diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,102025 @@
+{
+  "best_metric": 0.8059701492537313,
+  "best_model_checkpoint": "sketch-classifier/checkpoint-62930",
+  "epoch": 999.9967845659164,
+  "global_step": 155000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.06,
+      "learning_rate": 3.2258064516129035e-08,
+      "loss": 0.387,
+      "step": 10
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.451612903225807e-08,
+      "loss": 0.4275,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.67741935483871e-08,
+      "loss": 0.3798,
+      "step": 30
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.2903225806451614e-07,
+      "loss": 0.4312,
+      "step": 40
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6129032258064518e-07,
+      "loss": 0.3813,
+      "step": 50
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.935483870967742e-07,
+      "loss": 0.3515,
+      "step": 60
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.2580645161290325e-07,
+      "loss": 0.4242,
+      "step": 70
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.580645161290323e-07,
+      "loss": 0.4278,
+      "step": 80
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.9032258064516134e-07,
+      "loss": 0.4366,
+      "step": 90
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2258064516129035e-07,
+      "loss": 0.3937,
+      "step": 100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.548387096774194e-07,
+      "loss": 0.3621,
+      "step": 110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.870967741935484e-07,
+      "loss": 0.4241,
+      "step": 120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.1935483870967743e-07,
+      "loss": 0.4212,
+      "step": 130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.516129032258065e-07,
+      "loss": 0.4508,
+      "step": 140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.838709677419355e-07,
+      "loss": 0.4068,
+      "step": 150
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9598516821861267,
+      "eval_runtime": 0.8497,
+      "eval_samples_per_second": 236.552,
+      "eval_steps_per_second": 8.238,
+      "step": 155
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 5.161290322580646e-07,
+      "loss": 0.3503,
+      "step": 160
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 5.483870967741935e-07,
+      "loss": 0.3874,
+      "step": 170
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 5.806451612903227e-07,
+      "loss": 0.4261,
+      "step": 180
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.129032258064516e-07,
+      "loss": 0.3863,
+      "step": 190
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 6.451612903225807e-07,
+      "loss": 0.3655,
+      "step": 200
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.774193548387097e-07,
+      "loss": 0.4159,
+      "step": 210
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.096774193548388e-07,
+      "loss": 0.4044,
+      "step": 220
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.419354838709678e-07,
+      "loss": 0.3875,
+      "step": 230
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 7.741935483870968e-07,
+      "loss": 0.3972,
+      "step": 240
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.064516129032258e-07,
+      "loss": 0.393,
+      "step": 250
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.387096774193549e-07,
+      "loss": 0.3628,
+      "step": 260
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.709677419354838e-07,
+      "loss": 0.3807,
+      "step": 270
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.03225806451613e-07,
+      "loss": 0.3776,
+      "step": 280
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 9.35483870967742e-07,
+      "loss": 0.3714,
+      "step": 290
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.67741935483871e-07,
+      "loss": 0.4236,
+      "step": 300
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.406,
+      "step": 310
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9707355499267578,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 233.113,
+      "eval_steps_per_second": 8.118,
+      "step": 310
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.0322580645161291e-06,
+      "loss": 0.4828,
+      "step": 320
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.064516129032258e-06,
+      "loss": 0.3783,
+      "step": 330
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.096774193548387e-06,
+      "loss": 0.3926,
+      "step": 340
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.1290322580645162e-06,
+      "loss": 0.4248,
+      "step": 350
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.1612903225806454e-06,
+      "loss": 0.3764,
+      "step": 360
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.1935483870967743e-06,
+      "loss": 0.3403,
+      "step": 370
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.2258064516129033e-06,
+      "loss": 0.3929,
+      "step": 380
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2580645161290322e-06,
+      "loss": 0.3855,
+      "step": 390
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.2903225806451614e-06,
+      "loss": 0.4345,
+      "step": 400
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.3225806451612904e-06,
+      "loss": 0.4089,
+      "step": 410
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.3548387096774193e-06,
+      "loss": 0.4147,
+      "step": 420
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.3870967741935485e-06,
+      "loss": 0.3959,
+      "step": 430
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4193548387096776e-06,
+      "loss": 0.4203,
+      "step": 440
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.4516129032258066e-06,
+      "loss": 0.3475,
+      "step": 450
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.4838709677419356e-06,
+      "loss": 0.4188,
+      "step": 460
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9717937111854553,
+      "eval_runtime": 0.8498,
+      "eval_samples_per_second": 236.514,
+      "eval_steps_per_second": 8.237,
+      "step": 465
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 1.5161290322580645e-06,
+      "loss": 0.3883,
+      "step": 470
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 1.5483870967741937e-06,
+      "loss": 0.3109,
+      "step": 480
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.5806451612903229e-06,
+      "loss": 0.3849,
+      "step": 490
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.6129032258064516e-06,
+      "loss": 0.3703,
+      "step": 500
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.6451612903225808e-06,
+      "loss": 0.3736,
+      "step": 510
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.6774193548387097e-06,
+      "loss": 0.3827,
