": 9860
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 1.6518404907975462e-05,
+      "loss": 0.8517,
+      "step": 9870
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 1.65013633265167e-05,
+      "loss": 0.8244,
+      "step": 9880
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 1.6484321745057944e-05,
+      "loss": 0.8545,
+      "step": 9890
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 1.6467280163599182e-05,
+      "loss": 0.8412,
+      "step": 9900
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 1.6450238582140423e-05,
+      "loss": 0.8255,
+      "step": 9910
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 1.6433197000681664e-05,
+      "loss": 0.7694,
+      "step": 9920
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 1.6417859577368782e-05,
+      "loss": 0.8121,
+      "step": 9930
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 1.640081799591002e-05,
+      "loss": 0.9003,
+      "step": 9940
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 1.638377641445126e-05,
+      "loss": 0.8674,
+      "step": 9950
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 1.6366734832992502e-05,
+      "loss": 0.8598,
+      "step": 9960
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 1.6349693251533743e-05,
+      "loss": 0.9051,
+      "step": 9970
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 1.6332651670074984e-05,
+      "loss": 0.9068,
+      "step": 9980
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 1.6315610088616225e-05,
+      "loss": 0.8546,
+      "step": 9990
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 1.6298568507157463e-05,
+      "loss": 0.8834,
+      "step": 10000
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 1.6281526925698707e-05,
+      "loss": 0.857,
+      "step": 10010
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 1.6264485344239945e-05,
+      "loss": 0.8379,
+      "step": 10020
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 1.624744376278119e-05,
+      "loss": 0.7932,
+      "step": 10030
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 1.6230402181322427e-05,
+      "loss": 0.8767,
+      "step": 10040
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 1.6213360599863668e-05,
+      "loss": 0.8097,
+      "step": 10050
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 1.619631901840491e-05,
+      "loss": 0.8676,
+      "step": 10060
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 1.617927743694615e-05,
+      "loss": 0.8694,
+      "step": 10070
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 1.6162235855487388e-05,
+      "loss": 0.8292,
+      "step": 10080
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 1.6145194274028632e-05,
+      "loss": 0.8382,
+      "step": 10090
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 1.612815269256987e-05,
+      "loss": 0.8226,
+      "step": 10100
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 0.8172,
+      "step": 10110
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 1.6094069529652352e-05,
+      "loss": 0.822,
+      "step": 10120
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 1.607702794819359e-05,
+      "loss": 0.8366,
+      "step": 10130
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 1.6059986366734834e-05,
+      "loss": 0.8318,
+      "step": 10140
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 1.6042944785276072e-05,
+      "loss": 0.7786,
+      "step": 10150
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 1.6025903203817313e-05,
+      "loss": 0.8001,
+      "step": 10160
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 1.6008861622358554e-05,
+      "loss": 0.8224,
+      "step": 10170
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 1.5991820040899795e-05,
+      "loss": 0.8692,
+      "step": 10180
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 1.5974778459441036e-05,
+      "loss": 0.8834,
+      "step": 10190
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 1.5957736877982277e-05,
+      "loss": 0.8262,
+      "step": 10200
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 1.5940695296523515e-05,
+      "loss": 0.8306,
+      "step": 10210
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 1.592365371506476e-05,
+      "loss": 0.807,
+      "step": 10220
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 1.5906612133605997e-05,
+      "loss": 0.838,
+      "step": 10230
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 1.588957055214724e-05,
+      "loss": 0.8909,
+      "step": 10240
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 1.587252897068848e-05,
+      "loss": 0.8102,
+      "step": 10250
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 1.585548738922972e-05,
+      "loss": 0.881,
+      "step": 10260
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 1.583844580777096e-05,
+      "loss": 0.8135,
+      "step": 10270
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 1.5821404226312202e-05,
+      "loss": 0.8451,
+      "step": 10280
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 1.580436264485344e-05,
+      "loss": 0.7812,
+      "step": 10290
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 1.5787321063394685e-05,
+      "loss": 0.8435,
+      "step": 10300
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 1.5770279481935922e-05,
+      "loss": 0.8345,
+      "step": 10310
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 1.5753237900477167e-05,
+      "loss": 0.794,
+      "step": 10320
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 1.5736196319018404e-05,
+      "loss": 0.8025,
+      "step": 10330
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 1.5719154737559645e-05,
+      "loss": 0.8473,
+      "step": 10340
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 1.5702113156100887e-05,
+      "loss": 0.8414,
+      "step": 10350
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 1.5685071574642128e-05,
+      "loss": 0.8643,
+      "step": 10360
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 1.566802999318337e-05,
+      "loss": 0.844,
+      "step": 10370
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 1.565098841172461e-05,
+      "loss": 0.8945,
+      "step": 10380
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 1.5633946830265847e-05,
+      "loss": 0.8955,
+      "step": 10390
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 1.5616905248807092e-05,
+      "loss": 0.8486,
+      "step": 10400
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 1.559986366734833e-05,
+      "loss": 0.7623,
+      "step": 10410
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 1.558282208588957e-05,
+      "loss": 0.8219,
+      "step": 10420
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 1.556578050443081e-05,
+      "loss": 0.8094,
+      "step": 10430
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7171980930934528,
+      "eval_loss": 0.8215836882591248,
+      "eval_runtime": 308.4557,
+      "eval_samples_per_second": 135.329,
+      "eval_steps_per_second": 67.666,
+      "step": 10432
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.5548738922972053e-05,
+      "loss": 0.8696,
+      "step": 10440
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.5531697341513294e-05,
+      "loss": 0.7966,
+      "step": 10450
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.5514655760054535e-05,
+      "loss": 0.8139,
+      "step": 10460
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.5497614178595773e-05,
+      "loss": 0.8474,
+      "step": 10470
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.5480572597137017e-05,
+      "loss": 0.8876,
+      "step": 10480
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.5463531015678255e-05,
+      "loss": 0.8357,
+      "step": 10490
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.5446489434219496e-05,
+      "loss": 0.81,
+      "step": 10500
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.5429447852760737e-05,
+      "loss": 0.8274,
+      "step": 10510
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.5412406271301978e-05,
+      "loss": 0.865,
+      "step": 10520
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.539536468984322e-05,
+      "loss": 0.8336,
+      "step": 10530
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.537832310838446e-05,
+      "loss": 0.8015,
+      "step": 10540
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.5361281526925698e-05,
+      "loss": 0.8096,
+      "step": 10550
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.5344239945466942e-05,
+      "loss": 0.7821,
+      "step": 10560
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.532719836400818e-05,
+      "loss": 0.801,
+      "step": 10570
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.5310156782549424e-05,
+      "loss": 0.8175,
+      "step": 10580
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.5293115201090662e-05,
+      "loss": 0.7711,
+      "step": 10590
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.52760736196319e-05,
+      "loss": 0.801,
+      "step": 10600
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.5259032038173144e-05,
+      "loss": 0.8223,
+      "step": 10610
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.5241990456714382e-05,
+      "loss": 0.7755,
+      "step": 10620
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.5224948875255624e-05,
+      "loss": 0.7749,
+      "step": 10630
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.5207907293796864e-05,
+      "loss": 0.8588,
+      "step": 10640
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.5190865712338105e-05,
+      "loss": 0.8567,
+      "step": 10650
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.5173824130879344e-05,
+      "loss": 0.8156,
+      "step": 10660
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.5156782549420587e-05,
+      "loss": 0.8569,
+      "step": 10670
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.5139740967961826e-05,
+      "loss": 0.7925,
+      "step": 10680
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.5122699386503067e-05,
+      "loss": 0.8095,
+      "step": 10690
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.5105657805044307e-05,
+      "loss": 0.7862,
+      "step": 10700
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.508861622358555e-05,
+      "loss": 0.8625,
+      "step": 10710
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.5071574642126789e-05,
+      "loss": 0.8346,
+      "step": 10720
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.5054533060668032e-05,
+      "loss": 0.8157,
+      "step": 10730
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.503749147920927e-05,
+      "loss": 0.8436,
+      "step": 10740
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.5020449897750512e-05,
+      "loss": 0.8282,
+      "step": 10750
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.5003408316291751e-05,
+      "loss": 0.7819,
+      "step": 10760
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.4986366734832994e-05,
+      "loss": 0.849,
+      "step": 10770
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.4969325153374234e-05,
+      "loss": 0.8311,
+      "step": 10780
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.4952283571915475e-05,
+      "loss": 0.7758,
+      "step": 10790
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.4935241990456716e-05,
+      "loss": 0.8569,
+      "step": 10800
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.4918200408997955e-05,
+      "loss": 0.8759,
+      "step": 10810
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.4901158827539194e-05,
+      "loss": 0.7901,
+      "step": 10820
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.4884117246080436e-05,
+      "loss": 0.8448,
+      "step": 10830
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.4867075664621677e-05,
+      "loss": 0.825,
+      "step": 10840
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.4850034083162918e-05,
+      "loss": 0.8034,
+      "step": 10850
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.4832992501704157e-05,
+      "loss": 0.8153,
+      "step": 10860
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.4815950920245398e-05,
+      "loss": 0.8164,
+      "step": 10870
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.479890933878664e-05,
+      "loss": 0.8278,
+      "step": 10880
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.478186775732788e-05,
+      "loss": 0.7876,
+      "step": 10890
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.4764826175869121e-05,
+      "loss": 0.7893,
+      "step": 10900
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.474778459441036e-05,
+      "loss": 0.8114,
+      "step": 10910
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.4730743012951602e-05,
+      "loss": 0.8695,
+      "step": 10920
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.4713701431492843e-05,
+      "loss": 0.8058,
+      "step": 10930
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.4696659850034084e-05,
+      "loss": 0.8294,
+      "step": 10940
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.4679618268575323e-05,
+      "loss": 0.7996,
+      "step": 10950
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.4662576687116564e-05,
+      "loss": 0.8512,
+      "step": 10960
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.4645535105657805e-05,
+      "loss": 0.7914,
+      "step": 10970
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.4628493524199046e-05,
+      "loss": 0.8309,
+      "step": 10980
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.4611451942740286e-05,
+      "loss": 0.779,
+      "step": 10990
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.4594410361281527e-05,
+      "loss": 0.7786,
+      "step": 11000
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.4577368779822768e-05,
+      "loss": 0.7715,
+      "step": 11010
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.4560327198364009e-05,
+      "loss": 0.8158,
+      "step": 11020
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.4543285616905248e-05,
+      "loss": 0.8501,
+      "step": 11030
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.452624403544649e-05,
+      "loss": 0.8203,
+      "step": 11040
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.450920245398773e-05,
+      "loss": 0.832,
+      "step": 11050
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.4492160872528972e-05,
+      "loss": 0.8753,
+      "step": 11060
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.4475119291070213e-05,
+      "loss": 0.8788,
+      "step": 11070
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.4458077709611452e-05,
+      "loss": 0.831,
