diff --git "a/checkpoint-351000/trainer_state.json" "b/checkpoint-351000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-351000/trainer_state.json"
@@ -0,0 +1,5290 @@
+{
+  "best_metric": 0.9368385076522827,
+  "best_model_checkpoint": "/kaggle/output/checkpoint-322000",
+  "epoch": 14.30084745762712,
+  "eval_steps": 1000,
+  "global_step": 351000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7777777777777777e-11,
+      "loss": 1.1078,
+      "step": 1
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.7750000000000004e-08,
+      "loss": 1.128,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "eval_accuracy": 0.330938123752495,
+      "eval_loss": 1.1083118915557861,
+      "eval_runtime": 12.3979,
+      "eval_samples_per_second": 404.1,
+      "eval_steps_per_second": 50.573,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.5527777777777784e-08,
+      "loss": 1.1164,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "eval_accuracy": 0.3323353293413174,
+      "eval_loss": 1.102672815322876,
+      "eval_runtime": 12.2861,
+      "eval_samples_per_second": 407.778,
+      "eval_steps_per_second": 51.033,
+      "step": 2000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.327777777777778e-08,
+      "loss": 1.1158,
+      "step": 3000
+    },
+    {
+      "epoch": 0.12,
+      "eval_accuracy": 0.34211576846307384,
+      "eval_loss": 1.1024774312973022,
+      "eval_runtime": 12.3746,
+      "eval_samples_per_second": 404.863,
+      "eval_steps_per_second": 50.668,
+      "step": 3000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.1105555555555557e-07,
+      "loss": 1.1115,
+      "step": 4000
+    },
+    {
+      "epoch": 0.16,
+      "eval_accuracy": 0.34890219560878244,
+      "eval_loss": 1.1012382507324219,
+      "eval_runtime": 12.5405,
+      "eval_samples_per_second": 399.505,
+      "eval_steps_per_second": 49.998,
+      "step": 4000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.3880555555555558e-07,
+      "loss": 1.1141,
+      "step": 5000
+    },
+    {
+      "epoch": 0.2,
+      "eval_accuracy": 0.3534930139720559,
+      "eval_loss": 1.1002192497253418,
+      "eval_runtime": 12.3001,
+      "eval_samples_per_second": 407.314,
+      "eval_steps_per_second": 50.975,
+      "step": 5000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.6658333333333335e-07,
+      "loss": 1.112,
+      "step": 6000
+    },
+    {
+      "epoch": 0.24,
+      "eval_accuracy": 0.3524950099800399,
+      "eval_loss": 1.100056767463684,
+      "eval_runtime": 12.4062,
+      "eval_samples_per_second": 403.831,
+      "eval_steps_per_second": 50.539,
+      "step": 6000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9433333333333334e-07,
+      "loss": 1.1115,
+      "step": 7000
+    },
+    {
+      "epoch": 0.29,
+      "eval_accuracy": 0.34331337325349304,
+      "eval_loss": 1.098986029624939,
+      "eval_runtime": 12.4456,
+      "eval_samples_per_second": 402.551,
+      "eval_steps_per_second": 50.379,
+      "step": 7000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.2211111111111114e-07,
+      "loss": 1.109,
+      "step": 8000
+    },
+    {
+      "epoch": 0.33,
+      "eval_accuracy": 0.3347305389221557,
+      "eval_loss": 1.098678469657898,
+      "eval_runtime": 12.5978,
+      "eval_samples_per_second": 397.688,
+      "eval_steps_per_second": 49.771,
+      "step": 8000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.4986111111111113e-07,
+      "loss": 1.1117,
+      "step": 9000
+    },
+    {
+      "epoch": 0.37,
+      "eval_accuracy": 0.34311377245508984,
+      "eval_loss": 1.0984419584274292,
+      "eval_runtime": 12.313,
+      "eval_samples_per_second": 406.886,
+      "eval_steps_per_second": 50.922,
+      "step": 9000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.776388888888889e-07,
+      "loss": 1.1081,
+      "step": 10000
+    },
+    {
+      "epoch": 0.41,
+      "eval_accuracy": 0.35429141716566864,
+      "eval_loss": 1.0968743562698364,
+      "eval_runtime": 12.4088,
+      "eval_samples_per_second": 403.744,
+      "eval_steps_per_second": 50.528,
+      "step": 10000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.053888888888889e-07,
+      "loss": 1.107,
+      "step": 11000
+    },
+    {
+      "epoch": 0.45,
+      "eval_accuracy": 0.3722554890219561,
+      "eval_loss": 1.0950665473937988,
+      "eval_runtime": 12.5775,
+      "eval_samples_per_second": 398.33,
+      "eval_steps_per_second": 49.851,
+      "step": 11000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.331666666666667e-07,
+      "loss": 1.1057,
+      "step": 12000
+    },
+    {
+      "epoch": 0.49,
+      "eval_accuracy": 0.37544910179640717,
+      "eval_loss": 1.0940738916397095,
+      "eval_runtime": 12.5186,
+      "eval_samples_per_second": 400.203,
+      "eval_steps_per_second": 50.085,
+      "step": 12000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.609166666666667e-07,
+      "loss": 1.1069,
+      "step": 13000
+    },
+    {
+      "epoch": 0.53,
+      "eval_accuracy": 0.3401197604790419,
+      "eval_loss": 1.0937162637710571,
+      "eval_runtime": 12.2594,
+      "eval_samples_per_second": 408.666,
+      "eval_steps_per_second": 51.144,
+      "step": 13000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.886944444444445e-07,
+      "loss": 1.1046,
+      "step": 14000
+    },
+    {
+      "epoch": 0.57,
+      "eval_accuracy": 0.37604790419161677,
+      "eval_loss": 1.0926916599273682,
+      "eval_runtime": 12.256,
+      "eval_samples_per_second": 408.78,
+      "eval_steps_per_second": 51.159,
+      "step": 14000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.164444444444445e-07,
+      "loss": 1.1046,
+      "step": 15000
+    },
+    {
+      "epoch": 0.61,
+      "eval_accuracy": 0.39201596806387223,
+      "eval_loss": 1.0912976264953613,
+      "eval_runtime": 12.5019,
+      "eval_samples_per_second": 400.74,
+      "eval_steps_per_second": 50.152,
+      "step": 15000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.442222222222223e-07,
+      "loss": 1.1052,
+      "step": 16000
+    },
+    {
+      "epoch": 0.65,
+      "eval_accuracy": 0.393812375249501,
+      "eval_loss": 1.0901482105255127,
+      "eval_runtime": 12.3407,
+      "eval_samples_per_second": 405.973,
+      "eval_steps_per_second": 50.807,
+      "step": 16000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7197222222222224e-07,
+      "loss": 1.1011,
+      "step": 17000
+    },
+    {
+      "epoch": 0.69,
+      "eval_accuracy": 0.3932135728542914,
+      "eval_loss": 1.0901678800582886,
+      "eval_runtime": 12.3644,
+      "eval_samples_per_second": 405.197,
+      "eval_steps_per_second": 50.71,
+      "step": 17000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.997500000000001e-07,
+      "loss": 1.1,
+      "step": 18000
+    },
+    {
+      "epoch": 0.73,
+      "eval_accuracy": 0.38143712574850297,
+      "eval_loss": 1.0897998809814453,
+      "eval_runtime": 12.3516,
+      "eval_samples_per_second": 405.615,
+      "eval_steps_per_second": 50.763,
+      "step": 18000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.275277777777778e-07,
+      "loss": 1.1007,
+      "step": 19000
+    },
+    {
+      "epoch": 0.77,
+      "eval_accuracy": 0.40678642714570856,
+      "eval_loss": 1.0876121520996094,
+      "eval_runtime": 12.6608,
+      "eval_samples_per_second": 395.708,
+      "eval_steps_per_second": 49.523,
+      "step": 19000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.552777777777778e-07,
+      "loss": 1.1013,
+      "step": 20000
+    },
+    {
+      "epoch": 0.81,
+      "eval_accuracy": 0.4053892215568862,
+      "eval_loss": 1.0876350402832031,
+      "eval_runtime": 12.3243,
+      "eval_samples_per_second": 406.513,
+      "eval_steps_per_second": 50.875,
+      "step": 20000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.830555555555556e-07,
+      "loss": 1.0987,
+      "step": 21000
+    },
+    {
+      "epoch": 0.86,
+      "eval_accuracy": 0.40858283433133735,
+      "eval_loss": 1.085940957069397,
+      "eval_runtime": 12.3217,
+      "eval_samples_per_second": 406.598,
+      "eval_steps_per_second": 50.886,
+      "step": 21000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.108055555555556e-07,
+      "loss": 1.0986,
+      "step": 22000
+    },
+    {
+      "epoch": 0.9,
+      "eval_accuracy": 0.3972055888223553,
+      "eval_loss": 1.085978388786316,
+      "eval_runtime": 12.4285,
+      "eval_samples_per_second": 403.107,
+      "eval_steps_per_second": 50.449,
+      "step": 22000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.385833333333334e-07,
+      "loss": 1.0944,
+      "step": 23000
+    },
+    {
+      "epoch": 0.94,
+      "eval_accuracy": 0.41097804391217563,
+      "eval_loss": 1.0837496519088745,
+      "eval_runtime": 12.7175,
+      "eval_samples_per_second": 393.946,
+      "eval_steps_per_second": 49.302,
+      "step": 23000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.663611111111112e-07,
+      "loss": 1.095,
+      "step": 24000
+    },
+    {
+      "epoch": 0.98,
+      "eval_accuracy": 0.4121756487025948,
+      "eval_loss": 1.0820704698562622,
+      "eval_runtime": 12.3323,
+      "eval_samples_per_second": 406.251,
+      "eval_steps_per_second": 50.842,
+      "step": 24000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.941111111111112e-07,
+      "loss": 1.0915,
+      "step": 25000
+    },
+    {
+      "epoch": 1.02,
+      "eval_accuracy": 0.4177644710578842,
+      "eval_loss": 1.0762931108474731,
+      "eval_runtime": 12.3887,
+      "eval_samples_per_second": 404.399,
+      "eval_steps_per_second": 50.61,
+      "step": 25000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 7.218888888888889e-07,
+      "loss": 1.0889,
+      "step": 26000
+    },
+    {
+      "epoch": 1.06,
+      "eval_accuracy": 0.43253493013972055,
+      "eval_loss": 1.0695326328277588,
+      "eval_runtime": 12.5611,
+      "eval_samples_per_second": 398.85,
+      "eval_steps_per_second": 49.916,
+      "step": 26000
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.49638888888889e-07,
+      "loss": 1.0827,
+      "step": 27000
+    },
+    {
+      "epoch": 1.1,
+      "eval_accuracy": 0.4343313373253493,
+      "eval_loss": 1.0662548542022705,
+      "eval_runtime": 12.641,
+      "eval_samples_per_second": 396.328,
+      "eval_steps_per_second": 49.6,
+      "step": 27000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.774166666666668e-07,
+      "loss": 1.0765,
+      "step": 28000
+    },
+    {
+      "epoch": 1.14,
+      "eval_accuracy": 0.4315369261477046,
+      "eval_loss": 1.0609794855117798,
+      "eval_runtime": 12.4637,
+      "eval_samples_per_second": 401.967,
+      "eval_steps_per_second": 50.306,
+      "step": 28000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.051666666666667e-07,
+      "loss": 1.0736,
+      "step": 29000
+    },
+    {
+      "epoch": 1.18,
+      "eval_accuracy": 0.4377245508982036,
+      "eval_loss": 1.0587859153747559,
+      "eval_runtime": 12.3801,
+      "eval_samples_per_second": 404.683,
+      "eval_steps_per_second": 50.646,
+      "step": 29000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.329444444444445e-07,
+      "loss": 1.0726,
+      "step": 30000
+    },
+    {
+      "epoch": 1.22,
+      "eval_accuracy": 0.4249500998003992,
+      "eval_loss": 1.0559338331222534,
+      "eval_runtime": 12.5945,
+      "eval_samples_per_second": 397.792,
+      "eval_steps_per_second": 49.784,
+      "step": 30000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.606944444444445e-07,
+      "loss": 1.0751,
+      "step": 31000
+    },
+    {
+      "epoch": 1.26,
+      "eval_accuracy": 0.4427145708582834,
+      "eval_loss": 1.0576916933059692,
+      "eval_runtime": 12.6324,
+      "eval_samples_per_second": 396.599,
+      "eval_steps_per_second": 49.634,
+      "step": 31000
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.884722222222224e-07,
+      "loss": 1.0708,
+      "step": 32000
+    },
+    {
+      "epoch": 1.3,
+      "eval_accuracy": 0.42634730538922155,
+      "eval_loss": 1.0571542978286743,
+      "eval_runtime": 12.4696,
+      "eval_samples_per_second": 401.776,
+      "eval_steps_per_second": 50.282,
+      "step": 32000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.162222222222223e-07,
+      "loss": 1.0711,
+      "step": 33000
+    },
+    {
+      "epoch": 1.34,
+      "eval_accuracy": 0.4419161676646707,
+      "eval_loss": 1.0544849634170532,
+      "eval_runtime": 12.415,
+      "eval_samples_per_second": 403.544,
+      "eval_steps_per_second": 50.503,
+      "step": 33000
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.440000000000001e-07,
+      "loss": 1.0666,
+      "step": 34000
+    },
+    {
+      "epoch": 1.39,
+      "eval_accuracy": 0.43333333333333335,
+      "eval_loss": 1.05018150806427,
+      "eval_runtime": 12.6226,
+      "eval_samples_per_second": 396.906,
+      "eval_steps_per_second": 49.673,
+      "step": 34000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.7175e-07,
+      "loss": 1.0698,
+      "step": 35000
+    },
+    {
+      "epoch": 1.43,
+      "eval_accuracy": 0.4377245508982036,
+      "eval_loss": 1.0488406419754028,
+      "eval_runtime": 12.6246,
+      "eval_samples_per_second": 396.844,
+      "eval_steps_per_second": 49.665,
+      "step": 35000
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.995277777777778e-07,
+      "loss": 1.067,
+      "step": 36000
+    },
+    {
+      "epoch": 1.47,
+      "eval_accuracy": 0.4407185628742515,
+      "eval_loss": 1.047359824180603,
+      "eval_runtime": 12.4039,
+      "eval_samples_per_second": 403.906,
+      "eval_steps_per_second": 50.549,
+      "step": 36000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.027277777777778e-06,
+      "loss": 1.0636,
+      "step": 37000
+    },
+    {
+      "epoch": 1.51,
+      "eval_accuracy": 0.44451097804391215,
+      "eval_loss": 1.0455204248428345,
+      "eval_runtime": 12.3836,
+      "eval_samples_per_second": 404.569,
+      "eval_steps_per_second": 50.632,
+      "step": 37000
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.0550555555555557e-06,
+      "loss": 1.0658,
+      "step": 38000
+    },
+    {
+      "epoch": 1.55,
+      "eval_accuracy": 0.4407185628742515,
+      "eval_loss": 1.0461794137954712,
+      "eval_runtime": 12.5546,
+      "eval_samples_per_second": 399.057,
+      "eval_steps_per_second": 49.942,
+      "step": 38000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.0828333333333334e-06,
+      "loss": 1.0635,
+      "step": 39000
+    },
+    {
+      "epoch": 1.59,
+      "eval_accuracy": 0.4367265469061876,
+      "eval_loss": 1.0461270809173584,
+      "eval_runtime": 12.5509,
+      "eval_samples_per_second": 399.173,
+      "eval_steps_per_second": 49.956,
+      "step": 39000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.1105833333333335e-06,
+      "loss": 1.0596,
+      "step": 40000
+    },
+    {
+      "epoch": 1.63,
+      "eval_accuracy": 0.449500998003992,
+      "eval_loss": 1.0418727397918701,
+      "eval_runtime": 12.4044,
+      "eval_samples_per_second": 403.889,
+      "eval_steps_per_second": 50.547,
+      "step": 40000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.1383611111111113e-06,
+      "loss": 1.0595,
+      "step": 41000
+    },
+    {
+      "epoch": 1.67,
+      "eval_accuracy": 0.4409181636726547,
+      "eval_loss": 1.0417557954788208,
+      "eval_runtime": 12.3642,
+      "eval_samples_per_second": 405.202,
+      "eval_steps_per_second": 50.711,
+      "step": 41000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.1661111111111111e-06,
+      "loss": 1.054,
+      "step": 42000
+    },
+    {
+      "epoch": 1.71,
+      "eval_accuracy": 0.4357285429141717,
+      "eval_loss": 1.0383535623550415,
+      "eval_runtime": 12.3634,
+      "eval_samples_per_second": 405.23,
+      "eval_steps_per_second": 50.714,
+      "step": 42000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.193888888888889e-06,
+      "loss": 1.0543,
+      "step": 43000
+    },
+    {
+      "epoch": 1.75,
+      "eval_accuracy": 0.43812375249500995,
+      "eval_loss": 1.0463169813156128,
+      "eval_runtime": 12.6398,
+      "eval_samples_per_second": 396.366,
+      "eval_steps_per_second": 49.605,
+      "step": 43000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.221638888888889e-06,
+      "loss": 1.0579,