+      "step": 11080
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.4441036128152693e-05,
+      "loss": 0.8337,
+      "step": 11090
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4423994546693934e-05,
+      "loss": 0.8519,
+      "step": 11100
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4406952965235175e-05,
+      "loss": 0.8776,
+      "step": 11110
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.4389911383776415e-05,
+      "loss": 0.86,
+      "step": 11120
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.4372869802317656e-05,
+      "loss": 0.7805,
+      "step": 11130
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.4355828220858897e-05,
+      "loss": 0.8655,
+      "step": 11140
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.4338786639400138e-05,
+      "loss": 0.8638,
+      "step": 11150
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4321745057941377e-05,
+      "loss": 0.8578,
+      "step": 11160
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4304703476482618e-05,
+      "loss": 0.7936,
+      "step": 11170
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.428766189502386e-05,
+      "loss": 0.8029,
+      "step": 11180
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.42706203135651e-05,
+      "loss": 0.8411,
+      "step": 11190
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.425357873210634e-05,
+      "loss": 0.8382,
+      "step": 11200
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.423653715064758e-05,
+      "loss": 0.7663,
+      "step": 11210
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.4219495569188822e-05,
+      "loss": 0.8497,
+      "step": 11220
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.4202453987730063e-05,
+      "loss": 0.7816,
+      "step": 11230
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.4185412406271304e-05,
+      "loss": 0.8075,
+      "step": 11240
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.4168370824812543e-05,
+      "loss": 0.8626,
+      "step": 11250
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.4151329243353784e-05,
+      "loss": 0.8307,
+      "step": 11260
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.4134287661895024e-05,
+      "loss": 0.8287,
+      "step": 11270
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.4117246080436265e-05,
+      "loss": 0.8031,
+      "step": 11280
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.4100204498977504e-05,
+      "loss": 0.8382,
+      "step": 11290
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.4083162917518745e-05,
+      "loss": 0.8148,
+      "step": 11300
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.4066121336059986e-05,
+      "loss": 0.8323,
+      "step": 11310
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.4049079754601227e-05,
+      "loss": 0.8302,
+      "step": 11320
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.4032038173142467e-05,
+      "loss": 0.8288,
+      "step": 11330
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.4014996591683708e-05,
+      "loss": 0.8312,
+      "step": 11340
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.3997955010224949e-05,
+      "loss": 0.7778,
+      "step": 11350
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.398091342876619e-05,
+      "loss": 0.8143,
+      "step": 11360
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.396387184730743e-05,
+      "loss": 0.8002,
+      "step": 11370
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.394683026584867e-05,
+      "loss": 0.825,
+      "step": 11380
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.3929788684389911e-05,
+      "loss": 0.807,
+      "step": 11390
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.3912747102931152e-05,
+      "loss": 0.8365,
+      "step": 11400
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.3895705521472393e-05,
+      "loss": 0.831,
+      "step": 11410
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.3878663940013633e-05,
+      "loss": 0.7797,
+      "step": 11420
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.3861622358554874e-05,
+      "loss": 0.8522,
+      "step": 11430
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.3844580777096115e-05,
+      "loss": 0.8107,
+      "step": 11440
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.3827539195637356e-05,
+      "loss": 0.7894,
+      "step": 11450
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.3810497614178595e-05,
+      "loss": 0.835,
+      "step": 11460
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.3793456032719836e-05,
+      "loss": 0.8086,
+      "step": 11470
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.3776414451261078e-05,
+      "loss": 0.7721,
+      "step": 11480
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.3759372869802319e-05,
+      "loss": 0.8108,
+      "step": 11490
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.3742331288343558e-05,
+      "loss": 0.8243,
+      "step": 11500
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.3725289706884799e-05,
+      "loss": 0.7878,
+      "step": 11510
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.370824812542604e-05,
+      "loss": 0.8374,
+      "step": 11520
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.3691206543967281e-05,
+      "loss": 0.8457,
+      "step": 11530
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.367416496250852e-05,
+      "loss": 0.8157,
+      "step": 11540
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.3657123381049762e-05,
+      "loss": 0.7969,
+      "step": 11550
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.3640081799591003e-05,
+      "loss": 0.8237,
+      "step": 11560
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.362474437627812e-05,
+      "loss": 0.8615,
+      "step": 11570
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.3607702794819359e-05,
+      "loss": 0.8207,
+      "step": 11580
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.35906612133606e-05,
+      "loss": 0.8159,
+      "step": 11590
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.357361963190184e-05,
+      "loss": 0.8421,
+      "step": 11600
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.3556578050443082e-05,
+      "loss": 0.7963,
+      "step": 11610
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.3539536468984323e-05,
+      "loss": 0.8274,
+      "step": 11620
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.3522494887525562e-05,
+      "loss": 0.8429,
+      "step": 11630
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.3505453306066803e-05,
+      "loss": 0.7593,
+      "step": 11640
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.3488411724608044e-05,
+      "loss": 0.8045,
+      "step": 11650
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.3471370143149285e-05,
+      "loss": 0.7647,
+      "step": 11660
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.3454328561690525e-05,
+      "loss": 0.7956,
+      "step": 11670
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.3437286980231766e-05,
+      "loss": 0.779,
+      "step": 11680
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.3420245398773007e-05,
+      "loss": 0.7935,
+      "step": 11690
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.3403203817314248e-05,
+      "loss": 0.849,
+      "step": 11700
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.3386162235855487e-05,
+      "loss": 0.8254,
+      "step": 11710
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.3369120654396728e-05,
+      "loss": 0.803,
+      "step": 11720
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.335207907293797e-05,
+      "loss": 0.7733,
+      "step": 11730
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.7297990082169465,
+      "eval_loss": 0.8018476963043213,
+      "eval_runtime": 307.0701,
+      "eval_samples_per_second": 135.94,
+      "eval_steps_per_second": 67.971,
+      "step": 11736
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.333503749147921e-05,
+      "loss": 0.9283,
+      "step": 11740
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 1.331799591002045e-05,
+      "loss": 0.7649,
+      "step": 11750
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 1.3300954328561691e-05,
+      "loss": 0.732,
+      "step": 11760
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 1.3283912747102932e-05,
+      "loss": 0.8149,
+      "step": 11770
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 1.3266871165644173e-05,
+      "loss": 0.808,
+      "step": 11780
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 1.3249829584185414e-05,
+      "loss": 0.8311,
+      "step": 11790
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 1.3232788002726654e-05,
+      "loss": 0.7612,
+      "step": 11800
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 1.3215746421267895e-05,
+      "loss": 0.8216,
+      "step": 11810
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 1.3198704839809136e-05,
+      "loss": 0.81,
+      "step": 11820
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 1.3181663258350377e-05,
+      "loss": 0.8136,
+      "step": 11830
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 1.3164621676891616e-05,
+      "loss": 0.7976,
+      "step": 11840
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 1.3147580095432857e-05,
+      "loss": 0.8298,
+      "step": 11850
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 1.3130538513974097e-05,
+      "loss": 0.8521,
+      "step": 11860
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 1.3113496932515338e-05,
+      "loss": 0.6997,
+      "step": 11870
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 1.3096455351056577e-05,
+      "loss": 0.8282,
+      "step": 11880
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 1.3079413769597818e-05,
+      "loss": 0.7791,
+      "step": 11890
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 1.3062372188139059e-05,
+      "loss": 0.813,
+      "step": 11900
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 1.30453306066803e-05,
+      "loss": 0.7986,
+      "step": 11910
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 1.302828902522154e-05,
+      "loss": 0.7645,
+      "step": 11920
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 1.301124744376278e-05,
+      "loss": 0.813,
+      "step": 11930
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 1.2994205862304022e-05,
+      "loss": 0.8196,
+      "step": 11940
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 1.2977164280845263e-05,
+      "loss": 0.7833,
+      "step": 11950
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 1.2960122699386504e-05,
+      "loss": 0.8014,
+      "step": 11960
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 1.2943081117927743e-05,
+      "loss": 0.7849,
+      "step": 11970
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 1.2926039536468984e-05,
+      "loss": 0.8945,
+      "step": 11980
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 1.2908997955010225e-05,
+      "loss": 0.8875,
+      "step": 11990
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 1.2891956373551466e-05,
+      "loss": 0.874,
+      "step": 12000
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 1.2874914792092706e-05,
+      "loss": 0.7901,
+      "step": 12010
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 1.2857873210633947e-05,
+      "loss": 0.7978,
+      "step": 12020
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 1.2840831629175188e-05,
+      "loss": 0.7471,
+      "step": 12030
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 1.2823790047716429e-05,
+      "loss": 0.853,
+      "step": 12040
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 1.2806748466257668e-05,
+      "loss": 0.7622,
+      "step": 12050
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 1.278970688479891e-05,
+      "loss": 0.7749,
+      "step": 12060
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 1.277266530334015e-05,
+      "loss": 0.82,
+      "step": 12070
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 1.2755623721881391e-05,
+      "loss": 0.7758,
+      "step": 12080
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 1.273858214042263e-05,
+      "loss": 0.8552,
+      "step": 12090
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 1.2721540558963872e-05,
+      "loss": 0.837,
+      "step": 12100
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 1.2704498977505113e-05,
+      "loss": 0.7895,
+      "step": 12110
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 1.2687457396046354e-05,
+      "loss": 0.8489,
+      "step": 12120
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 1.2670415814587595e-05,
+      "loss": 0.7795,
+      "step": 12130
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 1.2653374233128834e-05,
+      "loss": 0.8417,
+      "step": 12140
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 1.2636332651670075e-05,
+      "loss": 0.7759,
+      "step": 12150
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.2619291070211317e-05,
+      "loss": 0.7988,
+      "step": 12160
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.2602249488752558e-05,
+      "loss": 0.798,
+      "step": 12170
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.2585207907293797e-05,
+      "loss": 0.8033,
+      "step": 12180
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.2568166325835038e-05,
+      "loss": 0.817,
+      "step": 12190
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.2551124744376279e-05,
+      "loss": 0.756,
+      "step": 12200
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.253408316291752e-05,
+      "loss": 0.7786,
+      "step": 12210
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.251704158145876e-05,
+      "loss": 0.7857,
+      "step": 12220
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.25e-05,
+      "loss": 0.8499,
+      "step": 12230
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.2482958418541242e-05,
+      "loss": 0.7713,
+      "step": 12240
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.2465916837082483e-05,
+      "loss": 0.776,
+      "step": 12250
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.2448875255623722e-05,
+      "loss": 0.8165,
+      "step": 12260
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.2431833674164963e-05,
+      "loss": 0.7849,
+      "step": 12270
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.2414792092706204e-05,
+      "loss": 0.7905,
+      "step": 12280
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.2397750511247445e-05,
+      "loss": 0.837,
+      "step": 12290
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.2380708929788686e-05,