+      "step": 44000
+    },
+    {
+      "epoch": 1.79,
+      "eval_accuracy": 0.4415169660678643,
+      "eval_loss": 1.0388957262039185,
+      "eval_runtime": 12.4542,
+      "eval_samples_per_second": 402.274,
+      "eval_steps_per_second": 50.344,
+      "step": 44000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.2494166666666668e-06,
+      "loss": 1.0561,
+      "step": 45000
+    },
+    {
+      "epoch": 1.83,
+      "eval_accuracy": 0.46187624750499,
+      "eval_loss": 1.031552791595459,
+      "eval_runtime": 12.4082,
+      "eval_samples_per_second": 403.764,
+      "eval_steps_per_second": 50.531,
+      "step": 45000
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2771666666666668e-06,
+      "loss": 1.0516,
+      "step": 46000
+    },
+    {
+      "epoch": 1.87,
+      "eval_accuracy": 0.4538922155688623,
+      "eval_loss": 1.0312999486923218,
+      "eval_runtime": 12.3695,
+      "eval_samples_per_second": 405.029,
+      "eval_steps_per_second": 50.689,
+      "step": 46000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3049444444444446e-06,
+      "loss": 1.0471,
+      "step": 47000
+    },
+    {
+      "epoch": 1.91,
+      "eval_accuracy": 0.46387225548902195,
+      "eval_loss": 1.027535319328308,
+      "eval_runtime": 12.5846,
+      "eval_samples_per_second": 398.107,
+      "eval_steps_per_second": 49.823,
+      "step": 47000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.3326944444444447e-06,
+      "loss": 1.054,
+      "step": 48000
+    },
+    {
+      "epoch": 1.96,
+      "eval_accuracy": 0.4568862275449102,
+      "eval_loss": 1.0308438539505005,
+      "eval_runtime": 12.5646,
+      "eval_samples_per_second": 398.74,
+      "eval_steps_per_second": 49.902,
+      "step": 48000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.3604722222222224e-06,
+      "loss": 1.0445,
+      "step": 49000
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.47045908183632734,
+      "eval_loss": 1.0230755805969238,
+      "eval_runtime": 12.4305,
+      "eval_samples_per_second": 403.041,
+      "eval_steps_per_second": 50.44,
+      "step": 49000
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.3882222222222223e-06,
+      "loss": 1.0479,
+      "step": 50000
+    },
+    {
+      "epoch": 2.04,
+      "eval_accuracy": 0.47305389221556887,
+      "eval_loss": 1.0201445817947388,
+      "eval_runtime": 12.3672,
+      "eval_samples_per_second": 405.103,
+      "eval_steps_per_second": 50.698,
+      "step": 50000
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.416e-06,
+      "loss": 1.0482,
+      "step": 51000
+    },
+    {
+      "epoch": 2.08,
+      "eval_accuracy": 0.4678642714570858,
+      "eval_loss": 1.0167714357376099,
+      "eval_runtime": 12.5982,
+      "eval_samples_per_second": 397.677,
+      "eval_steps_per_second": 49.769,
+      "step": 51000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.4437500000000002e-06,
+      "loss": 1.0406,
+      "step": 52000
+    },
+    {
+      "epoch": 2.12,
+      "eval_accuracy": 0.4722554890219561,
+      "eval_loss": 1.0175116062164307,
+      "eval_runtime": 12.6041,
+      "eval_samples_per_second": 397.49,
+      "eval_steps_per_second": 49.746,
+      "step": 52000
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.471527777777778e-06,
+      "loss": 1.0457,
+      "step": 53000
+    },
+    {
+      "epoch": 2.16,
+      "eval_accuracy": 0.4748502994011976,
+      "eval_loss": 1.010857343673706,
+      "eval_runtime": 12.3728,
+      "eval_samples_per_second": 404.921,
+      "eval_steps_per_second": 50.676,
+      "step": 53000
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.4993055555555557e-06,
+      "loss": 1.0399,
+      "step": 54000
+    },
+    {
+      "epoch": 2.2,
+      "eval_accuracy": 0.48522954091816367,
+      "eval_loss": 1.0106219053268433,
+      "eval_runtime": 12.3952,
+      "eval_samples_per_second": 404.188,
+      "eval_steps_per_second": 50.584,
+      "step": 54000
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.5270555555555558e-06,
+      "loss": 1.0412,
+      "step": 55000
+    },
+    {
+      "epoch": 2.24,
+      "eval_accuracy": 0.4750499001996008,
+      "eval_loss": 1.0111658573150635,
+      "eval_runtime": 12.4858,
+      "eval_samples_per_second": 401.257,
+      "eval_steps_per_second": 50.217,
+      "step": 55000
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.5548333333333335e-06,
+      "loss": 1.0347,
+      "step": 56000
+    },
+    {
+      "epoch": 2.28,
+      "eval_accuracy": 0.47924151696606787,
+      "eval_loss": 1.0121456384658813,
+      "eval_runtime": 12.6206,
+      "eval_samples_per_second": 396.97,
+      "eval_steps_per_second": 49.681,
+      "step": 56000
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.5825833333333334e-06,
+      "loss": 1.0324,
+      "step": 57000
+    },
+    {
+      "epoch": 2.32,
+      "eval_accuracy": 0.4738522954091816,
+      "eval_loss": 1.0100611448287964,
+      "eval_runtime": 12.3748,
+      "eval_samples_per_second": 404.857,
+      "eval_steps_per_second": 50.668,
+      "step": 57000
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.6103611111111112e-06,
+      "loss": 1.0346,
+      "step": 58000
+    },
+    {
+      "epoch": 2.36,
+      "eval_accuracy": 0.4694610778443114,
+      "eval_loss": 1.0124626159667969,
+      "eval_runtime": 12.4716,
+      "eval_samples_per_second": 401.711,
+      "eval_steps_per_second": 50.274,
+      "step": 58000
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.638111111111111e-06,
+      "loss": 1.0309,
+      "step": 59000
+    },
+    {
+      "epoch": 2.4,
+      "eval_accuracy": 0.48982035928143713,
+      "eval_loss": 1.0022122859954834,
+      "eval_runtime": 12.4734,
+      "eval_samples_per_second": 401.655,
+      "eval_steps_per_second": 50.267,
+      "step": 59000
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.665888888888889e-06,
+      "loss": 1.0351,
+      "step": 60000
+    },
+    {
+      "epoch": 2.44,
+      "eval_accuracy": 0.4802395209580838,
+      "eval_loss": 1.004167914390564,
+      "eval_runtime": 12.6287,
+      "eval_samples_per_second": 396.714,
+      "eval_steps_per_second": 49.649,
+      "step": 60000
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.693638888888889e-06,
+      "loss": 1.0312,
+      "step": 61000
+    },
+    {
+      "epoch": 2.49,
+      "eval_accuracy": 0.48323353293413174,
+      "eval_loss": 1.0065865516662598,
+      "eval_runtime": 12.4143,
+      "eval_samples_per_second": 403.567,
+      "eval_steps_per_second": 50.506,
+      "step": 61000
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.7214166666666666e-06,
+      "loss": 1.0329,
+      "step": 62000
+    },
+    {
+      "epoch": 2.53,
+      "eval_accuracy": 0.47924151696606787,
+      "eval_loss": 1.0041630268096924,
+      "eval_runtime": 12.4326,
+      "eval_samples_per_second": 402.974,
+      "eval_steps_per_second": 50.432,
+      "step": 62000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.7491666666666667e-06,
+      "loss": 1.0355,
+      "step": 63000
+    },
+    {
+      "epoch": 2.57,
+      "eval_accuracy": 0.49261477045908186,
+      "eval_loss": 1.0007619857788086,
+      "eval_runtime": 12.4051,
+      "eval_samples_per_second": 403.866,
+      "eval_steps_per_second": 50.544,
+      "step": 63000
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.7769444444444447e-06,
+      "loss": 1.0297,
+      "step": 64000
+    },
+    {
+      "epoch": 2.61,
+      "eval_accuracy": 0.4942115768463074,
+      "eval_loss": 0.9986936450004578,
+      "eval_runtime": 12.5727,
+      "eval_samples_per_second": 398.482,
+      "eval_steps_per_second": 49.87,
+      "step": 64000
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.8046944444444446e-06,
+      "loss": 1.0277,
+      "step": 65000
+    },
+    {
+      "epoch": 2.65,
+      "eval_accuracy": 0.49121756487025947,
+      "eval_loss": 1.0008972883224487,
+      "eval_runtime": 12.4544,
+      "eval_samples_per_second": 402.267,
+      "eval_steps_per_second": 50.344,
+      "step": 65000
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.8324722222222223e-06,
+      "loss": 1.0341,
+      "step": 66000
+    },
+    {
+      "epoch": 2.69,
+      "eval_accuracy": 0.499001996007984,
+      "eval_loss": 0.9953575134277344,
+      "eval_runtime": 12.4973,
+      "eval_samples_per_second": 400.886,
+      "eval_steps_per_second": 50.171,
+      "step": 66000
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8602222222222222e-06,
+      "loss": 1.0279,
+      "step": 67000
+    },
+    {
+      "epoch": 2.73,
+      "eval_accuracy": 0.493812375249501,
+      "eval_loss": 0.998134195804596,
+      "eval_runtime": 12.6589,
+      "eval_samples_per_second": 395.769,
+      "eval_steps_per_second": 49.53,
+      "step": 67000
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.8880000000000002e-06,
+      "loss": 1.03,
+      "step": 68000
+    },
+    {
+      "epoch": 2.77,
+      "eval_accuracy": 0.5051896207584831,
+      "eval_loss": 0.9947481751441956,
+      "eval_runtime": 12.7939,
+      "eval_samples_per_second": 391.594,
+      "eval_steps_per_second": 49.008,
+      "step": 68000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9157500000000003e-06,
+      "loss": 1.0283,
+      "step": 69000
+    },
+    {
+      "epoch": 2.81,
+      "eval_accuracy": 0.5001996007984032,
+      "eval_loss": 0.997833251953125,
+      "eval_runtime": 12.8129,
+      "eval_samples_per_second": 391.012,
+      "eval_steps_per_second": 48.935,
+      "step": 69000
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.943527777777778e-06,
+      "loss": 1.0288,
+      "step": 70000
+    },
+    {
+      "epoch": 2.85,
+      "eval_accuracy": 0.49181636726546907,
+      "eval_loss": 1.00169038772583,
+      "eval_runtime": 12.5936,
+      "eval_samples_per_second": 397.822,
+      "eval_steps_per_second": 49.787,
+      "step": 70000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.9712777777777777e-06,
+      "loss": 1.0238,
+      "step": 71000
+    },
+    {
+      "epoch": 2.89,
+      "eval_accuracy": 0.4944111776447106,
+      "eval_loss": 1.001316785812378,
+      "eval_runtime": 12.485,
+      "eval_samples_per_second": 401.282,
+      "eval_steps_per_second": 50.22,
+      "step": 71000
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.9990555555555557e-06,
+      "loss": 1.0268,
+      "step": 72000
+    },
+    {
+      "epoch": 2.93,
+      "eval_accuracy": 0.49241516966067866,
+      "eval_loss": 1.0005468130111694,
+      "eval_runtime": 12.501,
+      "eval_samples_per_second": 400.768,
+      "eval_steps_per_second": 50.156,
+      "step": 72000
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.026805555555556e-06,
+      "loss": 1.0266,
+      "step": 73000
+    },
+    {
+      "epoch": 2.97,
+      "eval_accuracy": 0.4940119760479042,
+      "eval_loss": 0.9998964071273804,
+      "eval_runtime": 12.6227,
+      "eval_samples_per_second": 396.904,
+      "eval_steps_per_second": 49.672,
+      "step": 73000
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0545833333333335e-06,
+      "loss": 1.0222,
+      "step": 74000
+    },
+    {
+      "epoch": 3.01,
+      "eval_accuracy": 0.501996007984032,
+      "eval_loss": 0.9927905201911926,
+      "eval_runtime": 12.7924,
+      "eval_samples_per_second": 391.64,
+      "eval_steps_per_second": 49.014,
+      "step": 74000
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0823333333333334e-06,
+      "loss": 1.0198,
+      "step": 75000
+    },
+    {
+      "epoch": 3.06,
+      "eval_accuracy": 0.5025948103792415,
+      "eval_loss": 0.9925487041473389,
+      "eval_runtime": 12.6297,
+      "eval_samples_per_second": 396.684,
+      "eval_steps_per_second": 49.645,
+      "step": 75000
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.1101111111111113e-06,
+      "loss": 1.0203,
+      "step": 76000
+    },
+    {
+      "epoch": 3.1,
+      "eval_accuracy": 0.5027944111776447,
+      "eval_loss": 0.9909006953239441,
+      "eval_runtime": 12.4265,
+      "eval_samples_per_second": 403.17,
+      "eval_steps_per_second": 50.457,
+      "step": 76000
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.137861111111111e-06,
+      "loss": 1.022,
+      "step": 77000
+    },
+    {
+      "epoch": 3.14,
+      "eval_accuracy": 0.48303393213572854,
+      "eval_loss": 1.010209083557129,
+      "eval_runtime": 12.3999,
+      "eval_samples_per_second": 404.037,
+      "eval_steps_per_second": 50.565,
+      "step": 77000
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.1656388888888888e-06,
+      "loss": 1.0133,
+      "step": 78000
+    },
+    {
+      "epoch": 3.18,
+      "eval_accuracy": 0.5075848303393213,
+      "eval_loss": 0.9907119274139404,
+      "eval_runtime": 12.4429,
+      "eval_samples_per_second": 402.64,
+      "eval_steps_per_second": 50.39,
+      "step": 78000
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.193388888888889e-06,
+      "loss": 1.0207,
+      "step": 79000
+    },
+    {
+      "epoch": 3.22,
+      "eval_accuracy": 0.4944111776447106,
+      "eval_loss": 1.0015305280685425,
+      "eval_runtime": 12.7143,
+      "eval_samples_per_second": 394.046,
+      "eval_steps_per_second": 49.315,
+      "step": 79000
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.221166666666667e-06,
+      "loss": 1.0172,
+      "step": 80000
+    },
+    {
+      "epoch": 3.26,
+      "eval_accuracy": 0.49540918163672654,
+      "eval_loss": 0.9991635680198669,
+      "eval_runtime": 12.7122,
+      "eval_samples_per_second": 394.109,
+      "eval_steps_per_second": 49.323,
+      "step": 80000
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.248916666666667e-06,
+      "loss": 1.0185,
+      "step": 81000
+    },
+    {
+      "epoch": 3.3,
+      "eval_accuracy": 0.49001996007984033,
+      "eval_loss": 1.0026898384094238,
+      "eval_runtime": 12.6101,
+      "eval_samples_per_second": 397.299,
+      "eval_steps_per_second": 49.722,
+      "step": 81000
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.2766944444444444e-06,
+      "loss": 1.0287,
+      "step": 82000
+    },
+    {
+      "epoch": 3.34,
+      "eval_accuracy": 0.4998003992015968,
+      "eval_loss": 0.9921432733535767,
+      "eval_runtime": 12.4639,
+      "eval_samples_per_second": 401.961,
+      "eval_steps_per_second": 50.305,
+      "step": 82000
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.3044444444444443e-06,
+      "loss": 1.0142,
+      "step": 83000
+    },
+    {
+      "epoch": 3.38,
+      "eval_accuracy": 0.5017964071856288,
+      "eval_loss": 0.9938461184501648,
+      "eval_runtime": 12.4125,
+      "eval_samples_per_second": 403.625,
+      "eval_steps_per_second": 50.514,
+      "step": 83000
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.3322222222222223e-06,
+      "loss": 1.0186,
+      "step": 84000
+    },
+    {
+      "epoch": 3.42,
+      "eval_accuracy": 0.5089820359281437,
+      "eval_loss": 0.9863815307617188,
+      "eval_runtime": 12.6556,
+      "eval_samples_per_second": 395.872,
+      "eval_steps_per_second": 49.543,
+      "step": 84000
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.3599722222222226e-06,
+      "loss": 1.0151,
+      "step": 85000
+    },
+    {
+      "epoch": 3.46,
+      "eval_accuracy": 0.5041916167664671,
+      "eval_loss": 0.9891662001609802,
+      "eval_runtime": 12.6529,
+      "eval_samples_per_second": 395.958,
+      "eval_steps_per_second": 49.554,
+      "step": 85000
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.38775e-06,
+      "loss": 1.0268,
+      "step": 86000
+    },
+    {
+      "epoch": 3.5,
+      "eval_accuracy": 0.5043912175648703,
+      "eval_loss": 0.9857804179191589,
+      "eval_runtime": 12.4197,
+      "eval_samples_per_second": 403.391,
+      "eval_steps_per_second": 50.484,
+      "step": 86000
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.4155e-06,
+      "loss": 1.0152,
+      "step": 87000
+    },
+    {
+      "epoch": 3.54,
+      "eval_accuracy": 0.5075848303393213,
+      "eval_loss": 0.9841725826263428,
+      "eval_runtime": 12.4152,
+      "eval_samples_per_second": 403.538,
+      "eval_steps_per_second": 50.503,
+      "step": 87000
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.443277777777778e-06,
+      "loss": 1.0201,
+      "step": 88000
+    },
+    {
+      "epoch": 3.59,
+      "eval_accuracy": 0.5065868263473053,