+      "loss": 0.823,
+      "step": 12300
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.2363667348329924e-05,
+      "loss": 0.7709,
+      "step": 12310
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.2346625766871165e-05,
+      "loss": 0.8507,
+      "step": 12320
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.2329584185412406e-05,
+      "loss": 0.7975,
+      "step": 12330
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.2312542603953647e-05,
+      "loss": 0.7458,
+      "step": 12340
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.2295501022494887e-05,
+      "loss": 0.7975,
+      "step": 12350
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.2278459441036128e-05,
+      "loss": 0.8277,
+      "step": 12360
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.2261417859577369e-05,
+      "loss": 0.825,
+      "step": 12370
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.224437627811861e-05,
+      "loss": 0.8568,
+      "step": 12380
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.2227334696659849e-05,
+      "loss": 0.8044,
+      "step": 12390
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.221029311520109e-05,
+      "loss": 0.8489,
+      "step": 12400
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.2193251533742331e-05,
+      "loss": 0.823,
+      "step": 12410
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.2176209952283572e-05,
+      "loss": 0.7955,
+      "step": 12420
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.2159168370824812e-05,
+      "loss": 0.8011,
+      "step": 12430
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.2142126789366053e-05,
+      "loss": 0.7956,
+      "step": 12440
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.2125085207907294e-05,
+      "loss": 0.8618,
+      "step": 12450
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.2108043626448535e-05,
+      "loss": 0.7471,
+      "step": 12460
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.2091002044989776e-05,
+      "loss": 0.8617,
+      "step": 12470
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2073960463531015e-05,
+      "loss": 0.7884,
+      "step": 12480
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2056918882072256e-05,
+      "loss": 0.8083,
+      "step": 12490
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2039877300613497e-05,
+      "loss": 0.8018,
+      "step": 12500
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2022835719154738e-05,
+      "loss": 0.755,
+      "step": 12510
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2005794137695978e-05,
+      "loss": 0.8412,
+      "step": 12520
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1988752556237219e-05,
+      "loss": 0.8322,
+      "step": 12530
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.197171097477846e-05,
+      "loss": 0.7998,
+      "step": 12540
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1954669393319701e-05,
+      "loss": 0.8032,
+      "step": 12550
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.193762781186094e-05,
+      "loss": 0.7388,
+      "step": 12560
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.1920586230402181e-05,
+      "loss": 0.8043,
+      "step": 12570
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.1903544648943423e-05,
+      "loss": 0.7467,
+      "step": 12580
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.1886503067484664e-05,
+      "loss": 0.734,
+      "step": 12590
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.1869461486025903e-05,
+      "loss": 0.8428,
+      "step": 12600
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.1852419904567144e-05,
+      "loss": 0.739,
+      "step": 12610
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 1.1835378323108385e-05,
+      "loss": 0.8105,
+      "step": 12620
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 1.1818336741649626e-05,
+      "loss": 0.7703,
+      "step": 12630
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 1.1801295160190867e-05,
+      "loss": 0.8545,
+      "step": 12640
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 1.1784253578732107e-05,
+      "loss": 0.7988,
+      "step": 12650
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 1.1767211997273348e-05,
+      "loss": 0.739,
+      "step": 12660
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 1.1750170415814589e-05,
+      "loss": 0.8134,
+      "step": 12670
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 1.173312883435583e-05,
+      "loss": 0.7641,
+      "step": 12680
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 1.171608725289707e-05,
+      "loss": 0.735,
+      "step": 12690
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 1.169904567143831e-05,
+      "loss": 0.7355,
+      "step": 12700
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 1.1682004089979551e-05,
+      "loss": 0.8125,
+      "step": 12710
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 1.1664962508520792e-05,
+      "loss": 0.7467,
+      "step": 12720
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 1.1647920927062032e-05,
+      "loss": 0.7885,
+      "step": 12730
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 1.1630879345603273e-05,
+      "loss": 0.7381,
+      "step": 12740
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 1.1613837764144514e-05,
+      "loss": 0.7532,
+      "step": 12750
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 1.1596796182685753e-05,
+      "loss": 0.799,
+      "step": 12760
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 1.1579754601226993e-05,
+      "loss": 0.7939,
+      "step": 12770
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 1.1562713019768234e-05,
+      "loss": 0.7745,
+      "step": 12780
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.1545671438309475e-05,
+      "loss": 0.7375,
+      "step": 12790
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.1528629856850716e-05,
+      "loss": 0.8336,
+      "step": 12800
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.1511588275391957e-05,
+      "loss": 0.7751,
+      "step": 12810
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.1494546693933196e-05,
+      "loss": 0.8069,
+      "step": 12820
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.1477505112474437e-05,
+      "loss": 0.7874,
+      "step": 12830
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.1460463531015678e-05,
+      "loss": 0.7628,
+      "step": 12840
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.144342194955692e-05,
+      "loss": 0.8091,
+      "step": 12850
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.1426380368098159e-05,
+      "loss": 0.7204,
+      "step": 12860
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.14093387866394e-05,
+      "loss": 0.7949,
+      "step": 12870
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.1392297205180641e-05,
+      "loss": 0.8377,
+      "step": 12880
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.1375255623721882e-05,
+      "loss": 0.8243,
+      "step": 12890
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.1358214042263121e-05,
+      "loss": 0.8249,
+      "step": 12900
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 1.1341172460804362e-05,
+      "loss": 0.7311,
+      "step": 12910
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 1.1324130879345603e-05,
+      "loss": 0.7715,
+      "step": 12920
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 1.1307089297886845e-05,
+      "loss": 0.7846,
+      "step": 12930
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 1.1290047716428084e-05,
+      "loss": 0.7336,
+      "step": 12940
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.1273006134969325e-05,
+      "loss": 0.7272,
+      "step": 12950
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.1255964553510566e-05,
+      "loss": 0.8281,
+      "step": 12960
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.1238922972051807e-05,
+      "loss": 0.822,
+      "step": 12970
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.1221881390593048e-05,
+      "loss": 0.8451,
+      "step": 12980
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1204839809134287e-05,
+      "loss": 0.8324,
+      "step": 12990
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.1187798227675529e-05,
+      "loss": 0.8126,
+      "step": 13000
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.117075664621677e-05,
+      "loss": 0.7574,
+      "step": 13010
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.115371506475801e-05,
+      "loss": 0.8003,
+      "step": 13020
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.113667348329925e-05,
+      "loss": 0.8233,
+      "step": 13030
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.1119631901840491e-05,
+      "loss": 0.7892,
+      "step": 13040
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.7425676161272549,
+      "eval_loss": 0.7517141103744507,
+      "eval_runtime": 311.0565,
+      "eval_samples_per_second": 134.197,
+      "eval_steps_per_second": 67.1,
+      "step": 13040
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 1.1102590320381732e-05,
+      "loss": 0.8305,
+      "step": 13050
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 1.1085548738922973e-05,
+      "loss": 0.8131,
+      "step": 13060
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 1.1068507157464213e-05,
+      "loss": 0.807,
+      "step": 13070
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 1.1051465576005454e-05,
+      "loss": 0.7837,
+      "step": 13080
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 1.1034423994546695e-05,
+      "loss": 0.8118,
+      "step": 13090
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 1.1017382413087936e-05,
+      "loss": 0.8264,
+      "step": 13100
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 1.1000340831629175e-05,
+      "loss": 0.7689,
+      "step": 13110
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 1.0983299250170416e-05,
+      "loss": 0.7932,
+      "step": 13120
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 1.0966257668711657e-05,
+      "loss": 0.7911,
+      "step": 13130
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 1.0949216087252898e-05,
+      "loss": 0.7766,
+      "step": 13140
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 1.093217450579414e-05,
+      "loss": 0.7948,
+      "step": 13150
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 1.0915132924335379e-05,
+      "loss": 0.8203,
+      "step": 13160
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 1.089809134287662e-05,
+      "loss": 0.8316,
+      "step": 13170
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 1.0881049761417861e-05,
+      "loss": 0.8023,
+      "step": 13180
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 1.0864008179959102e-05,
+      "loss": 0.8001,
+      "step": 13190
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 1.0846966598500341e-05,
+      "loss": 0.8304,
+      "step": 13200
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 1.0829925017041582e-05,
+      "loss": 0.7167,
+      "step": 13210
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 1.0812883435582822e-05,
+      "loss": 0.7357,
+      "step": 13220
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 1.0795841854124063e-05,
+      "loss": 0.7452,
+      "step": 13230
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 1.0778800272665302e-05,
+      "loss": 0.7949,
+      "step": 13240
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 1.0761758691206543e-05,
+      "loss": 0.753,
+      "step": 13250
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 1.0744717109747784e-05,
+      "loss": 0.8421,
+      "step": 13260
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 1.0727675528289025e-05,
+      "loss": 0.7798,
+      "step": 13270
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 1.0710633946830265e-05,
+      "loss": 0.767,
+      "step": 13280
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 1.0693592365371506e-05,
+      "loss": 0.7942,
+      "step": 13290
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 1.0676550783912747e-05,
+      "loss": 0.8141,
+      "step": 13300
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 1.0659509202453988e-05,
+      "loss": 0.7729,
+      "step": 13310
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 1.0642467620995229e-05,
+      "loss": 0.8095,
+      "step": 13320
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 1.0625426039536468e-05,
+      "loss": 0.7597,
+      "step": 13330
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 1.060838445807771e-05,
+      "loss": 0.8827,
+      "step": 13340
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 1.059134287661895e-05,
+      "loss": 0.8613,
+      "step": 13350
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 1.0574301295160192e-05,
+      "loss": 0.7873,
+      "step": 13360
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 1.0557259713701431e-05,
+      "loss": 0.8402,
+      "step": 13370
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 1.0540218132242672e-05,
+      "loss": 0.7513,
+      "step": 13380
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 1.0523176550783913e-05,
+      "loss": 0.7907,
+      "step": 13390
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 1.0506134969325154e-05,
+      "loss": 0.8213,
+      "step": 13400
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 1.0489093387866393e-05,
+      "loss": 0.7687,
+      "step": 13410
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 1.0472051806407635e-05,
+      "loss": 0.8077,
+      "step": 13420
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 1.0455010224948876e-05,
+      "loss": 0.7762,
+      "step": 13430
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 1.0437968643490117e-05,
+      "loss": 0.8407,
+      "step": 13440
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 1.0420927062031356e-05,
+      "loss": 0.7854,
+      "step": 13450
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 1.0403885480572597e-05,
+      "loss": 0.7896,
+      "step": 13460
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 1.0386843899113838e-05,
+      "loss": 0.8378,
+      "step": 13470
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 1.036980231765508e-05,
+      "loss": 0.7457,
+      "step": 13480
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 1.035276073619632e-05,
+      "loss": 0.8307,
+      "step": 13490
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 1.033571915473756e-05,
+      "loss": 0.8089,
+      "step": 13500
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 1.03186775732788e-05,
+      "loss": 0.8025,
+      "step": 13510
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 1.0301635991820042e-05,
+      "loss": 0.7999,
+      "step": 13520
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 1.0284594410361283e-05,
+      "loss": 0.8259,
+      "step": 13530
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 1.0267552828902522e-05,
+      "loss": 0.7528,
+      "step": 13540
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 1.0250511247443763e-05,
+      "loss": 0.7765,
+      "step": 13550
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 1.0233469665985004e-05,
+      "loss": 0.7767,
+      "step": 13560
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 1.0216428084526245e-05,
+      "loss": 0.774,
+      "step": 13570
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 1.0199386503067485e-05,
+      "loss": 0.7842,
+      "step": 13580
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 1.0182344921608726e-05,
+      "loss": 0.8339,
+      "step": 13590
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 1.0165303340149967e-05,
+      "loss": 0.7516,
+      "step": 13600
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 1.0148261758691208e-05,
+      "loss": 0.757,
+      "step": 13610
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 1.0131220177232447e-05,
+      "loss": 0.8159,
+      "step": 13620
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 1.0114178595773688e-05,
+      "loss": 0.8153,
+      "step": 13630
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 1.009713701431493e-05,
+      "loss": 0.7875,
+      "step": 13640
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 1.008009543285617e-05,
+      "loss": 0.7746,
+      "step": 13650
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 1.006305385139741e-05,
+      "loss": 0.7673,
+      "step": 13660
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 1.004601226993865e-05,
+      "loss": 0.8308,
+      "step": 13670
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 1.002897068847989e-05,
+      "loss": 0.7649,
+      "step": 13680
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 1.0011929107021131e-05,
+      "loss": 0.7553,
+      "step": 13690
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 9.994887525562372e-06,
+      "loss": 0.8222,
+      "step": 13700
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 9.977845944103612e-06,
+      "loss": 0.8145,
+      "step": 13710
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 9.960804362644853e-06,
+      "loss": 0.7587,
+      "step": 13720
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 9.943762781186094e-06,
+      "loss": 0.8305,
+      "step": 13730
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 9.926721199727335e-06,
+      "loss": 0.7589,
+      "step": 13740
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 9.909679618268574e-06,
+      "loss": 0.7977,
+      "step": 13750
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 9.892638036809815e-06,
+      "loss": 0.8363,
+      "step": 13760
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 9.875596455351057e-06,
+      "loss": 0.8263,
+      "step": 13770
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 9.858554873892298e-06,
+      "loss": 0.8386,
+      "step": 13780
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 9.841513292433537e-06,
+      "loss": 0.8094,
+      "step": 13790
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 9.824471710974778e-06,
+      "loss": 0.7401,
+      "step": 13800
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 9.807430129516019e-06,
+      "loss": 0.7615,
+      "step": 13810
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 9.79038854805726e-06,
+      "loss": 0.7803,
+      "step": 13820
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 9.773346966598501e-06,
+      "loss": 0.8398,
+      "step": 13830
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 9.75630538513974e-06,
+      "loss": 0.8256,
+      "step": 13840
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 9.739263803680982e-06,
+      "loss": 0.7747,
+      "step": 13850
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 9.722222222222223e-06,
+      "loss": 0.7543,
+      "step": 13860
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 9.705180640763464e-06,
+      "loss": 0.7872,
+      "step": 13870
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 9.688139059304703e-06,
+      "loss": 0.7775,
+      "step": 13880
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 9.671097477845944e-06,
+      "loss": 0.8072,
+      "step": 13890
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 9.654055896387185e-06,
+      "loss": 0.7305,
+      "step": 13900
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.637014314928426e-06,
+      "loss": 0.7538,
+      "step": 13910
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.619972733469666e-06,
+      "loss": 0.8287,
+      "step": 13920
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 9.602931152010907e-06,
+      "loss": 0.7899,
+      "step": 13930
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 9.585889570552148e-06,
+      "loss": 0.8109,
+      "step": 13940
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 9.568847989093389e-06,
+      "loss": 0.7612,
+      "step": 13950
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 9.551806407634628e-06,
+      "loss": 0.7746,
+      "step": 13960
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 9.53476482617587e-06,
+      "loss": 0.7482,
+      "step": 13970
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 9.51772324471711e-06,
+      "loss": 0.774,
+      "step": 13980
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 9.500681663258351e-06,
+      "loss": 0.7667,
+      "step": 13990
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 9.48364008179959e-06,
+      "loss": 0.7395,
+      "step": 14000
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 9.466598500340832e-06,
+      "loss": 0.758,
+      "step": 14010
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 9.449556918882073e-06,
+      "loss": 0.7952,
+      "step": 14020
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 9.432515337423314e-06,
+      "loss": 0.8122,
+      "step": 14030
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 9.415473755964555e-06,
+      "loss": 0.8136,
+      "step": 14040
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 9.398432174505794e-06,
+      "loss": 0.8201,
+      "step": 14050
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 9.381390593047035e-06,
+      "loss": 0.8034,
+      "step": 14060
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 9.364349011588277e-06,
+      "loss": 0.8241,
+      "step": 14070
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.347307430129518e-06,
+      "loss": 0.7367,
+      "step": 14080
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.330265848670757e-06,
+      "loss": 0.8135,
+      "step": 14090
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 9.313224267211998e-06,
+      "loss": 0.7583,
+      "step": 14100
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 9.296182685753239e-06,
+      "loss": 0.8368,
+      "step": 14110
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 9.27914110429448e-06,
+      "loss": 0.8164,
+      "step": 14120
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 9.262099522835718e-06,
+      "loss": 0.7151,
+      "step": 14130
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 9.245057941376959e-06,
+      "loss": 0.77,
+      "step": 14140
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 9.2280163599182e-06,
+      "loss": 0.8077,
+      "step": 14150
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 9.210974778459441e-06,
+      "loss": 0.8156,
+      "step": 14160
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 9.193933197000682e-06,
+      "loss": 0.747,
+      "step": 14170
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 9.176891615541921e-06,
+      "loss": 0.7577,
+      "step": 14180
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 9.159850034083163e-06,
+      "loss": 0.7784,
+      "step": 14190
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 9.142808452624404e-06,
+      "loss": 0.8209,
+      "step": 14200
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 9.125766871165645e-06,
+      "loss": 0.7644,
+      "step": 14210
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 9.108725289706884e-06,
+      "loss": 0.7547,
+      "step": 14220
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 9.091683708248125e-06,
+      "loss": 0.8402,
+      "step": 14230
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 9.074642126789366e-06,
+      "loss": 0.7263,
+      "step": 14240
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 9.057600545330607e-06,
+      "loss": 0.7802,
+      "step": 14250
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 9.040558963871847e-06,
+      "loss": 0.7682,
+      "step": 14260
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 9.023517382413088e-06,
+      "loss": 0.7886,
+      "step": 14270
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 9.006475800954329e-06,
+      "loss": 0.761,
+      "step": 14280
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 8.98943421949557e-06,
+      "loss": 0.7852,
+      "step": 14290
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 8.972392638036809e-06,
+      "loss": 0.7774,
+      "step": 14300
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 8.95535105657805e-06,
+      "loss": 0.7593,
+      "step": 14310
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 8.938309475119291e-06,
+      "loss": 0.7169,
+      "step": 14320
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 8.921267893660532e-06,
+      "loss": 0.797,
+      "step": 14330
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 8.904226312201772e-06,
+      "loss": 0.8736,
+      "step": 14340
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.7481733464293414,
+      "eval_loss": 0.7481608390808105,
+      "eval_runtime": 308.386,
+      "eval_samples_per_second": 135.36,
+      "eval_steps_per_second": 67.681,
+      "step": 14344
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 8.887184730743013e-06,
+      "loss": 0.8159,
+      "step": 14350
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 8.870143149284254e-06,
+      "loss": 0.746,
+      "step": 14360
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 8.853101567825495e-06,
+      "loss": 0.8024,
+      "step": 14370
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 8.836059986366736e-06,
+      "loss": 0.8298,
+      "step": 14380
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 8.819018404907975e-06,
+      "loss": 0.8022,
+      "step": 14390
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 8.801976823449216e-06,
+      "loss": 0.7408,
+      "step": 14400
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 8.784935241990457e-06,
+      "loss": 0.7694,
+      "step": 14410
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 8.767893660531698e-06,
+      "loss": 0.7639,
+      "step": 14420
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 8.750852079072938e-06,
+      "loss": 0.7767,
+      "step": 14430
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 8.733810497614179e-06,
+      "loss": 0.7738,
+      "step": 14440
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 8.71676891615542e-06,
+      "loss": 0.7859,
+      "step": 14450
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 8.699727334696661e-06,
+      "loss": 0.7883,
+      "step": 14460
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 8.6826857532379e-06,
+      "loss": 0.7462,
+      "step": 14470
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 8.665644171779141e-06,
+      "loss": 0.7651,
+      "step": 14480
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 8.648602590320383e-06,
+      "loss": 0.7527,
+      "step": 14490
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 8.631561008861624e-06,
+      "loss": 0.7703,
+      "step": 14500
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 8.614519427402863e-06,
+      "loss": 0.7692,
+      "step": 14510
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 8.597477845944104e-06,
+      "loss": 0.789,
+      "step": 14520
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 8.580436264485345e-06,
+      "loss": 0.7653,
+      "step": 14530
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 8.563394683026586e-06,
+      "loss": 0.748,
+      "step": 14540
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 8.546353101567827e-06,
+      "loss": 0.8126,
+      "step": 14550
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 8.529311520109067e-06,
+      "loss": 0.787,
+      "step": 14560
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 8.512269938650308e-06,
+      "loss": 0.7589,
+      "step": 14570
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 8.495228357191547e-06,
+      "loss": 0.7504,
+      "step": 14580
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 8.478186775732788e-06,
+      "loss": 0.7531,
+      "step": 14590
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 8.461145194274027e-06,
+      "loss": 0.7778,
+      "step": 14600
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 8.444103612815269e-06,
+      "loss": 0.8005,
+      "step": 14610
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 8.42706203135651e-06,
+      "loss": 0.7667,
+      "step": 14620
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 8.41002044989775e-06,
+      "loss": 0.693,
+      "step": 14630
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 8.39297886843899e-06,
+      "loss": 0.7671,
+      "step": 14640
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 8.375937286980231e-06,
+      "loss": 0.8129,
+      "step": 14650
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 8.358895705521472e-06,
+      "loss": 0.7351,
+      "step": 14660
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 8.341854124062713e-06,
+      "loss": 0.7628,
+      "step": 14670
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 8.324812542603953e-06,
+      "loss": 0.8285,
+      "step": 14680
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 8.307770961145194e-06,
+      "loss": 0.7607,
+      "step": 14690
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 8.290729379686435e-06,
+      "loss": 0.8104,
+      "step": 14700
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 8.273687798227676e-06,
+      "loss": 0.7647,
+      "step": 14710
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 8.256646216768917e-06,
+      "loss": 0.7298,
+      "step": 14720