+      "eval_loss": 0.9841102957725525,
+      "eval_runtime": 12.4716,
+      "eval_samples_per_second": 401.711,
+      "eval_steps_per_second": 50.274,
+      "step": 88000
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.471027777777778e-06,
+      "loss": 1.0113,
+      "step": 89000
+    },
+    {
+      "epoch": 3.63,
+      "eval_accuracy": 0.5079840319361277,
+      "eval_loss": 0.9851917028427124,
+      "eval_runtime": 12.6634,
+      "eval_samples_per_second": 395.627,
+      "eval_steps_per_second": 49.513,
+      "step": 89000
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.498805555555556e-06,
+      "loss": 1.0195,
+      "step": 90000
+    },
+    {
+      "epoch": 3.67,
+      "eval_accuracy": 0.5039920159680639,
+      "eval_loss": 0.9855921864509583,
+      "eval_runtime": 12.6226,
+      "eval_samples_per_second": 396.908,
+      "eval_steps_per_second": 49.673,
+      "step": 90000
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.5265555555555557e-06,
+      "loss": 1.0202,
+      "step": 91000
+    },
+    {
+      "epoch": 3.71,
+      "eval_accuracy": 0.5001996007984032,
+      "eval_loss": 0.9933446645736694,
+      "eval_runtime": 12.4993,
+      "eval_samples_per_second": 400.821,
+      "eval_steps_per_second": 50.163,
+      "step": 91000
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.5543333333333337e-06,
+      "loss": 1.0183,
+      "step": 92000
+    },
+    {
+      "epoch": 3.75,
+      "eval_accuracy": 0.5,
+      "eval_loss": 0.9928790926933289,
+      "eval_runtime": 12.4539,
+      "eval_samples_per_second": 402.285,
+      "eval_steps_per_second": 50.346,
+      "step": 92000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.5820833333333335e-06,
+      "loss": 1.02,
+      "step": 93000
+    },
+    {
+      "epoch": 3.79,
+      "eval_accuracy": 0.5011976047904192,
+      "eval_loss": 0.9867545962333679,
+      "eval_runtime": 12.4054,
+      "eval_samples_per_second": 403.855,
+      "eval_steps_per_second": 50.542,
+      "step": 93000
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.609861111111111e-06,
+      "loss": 1.0144,
+      "step": 94000
+    },
+    {
+      "epoch": 3.83,
+      "eval_accuracy": 0.5027944111776447,
+      "eval_loss": 0.9887515902519226,
+      "eval_runtime": 12.8101,
+      "eval_samples_per_second": 391.097,
+      "eval_steps_per_second": 48.946,
+      "step": 94000
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.6376111111111114e-06,
+      "loss": 1.0166,
+      "step": 95000
+    },
+    {
+      "epoch": 3.87,
+      "eval_accuracy": 0.5077844311377245,
+      "eval_loss": 0.9852330684661865,
+      "eval_runtime": 12.6394,
+      "eval_samples_per_second": 396.379,
+      "eval_steps_per_second": 49.607,
+      "step": 95000
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.6653888888888894e-06,
+      "loss": 1.0173,
+      "step": 96000
+    },
+    {
+      "epoch": 3.91,
+      "eval_accuracy": 0.5053892215568863,
+      "eval_loss": 0.9848083257675171,
+      "eval_runtime": 12.5421,
+      "eval_samples_per_second": 399.456,
+      "eval_steps_per_second": 49.992,
+      "step": 96000
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.6931388888888892e-06,
+      "loss": 1.0159,
+      "step": 97000
+    },
+    {
+      "epoch": 3.95,
+      "eval_accuracy": 0.5051896207584831,
+      "eval_loss": 0.9930784702301025,
+      "eval_runtime": 12.6498,
+      "eval_samples_per_second": 396.054,
+      "eval_steps_per_second": 49.566,
+      "step": 97000
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.7209166666666668e-06,
+      "loss": 1.0126,
+      "step": 98000
+    },
+    {
+      "epoch": 3.99,
+      "eval_accuracy": 0.5037924151696607,
+      "eval_loss": 0.9865526556968689,
+      "eval_runtime": 12.4969,
+      "eval_samples_per_second": 400.899,
+      "eval_steps_per_second": 50.172,
+      "step": 98000
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.7486666666666666e-06,
+      "loss": 1.0129,
+      "step": 99000
+    },
+    {
+      "epoch": 4.03,
+      "eval_accuracy": 0.5073852295409181,
+      "eval_loss": 0.9890230298042297,
+      "eval_runtime": 12.7759,
+      "eval_samples_per_second": 392.143,
+      "eval_steps_per_second": 49.077,
+      "step": 99000
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.7764444444444446e-06,
+      "loss": 1.0145,
+      "step": 100000
+    },
+    {
+      "epoch": 4.07,
+      "eval_accuracy": 0.506187624750499,
+      "eval_loss": 0.9834555387496948,
+      "eval_runtime": 12.6734,
+      "eval_samples_per_second": 395.316,
+      "eval_steps_per_second": 49.474,
+      "step": 100000
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.804194444444445e-06,
+      "loss": 1.0164,
+      "step": 101000
+    },
+    {
+      "epoch": 4.12,
+      "eval_accuracy": 0.5081836327345309,
+      "eval_loss": 0.9881391525268555,
+      "eval_runtime": 12.4161,
+      "eval_samples_per_second": 403.509,
+      "eval_steps_per_second": 50.499,
+      "step": 101000
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.8319722222222225e-06,
+      "loss": 1.0131,
+      "step": 102000
+    },
+    {
+      "epoch": 4.16,
+      "eval_accuracy": 0.5007984031936128,
+      "eval_loss": 0.9866752624511719,
+      "eval_runtime": 12.6211,
+      "eval_samples_per_second": 396.953,
+      "eval_steps_per_second": 49.679,
+      "step": 102000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.8597222222222223e-06,
+      "loss": 1.0172,
+      "step": 103000
+    },
+    {
+      "epoch": 4.2,
+      "eval_accuracy": 0.5117764471057884,
+      "eval_loss": 0.9857889413833618,
+      "eval_runtime": 12.4246,
+      "eval_samples_per_second": 403.233,
+      "eval_steps_per_second": 50.464,
+      "step": 103000
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.8875000000000003e-06,
+      "loss": 1.0118,
+      "step": 104000
+    },
+    {
+      "epoch": 4.24,
+      "eval_accuracy": 0.501996007984032,
+      "eval_loss": 0.9917896389961243,
+      "eval_runtime": 12.7707,
+      "eval_samples_per_second": 392.304,
+      "eval_steps_per_second": 49.097,
+      "step": 104000
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.91525e-06,
+      "loss": 1.0082,
+      "step": 105000
+    },
+    {
+      "epoch": 4.28,
+      "eval_accuracy": 0.4940119760479042,
+      "eval_loss": 1.0039118528366089,
+      "eval_runtime": 12.7128,
+      "eval_samples_per_second": 394.092,
+      "eval_steps_per_second": 49.32,
+      "step": 105000
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.9430277777777777e-06,
+      "loss": 1.006,
+      "step": 106000
+    },
+    {
+      "epoch": 4.32,
+      "eval_accuracy": 0.5023952095808383,
+      "eval_loss": 0.9883149862289429,
+      "eval_runtime": 12.7217,
+      "eval_samples_per_second": 393.814,
+      "eval_steps_per_second": 49.286,
+      "step": 106000
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.970777777777778e-06,
+      "loss": 1.008,
+      "step": 107000
+    },
+    {
+      "epoch": 4.36,
+      "eval_accuracy": 0.5023952095808383,
+      "eval_loss": 0.9926295876502991,
+      "eval_runtime": 12.4187,
+      "eval_samples_per_second": 403.424,
+      "eval_steps_per_second": 50.488,
+      "step": 107000
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.998555555555556e-06,
+      "loss": 1.0078,
+      "step": 108000
+    },
+    {
+      "epoch": 4.4,
+      "eval_accuracy": 0.5069860279441117,
+      "eval_loss": 0.9883684515953064,
+      "eval_runtime": 12.4562,
+      "eval_samples_per_second": 402.21,
+      "eval_steps_per_second": 50.336,
+      "step": 108000
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.026305555555556e-06,
+      "loss": 1.0052,
+      "step": 109000
+    },
+    {
+      "epoch": 4.44,
+      "eval_accuracy": 0.5083832335329341,
+      "eval_loss": 0.9893190264701843,
+      "eval_runtime": 12.4193,
+      "eval_samples_per_second": 403.404,
+      "eval_steps_per_second": 50.486,
+      "step": 109000
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.0540833333333334e-06,
+      "loss": 1.0102,
+      "step": 110000
+    },
+    {
+      "epoch": 4.48,
+      "eval_accuracy": 0.5059880239520959,
+      "eval_loss": 0.9876956939697266,
+      "eval_runtime": 12.7204,
+      "eval_samples_per_second": 393.854,
+      "eval_steps_per_second": 49.291,
+      "step": 110000
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.0818333333333333e-06,
+      "loss": 1.004,
+      "step": 111000
+    },
+    {
+      "epoch": 4.52,
+      "eval_accuracy": 0.5099800399201597,
+      "eval_loss": 0.9822403788566589,
+      "eval_runtime": 12.7671,
+      "eval_samples_per_second": 392.416,
+      "eval_steps_per_second": 49.111,
+      "step": 111000
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.1096111111111113e-06,
+      "loss": 1.016,
+      "step": 112000
+    },
+    {
+      "epoch": 4.56,
+      "eval_accuracy": 0.5075848303393213,
+      "eval_loss": 0.9813176393508911,
+      "eval_runtime": 12.5425,
+      "eval_samples_per_second": 399.442,
+      "eval_steps_per_second": 49.99,
+      "step": 112000
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.137361111111111e-06,
+      "loss": 1.0024,
+      "step": 113000
+    },
+    {
+      "epoch": 4.6,
+      "eval_accuracy": 0.5,
+      "eval_loss": 0.9851263165473938,
+      "eval_runtime": 12.718,
+      "eval_samples_per_second": 393.931,
+      "eval_steps_per_second": 49.3,
+      "step": 113000
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.165138888888889e-06,
+      "loss": 1.0096,
+      "step": 114000
+    },
+    {
+      "epoch": 4.64,
+      "eval_accuracy": 0.5101796407185629,
+      "eval_loss": 0.9791144132614136,
+      "eval_runtime": 12.4876,
+      "eval_samples_per_second": 401.199,
+      "eval_steps_per_second": 50.21,
+      "step": 114000
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.192888888888889e-06,
+      "loss": 1.0111,
+      "step": 115000
+    },
+    {
+      "epoch": 4.69,
+      "eval_accuracy": 0.5131736526946108,
+      "eval_loss": 0.9794335961341858,
+      "eval_runtime": 12.6803,
+      "eval_samples_per_second": 395.101,
+      "eval_steps_per_second": 49.447,
+      "step": 115000
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.220666666666667e-06,
+      "loss": 1.005,
+      "step": 116000
+    },
+    {
+      "epoch": 4.73,
+      "eval_accuracy": 0.506187624750499,
+      "eval_loss": 0.9822920560836792,
+      "eval_runtime": 12.7349,
+      "eval_samples_per_second": 393.408,
+      "eval_steps_per_second": 49.235,
+      "step": 116000
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.248416666666667e-06,
+      "loss": 1.0116,
+      "step": 117000
+    },
+    {
+      "epoch": 4.77,
+      "eval_accuracy": 0.5073852295409181,
+      "eval_loss": 0.9905573725700378,
+      "eval_runtime": 12.5859,
+      "eval_samples_per_second": 398.064,
+      "eval_steps_per_second": 49.818,
+      "step": 117000
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.276194444444445e-06,
+      "loss": 1.0108,
+      "step": 118000
+    },
+    {
+      "epoch": 4.81,
+      "eval_accuracy": 0.5121756487025948,
+      "eval_loss": 0.9814850687980652,
+      "eval_runtime": 12.5991,
+      "eval_samples_per_second": 397.649,
+      "eval_steps_per_second": 49.766,
+      "step": 118000
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.303944444444445e-06,
+      "loss": 1.0059,
+      "step": 119000
+    },
+    {
+      "epoch": 4.85,
+      "eval_accuracy": 0.5135728542914172,
+      "eval_loss": 0.9780199527740479,
+      "eval_runtime": 12.4558,
+      "eval_samples_per_second": 402.221,
+      "eval_steps_per_second": 50.338,
+      "step": 119000
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.331722222222222e-06,
+      "loss": 1.0135,
+      "step": 120000
+    },
+    {
+      "epoch": 4.89,
+      "eval_accuracy": 0.5039920159680639,
+      "eval_loss": 0.9968580603599548,
+      "eval_runtime": 12.6517,
+      "eval_samples_per_second": 395.994,
+      "eval_steps_per_second": 49.558,
+      "step": 120000
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.3595e-06,
+      "loss": 1.0023,
+      "step": 121000
+    },
+    {
+      "epoch": 4.93,
+      "eval_accuracy": 0.512375249500998,
+      "eval_loss": 0.9820229411125183,
+      "eval_runtime": 12.6842,
+      "eval_samples_per_second": 394.979,
+      "eval_steps_per_second": 49.431,
+      "step": 121000
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.38725e-06,
+      "loss": 1.0085,
+      "step": 122000
+    },
+    {
+      "epoch": 4.97,
+      "eval_accuracy": 0.5117764471057884,
+      "eval_loss": 0.9823063611984253,
+      "eval_runtime": 12.7386,
+      "eval_samples_per_second": 393.293,
+      "eval_steps_per_second": 49.22,
+      "step": 122000
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.415027777777778e-06,
+      "loss": 1.0076,
+      "step": 123000
+    },
+    {
+      "epoch": 5.01,
+      "eval_accuracy": 0.500998003992016,
+      "eval_loss": 0.9879907369613647,
+      "eval_runtime": 12.4926,
+      "eval_samples_per_second": 401.037,
+      "eval_steps_per_second": 50.19,
+      "step": 123000
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.442777777777778e-06,
+      "loss": 1.0005,
+      "step": 124000
+    },
+    {
+      "epoch": 5.05,
+      "eval_accuracy": 0.5093812375249501,
+      "eval_loss": 0.9808560609817505,
+      "eval_runtime": 12.4952,
+      "eval_samples_per_second": 400.955,
+      "eval_steps_per_second": 50.179,
+      "step": 124000
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.470555555555556e-06,
+      "loss": 1.0048,
+      "step": 125000
+    },
+    {
+      "epoch": 5.09,
+      "eval_accuracy": 0.5099800399201597,
+      "eval_loss": 0.981860876083374,
+      "eval_runtime": 12.4425,
+      "eval_samples_per_second": 402.652,
+      "eval_steps_per_second": 50.392,
+      "step": 125000
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.4983333333333334e-06,
+      "loss": 1.0049,
+      "step": 126000
+    },
+    {
+      "epoch": 5.13,
+      "eval_accuracy": 0.5105788423153692,
+      "eval_loss": 0.9849931001663208,
+      "eval_runtime": 12.6135,
+      "eval_samples_per_second": 397.192,
+      "eval_steps_per_second": 49.708,
+      "step": 126000
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.5261111111111114e-06,
+      "loss": 1.0056,
+      "step": 127000
+    },
+    {
+      "epoch": 5.17,
+      "eval_accuracy": 0.5049900199600799,
+      "eval_loss": 0.9838124513626099,
+      "eval_runtime": 12.4113,
+      "eval_samples_per_second": 403.663,
+      "eval_steps_per_second": 50.518,
+      "step": 127000
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.5538611111111113e-06,
+      "loss": 0.9995,
+      "step": 128000
+    },
+    {
+      "epoch": 5.22,
+      "eval_accuracy": 0.5099800399201597,
+      "eval_loss": 0.9798722863197327,
+      "eval_runtime": 12.3902,
+      "eval_samples_per_second": 404.351,
+      "eval_steps_per_second": 50.604,
+      "step": 128000
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.5816388888888893e-06,
+      "loss": 1.003,
+      "step": 129000
+    },
+    {
+      "epoch": 5.26,
+      "eval_accuracy": 0.5055888223552895,
+      "eval_loss": 0.9836646914482117,
+      "eval_runtime": 12.3384,
+      "eval_samples_per_second": 406.05,
+      "eval_steps_per_second": 50.817,
+      "step": 129000
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.609388888888889e-06,
+      "loss": 0.9956,
+      "step": 130000
+    },
+    {
+      "epoch": 5.3,
+      "eval_accuracy": 0.5079840319361277,
+      "eval_loss": 0.9859731793403625,
+      "eval_runtime": 12.6442,
+      "eval_samples_per_second": 396.23,
+      "eval_steps_per_second": 49.588,
+      "step": 130000
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6371666666666667e-06,
+      "loss": 0.9989,
+      "step": 131000
+    },
+    {
+      "epoch": 5.34,
+      "eval_accuracy": 0.5121756487025948,
+      "eval_loss": 0.9777482151985168,
+      "eval_runtime": 12.5026,
+      "eval_samples_per_second": 400.716,
+      "eval_steps_per_second": 50.149,
+      "step": 131000
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6649166666666665e-06,
+      "loss": 0.9992,
+      "step": 132000
+    },
+    {
+      "epoch": 5.38,
+      "eval_accuracy": 0.5131736526946108,
+      "eval_loss": 0.9786902666091919,
+      "eval_runtime": 12.4113,
+      "eval_samples_per_second": 403.664,
+      "eval_steps_per_second": 50.518,
+      "step": 132000
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6926666666666673e-06,
+      "loss": 1.0033,
+      "step": 133000
+    },
+    {
+      "epoch": 5.42,
+      "eval_accuracy": 0.5103792415169661,
+      "eval_loss": 0.9823509454727173,
+      "eval_runtime": 12.3681,
+      "eval_samples_per_second": 405.073,
+      "eval_steps_per_second": 50.695,
+      "step": 133000
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.7204444444444444e-06,
+      "loss": 1.001,
+      "step": 134000
+    },
+    {
+      "epoch": 5.46,
+      "eval_accuracy": 0.5055888223552895,
+      "eval_loss": 0.980103075504303,
+      "eval_runtime": 12.658,
+      "eval_samples_per_second": 395.796,
+      "eval_steps_per_second": 49.534,
+      "step": 134000
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.7482222222222224e-06,
+      "loss": 1.0074,
+      "step": 135000
+    },
+    {
+      "epoch": 5.5,
+      "eval_accuracy": 0.5067864271457085,
+      "eval_loss": 0.9919066429138184,
+      "eval_runtime": 12.669,
+      "eval_samples_per_second": 395.452,
+      "eval_steps_per_second": 49.491,
+      "step": 135000
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.7760000000000004e-06,
+      "loss": 1.0056,
+      "step": 136000
+    },
+    {
+      "epoch": 5.54,
+      "eval_accuracy": 0.5117764471057884,
+      "eval_loss": 0.9795469641685486,
+      "eval_runtime": 12.3923,
+      "eval_samples_per_second": 404.284,
+      "eval_steps_per_second": 50.596,
+      "step": 136000
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.8037500000000002e-06,
+      "loss": 1.0029,
+      "step": 137000
+    },
+    {
+      "epoch": 5.58,
+      "eval_accuracy": 0.5101796407185629,
+      "eval_loss": 0.9805279970169067,
+      "eval_runtime": 12.4005,
+      "eval_samples_per_second": 404.015,
+      "eval_steps_per_second": 50.562,
+      "step": 137000
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.831527777777778e-06,
+      "loss": 1.0018,
+      "step": 138000
+    },
+    {
+      "epoch": 5.62,
+      "eval_accuracy": 0.5015968063872256,
+      "eval_loss": 0.9934979677200317,
+      "eval_runtime": 12.6438,
+      "eval_samples_per_second": 396.24,
+      "eval_steps_per_second": 49.589,
+      "step": 138000
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.859277777777778e-06,
+      "loss": 0.9985,
+      "step": 139000
+    },
+    {
+      "epoch": 5.66,
+      "eval_accuracy": 0.5083832335329341,
+      "eval_loss": 0.984520435333252,
+      "eval_runtime": 12.6399,
+      "eval_samples_per_second": 396.364,
+      "eval_steps_per_second": 49.605,
+      "step": 139000
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.887055555555556e-06,
+      "loss": 1.0007,
+      "step": 140000
+    },
+    {
+      "epoch": 5.7,
+      "eval_accuracy": 0.5037924151696607,
+      "eval_loss": 0.9855640530586243,
+      "eval_runtime": 12.4087,
+      "eval_samples_per_second": 403.748,
+      "eval_steps_per_second": 50.529,
+      "step": 140000
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.9148055555555555e-06,
+      "loss": 1.0037,
+      "step": 141000
+    },
+    {
+      "epoch": 5.74,
+      "eval_accuracy": 0.5117764471057884,
+      "eval_loss": 0.982020914554596,
+      "eval_runtime": 12.3775,
+      "eval_samples_per_second": 404.766,
+      "eval_steps_per_second": 50.656,
+      "step": 141000
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.9425833333333335e-06,
+      "loss": 1.0006,
+      "step": 142000
+    },
+    {
+      "epoch": 5.79,
+      "eval_accuracy": 0.5105788423153692,
+      "eval_loss": 0.9820439219474792,
+      "eval_runtime": 12.5661,
+      "eval_samples_per_second": 398.693,
+      "eval_steps_per_second": 49.896,
+      "step": 142000
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.970333333333333e-06,
+      "loss": 1.0027,
+      "step": 143000
+    },
+    {
+      "epoch": 5.83,
+      "eval_accuracy": 0.513373253493014,
+      "eval_loss": 0.984400749206543,
+      "eval_runtime": 12.5965,
+      "eval_samples_per_second": 397.729,
+      "eval_steps_per_second": 49.776,
+      "step": 143000
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.998111111111111e-06,
+      "loss": 1.0027,
+      "step": 144000
+    },
+    {
+      "epoch": 5.87,
+      "eval_accuracy": 0.5129740518962076,
+      "eval_loss": 0.9762855172157288,
+      "eval_runtime": 12.3769,
+      "eval_samples_per_second": 404.787,
+      "eval_steps_per_second": 50.659,
+      "step": 144000
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.025861111111111e-06,
+      "loss": 1.0023,
+      "step": 145000
+    },
+    {
+      "epoch": 5.91,
+      "eval_accuracy": 0.51437125748503,
+      "eval_loss": 0.984603226184845,
+      "eval_runtime": 12.3446,
+      "eval_samples_per_second": 405.846,
+      "eval_steps_per_second": 50.792,
+      "step": 145000
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.053638888888889e-06,
+      "loss": 1.0029,
+      "step": 146000
+    },
+    {
+      "epoch": 5.95,
+      "eval_accuracy": 0.5137724550898204,
+      "eval_loss": 0.9721319079399109,
+      "eval_runtime": 12.3601,
+      "eval_samples_per_second": 405.337,
+      "eval_steps_per_second": 50.728,
+      "step": 146000
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.081361111111112e-06,
+      "loss": 1.0032,
+      "step": 147000
+    },
+    {
+      "epoch": 5.99,
+      "eval_accuracy": 0.513373253493014,
+      "eval_loss": 0.9780867695808411,
+      "eval_runtime": 12.5925,
+      "eval_samples_per_second": 397.855,
+      "eval_steps_per_second": 49.791,
+      "step": 147000
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.109138888888889e-06,
+      "loss": 0.9902,
+      "step": 148000
+    },
+    {
+      "epoch": 6.03,
+      "eval_accuracy": 0.513373253493014,
+      "eval_loss": 0.9827443957328796,
+      "eval_runtime": 12.3726,
+      "eval_samples_per_second": 404.928,
+      "eval_steps_per_second": 50.677,
+      "step": 148000
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.136916666666667e-06,
+      "loss": 0.9942,
+      "step": 149000
+    },
+    {
+      "epoch": 6.07,
+      "eval_accuracy": 0.5181636726546907,
+      "eval_loss": 0.9756249189376831,
+      "eval_runtime": 12.3491,
+      "eval_samples_per_second": 405.696,
+      "eval_steps_per_second": 50.773,
+      "step": 149000
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.164694444444445e-06,
+      "loss": 0.9959,
+      "step": 150000
+    },
+    {
+      "epoch": 6.11,
+      "eval_accuracy": 0.5163672654690619,
+      "eval_loss": 0.9778909087181091,
+      "eval_runtime": 12.3406,
+      "eval_samples_per_second": 405.978,
+      "eval_steps_per_second": 50.808,
+      "step": 150000
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.192444444444445e-06,
+      "loss": 0.9958,
+      "step": 151000
+    },
+    {
+      "epoch": 6.15,
+      "eval_accuracy": 0.5091816367265469,
+      "eval_loss": 0.9797512292861938,
+      "eval_runtime": 12.6166,
+      "eval_samples_per_second": 397.094,
+      "eval_steps_per_second": 49.696,
+      "step": 151000
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.220222222222223e-06,
+      "loss": 0.9972,
+      "step": 152000
+    },
+    {
+      "epoch": 6.19,
+      "eval_accuracy": 0.5091816367265469,
+      "eval_loss": 0.9767274260520935,
+      "eval_runtime": 12.3395,
+      "eval_samples_per_second": 406.014,
+      "eval_steps_per_second": 50.813,
+      "step": 152000
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.2479722222222225e-06,
+      "loss": 0.9966,
+      "step": 153000
+    },
+    {
+      "epoch": 6.23,
+      "eval_accuracy": 0.5081836327345309,
+      "eval_loss": 0.9737312197685242,
+      "eval_runtime": 12.4793,
+      "eval_samples_per_second": 401.466,
+      "eval_steps_per_second": 50.243,
+      "step": 153000
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.27575e-06,
+      "loss": 0.9976,
+      "step": 154000
+    },
+    {
+      "epoch": 6.27,
+      "eval_accuracy": 0.512375249500998,
+      "eval_loss": 0.98174649477005,
+      "eval_runtime": 12.3506,
+      "eval_samples_per_second": 405.649,
+      "eval_steps_per_second": 50.767,
+      "step": 154000
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.3034999999999995e-06,
+      "loss": 0.9944,
+      "step": 155000
+    },
+    {
+      "epoch": 6.32,
+      "eval_accuracy": 0.5131736526946108,
+      "eval_loss": 0.9855016469955444,
+      "eval_runtime": 12.6363,
+      "eval_samples_per_second": 396.478,
+      "eval_steps_per_second": 49.619,
+      "step": 155000
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 4.3312777777777775e-06,
+      "loss": 0.9941,
+      "step": 156000
+    },
+    {
+      "epoch": 6.36,
+      "eval_accuracy": 0.5137724550898204,
+      "eval_loss": 0.9729366302490234,
+      "eval_runtime": 12.4845,
+      "eval_samples_per_second": 401.297,
+      "eval_steps_per_second": 50.222,
+      "step": 156000
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.359027777777778e-06,
+      "loss": 0.9915,
+      "step": 157000
+    },
+    {
+      "epoch": 6.4,
+      "eval_accuracy": 0.5139720558882236,
+      "eval_loss": 0.9848944544792175,
+      "eval_runtime": 12.3962,
+      "eval_samples_per_second": 404.155,
+      "eval_steps_per_second": 50.58,
+      "step": 157000
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.386805555555556e-06,
+      "loss": 0.9917,
+      "step": 158000
+    },
+    {
+      "epoch": 6.44,
+      "eval_accuracy": 0.5149700598802395,
+      "eval_loss": 0.9808858633041382,
+      "eval_runtime": 12.3333,
+      "eval_samples_per_second": 406.217,
+      "eval_steps_per_second": 50.838,
+      "step": 158000
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 4.414555555555556e-06,
+      "loss": 0.9879,
+      "step": 159000
+    },
+    {
+      "epoch": 6.48,
+      "eval_accuracy": 0.5127744510978044,
+      "eval_loss": 0.9762862920761108,
+      "eval_runtime": 12.6161,
+      "eval_samples_per_second": 397.11,
+      "eval_steps_per_second": 49.698,
+      "step": 159000
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 4.442333333333334e-06,
+      "loss": 0.9945,
+      "step": 160000
+    },
+    {
+      "epoch": 6.52,
+      "eval_accuracy": 0.5157684630738523,
+      "eval_loss": 0.9736975431442261,
+      "eval_runtime": 12.5588,
+      "eval_samples_per_second": 398.923,
+      "eval_steps_per_second": 49.925,
+      "step": 160000
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 4.470083333333334e-06,
+      "loss": 0.996,
+      "step": 161000
+    },
+    {
+      "epoch": 6.56,
+      "eval_accuracy": 0.5191616766467065,
+      "eval_loss": 0.9773131608963013,
+      "eval_runtime": 12.391,
+      "eval_samples_per_second": 404.325,
+      "eval_steps_per_second": 50.601,
+      "step": 161000
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.497861111111111e-06,
+      "loss": 1.0038,
+      "step": 162000
+    },
+    {
+      "epoch": 6.6,
+      "eval_accuracy": 0.5189620758483033,
+      "eval_loss": 0.976509690284729,
+      "eval_runtime": 12.3726,
+      "eval_samples_per_second": 404.926,
+      "eval_steps_per_second": 50.676,
+      "step": 162000
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 4.525611111111111e-06,
+      "loss": 1.0001,
+      "step": 163000
+    },
+    {
+      "epoch": 6.64,
+      "eval_accuracy": 0.5179640718562875,
+      "eval_loss": 0.9701054096221924,
+      "eval_runtime": 12.6463,
+      "eval_samples_per_second": 396.162,
+      "eval_steps_per_second": 49.58,
+      "step": 163000
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.553388888888889e-06,
+      "loss": 0.9993,
+      "step": 164000
+    },
+    {
+      "epoch": 6.68,
+      "eval_accuracy": 0.51437125748503,
+      "eval_loss": 0.97454833984375,
+      "eval_runtime": 12.6092,
+      "eval_samples_per_second": 397.327,
+      "eval_steps_per_second": 49.725,
+      "step": 164000
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 4.581166666666667e-06,
+      "loss": 0.9999,
+      "step": 165000
+    },
+    {
+      "epoch": 6.72,
+      "eval_accuracy": 0.5077844311377245,
+      "eval_loss": 0.9879518151283264,
+      "eval_runtime": 12.3951,
+      "eval_samples_per_second": 404.19,
+      "eval_steps_per_second": 50.584,
+      "step": 165000
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.608916666666667e-06,
+      "loss": 0.994,
+      "step": 166000
+    },
+    {
+      "epoch": 6.76,
+      "eval_accuracy": 0.5121756487025948,
+      "eval_loss": 0.9857090711593628,
+      "eval_runtime": 12.413,
+      "eval_samples_per_second": 403.608,
+      "eval_steps_per_second": 50.511,
+      "step": 166000
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 4.636694444444445e-06,
+      "loss": 1.0017,
+      "step": 167000
+    },
+    {
+      "epoch": 6.8,
+      "eval_accuracy": 0.5189620758483033,
+      "eval_loss": 0.9743807315826416,
+      "eval_runtime": 12.5263,
+      "eval_samples_per_second": 399.958,
+      "eval_steps_per_second": 50.055,
+      "step": 167000
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.664472222222222e-06,
+      "loss": 0.9942,
+      "step": 168000
+    },
+    {
+      "epoch": 6.84,
+      "eval_accuracy": 0.5179640718562875,
+      "eval_loss": 0.9760473370552063,
+      "eval_runtime": 12.6417,
+      "eval_samples_per_second": 396.307,
+      "eval_steps_per_second": 49.598,
+      "step": 168000
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 4.6922222222222226e-06,
+      "loss": 0.9939,
+      "step": 169000
+    },
+    {
+      "epoch": 6.89,
+      "eval_accuracy": 0.5145708582834332,
+      "eval_loss": 0.9860849380493164,
+      "eval_runtime": 12.3718,
+      "eval_samples_per_second": 404.954,
+      "eval_steps_per_second": 50.68,
+      "step": 169000
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.72e-06,
+      "loss": 0.9939,
+      "step": 170000
+    },
+    {
+      "epoch": 6.93,
+      "eval_accuracy": 0.51437125748503,
+      "eval_loss": 0.9720783233642578,
+      "eval_runtime": 12.3506,
+      "eval_samples_per_second": 405.649,
+      "eval_steps_per_second": 50.767,
+      "step": 170000
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.747777777777778e-06,
+      "loss": 0.9904,
+      "step": 171000
+    },
+    {
+      "epoch": 6.97,
+      "eval_accuracy": 0.5199600798403193,
+      "eval_loss": 0.968160092830658,
+      "eval_runtime": 12.3653,
+      "eval_samples_per_second": 405.167,
+      "eval_steps_per_second": 50.707,
+      "step": 171000
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.775555555555556e-06,
+      "loss": 0.9944,
+      "step": 172000
+    },
+    {
+      "epoch": 7.01,
+      "eval_accuracy": 0.5169660678642715,
+      "eval_loss": 0.9733211994171143,
+      "eval_runtime": 12.6298,
+      "eval_samples_per_second": 396.682,
+      "eval_steps_per_second": 49.645,
+      "step": 172000
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 4.803305555555556e-06,
+      "loss": 0.9962,
+      "step": 173000
+    },
+    {
+      "epoch": 7.05,
+      "eval_accuracy": 0.5209580838323353,
+      "eval_loss": 0.9708949327468872,
+      "eval_runtime": 12.646,
+      "eval_samples_per_second": 396.172,
+      "eval_steps_per_second": 49.581,
+      "step": 173000
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.831055555555556e-06,
+      "loss": 0.987,
+      "step": 174000
+    },
+    {
+      "epoch": 7.09,
+      "eval_accuracy": 0.5235528942115768,
+      "eval_loss": 0.9703502058982849,
+      "eval_runtime": 12.4251,
+      "eval_samples_per_second": 403.216,
+      "eval_steps_per_second": 50.462,
+      "step": 174000
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.858833333333333e-06,
+      "loss": 0.9918,
+      "step": 175000
+    },
+    {
+      "epoch": 7.13,
+      "eval_accuracy": 0.5189620758483033,
+      "eval_loss": 0.9747355580329895,
+      "eval_runtime": 12.4108,
+      "eval_samples_per_second": 403.68,
+      "eval_steps_per_second": 50.52,
+      "step": 175000
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 4.886611111111111e-06,
+      "loss": 0.9927,
+      "step": 176000
+    },
+    {
+      "epoch": 7.17,
+      "eval_accuracy": 0.5129740518962076,
+      "eval_loss": 0.9888993501663208,
+      "eval_runtime": 12.3966,
+      "eval_samples_per_second": 404.144,