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 8.239604635310156e-06,
+      "loss": 0.7786,
+      "step": 14730
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 8.222563053851397e-06,
+      "loss": 0.7807,
+      "step": 14740
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 8.205521472392638e-06,
+      "loss": 0.7664,
+      "step": 14750
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 8.18847989093388e-06,
+      "loss": 0.7996,
+      "step": 14760
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 8.171438309475119e-06,
+      "loss": 0.8252,
+      "step": 14770
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 8.15439672801636e-06,
+      "loss": 0.7686,
+      "step": 14780
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 8.137355146557601e-06,
+      "loss": 0.7778,
+      "step": 14790
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 8.120313565098842e-06,
+      "loss": 0.8013,
+      "step": 14800
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 8.103271983640081e-06,
+      "loss": 0.7604,
+      "step": 14810
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 8.086230402181322e-06,
+      "loss": 0.7867,
+      "step": 14820
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 8.069188820722563e-06,
+      "loss": 0.7928,
+      "step": 14830
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 8.052147239263804e-06,
+      "loss": 0.7806,
+      "step": 14840
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 8.035105657805044e-06,
+      "loss": 0.721,
+      "step": 14850
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 8.018064076346285e-06,
+      "loss": 0.8207,
+      "step": 14860
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 8.001022494887526e-06,
+      "loss": 0.803,
+      "step": 14870
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 7.983980913428767e-06,
+      "loss": 0.8167,
+      "step": 14880
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 7.966939331970008e-06,
+      "loss": 0.7249,
+      "step": 14890
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 7.949897750511247e-06,
+      "loss": 0.7559,
+      "step": 14900
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 7.932856169052489e-06,
+      "loss": 0.7577,
+      "step": 14910
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 7.91581458759373e-06,
+      "loss": 0.7632,
+      "step": 14920
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 7.89877300613497e-06,
+      "loss": 0.793,
+      "step": 14930
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 7.88173142467621e-06,
+      "loss": 0.6827,
+      "step": 14940
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 7.864689843217451e-06,
+      "loss": 0.7822,
+      "step": 14950
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 7.847648261758692e-06,
+      "loss": 0.7779,
+      "step": 14960
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 7.830606680299933e-06,
+      "loss": 0.7757,
+      "step": 14970
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 7.813565098841173e-06,
+      "loss": 0.7572,
+      "step": 14980
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 7.796523517382414e-06,
+      "loss": 0.7549,
+      "step": 14990
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 7.779481935923655e-06,
+      "loss": 0.7785,
+      "step": 15000
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.762440354464896e-06,
+      "loss": 0.7695,
+      "step": 15010
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 7.745398773006135e-06,
+      "loss": 0.7538,
+      "step": 15020
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 7.728357191547375e-06,
+      "loss": 0.7485,
+      "step": 15030
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 7.711315610088616e-06,
+      "loss": 0.7648,
+      "step": 15040
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 7.694274028629857e-06,
+      "loss": 0.815,
+      "step": 15050
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 7.677232447171098e-06,
+      "loss": 0.7589,
+      "step": 15060
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 7.660190865712337e-06,
+      "loss": 0.7636,
+      "step": 15070
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 7.643149284253578e-06,
+      "loss": 0.8084,
+      "step": 15080
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 7.626107702794819e-06,
+      "loss": 0.7974,
+      "step": 15090
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 7.609066121336059e-06,
+      "loss": 0.7813,
+      "step": 15100
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 7.5920245398773005e-06,
+      "loss": 0.7649,
+      "step": 15110
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 7.574982958418541e-06,
+      "loss": 0.7393,
+      "step": 15120
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 7.557941376959782e-06,
+      "loss": 0.7347,
+      "step": 15130
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 7.540899795501022e-06,
+      "loss": 0.7636,
+      "step": 15140
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 7.523858214042263e-06,
+      "loss": 0.7704,
+      "step": 15150
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 7.506816632583503e-06,
+      "loss": 0.7621,
+      "step": 15160
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 7.489775051124744e-06,
+      "loss": 0.7738,
+      "step": 15170
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 7.4727334696659845e-06,
+      "loss": 0.7882,
+      "step": 15180
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 7.455691888207226e-06,
+      "loss": 0.792,
+      "step": 15190
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 7.438650306748467e-06,
+      "loss": 0.711,
+      "step": 15200
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 7.421608725289707e-06,
+      "loss": 0.7739,
+      "step": 15210
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 7.404567143830948e-06,
+      "loss": 0.8079,
+      "step": 15220
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 7.387525562372188e-06,
+      "loss": 0.7994,
+      "step": 15230
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 7.370483980913429e-06,
+      "loss": 0.8045,
+      "step": 15240
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 7.3534423994546694e-06,
+      "loss": 0.7866,
+      "step": 15250
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 7.3364008179959105e-06,
+      "loss": 0.7313,
+      "step": 15260
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 7.319359236537151e-06,
+      "loss": 0.787,
+      "step": 15270
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 7.302317655078392e-06,
+      "loss": 0.7644,
+      "step": 15280
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 7.285276073619632e-06,
+      "loss": 0.8043,
+      "step": 15290
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 7.268234492160873e-06,
+      "loss": 0.8052,
+      "step": 15300
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 7.251192910702113e-06,
+      "loss": 0.7864,
+      "step": 15310
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 7.234151329243354e-06,
+      "loss": 0.7572,
+      "step": 15320
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 7.2171097477845945e-06,
+      "loss": 0.7533,
+      "step": 15330
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 7.200068166325836e-06,
+      "loss": 0.8065,
+      "step": 15340
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 7.183026584867076e-06,
+      "loss": 0.7221,
+      "step": 15350
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 7.165985003408317e-06,
+      "loss": 0.8084,
+      "step": 15360
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 7.148943421949557e-06,
+      "loss": 0.6837,
+      "step": 15370
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 7.131901840490797e-06,
+      "loss": 0.7272,
+      "step": 15380
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 7.114860259032038e-06,
+      "loss": 0.7843,
+      "step": 15390
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 7.097818677573279e-06,
+      "loss": 0.8315,
+      "step": 15400
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 7.08077709611452e-06,
+      "loss": 0.7944,
+      "step": 15410
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 7.06373551465576e-06,
+      "loss": 0.7041,
+      "step": 15420
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 7.046693933197001e-06,
+      "loss": 0.7999,
+      "step": 15430
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 7.029652351738241e-06,
+      "loss": 0.8524,
+      "step": 15440
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 7.012610770279482e-06,
+      "loss": 0.7569,
+      "step": 15450
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 6.9955691888207224e-06,
+      "loss": 0.77,
+      "step": 15460
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 6.9785276073619635e-06,
+      "loss": 0.7803,
+      "step": 15470
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 6.961486025903204e-06,
+      "loss": 0.7407,
+      "step": 15480
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 6.944444444444445e-06,
+      "loss": 0.7604,
+      "step": 15490
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 6.927402862985685e-06,
+      "loss": 0.7106,
+      "step": 15500
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 6.910361281526926e-06,
+      "loss": 0.7643,
+      "step": 15510
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 6.893319700068166e-06,
+      "loss": 0.7635,
+      "step": 15520
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 6.876278118609407e-06,
+      "loss": 0.7425,
+      "step": 15530
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 6.8592365371506476e-06,
+      "loss": 0.7849,
+      "step": 15540
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 6.842194955691889e-06,
+      "loss": 0.7734,
+      "step": 15550
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 6.82515337423313e-06,
+      "loss": 0.7469,
+      "step": 15560
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 6.80811179277437e-06,
+      "loss": 0.7827,
+      "step": 15570
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 6.791070211315611e-06,
+      "loss": 0.7833,
+      "step": 15580
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 6.774028629856851e-06,
+      "loss": 0.7861,
+      "step": 15590
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 6.756987048398091e-06,
+      "loss": 0.8117,
+      "step": 15600
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 6.739945466939332e-06,
+      "loss": 0.7734,
+      "step": 15610
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 6.722903885480573e-06,
+      "loss": 0.7362,
+      "step": 15620
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 6.705862304021813e-06,
+      "loss": 0.7801,
+      "step": 15630
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 6.688820722563054e-06,
+      "loss": 0.7035,
+      "step": 15640
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.7487722492393935,
+      "eval_loss": 0.7729619145393372,
+      "eval_runtime": 309.5558,
+      "eval_samples_per_second": 134.848,
+      "eval_steps_per_second": 67.426,
+      "step": 15648
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 6.671779141104294e-06,
+      "loss": 0.7553,
+      "step": 15650
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 6.654737559645535e-06,
+      "loss": 0.7588,
+      "step": 15660
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 6.6376959781867754e-06,
+      "loss": 0.7583,
+      "step": 15670
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 6.6206543967280165e-06,
+      "loss": 0.7455,
+      "step": 15680
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 6.603612815269257e-06,
+      "loss": 0.7305,
+      "step": 15690
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 6.586571233810498e-06,
+      "loss": 0.772,
+      "step": 15700
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 6.571233810497614e-06,
+      "loss": 0.7644,
+      "step": 15710
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 6.5541922290388555e-06,
+      "loss": 0.7866,
+      "step": 15720
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 6.537150647580096e-06,
+      "loss": 0.7389,
+      "step": 15730
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 6.520109066121336e-06,
+      "loss": 0.7237,
+      "step": 15740
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 6.503067484662576e-06,
+      "loss": 0.7315,
+      "step": 15750
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 6.486025903203817e-06,
+      "loss": 0.7371,
+      "step": 15760
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 6.468984321745058e-06,
+      "loss": 0.7157,
+      "step": 15770
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 6.4519427402862984e-06,
+      "loss": 0.8073,
+      "step": 15780
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 6.4349011588275395e-06,
+      "loss": 0.763,
+      "step": 15790
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 6.41785957736878e-06,
+      "loss": 0.7828,
+      "step": 15800
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 6.400817995910021e-06,
+      "loss": 0.7713,
+      "step": 15810
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 6.383776414451261e-06,
+      "loss": 0.7705,
+      "step": 15820
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 6.366734832992502e-06,
+      "loss": 0.8068,
+      "step": 15830
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 6.349693251533742e-06,
+      "loss": 0.7472,
+      "step": 15840
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 6.332651670074983e-06,
+      "loss": 0.7884,
+      "step": 15850
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 6.3156100886162236e-06,
+      "loss": 0.7791,
+      "step": 15860
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 6.298568507157465e-06,
+      "loss": 0.7639,
+      "step": 15870
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 6.281526925698705e-06,
+      "loss": 0.8227,
+      "step": 15880
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 6.264485344239946e-06,
+      "loss": 0.7436,
+      "step": 15890
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 6.247443762781186e-06,
+      "loss": 0.7319,
+      "step": 15900
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 6.230402181322427e-06,
+      "loss": 0.7131,
+      "step": 15910
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 6.213360599863667e-06,
+      "loss": 0.7131,
+      "step": 15920
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 6.1963190184049085e-06,
+      "loss": 0.7659,
+      "step": 15930
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 6.1792774369461495e-06,
+      "loss": 0.7921,
+      "step": 15940
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 6.16223585548739e-06,
+      "loss": 0.7246,
+      "step": 15950
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 6.14519427402863e-06,
+      "loss": 0.7552,
+      "step": 15960
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 6.12815269256987e-06,
+      "loss": 0.775,
+      "step": 15970
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 0.7376,
+      "step": 15980
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 6.0940695296523515e-06,
+      "loss": 0.7099,
+      "step": 15990
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 6.0770279481935925e-06,
+      "loss": 0.7629,
+      "step": 16000
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 6.059986366734833e-06,
+      "loss": 0.7765,
+      "step": 16010
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 6.042944785276074e-06,
+      "loss": 0.7369,
+      "step": 16020
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 6.025903203817314e-06,
+      "loss": 0.7194,
+      "step": 16030
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 6.008861622358555e-06,
+      "loss": 0.7459,
+      "step": 16040
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 5.991820040899795e-06,
+      "loss": 0.7639,
+      "step": 16050
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 5.974778459441036e-06,
+      "loss": 0.8181,
+      "step": 16060
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 5.9577368779822766e-06,
+      "loss": 0.7205,
+      "step": 16070
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 5.940695296523518e-06,
+      "loss": 0.7648,
+      "step": 16080
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 5.923653715064758e-06,
+      "loss": 0.7516,
+      "step": 16090
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 5.906612133605999e-06,
+      "loss": 0.7774,
+      "step": 16100
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 5.88957055214724e-06,
+      "loss": 0.6938,
+      "step": 16110
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 5.87252897068848e-06,
+      "loss": 0.808,
+      "step": 16120
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 5.855487389229721e-06,
+      "loss": 0.7999,
+      "step": 16130
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 5.8384458077709615e-06,
+      "loss": 0.7431,
+      "step": 16140
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 5.8214042263122025e-06,
+      "loss": 0.7346,
+      "step": 16150
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 5.804362644853443e-06,
+      "loss": 0.7394,
+      "step": 16160
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 5.787321063394684e-06,
+      "loss": 0.7674,
+      "step": 16170
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 5.770279481935924e-06,
+      "loss": 0.793,
+      "step": 16180
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 5.753237900477164e-06,
+      "loss": 0.7205,
+      "step": 16190
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 5.7361963190184045e-06,
+      "loss": 0.7861,
+      "step": 16200
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 5.7191547375596455e-06,
+      "loss": 0.7682,
+      "step": 16210
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 5.702113156100886e-06,
+      "loss": 0.7318,
+      "step": 16220
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 5.685071574642127e-06,
+      "loss": 0.7646,
+      "step": 16230
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 5.668029993183367e-06,
+      "loss": 0.793,
+      "step": 16240
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 5.650988411724608e-06,
+      "loss": 0.7705,
+      "step": 16250
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 5.633946830265848e-06,
+      "loss": 0.7395,
+      "step": 16260
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 5.616905248807089e-06,
+      "loss": 0.7382,
+      "step": 16270
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 5.59986366734833e-06,
+      "loss": 0.7341,
+      "step": 16280
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 5.582822085889571e-06,
+      "loss": 0.7519,
+      "step": 16290
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 5.565780504430812e-06,
+      "loss": 0.7389,
+      "step": 16300
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 5.548738922972052e-06,
+      "loss": 0.7619,
+      "step": 16310
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 5.531697341513293e-06,
+      "loss": 0.7674,
+      "step": 16320
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 5.514655760054533e-06,
+      "loss": 0.7741,
+      "step": 16330
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 5.497614178595774e-06,
+      "loss": 0.761,
+      "step": 16340
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 5.4805725971370145e-06,
+      "loss": 0.7799,
+      "step": 16350
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 5.4635310156782555e-06,
+      "loss": 0.7736,
+      "step": 16360
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 5.446489434219496e-06,
+      "loss": 0.8165,
+      "step": 16370
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 5.429447852760737e-06,
+      "loss": 0.7582,
+      "step": 16380
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 5.412406271301977e-06,
+      "loss": 0.7728,
+      "step": 16390
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 5.395364689843218e-06,
+      "loss": 0.7999,
+      "step": 16400
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 5.378323108384458e-06,
+      "loss": 0.805,
+      "step": 16410
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 5.3612815269256985e-06,
+      "loss": 0.8066,
+      "step": 16420
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 5.344239945466939e-06,
+      "loss": 0.6918,
+      "step": 16430
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 5.32719836400818e-06,
+      "loss": 0.7412,
+      "step": 16440
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 5.310156782549421e-06,
+      "loss": 0.7463,
+      "step": 16450
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 5.293115201090661e-06,
+      "loss": 0.7482,
+      "step": 16460
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 5.276073619631902e-06,
+      "loss": 0.7871,
+      "step": 16470
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 5.259032038173142e-06,
+      "loss": 0.7691,
+      "step": 16480
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 5.241990456714383e-06,
+      "loss": 0.7703,
+      "step": 16490
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 5.224948875255624e-06,
+      "loss": 0.7806,
+      "step": 16500
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 5.207907293796865e-06,
+      "loss": 0.7653,
+      "step": 16510
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 5.190865712338105e-06,
+      "loss": 0.7787,
+      "step": 16520
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 5.173824130879346e-06,
+      "loss": 0.6742,
+      "step": 16530
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 5.156782549420586e-06,
+      "loss": 0.7753,
+      "step": 16540
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 5.139740967961827e-06,
+      "loss": 0.8135,
+      "step": 16550
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 5.1226993865030675e-06,
+      "loss": 0.7775,
+      "step": 16560
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 5.1056578050443085e-06,
+      "loss": 0.7667,
+      "step": 16570
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 5.088616223585549e-06,
+      "loss": 0.7787,
+      "step": 16580
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 5.07157464212679e-06,
+      "loss": 0.7386,
+      "step": 16590
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 5.05453306066803e-06,
+      "loss": 0.746,
+      "step": 16600
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 5.037491479209271e-06,
+      "loss": 0.7722,
+      "step": 16610
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 5.020449897750511e-06,
+      "loss": 0.812,
+      "step": 16620
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 5.003408316291752e-06,
+      "loss": 0.7605,
+      "step": 16630
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.986366734832993e-06,
+      "loss": 0.7965,
+      "step": 16640
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 4.969325153374233e-06,
+      "loss": 0.706,
+      "step": 16650
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.952283571915474e-06,
+      "loss": 0.7306,
+      "step": 16660
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.935241990456714e-06,
+      "loss": 0.7133,
+      "step": 16670
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.918200408997955e-06,
+      "loss": 0.751,
+      "step": 16680
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 4.901158827539195e-06,
+      "loss": 0.7787,
+      "step": 16690
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.884117246080436e-06,
+      "loss": 0.7462,
+      "step": 16700
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.867075664621677e-06,
+      "loss": 0.8488,
+      "step": 16710
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.850034083162918e-06,
+      "loss": 0.7439,
+      "step": 16720
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 4.832992501704158e-06,
+      "loss": 0.7281,
+      "step": 16730
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.815950920245399e-06,
+      "loss": 0.6796,
+      "step": 16740
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.798909338786639e-06,
+      "loss": 0.7354,
+      "step": 16750
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.78186775732788e-06,
+      "loss": 0.7547,
+      "step": 16760
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 4.7648261758691205e-06,
+      "loss": 0.7333,
+      "step": 16770
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.7477845944103615e-06,
+      "loss": 0.8166,
+      "step": 16780
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.730743012951602e-06,
+      "loss": 0.764,
+      "step": 16790
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.713701431492843e-06,
+      "loss": 0.7792,
+      "step": 16800
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 4.696659850034084e-06,
+      "loss": 0.7089,
+      "step": 16810
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.679618268575324e-06,
+      "loss": 0.7608,
+      "step": 16820
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.662576687116565e-06,
+      "loss": 0.7466,
+      "step": 16830
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.645535105657805e-06,
+      "loss": 0.7375,
+      "step": 16840
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 4.6284935241990464e-06,
+      "loss": 0.7645,
+      "step": 16850
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.611451942740287e-06,
+      "loss": 0.6682,
+      "step": 16860
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.594410361281527e-06,
+      "loss": 0.7698,
+      "step": 16870
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.577368779822767e-06,
+      "loss": 0.7305,
+      "step": 16880
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.560327198364008e-06,
+      "loss": 0.6571,
+      "step": 16890
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.543285616905248e-06,
+      "loss": 0.7698,
+      "step": 16900
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.526244035446489e-06,
+      "loss": 0.7541,
+      "step": 16910
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.50920245398773e-06,
+      "loss": 0.7455,
+      "step": 16920
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.492160872528971e-06,
+      "loss": 0.7243,
+      "step": 16930
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 4.475119291070211e-06,
+      "loss": 0.7715,
+      "step": 16940
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 4.458077709611452e-06,
+      "loss": 0.7361,
+      "step": 16950
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.7510241238051889,
+      "eval_loss": 0.7676522731781006,
+      "eval_runtime": 305.6662,
+      "eval_samples_per_second": 136.564,
+      "eval_steps_per_second": 68.284,
+      "step": 16952
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.441036128152692e-06,
+      "loss": 0.8415,
+      "step": 16960
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.423994546693933e-06,
+      "loss": 0.7245,
+      "step": 16970
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 4.406952965235174e-06,
+      "loss": 0.7709,
+      "step": 16980
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.3899113837764145e-06,
+      "loss": 0.7723,
+      "step": 16990
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.372869802317656e-06,
+      "loss": 0.799,
+      "step": 17000
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.355828220858896e-06,
+      "loss": 0.6982,
+      "step": 17010
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 4.338786639400137e-06,
+      "loss": 0.741,
+      "step": 17020
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.321745057941377e-06,
+      "loss": 0.7451,
+      "step": 17030
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.304703476482618e-06,
+      "loss": 0.7342,
+      "step": 17040
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 4.287661895023858e-06,
+      "loss": 0.7048,
+      "step": 17050
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 4.2706203135650994e-06,
+      "loss": 0.7065,
+      "step": 17060
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.25357873210634e-06,
+      "loss": 0.7508,
+      "step": 17070
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.236537150647581e-06,
+      "loss": 0.811,
+      "step": 17080
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.219495569188821e-06,
+      "loss": 0.7547,
+      "step": 17090
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.202453987730061e-06,
+      "loss": 0.7736,
+      "step": 17100
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.185412406271301e-06,
+      "loss": 0.7466,
+      "step": 17110
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.168370824812542e-06,
+      "loss": 0.7226,
+      "step": 17120
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.151329243353783e-06,
+      "loss": 0.7686,
+      "step": 17130
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.134287661895024e-06,
+      "loss": 0.79,
+      "step": 17140
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.117246080436265e-06,
+      "loss": 0.7689,
+      "step": 17150