+      "eval_steps_per_second": 50.579,
+      "step": 176000
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 4.914388888888889e-06,
+      "loss": 0.9852,
+      "step": 177000
+    },
+    {
+      "epoch": 7.21,
+      "eval_accuracy": 0.5179640718562875,
+      "eval_loss": 0.976629376411438,
+      "eval_runtime": 12.6228,
+      "eval_samples_per_second": 396.9,
+      "eval_steps_per_second": 49.672,
+      "step": 177000
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.942138888888889e-06,
+      "loss": 0.9903,
+      "step": 178000
+    },
+    {
+      "epoch": 7.25,
+      "eval_accuracy": 0.5153692614770459,
+      "eval_loss": 0.9784625768661499,
+      "eval_runtime": 12.5418,
+      "eval_samples_per_second": 399.466,
+      "eval_steps_per_second": 49.993,
+      "step": 178000
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 4.969888888888889e-06,
+      "loss": 0.9847,
+      "step": 179000
+    },
+    {
+      "epoch": 7.29,
+      "eval_accuracy": 0.5149700598802395,
+      "eval_loss": 0.9706389904022217,
+      "eval_runtime": 12.4409,
+      "eval_samples_per_second": 402.703,
+      "eval_steps_per_second": 50.398,
+      "step": 179000
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.997666666666667e-06,
+      "loss": 0.9903,
+      "step": 180000
+    },
+    {
+      "epoch": 7.33,
+      "eval_accuracy": 0.5227544910179641,
+      "eval_loss": 0.9752748608589172,
+      "eval_runtime": 12.4359,
+      "eval_samples_per_second": 402.866,
+      "eval_steps_per_second": 50.419,
+      "step": 180000
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 5.025444444444444e-06,
+      "loss": 0.9865,
+      "step": 181000
+    },
+    {
+      "epoch": 7.37,
+      "eval_accuracy": 0.5181636726546907,
+      "eval_loss": 0.9852616786956787,
+      "eval_runtime": 12.5661,
+      "eval_samples_per_second": 398.693,
+      "eval_steps_per_second": 49.896,
+      "step": 181000
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 5.053194444444444e-06,
+      "loss": 0.9908,
+      "step": 182000
+    },
+    {
+      "epoch": 7.42,
+      "eval_accuracy": 0.5139720558882236,
+      "eval_loss": 0.972352147102356,
+      "eval_runtime": 12.6901,
+      "eval_samples_per_second": 394.796,
+      "eval_steps_per_second": 49.409,
+      "step": 182000
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 5.080972222222222e-06,
+      "loss": 0.9939,
+      "step": 183000
+    },
+    {
+      "epoch": 7.46,
+      "eval_accuracy": 0.5201596806387225,
+      "eval_loss": 0.9690868258476257,
+      "eval_runtime": 12.5132,
+      "eval_samples_per_second": 400.378,
+      "eval_steps_per_second": 50.107,
+      "step": 183000
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 5.10875e-06,
+      "loss": 0.9858,
+      "step": 184000
+    },
+    {
+      "epoch": 7.5,
+      "eval_accuracy": 0.5215568862275449,
+      "eval_loss": 0.966168999671936,
+      "eval_runtime": 12.4268,
+      "eval_samples_per_second": 403.161,
+      "eval_steps_per_second": 50.455,
+      "step": 184000
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 5.136500000000001e-06,
+      "loss": 0.9916,
+      "step": 185000
+    },
+    {
+      "epoch": 7.54,
+      "eval_accuracy": 0.5235528942115768,
+      "eval_loss": 0.9717623591423035,
+      "eval_runtime": 12.4114,
+      "eval_samples_per_second": 403.662,
+      "eval_steps_per_second": 50.518,
+      "step": 185000
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 5.164277777777779e-06,
+      "loss": 0.9853,
+      "step": 186000
+    },
+    {
+      "epoch": 7.58,
+      "eval_accuracy": 0.5125748502994012,
+      "eval_loss": 0.9818634986877441,
+      "eval_runtime": 12.678,
+      "eval_samples_per_second": 395.174,
+      "eval_steps_per_second": 49.456,
+      "step": 186000
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 5.1920277777777786e-06,
+      "loss": 0.9838,
+      "step": 187000
+    },
+    {
+      "epoch": 7.62,
+      "eval_accuracy": 0.5223552894211577,
+      "eval_loss": 0.9702249765396118,
+      "eval_runtime": 12.6336,
+      "eval_samples_per_second": 396.563,
+      "eval_steps_per_second": 49.63,
+      "step": 187000
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 5.219805555555556e-06,
+      "loss": 0.9841,
+      "step": 188000
+    },
+    {
+      "epoch": 7.66,
+      "eval_accuracy": 0.5171656686626747,
+      "eval_loss": 0.9767837524414062,
+      "eval_runtime": 12.4043,
+      "eval_samples_per_second": 403.893,
+      "eval_steps_per_second": 50.547,
+      "step": 188000
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 5.247583333333334e-06,
+      "loss": 0.9789,
+      "step": 189000
+    },
+    {
+      "epoch": 7.7,
+      "eval_accuracy": 0.5225548902195609,
+      "eval_loss": 0.9727286696434021,
+      "eval_runtime": 12.4433,
+      "eval_samples_per_second": 402.627,
+      "eval_steps_per_second": 50.389,
+      "step": 189000
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 5.2753333333333335e-06,
+      "loss": 0.9882,
+      "step": 190000
+    },
+    {
+      "epoch": 7.74,
+      "eval_accuracy": 0.5137724550898204,
+      "eval_loss": 0.9826129674911499,
+      "eval_runtime": 12.4541,
+      "eval_samples_per_second": 402.279,
+      "eval_steps_per_second": 50.345,
+      "step": 190000
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 5.3031111111111115e-06,
+      "loss": 0.9891,
+      "step": 191000
+    },
+    {
+      "epoch": 7.78,
+      "eval_accuracy": 0.5047904191616767,
+      "eval_loss": 0.9946686625480652,
+      "eval_runtime": 12.6916,
+      "eval_samples_per_second": 394.75,
+      "eval_steps_per_second": 49.403,
+      "step": 191000
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 5.330861111111111e-06,
+      "loss": 0.9862,
+      "step": 192000
+    },
+    {
+      "epoch": 7.82,
+      "eval_accuracy": 0.5203592814371257,
+      "eval_loss": 0.9891188144683838,
+      "eval_runtime": 12.6632,
+      "eval_samples_per_second": 395.634,
+      "eval_steps_per_second": 49.513,
+      "step": 192000
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 5.358638888888889e-06,
+      "loss": 0.9892,
+      "step": 193000
+    },
+    {
+      "epoch": 7.86,
+      "eval_accuracy": 0.5165668662674651,
+      "eval_loss": 0.9784462451934814,
+      "eval_runtime": 12.4501,
+      "eval_samples_per_second": 402.406,
+      "eval_steps_per_second": 50.361,
+      "step": 193000
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 5.3864166666666665e-06,
+      "loss": 0.9846,
+      "step": 194000
+    },
+    {
+      "epoch": 7.9,
+      "eval_accuracy": 0.5193612774451097,
+      "eval_loss": 0.9772787690162659,
+      "eval_runtime": 12.4622,
+      "eval_samples_per_second": 402.015,
+      "eval_steps_per_second": 50.312,
+      "step": 194000
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 5.4141944444444445e-06,
+      "loss": 0.9914,
+      "step": 195000
+    },
+    {
+      "epoch": 7.94,
+      "eval_accuracy": 0.5179640718562875,
+      "eval_loss": 0.9673308730125427,
+      "eval_runtime": 12.4199,
+      "eval_samples_per_second": 403.385,
+      "eval_steps_per_second": 50.483,
+      "step": 195000
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 5.441944444444444e-06,
+      "loss": 0.9901,
+      "step": 196000
+    },
+    {
+      "epoch": 7.99,
+      "eval_accuracy": 0.5157684630738523,
+      "eval_loss": 0.97174072265625,
+      "eval_runtime": 12.7019,
+      "eval_samples_per_second": 394.429,
+      "eval_steps_per_second": 49.363,
+      "step": 196000
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.469722222222222e-06,
+      "loss": 0.9885,
+      "step": 197000
+    },
+    {
+      "epoch": 8.03,
+      "eval_accuracy": 0.5209580838323353,
+      "eval_loss": 0.9727852940559387,
+      "eval_runtime": 12.6818,
+      "eval_samples_per_second": 395.053,
+      "eval_steps_per_second": 49.441,
+      "step": 197000
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.497472222222222e-06,
+      "loss": 0.9791,
+      "step": 198000
+    },
+    {
+      "epoch": 8.07,
+      "eval_accuracy": 0.5245508982035928,
+      "eval_loss": 0.9674847722053528,
+      "eval_runtime": 12.4599,
+      "eval_samples_per_second": 402.091,
+      "eval_steps_per_second": 50.322,
+      "step": 198000
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.52525e-06,
+      "loss": 0.9754,
+      "step": 199000
+    },
+    {
+      "epoch": 8.11,
+      "eval_accuracy": 0.5175648702594811,
+      "eval_loss": 0.9784180521965027,
+      "eval_runtime": 12.43,
+      "eval_samples_per_second": 403.056,
+      "eval_steps_per_second": 50.442,
+      "step": 199000
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.552972222222223e-06,
+      "loss": 0.982,
+      "step": 200000
+    },
+    {
+      "epoch": 8.15,
+      "eval_accuracy": 0.5211576846307385,
+      "eval_loss": 0.9805364012718201,
+      "eval_runtime": 12.4275,
+      "eval_samples_per_second": 403.139,
+      "eval_steps_per_second": 50.453,
+      "step": 200000
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.580750000000001e-06,
+      "loss": 0.9851,
+      "step": 201000
+    },
+    {
+      "epoch": 8.19,
+      "eval_accuracy": 0.5241516966067864,
+      "eval_loss": 0.9727270007133484,
+      "eval_runtime": 12.6938,
+      "eval_samples_per_second": 394.682,
+      "eval_steps_per_second": 49.394,
+      "step": 201000
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.608527777777778e-06,
+      "loss": 0.9819,
+      "step": 202000
+    },
+    {
+      "epoch": 8.23,
+      "eval_accuracy": 0.5211576846307385,
+      "eval_loss": 0.9697861075401306,
+      "eval_runtime": 12.652,
+      "eval_samples_per_second": 395.984,
+      "eval_steps_per_second": 49.557,
+      "step": 202000
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.636305555555556e-06,
+      "loss": 0.9809,
+      "step": 203000
+    },
+    {
+      "epoch": 8.27,
+      "eval_accuracy": 0.5221556886227545,
+      "eval_loss": 0.9775295257568359,
+      "eval_runtime": 12.4367,
+      "eval_samples_per_second": 402.841,
+      "eval_steps_per_second": 50.415,
+      "step": 203000
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.664055555555556e-06,
+      "loss": 0.9746,
+      "step": 204000
+    },
+    {
+      "epoch": 8.31,
+      "eval_accuracy": 0.5171656686626747,
+      "eval_loss": 0.9735843539237976,
+      "eval_runtime": 12.4431,
+      "eval_samples_per_second": 402.633,
+      "eval_steps_per_second": 50.389,
+      "step": 204000
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 5.691833333333334e-06,
+      "loss": 0.9806,
+      "step": 205000
+    },
+    {
+      "epoch": 8.35,
+      "eval_accuracy": 0.5189620758483033,
+      "eval_loss": 0.9673370122909546,
+      "eval_runtime": 12.447,
+      "eval_samples_per_second": 402.505,
+      "eval_steps_per_second": 50.373,
+      "step": 205000
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 5.719611111111112e-06,
+      "loss": 0.9793,
+      "step": 206000
+    },
+    {
+      "epoch": 8.39,
+      "eval_accuracy": 0.5233532934131736,
+      "eval_loss": 0.9679150581359863,
+      "eval_runtime": 12.6348,
+      "eval_samples_per_second": 396.524,
+      "eval_steps_per_second": 49.625,
+      "step": 206000
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 5.7473611111111115e-06,
+      "loss": 0.9898,
+      "step": 207000
+    },
+    {
+      "epoch": 8.43,
+      "eval_accuracy": 0.5269461077844312,
+      "eval_loss": 0.966350793838501,
+      "eval_runtime": 12.6698,
+      "eval_samples_per_second": 395.427,
+      "eval_steps_per_second": 49.488,
+      "step": 207000
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 5.775138888888889e-06,
+      "loss": 0.9825,
+      "step": 208000
+    },
+    {
+      "epoch": 8.47,
+      "eval_accuracy": 0.5135728542914172,
+      "eval_loss": 0.9796927571296692,
+      "eval_runtime": 12.4433,
+      "eval_samples_per_second": 402.626,
+      "eval_steps_per_second": 50.389,
+      "step": 208000
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 5.8028888888888885e-06,
+      "loss": 0.9849,
+      "step": 209000
+    },
+    {
+      "epoch": 8.52,
+      "eval_accuracy": 0.5331337325349301,
+      "eval_loss": 0.965312659740448,
+      "eval_runtime": 12.4638,
+      "eval_samples_per_second": 401.964,
+      "eval_steps_per_second": 50.306,
+      "step": 209000
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 5.8306666666666665e-06,
+      "loss": 0.9741,
+      "step": 210000
+    },
+    {
+      "epoch": 8.56,
+      "eval_accuracy": 0.5323353293413173,
+      "eval_loss": 0.9701557755470276,
+      "eval_runtime": 12.4564,
+      "eval_samples_per_second": 402.204,
+      "eval_steps_per_second": 50.336,
+      "step": 210000
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 5.8584444444444445e-06,
+      "loss": 0.9741,
+      "step": 211000
+    },
+    {
+      "epoch": 8.6,
+      "eval_accuracy": 0.5179640718562875,
+      "eval_loss": 0.9906592965126038,
+      "eval_runtime": 12.6499,
+      "eval_samples_per_second": 396.05,
+      "eval_steps_per_second": 49.566,
+      "step": 211000
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 5.886194444444445e-06,
+      "loss": 0.984,
+      "step": 212000
+    },
+    {
+      "epoch": 8.64,
+      "eval_accuracy": 0.5251497005988024,
+      "eval_loss": 0.9637042284011841,
+      "eval_runtime": 12.5625,
+      "eval_samples_per_second": 398.806,
+      "eval_steps_per_second": 49.91,
+      "step": 212000
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 5.913972222222223e-06,
+      "loss": 0.9832,
+      "step": 213000
+    },
+    {
+      "epoch": 8.68,
+      "eval_accuracy": 0.5265469061876248,
+      "eval_loss": 0.9600540399551392,
+      "eval_runtime": 12.4666,
+      "eval_samples_per_second": 401.875,
+      "eval_steps_per_second": 50.295,
+      "step": 213000
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 5.94175e-06,
+      "loss": 0.9781,
+      "step": 214000
+    },
+    {
+      "epoch": 8.72,
+      "eval_accuracy": 0.524750499001996,
+      "eval_loss": 0.9645100235939026,
+      "eval_runtime": 12.4436,
+      "eval_samples_per_second": 402.617,
+      "eval_steps_per_second": 50.387,
+      "step": 214000
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 5.969527777777778e-06,
+      "loss": 0.9752,
+      "step": 215000
+    },
+    {
+      "epoch": 8.76,
+      "eval_accuracy": 0.5279441117764471,
+      "eval_loss": 0.9634738564491272,
+      "eval_runtime": 12.6681,
+      "eval_samples_per_second": 395.481,
+      "eval_steps_per_second": 49.494,
+      "step": 215000
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 5.99725e-06,
+      "loss": 0.9782,
+      "step": 216000
+    },
+    {
+      "epoch": 8.8,
+      "eval_accuracy": 0.5285429141716567,
+      "eval_loss": 0.9649245738983154,
+      "eval_runtime": 12.6661,
+      "eval_samples_per_second": 395.544,
+      "eval_steps_per_second": 49.502,
+      "step": 216000
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 6.025027777777778e-06,
+      "loss": 0.9726,
+      "step": 217000
+    },
+    {
+      "epoch": 8.84,
+      "eval_accuracy": 0.5231536926147704,
+      "eval_loss": 0.9692424535751343,
+      "eval_runtime": 12.4309,
+      "eval_samples_per_second": 403.028,
+      "eval_steps_per_second": 50.439,
+      "step": 217000
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 6.052805555555556e-06,
+      "loss": 0.9817,
+      "step": 218000
+    },
+    {
+      "epoch": 8.88,
+      "eval_accuracy": 0.5209580838323353,
+      "eval_loss": 0.9694675803184509,
+      "eval_runtime": 12.4379,
+      "eval_samples_per_second": 402.8,
+      "eval_steps_per_second": 50.41,
+      "step": 218000
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 6.080555555555556e-06,
+      "loss": 0.9771,
+      "step": 219000
+    },
+    {
+      "epoch": 8.92,
+      "eval_accuracy": 0.5199600798403193,
+      "eval_loss": 0.9674931168556213,
+      "eval_runtime": 12.5126,
+      "eval_samples_per_second": 400.395,
+      "eval_steps_per_second": 50.109,
+      "step": 219000
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 6.108333333333334e-06,
+      "loss": 0.9779,
+      "step": 220000
+    },
+    {
+      "epoch": 8.96,
+      "eval_accuracy": 0.5271457085828344,
+      "eval_loss": 0.9628773927688599,
+      "eval_runtime": 12.6781,