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.100204498977505e-06,
+      "loss": 0.7819,
+      "step": 17160
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 4.083162917518746e-06,
+      "loss": 0.73,
+      "step": 17170
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 4.066121336059986e-06,
+      "loss": 0.7194,
+      "step": 17180
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.049079754601227e-06,
+      "loss": 0.8154,
+      "step": 17190
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.0320381731424675e-06,
+      "loss": 0.7288,
+      "step": 17200
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 4.014996591683709e-06,
+      "loss": 0.7513,
+      "step": 17210
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 3.997955010224949e-06,
+      "loss": 0.7705,
+      "step": 17220
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 3.98091342876619e-06,
+      "loss": 0.677,
+      "step": 17230
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 3.96387184730743e-06,
+      "loss": 0.7307,
+      "step": 17240
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 3.946830265848671e-06,
+      "loss": 0.7598,
+      "step": 17250
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 3.929788684389911e-06,
+      "loss": 0.7728,
+      "step": 17260
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 3.9127471029311524e-06,
+      "loss": 0.7879,
+      "step": 17270
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 3.895705521472393e-06,
+      "loss": 0.707,
+      "step": 17280
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 3.878663940013634e-06,
+      "loss": 0.6973,
+      "step": 17290
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 3.861622358554874e-06,
+      "loss": 0.7903,
+      "step": 17300
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 3.844580777096115e-06,
+      "loss": 0.7881,
+      "step": 17310
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 3.827539195637356e-06,
+      "loss": 0.7273,
+      "step": 17320
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 3.8104976141785954e-06,
+      "loss": 0.7945,
+      "step": 17330
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 3.793456032719836e-06,
+      "loss": 0.7336,
+      "step": 17340
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 3.7764144512610767e-06,
+      "loss": 0.7424,
+      "step": 17350
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 3.7593728698023173e-06,
+      "loss": 0.7609,
+      "step": 17360
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 3.7423312883435584e-06,
+      "loss": 0.7589,
+      "step": 17370
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 3.7252897068847986e-06,
+      "loss": 0.7262,
+      "step": 17380
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 3.7082481254260393e-06,
+      "loss": 0.7484,
+      "step": 17390
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 3.6912065439672803e-06,
+      "loss": 0.7357,
+      "step": 17400
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 3.674164962508521e-06,
+      "loss": 0.7789,
+      "step": 17410
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 3.6571233810497616e-06,
+      "loss": 0.7289,
+      "step": 17420
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 3.6400817995910022e-06,
+      "loss": 0.7886,
+      "step": 17430
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 3.623040218132243e-06,
+      "loss": 0.6666,
+      "step": 17440
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 3.6059986366734835e-06,
+      "loss": 0.8058,
+      "step": 17450
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 3.588957055214724e-06,
+      "loss": 0.778,
+      "step": 17460
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 3.571915473755965e-06,
+      "loss": 0.7468,
+      "step": 17470
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 3.5548738922972054e-06,
+      "loss": 0.7603,
+      "step": 17480
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 3.537832310838446e-06,
+      "loss": 0.7456,
+      "step": 17490
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 3.5207907293796863e-06,
+      "loss": 0.7258,
+      "step": 17500
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 3.503749147920927e-06,
+      "loss": 0.7364,
+      "step": 17510
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 3.4867075664621676e-06,
+      "loss": 0.6973,
+      "step": 17520
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 3.469665985003408e-06,
+      "loss": 0.713,
+      "step": 17530
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 3.452624403544649e-06,
+      "loss": 0.6919,
+      "step": 17540
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 3.4355828220858895e-06,
+      "loss": 0.8061,
+      "step": 17550
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 3.41854124062713e-06,
+      "loss": 0.7541,
+      "step": 17560
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 3.401499659168371e-06,
+      "loss": 0.7406,
+      "step": 17570
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 3.384458077709612e-06,
+      "loss": 0.7036,
+      "step": 17580
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 3.3674164962508525e-06,
+      "loss": 0.7314,
+      "step": 17590
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 3.350374914792093e-06,
+      "loss": 0.7158,
+      "step": 17600
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.7838,
+      "step": 17610
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 3.316291751874574e-06,
+      "loss": 0.7624,
+      "step": 17620
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 3.2992501704158146e-06,
+      "loss": 0.7356,
+      "step": 17630
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 3.2822085889570552e-06,
+      "loss": 0.7513,
+      "step": 17640
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 3.265167007498296e-06,
+      "loss": 0.7545,
+      "step": 17650
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 3.2481254260395365e-06,
+      "loss": 0.6999,
+      "step": 17660
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 3.231083844580777e-06,
+      "loss": 0.7116,
+      "step": 17670
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 3.214042263122018e-06,
+      "loss": 0.7679,
+      "step": 17680
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 3.1970006816632584e-06,
+      "loss": 0.7721,
+      "step": 17690
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 3.179959100204499e-06,
+      "loss": 0.7736,
+      "step": 17700
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 3.1629175187457397e-06,
+      "loss": 0.7945,
+      "step": 17710
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 3.14587593728698e-06,
+      "loss": 0.8224,
+      "step": 17720
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 3.1288343558282206e-06,
+      "loss": 0.7954,
+      "step": 17730
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 3.1117927743694616e-06,
+      "loss": 0.7191,
+      "step": 17740
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 3.096455351056578e-06,
+      "loss": 0.7584,
+      "step": 17750
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 3.0794137695978185e-06,
+      "loss": 0.745,
+      "step": 17760
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 3.062372188139059e-06,
+      "loss": 0.7082,
+      "step": 17770
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 3.0453306066802997e-06,
+      "loss": 0.7523,
+      "step": 17780
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 3.028289025221541e-06,
+      "loss": 0.7169,
+      "step": 17790
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 3.0112474437627814e-06,
+      "loss": 0.7709,
+      "step": 17800
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.994205862304022e-06,
+      "loss": 0.7897,
+      "step": 17810
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.9771642808452627e-06,
+      "loss": 0.7276,
+      "step": 17820
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.9601226993865034e-06,
+      "loss": 0.777,
+      "step": 17830
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.943081117927744e-06,
+      "loss": 0.7312,
+      "step": 17840
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.9260395364689846e-06,
+      "loss": 0.6812,
+      "step": 17850
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.908997955010225e-06,
+      "loss": 0.7237,
+      "step": 17860
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.8919563735514655e-06,
+      "loss": 0.7103,
+      "step": 17870
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.874914792092706e-06,
+      "loss": 0.7461,
+      "step": 17880
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.8578732106339468e-06,
+      "loss": 0.7263,
+      "step": 17890
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.8408316291751874e-06,
+      "loss": 0.6706,
+      "step": 17900
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.823790047716428e-06,
+      "loss": 0.7347,
+      "step": 17910
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.8067484662576687e-06,
+      "loss": 0.7344,
+      "step": 17920
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.7897068847989093e-06,
+      "loss": 0.7849,
+      "step": 17930
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.77266530334015e-06,
+      "loss": 0.779,
+      "step": 17940
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.7556237218813906e-06,
+      "loss": 0.7517,
+      "step": 17950
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.7385821404226317e-06,
+      "loss": 0.7596,
+      "step": 17960
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.721540558963872e-06,
+      "loss": 0.7379,
+      "step": 17970
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.7044989775051125e-06,
+      "loss": 0.7425,
+      "step": 17980
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.687457396046353e-06,
+      "loss": 0.7405,
+      "step": 17990
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.670415814587594e-06,
+      "loss": 0.8089,
+      "step": 18000
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.6533742331288345e-06,
+      "loss": 0.6523,
+      "step": 18010
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.636332651670075e-06,
+      "loss": 0.7522,
+      "step": 18020
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.6192910702113157e-06,
+      "loss": 0.7518,
+      "step": 18030
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.6022494887525564e-06,
+      "loss": 0.7536,
+      "step": 18040
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.585207907293797e-06,
+      "loss": 0.7448,
+      "step": 18050
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.5681663258350376e-06,
+      "loss": 0.7135,
+      "step": 18060
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.5511247443762783e-06,
+      "loss": 0.7462,
+      "step": 18070
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.534083162917519e-06,
+      "loss": 0.6735,
+      "step": 18080
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.517041581458759e-06,
+      "loss": 0.7791,
+      "step": 18090
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.4999999999999998e-06,
+      "loss": 0.7464,
+      "step": 18100
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.4829584185412404e-06,
+      "loss": 0.7651,
+      "step": 18110
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.465916837082481e-06,
+      "loss": 0.7408,
+      "step": 18120
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.448875255623722e-06,
+      "loss": 0.7626,
+      "step": 18130
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.4318336741649628e-06,
+      "loss": 0.7414,
+      "step": 18140
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.4147920927062034e-06,
+      "loss": 0.7833,
+      "step": 18150
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.397750511247444e-06,
+      "loss": 0.733,
+      "step": 18160
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.3807089297886847e-06,
+      "loss": 0.7799,
+      "step": 18170
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.3636673483299253e-06,
+      "loss": 0.7348,
+      "step": 18180
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.346625766871166e-06,
+      "loss": 0.7933,
+      "step": 18190
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.329584185412406e-06,
+      "loss": 0.7737,
+      "step": 18200
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.312542603953647e-06,
+      "loss": 0.7061,
+      "step": 18210
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.2955010224948875e-06,
+      "loss": 0.776,
+      "step": 18220
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.278459441036128e-06,
+      "loss": 0.7045,
+      "step": 18230
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.2614178595773687e-06,
+      "loss": 0.7119,
+      "step": 18240
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.2443762781186094e-06,
+      "loss": 0.7808,
+      "step": 18250
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.7511918165920034,
+      "eval_loss": 0.7764942646026611,
+      "eval_runtime": 309.073,
+      "eval_samples_per_second": 135.059,
+      "eval_steps_per_second": 67.531,
+      "step": 18256
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.22733469665985e-06,
+      "loss": 0.8634,
+      "step": 18260
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.2102931152010906e-06,
+      "loss": 0.689,
+      "step": 18270
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1932515337423313e-06,
+      "loss": 0.7903,
+      "step": 18280
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.176209952283572e-06,
+      "loss": 0.7453,
+      "step": 18290
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1591683708248126e-06,
+      "loss": 0.7566,
+      "step": 18300
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.142126789366053e-06,
+      "loss": 0.7648,
+      "step": 18310
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.125085207907294e-06,
+      "loss": 0.7727,
+      "step": 18320
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1080436264485345e-06,
+      "loss": 0.657,
+      "step": 18330
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.091002044989775e-06,
+      "loss": 0.7296,
+      "step": 18340
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.0739604635310158e-06,
+      "loss": 0.7024,
+      "step": 18350
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.0569188820722564e-06,
+      "loss": 0.7229,
+      "step": 18360
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.039877300613497e-06,
+      "loss": 0.7341,
+      "step": 18370