+      "eval_samples_per_second": 395.17,
+      "eval_steps_per_second": 49.455,
+      "step": 220000
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 6.136111111111111e-06,
+      "loss": 0.9817,
+      "step": 221000
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.5259481037924152,
+      "eval_loss": 0.9672388434410095,
+      "eval_runtime": 12.6468,
+      "eval_samples_per_second": 396.147,
+      "eval_steps_per_second": 49.578,
+      "step": 221000
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 6.163888888888889e-06,
+      "loss": 0.9732,
+      "step": 222000
+    },
+    {
+      "epoch": 9.04,
+      "eval_accuracy": 0.5163672654690619,
+      "eval_loss": 0.9789386987686157,
+      "eval_runtime": 12.4028,
+      "eval_samples_per_second": 403.941,
+      "eval_steps_per_second": 50.553,
+      "step": 222000
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 6.191638888888889e-06,
+      "loss": 0.9687,
+      "step": 223000
+    },
+    {
+      "epoch": 9.09,
+      "eval_accuracy": 0.5313373253493014,
+      "eval_loss": 0.9697375893592834,
+      "eval_runtime": 12.4177,
+      "eval_samples_per_second": 403.456,
+      "eval_steps_per_second": 50.492,
+      "step": 223000
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 6.219416666666667e-06,
+      "loss": 0.971,
+      "step": 224000
+    },
+    {
+      "epoch": 9.13,
+      "eval_accuracy": 0.5217564870259481,
+      "eval_loss": 0.9790107011795044,
+      "eval_runtime": 12.4617,
+      "eval_samples_per_second": 402.032,
+      "eval_steps_per_second": 50.314,
+      "step": 224000
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 6.2471666666666675e-06,
+      "loss": 0.9685,
+      "step": 225000
+    },
+    {
+      "epoch": 9.17,
+      "eval_accuracy": 0.5175648702594811,
+      "eval_loss": 0.9818844795227051,
+      "eval_runtime": 12.6711,
+      "eval_samples_per_second": 395.389,
+      "eval_steps_per_second": 49.483,
+      "step": 225000
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 6.274944444444445e-06,
+      "loss": 0.9685,
+      "step": 226000
+    },
+    {
+      "epoch": 9.21,
+      "eval_accuracy": 0.5251497005988024,
+      "eval_loss": 0.9716070890426636,
+      "eval_runtime": 12.6653,
+      "eval_samples_per_second": 395.568,
+      "eval_steps_per_second": 49.505,
+      "step": 226000
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 6.3026944444444445e-06,
+      "loss": 0.9708,
+      "step": 227000
+    },
+    {
+      "epoch": 9.25,
+      "eval_accuracy": 0.5199600798403193,
+      "eval_loss": 0.9787278771400452,
+      "eval_runtime": 12.387,
+      "eval_samples_per_second": 404.455,
+      "eval_steps_per_second": 50.617,
+      "step": 227000
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 6.330444444444444e-06,
+      "loss": 0.9744,
+      "step": 228000
+    },
+    {
+      "epoch": 9.29,
+      "eval_accuracy": 0.5289421157684631,
+      "eval_loss": 0.9589318633079529,
+      "eval_runtime": 12.3845,
+      "eval_samples_per_second": 404.539,
+      "eval_steps_per_second": 50.628,
+      "step": 228000
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.358222222222222e-06,
+      "loss": 0.9647,
+      "step": 229000
+    },
+    {
+      "epoch": 9.33,
+      "eval_accuracy": 0.5273453093812375,
+      "eval_loss": 0.9603100419044495,
+      "eval_runtime": 12.492,
+      "eval_samples_per_second": 401.055,
+      "eval_steps_per_second": 50.192,
+      "step": 229000
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.386e-06,
+      "loss": 0.9711,
+      "step": 230000
+    },
+    {
+      "epoch": 9.37,
+      "eval_accuracy": 0.5233532934131736,
+      "eval_loss": 0.9708688259124756,
+      "eval_runtime": 12.6713,
+      "eval_samples_per_second": 395.382,
+      "eval_steps_per_second": 49.482,
+      "step": 230000
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 6.413777777777778e-06,
+      "loss": 0.9689,
+      "step": 231000
+    },
+    {
+      "epoch": 9.41,
+      "eval_accuracy": 0.5177644710578843,
+      "eval_loss": 0.9788658022880554,
+      "eval_runtime": 12.8708,
+      "eval_samples_per_second": 389.254,
+      "eval_steps_per_second": 48.715,
+      "step": 231000
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 6.441527777777778e-06,
+      "loss": 0.966,
+      "step": 232000
+    },
+    {
+      "epoch": 9.45,
+      "eval_accuracy": 0.5313373253493014,
+      "eval_loss": 0.9640158414840698,
+      "eval_runtime": 12.4391,
+      "eval_samples_per_second": 402.763,
+      "eval_steps_per_second": 50.406,
+      "step": 232000
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 6.469305555555556e-06,
+      "loss": 0.9687,
+      "step": 233000
+    },
+    {
+      "epoch": 9.49,
+      "eval_accuracy": 0.5319361277445109,
+      "eval_loss": 0.9611833095550537,
+      "eval_runtime": 12.411,
+      "eval_samples_per_second": 403.673,
+      "eval_steps_per_second": 50.52,
+      "step": 233000
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 6.497055555555556e-06,
+      "loss": 0.976,
+      "step": 234000
+    },
+    {
+      "epoch": 9.53,
+      "eval_accuracy": 0.526746506986028,
+      "eval_loss": 0.9661922454833984,
+      "eval_runtime": 12.4228,
+      "eval_samples_per_second": 403.29,
+      "eval_steps_per_second": 50.472,
+      "step": 234000
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 6.524833333333334e-06,
+      "loss": 0.975,
+      "step": 235000
+    },
+    {
+      "epoch": 9.57,
+      "eval_accuracy": 0.5271457085828344,
+      "eval_loss": 0.9553897976875305,
+      "eval_runtime": 12.6442,
+      "eval_samples_per_second": 396.228,
+      "eval_steps_per_second": 49.588,
+      "step": 235000
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 6.552583333333334e-06,
+      "loss": 0.9723,
+      "step": 236000
+    },
+    {
+      "epoch": 9.62,
+      "eval_accuracy": 0.5345309381237525,
+      "eval_loss": 0.9583929777145386,
+      "eval_runtime": 12.5664,
+      "eval_samples_per_second": 398.683,
+      "eval_steps_per_second": 49.895,
+      "step": 236000
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 6.580333333333335e-06,
+      "loss": 0.9732,
+      "step": 237000
+    },
+    {
+      "epoch": 9.66,
+      "eval_accuracy": 0.5401197604790419,
+      "eval_loss": 0.9524350762367249,
+      "eval_runtime": 12.4458,
+      "eval_samples_per_second": 402.544,
+      "eval_steps_per_second": 50.378,
+      "step": 237000
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 6.608111111111111e-06,
+      "loss": 0.9677,
+      "step": 238000
+    },
+    {
+      "epoch": 9.7,
+      "eval_accuracy": 0.5331337325349301,
+      "eval_loss": 0.9517861008644104,
+      "eval_runtime": 12.4885,
+      "eval_samples_per_second": 401.17,
+      "eval_steps_per_second": 50.206,
+      "step": 238000
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 6.635888888888889e-06,
+      "loss": 0.9738,
+      "step": 239000
+    },
+    {
+      "epoch": 9.74,
+      "eval_accuracy": 0.5297405189620759,
+      "eval_loss": 0.952104389667511,
+      "eval_runtime": 12.4878,
+      "eval_samples_per_second": 401.193,
+      "eval_steps_per_second": 50.209,
+      "step": 239000
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.663638888888889e-06,
+      "loss": 0.9736,
+      "step": 240000
+    },
+    {
+      "epoch": 9.78,
+      "eval_accuracy": 0.5359281437125748,
+      "eval_loss": 0.9539325833320618,
+      "eval_runtime": 12.6975,
+      "eval_samples_per_second": 394.567,
+      "eval_steps_per_second": 49.38,
+      "step": 240000
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 6.691416666666667e-06,
+      "loss": 0.9724,
+      "step": 241000
+    },
+    {
+      "epoch": 9.82,
+      "eval_accuracy": 0.5261477045908184,
+      "eval_loss": 0.9607310891151428,
+      "eval_runtime": 12.6091,
+      "eval_samples_per_second": 397.331,
+      "eval_steps_per_second": 49.726,
+      "step": 241000
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 6.719194444444445e-06,
+      "loss": 0.9662,
+      "step": 242000
+    },
+    {
+      "epoch": 9.86,
+      "eval_accuracy": 0.5199600798403193,
+      "eval_loss": 0.9779484272003174,
+      "eval_runtime": 12.5955,
+      "eval_samples_per_second": 397.761,
+      "eval_steps_per_second": 49.78,
+      "step": 242000
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 6.746972222222223e-06,
+      "loss": 0.9659,
+      "step": 243000
+    },
+    {
+      "epoch": 9.9,
+      "eval_accuracy": 0.5305389221556887,
+      "eval_loss": 0.9552076458930969,
+      "eval_runtime": 12.4354,
+      "eval_samples_per_second": 402.881,
+      "eval_steps_per_second": 50.42,
+      "step": 243000
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.7747222222222225e-06,
+      "loss": 0.9735,
+      "step": 244000
+    },
+    {
+      "epoch": 9.94,
+      "eval_accuracy": 0.525748502994012,
+      "eval_loss": 0.9630972146987915,
+      "eval_runtime": 12.426,
+      "eval_samples_per_second": 403.186,
+      "eval_steps_per_second": 50.459,
+      "step": 244000
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.802472222222222e-06,
+      "loss": 0.9636,
+      "step": 245000
+    },
+    {
+      "epoch": 9.98,
+      "eval_accuracy": 0.5355289421157685,
+      "eval_loss": 0.9556819796562195,
+      "eval_runtime": 12.661,
+      "eval_samples_per_second": 395.702,
+      "eval_steps_per_second": 49.522,
+      "step": 245000
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 6.83025e-06,
+      "loss": 0.9635,
+      "step": 246000
+    },
+    {
+      "epoch": 10.02,
+      "eval_accuracy": 0.5277445109780439,
+      "eval_loss": 0.9583378434181213,
+      "eval_runtime": 12.7061,
+      "eval_samples_per_second": 394.3,
+      "eval_steps_per_second": 49.347,
+      "step": 246000
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 6.858e-06,
+      "loss": 0.9622,
+      "step": 247000
+    },
+    {
+      "epoch": 10.06,
+      "eval_accuracy": 0.5305389221556887,
+      "eval_loss": 0.9524170160293579,
+      "eval_runtime": 12.4634,
+      "eval_samples_per_second": 401.978,
+      "eval_steps_per_second": 50.307,
+      "step": 247000
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 6.885777777777778e-06,
+      "loss": 0.9589,
+      "step": 248000
+    },
+    {
+      "epoch": 10.1,
+      "eval_accuracy": 0.5339321357285429,
+      "eval_loss": 0.9549177289009094,
+      "eval_runtime": 12.4705,
+      "eval_samples_per_second": 401.747,
+      "eval_steps_per_second": 50.279,
+      "step": 248000
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 6.913555555555556e-06,
+      "loss": 0.9572,
+      "step": 249000
+    },
+    {
+      "epoch": 10.15,
+      "eval_accuracy": 0.5167664670658683,
+      "eval_loss": 0.9746455550193787,
+      "eval_runtime": 12.3706,
+      "eval_samples_per_second": 404.993,
+      "eval_steps_per_second": 50.685,
+      "step": 249000
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 6.941305555555556e-06,
+      "loss": 0.9698,
+      "step": 250000
+    },
+    {
+      "epoch": 10.19,
+      "eval_accuracy": 0.5239520958083832,
+      "eval_loss": 0.9658505320549011,
+      "eval_runtime": 12.6516,
+      "eval_samples_per_second": 395.999,
+      "eval_steps_per_second": 49.559,
+      "step": 250000
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 6.969083333333333e-06,
+      "loss": 0.961,
+      "step": 251000
+    },
+    {
+      "epoch": 10.23,
+      "eval_accuracy": 0.5323353293413173,
+      "eval_loss": 0.9618701934814453,
+      "eval_runtime": 12.8731,
+      "eval_samples_per_second": 389.185,
+      "eval_steps_per_second": 48.706,
+      "step": 251000
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 6.996861111111111e-06,
+      "loss": 0.9626,
+      "step": 252000
+    },
+    {
+      "epoch": 10.27,
+      "eval_accuracy": 0.5307385229540919,
+      "eval_loss": 0.9668095707893372,
+      "eval_runtime": 12.4181,
+      "eval_samples_per_second": 403.444,
+      "eval_steps_per_second": 50.491,
+      "step": 252000
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 7.024611111111111e-06,
+      "loss": 0.9664,
+      "step": 253000
+    },
+    {
+      "epoch": 10.31,
+      "eval_accuracy": 0.5273453093812375,
+      "eval_loss": 0.9583172798156738,
+      "eval_runtime": 12.4319,
+      "eval_samples_per_second": 402.995,
+      "eval_steps_per_second": 50.435,
+      "step": 253000
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 7.052388888888889e-06,
+      "loss": 0.9655,
+      "step": 254000
+    },
+    {
+      "epoch": 10.35,
+      "eval_accuracy": 0.5315369261477045,
+      "eval_loss": 0.961645245552063,
+      "eval_runtime": 12.4116,
+      "eval_samples_per_second": 403.655,
+      "eval_steps_per_second": 50.517,
+      "step": 254000
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 7.080166666666667e-06,
+      "loss": 0.9608,
+      "step": 255000
+    },
+    {
+      "epoch": 10.39,
+      "eval_accuracy": 0.5379241516966068,
+      "eval_loss": 0.9561013579368591,
+      "eval_runtime": 12.719,
+      "eval_samples_per_second": 393.9,
+      "eval_steps_per_second": 49.296,
+      "step": 255000
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 7.107944444444445e-06,
+      "loss": 0.9561,
+      "step": 256000
+    },
+    {
+      "epoch": 10.43,
+      "eval_accuracy": 0.5253493013972056,
+      "eval_loss": 0.9628321528434753,
+      "eval_runtime": 12.6745,
+      "eval_samples_per_second": 395.283,
+      "eval_steps_per_second": 49.47,
+      "step": 256000
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 7.135666666666667e-06,
+      "loss": 0.9591,
+      "step": 257000
+    },
+    {
+      "epoch": 10.47,
+      "eval_accuracy": 0.5395209580838324,
+      "eval_loss": 0.9499542117118835,
+      "eval_runtime": 12.4075,
+      "eval_samples_per_second": 403.787,
+      "eval_steps_per_second": 50.534,
+      "step": 257000
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 7.163444444444445e-06,
+      "loss": 0.9665,
+      "step": 258000
+    },
+    {
+      "epoch": 10.51,
+      "eval_accuracy": 0.5279441117764471,
+      "eval_loss": 0.9630445241928101,
+      "eval_runtime": 12.5196,
+      "eval_samples_per_second": 400.174,
+      "eval_steps_per_second": 50.082,
+      "step": 258000
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 7.1911944444444446e-06,
+      "loss": 0.9652,
+      "step": 259000
+    },
+    {
+      "epoch": 10.55,
+      "eval_accuracy": 0.5341317365269461,
+      "eval_loss": 0.9542832374572754,
+      "eval_runtime": 12.4282,
+      "eval_samples_per_second": 403.116,
+      "eval_steps_per_second": 50.45,
+      "step": 259000
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 7.2189722222222226e-06,
+      "loss": 0.9604,
+      "step": 260000
+    },
+    {
+      "epoch": 10.59,
+      "eval_accuracy": 0.5303393213572855,
+      "eval_loss": 0.9566397666931152,
+      "eval_runtime": 12.6551,
+      "eval_samples_per_second": 395.889,
+      "eval_steps_per_second": 49.545,
+      "step": 260000
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 7.2467500000000005e-06,
+      "loss": 0.9535,
+      "step": 261000
+    },
+    {
+      "epoch": 10.63,
+      "eval_accuracy": 0.5387225548902196,
+      "eval_loss": 0.9569635391235352,
+      "eval_runtime": 12.6484,
+      "eval_samples_per_second": 396.098,
+      "eval_steps_per_second": 49.571,
+      "step": 261000
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 7.2745e-06,
+      "loss": 0.961,
+      "step": 262000
+    },
+    {
+      "epoch": 10.67,
+      "eval_accuracy": 0.5253493013972056,
+      "eval_loss": 0.96352618932724,
+      "eval_runtime": 12.4534,
+      "eval_samples_per_second": 402.299,
+      "eval_steps_per_second": 50.348,
+      "step": 262000
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 7.302277777777778e-06,
+      "loss": 0.9585,
+      "step": 263000
+    },
+    {
+      "epoch": 10.72,
+      "eval_accuracy": 0.5313373253493014,
+      "eval_loss": 0.957420289516449,
+      "eval_runtime": 12.4472,
+      "eval_samples_per_second": 402.502,
+      "eval_steps_per_second": 50.373,
+      "step": 263000
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 7.3300555555555555e-06,
+      "loss": 0.957,
+      "step": 264000
+    },
+    {
+      "epoch": 10.76,
+      "eval_accuracy": 0.5261477045908184,
+      "eval_loss": 0.9583243727684021,
+      "eval_runtime": 12.4481,
+      "eval_samples_per_second": 402.47,
+      "eval_steps_per_second": 50.369,
+      "step": 264000
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 7.3578333333333335e-06,
+      "loss": 0.9555,
+      "step": 265000
+    },
+    {
+      "epoch": 10.8,
+      "eval_accuracy": 0.5399201596806387,
+      "eval_loss": 0.9512184858322144,
+      "eval_runtime": 12.6707,
+      "eval_samples_per_second": 395.401,
+      "eval_steps_per_second": 49.484,
+      "step": 265000
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 7.385583333333333e-06,
+      "loss": 0.9662,
+      "step": 266000
+    },
+    {
+      "epoch": 10.84,
+      "eval_accuracy": 0.5313373253493014,
+      "eval_loss": 0.9524006843566895,
+      "eval_runtime": 12.8095,
+      "eval_samples_per_second": 391.115,
+      "eval_steps_per_second": 48.948,
+      "step": 266000
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 7.413361111111111e-06,
+      "loss": 0.9618,
+      "step": 267000
+    },
+    {
+      "epoch": 10.88,
+      "eval_accuracy": 0.5333333333333333,
+      "eval_loss": 0.9507670998573303,
+      "eval_runtime": 12.6773,
+      "eval_samples_per_second": 395.196,
+      "eval_steps_per_second": 49.459,
+      "step": 267000
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 7.441111111111111e-06,
+      "loss": 0.961,
+      "step": 268000
+    },
+    {
+      "epoch": 10.92,
+      "eval_accuracy": 0.5299401197604791,
+      "eval_loss": 0.9538514018058777,
+      "eval_runtime": 12.5073,
+      "eval_samples_per_second": 400.567,
+      "eval_steps_per_second": 50.131,
+      "step": 268000
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 7.468888888888889e-06,
+      "loss": 0.9639,
+      "step": 269000
+    },
+    {
+      "epoch": 10.96,
+      "eval_accuracy": 0.5369261477045908,
+      "eval_loss": 0.948660671710968,
+      "eval_runtime": 12.4387,
+      "eval_samples_per_second": 402.776,
+      "eval_steps_per_second": 50.407,
+      "step": 269000
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 7.496611111111111e-06,
+      "loss": 0.9586,
+      "step": 270000
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.5385229540918164,
+      "eval_loss": 0.9449055194854736,
+      "eval_runtime": 12.452,
+      "eval_samples_per_second": 402.345,
+      "eval_steps_per_second": 50.353,
+      "step": 270000
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 7.524388888888889e-06,
+      "loss": 0.9547,
+      "step": 271000
+    },
+    {
+      "epoch": 11.04,
+      "eval_accuracy": 0.5353293413173653,
+      "eval_loss": 0.9503560066223145,
+      "eval_runtime": 12.7025,
+      "eval_samples_per_second": 394.409,
+      "eval_steps_per_second": 49.36,
+      "step": 271000
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 7.552166666666667e-06,
+      "loss": 0.9474,
+      "step": 272000
+    },
+    {
+      "epoch": 11.08,
+      "eval_accuracy": 0.5111776447105788,
+      "eval_loss": 0.9692345857620239,
+      "eval_runtime": 12.688,
+      "eval_samples_per_second": 394.863,
+      "eval_steps_per_second": 49.417,
+      "step": 272000
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 7.579944444444445e-06,
+      "loss": 0.9585,
+      "step": 273000
+    },
+    {
+      "epoch": 11.12,
+      "eval_accuracy": 0.5327345309381237,
+      "eval_loss": 0.9541236162185669,
+      "eval_runtime": 12.4677,
+      "eval_samples_per_second": 401.84,
+      "eval_steps_per_second": 50.29,
+      "step": 273000
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 7.607694444444445e-06,
+      "loss": 0.95,
+      "step": 274000
+    },
+    {
+      "epoch": 11.16,
+      "eval_accuracy": 0.5121756487025948,
+      "eval_loss": 0.9793220162391663,
+      "eval_runtime": 12.5506,
+      "eval_samples_per_second": 399.183,
+      "eval_steps_per_second": 49.958,
+      "step": 274000
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 7.635472222222223e-06,
+      "loss": 0.9499,
+      "step": 275000
+    },
+    {
+      "epoch": 11.2,
+      "eval_accuracy": 0.5387225548902196,
+      "eval_loss": 0.9451339244842529,
+      "eval_runtime": 12.469,
+      "eval_samples_per_second": 401.798,
+      "eval_steps_per_second": 50.285,
+      "step": 275000
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 7.663222222222223e-06,
+      "loss": 0.9534,
+      "step": 276000
+    },
+    {
+      "epoch": 11.25,
+      "eval_accuracy": 0.5367265469061876,
+      "eval_loss": 0.9460021257400513,
+      "eval_runtime": 12.618,
+      "eval_samples_per_second": 397.053,
+      "eval_steps_per_second": 49.691,
+      "step": 276000
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 7.691000000000001e-06,
+      "loss": 0.9516,
+      "step": 277000
+    },
+    {
+      "epoch": 11.29,
+      "eval_accuracy": 0.5331337325349301,
+      "eval_loss": 0.9560016393661499,
+      "eval_runtime": 12.704,
+      "eval_samples_per_second": 394.364,
+      "eval_steps_per_second": 49.355,
+      "step": 277000
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 7.718777777777777e-06,
+      "loss": 0.956,
+      "step": 278000
+    },
+    {
+      "epoch": 11.33,
+      "eval_accuracy": 0.5245508982035928,
+      "eval_loss": 0.9636676907539368,
+      "eval_runtime": 12.6995,
+      "eval_samples_per_second": 394.502,
+      "eval_steps_per_second": 49.372,
+      "step": 278000
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 7.746527777777777e-06,
+      "loss": 0.9527,
+      "step": 279000
+    },
+    {
+      "epoch": 11.37,
+      "eval_accuracy": 0.5377245508982036,
+      "eval_loss": 0.9432642459869385,
+      "eval_runtime": 12.4866,
+      "eval_samples_per_second": 401.229,
+      "eval_steps_per_second": 50.214,
+      "step": 279000
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 7.774305555555556e-06,
+      "loss": 0.9508,
+      "step": 280000
+    },
+    {
+      "epoch": 11.41,
+      "eval_accuracy": 0.524750499001996,
+      "eval_loss": 0.9705446362495422,
+      "eval_runtime": 12.4449,
+      "eval_samples_per_second": 402.573,
+      "eval_steps_per_second": 50.382,
+      "step": 280000
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 7.802083333333333e-06,
+      "loss": 0.9563,
+      "step": 281000
+    },
+    {
+      "epoch": 11.45,
+      "eval_accuracy": 0.5377245508982036,
+      "eval_loss": 0.9441819190979004,
+      "eval_runtime": 12.4064,
+      "eval_samples_per_second": 403.823,
+      "eval_steps_per_second": 50.538,
+      "step": 281000
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 7.829861111111112e-06,
+      "loss": 0.9463,
+      "step": 282000
+    },
+    {
+      "epoch": 11.49,
+      "eval_accuracy": 0.5339321357285429,
+      "eval_loss": 0.9488281011581421,
+      "eval_runtime": 12.7117,
+      "eval_samples_per_second": 394.124,
+      "eval_steps_per_second": 49.325,
+      "step": 282000
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 7.857611111111111e-06,
+      "loss": 0.9588,
+      "step": 283000
+    },
+    {
+      "epoch": 11.53,
+      "eval_accuracy": 0.5331337325349301,
+      "eval_loss": 0.9450494647026062,
+      "eval_runtime": 12.6548,
+      "eval_samples_per_second": 395.898,
+      "eval_steps_per_second": 49.547,
+      "step": 283000
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 7.88538888888889e-06,
+      "loss": 0.9406,
+      "step": 284000
+    },
+    {
+      "epoch": 11.57,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.9425448775291443,
+      "eval_runtime": 12.4136,
+      "eval_samples_per_second": 403.589,
+      "eval_steps_per_second": 50.509,
+      "step": 284000
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 7.91313888888889e-06,
+      "loss": 0.9522,
+      "step": 285000
+    },
+    {
+      "epoch": 11.61,
+      "eval_accuracy": 0.5293413173652695,
+      "eval_loss": 0.9536769986152649,
+      "eval_runtime": 12.4749,
+      "eval_samples_per_second": 401.607,
+      "eval_steps_per_second": 50.261,
+      "step": 285000
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 7.940916666666667e-06,
+      "loss": 0.9546,
+      "step": 286000
+    },
+    {
+      "epoch": 11.65,
+      "eval_accuracy": 0.5253493013972056,
+      "eval_loss": 0.9613157510757446,
+      "eval_runtime": 12.4466,
+      "eval_samples_per_second": 402.521,
+      "eval_steps_per_second": 50.375,
+      "step": 286000
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 7.968666666666667e-06,
+      "loss": 0.9595,
+      "step": 287000
+    },
+    {
+      "epoch": 11.69,
+      "eval_accuracy": 0.5337325349301397,
+      "eval_loss": 0.9555002450942993,
+      "eval_runtime": 12.7808,
+      "eval_samples_per_second": 391.994,
+      "eval_steps_per_second": 49.058,
+      "step": 287000
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 7.996444444444446e-06,
+      "loss": 0.9525,
+      "step": 288000
+    },
+    {
+      "epoch": 11.73,
+      "eval_accuracy": 0.5265469061876248,
+      "eval_loss": 0.9589730501174927,
+      "eval_runtime": 12.6944,
+      "eval_samples_per_second": 394.661,
+      "eval_steps_per_second": 49.392,
+      "step": 288000
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 8.024194444444446e-06,
+      "loss": 0.9579,
+      "step": 289000
+    },
+    {
+      "epoch": 11.77,
+      "eval_accuracy": 0.52375249500998,
+      "eval_loss": 0.960227906703949,
+      "eval_runtime": 12.4913,
+      "eval_samples_per_second": 401.08,
+      "eval_steps_per_second": 50.195,
+      "step": 289000
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 8.051972222222223e-06,
+      "loss": 0.9514,
+      "step": 290000
+    },
+    {
+      "epoch": 11.82,
+      "eval_accuracy": 0.5377245508982036,
+      "eval_loss": 0.9468502402305603,
+      "eval_runtime": 12.4564,
+      "eval_samples_per_second": 402.203,
+      "eval_steps_per_second": 50.336,
+      "step": 290000
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 8.07975e-06,
+      "loss": 0.9555,
+      "step": 291000
+    },
+    {
+      "epoch": 11.86,
+      "eval_accuracy": 0.542315369261477,
+      "eval_loss": 0.9452052712440491,
+      "eval_runtime": 12.4859,
+      "eval_samples_per_second": 401.252,
+      "eval_steps_per_second": 50.217,
+      "step": 291000
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 8.1075e-06,
+      "loss": 0.9459,
+      "step": 292000
+    },
+    {
+      "epoch": 11.9,
+      "eval_accuracy": 0.5291417165668663,
+      "eval_loss": 0.9664433002471924,
+      "eval_runtime": 12.4925,
+      "eval_samples_per_second": 401.041,
+      "eval_steps_per_second": 50.19,
+      "step": 292000
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 8.135277777777777e-06,
+      "loss": 0.9563,
+      "step": 293000
+    },
+    {
+      "epoch": 11.94,
+      "eval_accuracy": 0.5351297405189621,
+      "eval_loss": 0.9565455317497253,
+      "eval_runtime": 12.7596,
+      "eval_samples_per_second": 392.645,
+      "eval_steps_per_second": 49.139,
+      "step": 293000
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 8.163027777777779e-06,
+      "loss": 0.949,
+      "step": 294000
+    },
+    {
+      "epoch": 11.98,
+      "eval_accuracy": 0.5439121756487026,
+      "eval_loss": 0.9515088796615601,
+      "eval_runtime": 12.7274,
+      "eval_samples_per_second": 393.638,
+      "eval_steps_per_second": 49.264,
+      "step": 294000
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 8.190805555555556e-06,
+      "loss": 0.9507,
+      "step": 295000
+    },
+    {
+      "epoch": 12.02,
+      "eval_accuracy": 0.536127744510978,
+      "eval_loss": 0.9496365785598755,
+      "eval_runtime": 12.4581,
+      "eval_samples_per_second": 402.149,
+      "eval_steps_per_second": 50.329,
+      "step": 295000
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 8.218555555555556e-06,
+      "loss": 0.9428,
+      "step": 296000
+    },
+    {
+      "epoch": 12.06,
+      "eval_accuracy": 0.5303393213572855,
+      "eval_loss": 0.9757717251777649,
+      "eval_runtime": 12.4508,
+      "eval_samples_per_second": 402.383,
+      "eval_steps_per_second": 50.358,
+      "step": 296000
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 8.246333333333334e-06,
+      "loss": 0.937,
+      "step": 297000
+    },
+    {
+      "epoch": 12.1,
+      "eval_accuracy": 0.5259481037924152,
+      "eval_loss": 0.978837251663208,
+      "eval_runtime": 12.5049,
+      "eval_samples_per_second": 400.643,
+      "eval_steps_per_second": 50.14,
+      "step": 297000
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 8.274083333333334e-06,
+      "loss": 0.9449,
+      "step": 298000
+    },
+    {
+      "epoch": 12.14,
+      "eval_accuracy": 0.5285429141716567,
+      "eval_loss": 0.9536109566688538,
+      "eval_runtime": 12.7697,
+      "eval_samples_per_second": 392.334,
+      "eval_steps_per_second": 49.1,
+      "step": 298000
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 8.301861111111111e-06,
+      "loss": 0.9424,
+      "step": 299000
+    },
+    {
+      "epoch": 12.18,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.9432350397109985,
+      "eval_runtime": 12.8344,
+      "eval_samples_per_second": 390.357,
+      "eval_steps_per_second": 48.853,
+      "step": 299000
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 8.329611111111111e-06,
+      "loss": 0.9346,
+      "step": 300000
+    },
+    {
+      "epoch": 12.22,
+      "eval_accuracy": 0.530938123752495,
+      "eval_loss": 0.9528337121009827,
+      "eval_runtime": 12.847,
+      "eval_samples_per_second": 389.973,
+      "eval_steps_per_second": 48.805,
+      "step": 300000
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 8.35738888888889e-06,
+      "loss": 0.9452,
+      "step": 301000
+    },
+    {
+      "epoch": 12.26,
+      "eval_accuracy": 0.5245508982035928,
+      "eval_loss": 0.9617663025856018,
+      "eval_runtime": 12.5528,
+      "eval_samples_per_second": 399.115,
+      "eval_steps_per_second": 49.949,
+      "step": 301000
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 8.385111111111111e-06,
+      "loss": 0.9452,
+      "step": 302000
+    },
+    {
+      "epoch": 12.3,
+      "eval_accuracy": 0.5275449101796407,
+      "eval_loss": 0.9557837247848511,
+      "eval_runtime": 12.5062,
+      "eval_samples_per_second": 400.602,
+      "eval_steps_per_second": 50.135,
+      "step": 302000
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 8.41288888888889e-06,
+      "loss": 0.9445,
+      "step": 303000
+    },
+    {
+      "epoch": 12.35,
+      "eval_accuracy": 0.5225548902195609,
+      "eval_loss": 0.9665641188621521,
+      "eval_runtime": 12.5422,
+      "eval_samples_per_second": 399.45,
+      "eval_steps_per_second": 49.991,
+      "step": 303000
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 8.440666666666667e-06,
+      "loss": 0.9414,
+      "step": 304000
+    },
+    {
+      "epoch": 12.39,
+      "eval_accuracy": 0.5305389221556887,
+      "eval_loss": 0.9500705599784851,
+      "eval_runtime": 12.7449,
+      "eval_samples_per_second": 393.099,
+      "eval_steps_per_second": 49.196,
+      "step": 304000
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 8.468444444444446e-06,
+      "loss": 0.9494,
+      "step": 305000
+    },
+    {
+      "epoch": 12.43,
+      "eval_accuracy": 0.536127744510978,
+      "eval_loss": 0.952370285987854,
+      "eval_runtime": 12.8032,
+      "eval_samples_per_second": 391.309,
+      "eval_steps_per_second": 48.972,
+      "step": 305000
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 8.496194444444446e-06,
+      "loss": 0.9474,
+      "step": 306000
+    },
+    {
+      "epoch": 12.47,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.9470945596694946,
+      "eval_runtime": 12.7933,
+      "eval_samples_per_second": 391.61,
+      "eval_steps_per_second": 49.01,
+      "step": 306000
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 8.523972222222223e-06,
+      "loss": 0.9438,
+      "step": 307000
+    },
+    {
+      "epoch": 12.51,
+      "eval_accuracy": 0.5375249500998004,
+      "eval_loss": 0.9444052577018738,
+      "eval_runtime": 12.5386,
+      "eval_samples_per_second": 399.565,
+      "eval_steps_per_second": 50.005,
+      "step": 307000
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 8.551694444444445e-06,
+      "loss": 0.9439,
+      "step": 308000
+    },
+    {
+      "epoch": 12.55,
+      "eval_accuracy": 0.5283433133732535,
+      "eval_loss": 0.9567458033561707,