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.0228357191547377e-06,
+      "loss": 0.7707,
+      "step": 18380
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.0057941376959783e-06,
+      "loss": 0.708,
+      "step": 18390
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.988752556237219e-06,
+      "loss": 0.7605,
+      "step": 18400
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.9717109747784596e-06,
+      "loss": 0.7309,
+      "step": 18410
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.9546693933197002e-06,
+      "loss": 0.746,
+      "step": 18420
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.9376278118609405e-06,
+      "loss": 0.7619,
+      "step": 18430
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.920586230402181e-06,
+      "loss": 0.7893,
+      "step": 18440
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.903544648943422e-06,
+      "loss": 0.7514,
+      "step": 18450
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.8865030674846626e-06,
+      "loss": 0.7031,
+      "step": 18460
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.8694614860259032e-06,
+      "loss": 0.7663,
+      "step": 18470
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.8524199045671439e-06,
+      "loss": 0.7959,
+      "step": 18480
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.8353783231083845e-06,
+      "loss": 0.7158,
+      "step": 18490
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.8183367416496251e-06,
+      "loss": 0.6885,
+      "step": 18500
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.8012951601908656e-06,
+      "loss": 0.7124,
+      "step": 18510
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.7842535787321064e-06,
+      "loss": 0.7218,
+      "step": 18520
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.767211997273347e-06,
+      "loss": 0.6939,
+      "step": 18530
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.7501704158145877e-06,
+      "loss": 0.7711,
+      "step": 18540
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.7331288343558283e-06,
+      "loss": 0.7044,
+      "step": 18550
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.716087252897069e-06,
+      "loss": 0.6821,
+      "step": 18560
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.6990456714383094e-06,
+      "loss": 0.6766,
+      "step": 18570
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.68200408997955e-06,
+      "loss": 0.7228,
+      "step": 18580
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.6649625085207907e-06,
+      "loss": 0.6671,
+      "step": 18590
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.6479209270620315e-06,
+      "loss": 0.7403,
+      "step": 18600
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.6308793456032722e-06,
+      "loss": 0.7248,
+      "step": 18610
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.6138377641445126e-06,
+      "loss": 0.7729,
+      "step": 18620
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.5967961826857532e-06,
+      "loss": 0.7139,
+      "step": 18630
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.5797546012269939e-06,
+      "loss": 0.728,
+      "step": 18640
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.5627130197682345e-06,
+      "loss": 0.7886,
+      "step": 18650
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.5456714383094752e-06,
+      "loss": 0.7495,
+      "step": 18660
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.5286298568507158e-06,
+      "loss": 0.7445,
+      "step": 18670
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.5115882753919562e-06,
+      "loss": 0.7044,
+      "step": 18680
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.494546693933197e-06,
+      "loss": 0.7799,
+      "step": 18690
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4775051124744377e-06,
+      "loss": 0.6863,
+      "step": 18700
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4604635310156784e-06,
+      "loss": 0.7484,
+      "step": 18710
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.443421949556919e-06,
+      "loss": 0.7498,
+      "step": 18720
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4263803680981596e-06,
+      "loss": 0.7772,
+      "step": 18730
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4093387866394e-06,
+      "loss": 0.7765,
+      "step": 18740
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.3922972051806407e-06,
+      "loss": 0.7141,
+      "step": 18750
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.3752556237218813e-06,
+      "loss": 0.8163,
+      "step": 18760
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.358214042263122e-06,
+      "loss": 0.7491,
+      "step": 18770
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.3411724608043628e-06,
+      "loss": 0.7195,
+      "step": 18780
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.3241308793456035e-06,
+      "loss": 0.7725,
+      "step": 18790
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.307089297886844e-06,
+      "loss": 0.727,
+      "step": 18800
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.2900477164280845e-06,
+      "loss": 0.7155,
+      "step": 18810
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.2730061349693252e-06,
+      "loss": 0.733,
+      "step": 18820
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.2559645535105658e-06,
+      "loss": 0.763,
+      "step": 18830
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.2389229720518065e-06,
+      "loss": 0.7589,
+      "step": 18840
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.2218813905930469e-06,
+      "loss": 0.7784,
+      "step": 18850
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.2048398091342877e-06,
+      "loss": 0.6912,
+      "step": 18860
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.1877982276755284e-06,
+      "loss": 0.7437,
+      "step": 18870
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.170756646216769e-06,
+      "loss": 0.7583,
+      "step": 18880
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.1537150647580097e-06,
+      "loss": 0.7193,
+      "step": 18890
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.1366734832992503e-06,
+      "loss": 0.7655,
+      "step": 18900
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.1196319018404907e-06,
+      "loss": 0.7586,
+      "step": 18910
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.1025903203817314e-06,
+      "loss": 0.7065,
+      "step": 18920
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.085548738922972e-06,
+      "loss": 0.7404,
+      "step": 18930
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.0685071574642126e-06,
+      "loss": 0.691,
+      "step": 18940
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.0514655760054535e-06,
+      "loss": 0.7284,
+      "step": 18950
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.0344239945466941e-06,
+      "loss": 0.7216,
+      "step": 18960
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.0190865712338105e-06,
+      "loss": 0.6977,
+      "step": 18970
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.0020449897750512e-06,
+      "loss": 0.6854,
+      "step": 18980
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 9.850034083162916e-07,
+      "loss": 0.76,
+      "step": 18990
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 9.679618268575324e-07,
+      "loss": 0.6916,
+      "step": 19000
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 9.50920245398773e-07,
+      "loss": 0.7113,
+      "step": 19010
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.338786639400137e-07,
+      "loss": 0.7467,
+      "step": 19020
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.168370824812543e-07,
+      "loss": 0.7814,
+      "step": 19030
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.997955010224949e-07,
+      "loss": 0.8037,
+      "step": 19040
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 8.827539195637355e-07,
+      "loss": 0.767,
+      "step": 19050
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.657123381049762e-07,
+      "loss": 0.662,
+      "step": 19060
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.486707566462168e-07,
+      "loss": 0.7162,
+      "step": 19070
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.316291751874575e-07,
+      "loss": 0.8016,
+      "step": 19080
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.14587593728698e-07,
+      "loss": 0.7118,
+      "step": 19090
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 7.975460122699386e-07,
+      "loss": 0.7519,
+      "step": 19100
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 7.805044308111794e-07,
+      "loss": 0.7184,
+      "step": 19110
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 7.634628493524199e-07,
+      "loss": 0.7926,
+      "step": 19120
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 7.464212678936605e-07,
+      "loss": 0.7689,
+      "step": 19130
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 7.293796864349011e-07,
+      "loss": 0.7286,
+      "step": 19140
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 7.123381049761418e-07,
+      "loss": 0.7351,
+      "step": 19150
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 6.952965235173825e-07,
+      "loss": 0.7235,
+      "step": 19160
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 6.78254942058623e-07,
+      "loss": 0.7183,
+      "step": 19170
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 6.612133605998636e-07,
+      "loss": 0.7438,
+      "step": 19180
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 6.441717791411044e-07,
+      "loss": 0.753,
+      "step": 19190
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 6.271301976823449e-07,
+      "loss": 0.7386,
+      "step": 19200
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 6.100886162235856e-07,
+      "loss": 0.8085,
+      "step": 19210
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 5.930470347648262e-07,
+      "loss": 0.6876,
+      "step": 19220
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 5.760054533060668e-07,
+      "loss": 0.7679,
+      "step": 19230
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 5.589638718473075e-07,
+      "loss": 0.7035,
+      "step": 19240
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 5.419222903885481e-07,
+      "loss": 0.7267,
+      "step": 19250
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 5.248807089297886e-07,
+      "loss": 0.7026,
+      "step": 19260
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 5.078391274710293e-07,
+      "loss": 0.7405,
+      "step": 19270
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.9079754601227e-07,
+      "loss": 0.7591,
+      "step": 19280
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.7375596455351057e-07,
+      "loss": 0.7281,
+      "step": 19290
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 4.567143830947512e-07,
+      "loss": 0.6897,
+      "step": 19300
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.3967280163599184e-07,
+      "loss": 0.7942,
+      "step": 19310
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.2263122017723243e-07,
+      "loss": 0.8101,
+      "step": 19320
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.055896387184731e-07,
+      "loss": 0.8037,
+      "step": 19330
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 3.885480572597137e-07,
+      "loss": 0.7616,
+      "step": 19340
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 3.7150647580095435e-07,
+      "loss": 0.6892,
+      "step": 19350
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 3.5446489434219494e-07,
+      "loss": 0.7434,
+      "step": 19360
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 3.3742331288343563e-07,
+      "loss": 0.7315,
+      "step": 19370
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 3.203817314246762e-07,
+      "loss": 0.724,
+      "step": 19380
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 3.033401499659168e-07,
+      "loss": 0.7304,
+      "step": 19390
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.862985685071575e-07,
+      "loss": 0.7184,
+      "step": 19400
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.692569870483981e-07,
+      "loss": 0.7777,
+      "step": 19410
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.522154055896387e-07,
+      "loss": 0.7664,
+      "step": 19420
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.3517382413087938e-07,
+      "loss": 0.6895,
+      "step": 19430
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1813224267212e-07,
+      "loss": 0.7081,
+      "step": 19440
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.010906612133606e-07,
+      "loss": 0.7956,
+      "step": 19450
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.8404907975460125e-07,
+      "loss": 0.6856,
+      "step": 19460
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.6700749829584184e-07,
+      "loss": 0.7017,
+      "step": 19470
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.4996591683708247e-07,
+      "loss": 0.7615,
+      "step": 19480
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.3292433537832311e-07,
+      "loss": 0.7505,
+      "step": 19490
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.1588275391956373e-07,
+      "loss": 0.7866,
+      "step": 19500
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 9.884117246080437e-08,
+      "loss": 0.7017,
+      "step": 19510
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 8.179959100204499e-08,
+      "loss": 0.7109,
+      "step": 19520
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 6.475800954328562e-08,
+      "loss": 0.7251,
+      "step": 19530
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.7716428084526244e-08,
+      "loss": 0.7563,
+      "step": 19540
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 3.067484662576688e-08,
+      "loss": 0.7924,
+      "step": 19550
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.36332651670075e-08,
+      "loss": 0.7359,
+      "step": 19560
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.7569891957933067,
+      "eval_loss": 0.756607711315155,
+      "eval_runtime": 307.9711,
+      "eval_samples_per_second": 135.542,
+      "eval_steps_per_second": 67.773,
+      "step": 19560
+    },
+    {
+      "epoch": 15.0,
+      "step": 19560,
+      "total_flos": 3.434575406726067e+19,
+      "train_loss": 0.92946919892707,
+      "train_runtime": 14238.5964,
+      "train_samples_per_second": 175.898,
+      "train_steps_per_second": 1.374
+    }
+  ],
+  "max_steps": 19560,
+  "num_train_epochs": 15,
+  "total_flos": 3.434575406726067e+19,
+  "trial_name": null,
+  "trial_params": null
+}