+      "eval_runtime": 12.447,
+      "eval_samples_per_second": 402.505,
+      "eval_steps_per_second": 50.373,
+      "step": 308000
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 8.579472222222222e-06,
+      "loss": 0.9503,
+      "step": 309000
+    },
+    {
+      "epoch": 12.59,
+      "eval_accuracy": 0.5311377245508982,
+      "eval_loss": 0.9504968523979187,
+      "eval_runtime": 12.5195,
+      "eval_samples_per_second": 400.175,
+      "eval_steps_per_second": 50.082,
+      "step": 309000
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 8.60725e-06,
+      "loss": 0.9483,
+      "step": 310000
+    },
+    {
+      "epoch": 12.63,
+      "eval_accuracy": 0.5305389221556887,
+      "eval_loss": 0.9485623240470886,
+      "eval_runtime": 12.7029,
+      "eval_samples_per_second": 394.399,
+      "eval_steps_per_second": 49.359,
+      "step": 310000
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 8.635027777777778e-06,
+      "loss": 0.9439,
+      "step": 311000
+    },
+    {
+      "epoch": 12.67,
+      "eval_accuracy": 0.536127744510978,
+      "eval_loss": 0.9592593312263489,
+      "eval_runtime": 12.7265,
+      "eval_samples_per_second": 393.668,
+      "eval_steps_per_second": 49.267,
+      "step": 311000
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 8.662777777777778e-06,
+      "loss": 0.9468,
+      "step": 312000
+    },
+    {
+      "epoch": 12.71,
+      "eval_accuracy": 0.5421157684630739,
+      "eval_loss": 0.9490076303482056,
+      "eval_runtime": 12.4775,
+      "eval_samples_per_second": 401.522,
+      "eval_steps_per_second": 50.25,
+      "step": 312000
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 8.690555555555555e-06,
+      "loss": 0.9513,
+      "step": 313000
+    },
+    {
+      "epoch": 12.75,
+      "eval_accuracy": 0.5341317365269461,
+      "eval_loss": 0.9534984827041626,
+      "eval_runtime": 12.4931,
+      "eval_samples_per_second": 401.022,
+      "eval_steps_per_second": 50.188,
+      "step": 313000
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 8.718305555555555e-06,
+      "loss": 0.9444,
+      "step": 314000
+    },
+    {
+      "epoch": 12.79,
+      "eval_accuracy": 0.5367265469061876,
+      "eval_loss": 0.9456366300582886,
+      "eval_runtime": 12.4767,
+      "eval_samples_per_second": 401.55,
+      "eval_steps_per_second": 50.254,
+      "step": 314000
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 8.746083333333334e-06,
+      "loss": 0.9483,
+      "step": 315000
+    },
+    {
+      "epoch": 12.83,
+      "eval_accuracy": 0.5365269461077844,
+      "eval_loss": 0.945328950881958,
+      "eval_runtime": 12.5135,
+      "eval_samples_per_second": 400.368,
+      "eval_steps_per_second": 50.106,
+      "step": 315000
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 8.773833333333334e-06,
+      "loss": 0.9377,
+      "step": 316000
+    },
+    {
+      "epoch": 12.87,
+      "eval_accuracy": 0.5455089820359281,
+      "eval_loss": 0.9395766854286194,
+      "eval_runtime": 12.7254,
+      "eval_samples_per_second": 393.702,
+      "eval_steps_per_second": 49.272,
+      "step": 316000
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 8.801611111111111e-06,
+      "loss": 0.9528,
+      "step": 317000
+    },
+    {
+      "epoch": 12.92,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.9379430413246155,
+      "eval_runtime": 12.7499,
+      "eval_samples_per_second": 392.944,
+      "eval_steps_per_second": 49.177,
+      "step": 317000
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 8.829361111111113e-06,
+      "loss": 0.9503,
+      "step": 318000
+    },
+    {
+      "epoch": 12.96,
+      "eval_accuracy": 0.5363273453093812,
+      "eval_loss": 0.9473833441734314,
+      "eval_runtime": 12.4854,
+      "eval_samples_per_second": 401.268,
+      "eval_steps_per_second": 50.219,
+      "step": 318000
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 8.85713888888889e-06,
+      "loss": 0.9485,
+      "step": 319000
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.5387225548902196,
+      "eval_loss": 0.95295250415802,
+      "eval_runtime": 12.4379,
+      "eval_samples_per_second": 402.8,
+      "eval_steps_per_second": 50.41,
+      "step": 319000
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 8.88488888888889e-06,
+      "loss": 0.9394,
+      "step": 320000
+    },
+    {
+      "epoch": 13.04,
+      "eval_accuracy": 0.5427145708582835,
+      "eval_loss": 0.9406971335411072,
+      "eval_runtime": 12.7195,
+      "eval_samples_per_second": 393.884,
+      "eval_steps_per_second": 49.295,
+      "step": 320000
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 8.912666666666669e-06,
+      "loss": 0.9317,
+      "step": 321000
+    },
+    {
+      "epoch": 13.08,
+      "eval_accuracy": 0.542315369261477,
+      "eval_loss": 0.9397866129875183,
+      "eval_runtime": 12.6428,
+      "eval_samples_per_second": 396.272,
+      "eval_steps_per_second": 49.593,
+      "step": 321000
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 8.940416666666668e-06,
+      "loss": 0.9421,
+      "step": 322000
+    },
+    {
+      "epoch": 13.12,
+      "eval_accuracy": 0.5441117764471058,
+      "eval_loss": 0.9368385076522827,
+      "eval_runtime": 12.7394,
+      "eval_samples_per_second": 393.269,
+      "eval_steps_per_second": 49.218,
+      "step": 322000
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 8.968194444444444e-06,
+      "loss": 0.9328,
+      "step": 323000
+    },
+    {
+      "epoch": 13.16,
+      "eval_accuracy": 0.5451097804391217,
+      "eval_loss": 0.9509682059288025,
+      "eval_runtime": 12.7056,
+      "eval_samples_per_second": 394.315,
+      "eval_steps_per_second": 49.348,
+      "step": 323000
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 8.995944444444444e-06,
+      "loss": 0.9354,
+      "step": 324000
+    },
+    {
+      "epoch": 13.2,
+      "eval_accuracy": 0.5415169660678643,
+      "eval_loss": 0.9415899515151978,
+      "eval_runtime": 12.4566,
+      "eval_samples_per_second": 402.197,
+      "eval_steps_per_second": 50.335,
+      "step": 324000
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 9.023722222222223e-06,
+      "loss": 0.9293,
+      "step": 325000
+    },
+    {
+      "epoch": 13.24,
+      "eval_accuracy": 0.5365269461077844,
+      "eval_loss": 0.9573070406913757,
+      "eval_runtime": 12.4547,
+      "eval_samples_per_second": 402.257,
+      "eval_steps_per_second": 50.342,
+      "step": 325000
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 9.051472222222222e-06,
+      "loss": 0.9337,
+      "step": 326000
+    },
+    {
+      "epoch": 13.28,
+      "eval_accuracy": 0.5305389221556887,
+      "eval_loss": 0.9645636677742004,
+      "eval_runtime": 12.682,
+      "eval_samples_per_second": 395.049,
+      "eval_steps_per_second": 49.44,
+      "step": 326000
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 9.07925e-06,
+      "loss": 0.9411,
+      "step": 327000
+    },
+    {
+      "epoch": 13.32,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.9383572936058044,
+      "eval_runtime": 12.8095,
+      "eval_samples_per_second": 391.116,
+      "eval_steps_per_second": 48.948,
+      "step": 327000
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 9.107000000000001e-06,
+      "loss": 0.9314,
+      "step": 328000
+    },
+    {
+      "epoch": 13.36,
+      "eval_accuracy": 0.5357285429141716,
+      "eval_loss": 0.9546739459037781,
+      "eval_runtime": 12.7698,
+      "eval_samples_per_second": 392.332,
+      "eval_steps_per_second": 49.1,
+      "step": 328000
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 9.134777777777778e-06,
+      "loss": 0.9397,
+      "step": 329000
+    },
+    {
+      "epoch": 13.4,
+      "eval_accuracy": 0.5411177644710579,
+      "eval_loss": 0.9408701062202454,
+      "eval_runtime": 12.6085,
+      "eval_samples_per_second": 397.35,
+      "eval_steps_per_second": 49.728,
+      "step": 329000
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 9.162527777777778e-06,
+      "loss": 0.9364,
+      "step": 330000
+    },
+    {
+      "epoch": 13.45,
+      "eval_accuracy": 0.5421157684630739,
+      "eval_loss": 0.9512919187545776,
+      "eval_runtime": 12.4482,
+      "eval_samples_per_second": 402.467,
+      "eval_steps_per_second": 50.369,
+      "step": 330000
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 9.190305555555557e-06,
+      "loss": 0.939,
+      "step": 331000
+    },
+    {
+      "epoch": 13.49,
+      "eval_accuracy": 0.5407185628742515,
+      "eval_loss": 0.9456509947776794,
+      "eval_runtime": 12.433,
+      "eval_samples_per_second": 402.959,
+      "eval_steps_per_second": 50.43,
+      "step": 331000
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 9.218055555555557e-06,
+      "loss": 0.9343,
+      "step": 332000
+    },
+    {
+      "epoch": 13.53,
+      "eval_accuracy": 0.5403193612774451,
+      "eval_loss": 0.9462023377418518,
+      "eval_runtime": 12.4252,
+      "eval_samples_per_second": 403.212,
+      "eval_steps_per_second": 50.462,
+      "step": 332000
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 9.245805555555557e-06,
+      "loss": 0.9386,
+      "step": 333000
+    },
+    {
+      "epoch": 13.57,
+      "eval_accuracy": 0.5411177644710579,
+      "eval_loss": 0.9454513788223267,
+      "eval_runtime": 12.743,
+      "eval_samples_per_second": 393.157,
+      "eval_steps_per_second": 49.204,
+      "step": 333000
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 9.273583333333334e-06,
+      "loss": 0.9418,
+      "step": 334000
+    },
+    {
+      "epoch": 13.61,
+      "eval_accuracy": 0.5395209580838324,
+      "eval_loss": 0.9482743740081787,
+      "eval_runtime": 12.679,
+      "eval_samples_per_second": 395.142,
+      "eval_steps_per_second": 49.452,
+      "step": 334000
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 9.301361111111113e-06,
+      "loss": 0.9382,
+      "step": 335000
+    },
+    {
+      "epoch": 13.65,
+      "eval_accuracy": 0.5293413173652695,
+      "eval_loss": 0.9673813581466675,
+      "eval_runtime": 12.5698,
+      "eval_samples_per_second": 398.573,
+      "eval_steps_per_second": 49.881,
+      "step": 335000
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.329138888888888e-06,
+      "loss": 0.9381,
+      "step": 336000
+    },
+    {
+      "epoch": 13.69,
+      "eval_accuracy": 0.5425149700598803,
+      "eval_loss": 0.9426034688949585,
+      "eval_runtime": 12.4664,
+      "eval_samples_per_second": 401.88,
+      "eval_steps_per_second": 50.295,
+      "step": 336000
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.356888888888888e-06,
+      "loss": 0.9411,
+      "step": 337000
+    },
+    {
+      "epoch": 13.73,
+      "eval_accuracy": 0.5355289421157685,
+      "eval_loss": 0.9570037722587585,
+      "eval_runtime": 12.5137,
+      "eval_samples_per_second": 400.361,
+      "eval_steps_per_second": 50.105,
+      "step": 337000
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.384666666666667e-06,
+      "loss": 0.9381,
+      "step": 338000
+    },
+    {
+      "epoch": 13.77,
+      "eval_accuracy": 0.5369261477045908,
+      "eval_loss": 0.9421893358230591,
+      "eval_runtime": 12.8594,
+      "eval_samples_per_second": 389.598,
+      "eval_steps_per_second": 48.758,
+      "step": 338000
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.412416666666667e-06,
+      "loss": 0.9412,
+      "step": 339000
+    },
+    {
+      "epoch": 13.81,
+      "eval_accuracy": 0.5427145708582835,
+      "eval_loss": 0.9445858001708984,
+      "eval_runtime": 12.7414,
+      "eval_samples_per_second": 393.207,
+      "eval_steps_per_second": 49.21,
+      "step": 339000
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.440194444444445e-06,
+      "loss": 0.9381,
+      "step": 340000
+    },
+    {
+      "epoch": 13.85,
+      "eval_accuracy": 0.5377245508982036,
+      "eval_loss": 0.9491099119186401,
+      "eval_runtime": 12.7302,
+      "eval_samples_per_second": 393.553,
+      "eval_steps_per_second": 49.253,
+      "step": 340000
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.467916666666666e-06,
+      "loss": 0.9302,
+      "step": 341000
+    },
+    {
+      "epoch": 13.89,
+      "eval_accuracy": 0.5405189620758483,
+      "eval_loss": 0.938602864742279,
+      "eval_runtime": 12.5313,
+      "eval_samples_per_second": 399.798,
+      "eval_steps_per_second": 50.035,
+      "step": 341000
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.495694444444445e-06,
+      "loss": 0.9318,
+      "step": 342000
+    },
+    {
+      "epoch": 13.93,
+      "eval_accuracy": 0.5335329341317365,
+      "eval_loss": 0.9562578201293945,
+      "eval_runtime": 12.5554,
+      "eval_samples_per_second": 399.03,
+      "eval_steps_per_second": 49.938,
+      "step": 342000
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.523472222222222e-06,
+      "loss": 0.9429,
+      "step": 343000
+    },
+    {
+      "epoch": 13.97,
+      "eval_accuracy": 0.5321357285429141,
+      "eval_loss": 0.9488294720649719,
+      "eval_runtime": 12.4628,
+      "eval_samples_per_second": 401.996,
+      "eval_steps_per_second": 50.31,
+      "step": 343000
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.551250000000001e-06,
+      "loss": 0.9303,
+      "step": 344000
+    },
+    {
+      "epoch": 14.02,
+      "eval_accuracy": 0.542315369261477,
+      "eval_loss": 0.9512555003166199,
+      "eval_runtime": 12.6385,
+      "eval_samples_per_second": 396.408,
+      "eval_steps_per_second": 49.61,
+      "step": 344000
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.579000000000001e-06,
+      "loss": 0.9142,
+      "step": 345000
+    },
+    {
+      "epoch": 14.06,
+      "eval_accuracy": 0.5459081836327345,
+      "eval_loss": 0.9470293521881104,
+      "eval_runtime": 12.7265,
+      "eval_samples_per_second": 393.668,
+      "eval_steps_per_second": 49.267,
+      "step": 345000
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.606777777777778e-06,
+      "loss": 0.9251,
+      "step": 346000
+    },
+    {
+      "epoch": 14.1,
+      "eval_accuracy": 0.530938123752495,
+      "eval_loss": 0.9578275084495544,
+      "eval_runtime": 13.1592,
+      "eval_samples_per_second": 380.723,
+      "eval_steps_per_second": 47.647,
+      "step": 346000
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.634527777777778e-06,
+      "loss": 0.9276,
+      "step": 347000
+    },
+    {
+      "epoch": 14.14,
+      "eval_accuracy": 0.5373253493013972,
+      "eval_loss": 0.950298547744751,
+      "eval_runtime": 12.4421,
+      "eval_samples_per_second": 402.666,
+      "eval_steps_per_second": 50.394,
+      "step": 347000
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.662305555555557e-06,
+      "loss": 0.9227,
+      "step": 348000
+    },
+    {
+      "epoch": 14.18,
+      "eval_accuracy": 0.5445109780439121,
+      "eval_loss": 0.9493466019630432,
+      "eval_runtime": 12.5973,
+      "eval_samples_per_second": 397.706,
+      "eval_steps_per_second": 49.773,
+      "step": 348000
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.690055555555557e-06,
+      "loss": 0.9239,
+      "step": 349000
+    },
+    {
+      "epoch": 14.22,
+      "eval_accuracy": 0.536127744510978,
+      "eval_loss": 0.9692854285240173,
+      "eval_runtime": 12.4548,
+      "eval_samples_per_second": 402.254,
+      "eval_steps_per_second": 50.342,
+      "step": 349000
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.717833333333334e-06,
+      "loss": 0.9267,
+      "step": 350000
+    },
+    {
+      "epoch": 14.26,
+      "eval_accuracy": 0.53812375249501,
+      "eval_loss": 0.9507613778114319,
+      "eval_runtime": 12.5451,
+      "eval_samples_per_second": 399.359,
+      "eval_steps_per_second": 49.98,
+      "step": 350000
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.745583333333334e-06,
+      "loss": 0.9282,
+      "step": 351000
+    },
+    {
+      "epoch": 14.3,
+      "eval_accuracy": 0.5389221556886228,
+      "eval_loss": 0.9410892128944397,
+      "eval_runtime": 12.7548,
+      "eval_samples_per_second": 392.794,
+      "eval_steps_per_second": 49.158,
+      "step": 351000
+    }
+  ],
+  "logging_steps": 1000,
+  "max_steps": 10000000,
+  "num_train_epochs": 408,
+  "save_steps": 1000,
+  "total_flos": 1.2230172761495962e+17,
+  "trial_name": null,
+  "trial_params": null
+}