+      "step": 520
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.7096774193548389e-06,
+      "loss": 0.3978,
+      "step": 530
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.7419354838709676e-06,
+      "loss": 0.3363,
+      "step": 540
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.7741935483870968e-06,
+      "loss": 0.3788,
+      "step": 550
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.806451612903226e-06,
+      "loss": 0.3705,
+      "step": 560
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.838709677419355e-06,
+      "loss": 0.4358,
+      "step": 570
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.870967741935484e-06,
+      "loss": 0.345,
+      "step": 580
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.9032258064516128e-06,
+      "loss": 0.3864,
+      "step": 590
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.935483870967742e-06,
+      "loss": 0.3881,
+      "step": 600
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.967741935483871e-06,
+      "loss": 0.3775,
+      "step": 610
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.3231,
+      "step": 620
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9448372721672058,
+      "eval_runtime": 0.8656,
+      "eval_samples_per_second": 232.221,
+      "eval_steps_per_second": 8.087,
+      "step": 620
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.032258064516129e-06,
+      "loss": 0.3734,
+      "step": 630
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.0645161290322582e-06,
+      "loss": 0.39,
+      "step": 640
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.096774193548387e-06,
+      "loss": 0.3951,
+      "step": 650
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.129032258064516e-06,
+      "loss": 0.4064,
+      "step": 660
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.1612903225806455e-06,
+      "loss": 0.4418,
+      "step": 670
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.193548387096774e-06,
+      "loss": 0.4253,
+      "step": 680
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.2258064516129034e-06,
+      "loss": 0.3744,
+      "step": 690
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.2580645161290324e-06,
+      "loss": 0.4139,
+      "step": 700
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.2903225806451614e-06,
+      "loss": 0.4136,
+      "step": 710
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.3225806451612907e-06,
+      "loss": 0.332,
+      "step": 720
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.3548387096774193e-06,
+      "loss": 0.3626,
+      "step": 730
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.3870967741935486e-06,
+      "loss": 0.364,
+      "step": 740
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.4193548387096776e-06,
+      "loss": 0.3864,
+      "step": 750
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.4516129032258066e-06,
+      "loss": 0.39,
+      "step": 760
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.4838709677419355e-06,
+      "loss": 0.3606,
+      "step": 770
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.965462327003479,
+      "eval_runtime": 0.8298,
+      "eval_samples_per_second": 242.233,
+      "eval_steps_per_second": 8.436,
+      "step": 775
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.5161290322580645e-06,
+      "loss": 0.3716,
+      "step": 780
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.5483870967741934e-06,
+      "loss": 0.412,
+      "step": 790
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.580645161290323e-06,
+      "loss": 0.3573,
+      "step": 800
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.6129032258064518e-06,
+      "loss": 0.4159,
+      "step": 810
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.6451612903225807e-06,
+      "loss": 0.4179,
+      "step": 820
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.67741935483871e-06,
+      "loss": 0.3611,
+      "step": 830
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.7096774193548386e-06,
+      "loss": 0.4093,
+      "step": 840
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.741935483870968e-06,
+      "loss": 0.3899,
+      "step": 850
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.774193548387097e-06,
+      "loss": 0.3193,
+      "step": 860
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.806451612903226e-06,
+      "loss": 0.4153,
+      "step": 870
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.8387096774193553e-06,
+      "loss": 0.3429,
+      "step": 880
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.870967741935484e-06,
+      "loss": 0.3805,
+      "step": 890
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.903225806451613e-06,
+      "loss": 0.3357,
+      "step": 900
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.935483870967742e-06,
+      "loss": 0.4365,
+      "step": 910
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.967741935483871e-06,
+      "loss": 0.375,
+      "step": 920
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3e-06,
+      "loss": 0.4406,
+      "step": 930
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9668756723403931,
+      "eval_runtime": 0.9024,
+      "eval_samples_per_second": 222.729,
+      "eval_steps_per_second": 7.757,
+      "step": 930
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.032258064516129e-06,
+      "loss": 0.4183,
+      "step": 940
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.064516129032258e-06,
+      "loss": 0.3177,
+      "step": 950
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.0967741935483874e-06,
+      "loss": 0.3799,
+      "step": 960
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.129032258064516e-06,
+      "loss": 0.3762,
+      "step": 970
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.1612903225806457e-06,
+      "loss": 0.3292,
+      "step": 980
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.1935483870967747e-06,
+      "loss": 0.3632,
+      "step": 990
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.225806451612903e-06,
+      "loss": 0.3387,
+      "step": 1000
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.258064516129032e-06,
+      "loss": 0.333,
+      "step": 1010
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.2903225806451615e-06,
+      "loss": 0.3422,
+      "step": 1020
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3225806451612905e-06,
+      "loss": 0.3666,
+      "step": 1030
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3548387096774194e-06,
+      "loss": 0.3363,
+      "step": 1040
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.387096774193549e-06,
+      "loss": 0.3963,
+      "step": 1050
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.4193548387096778e-06,
+      "loss": 0.4009,
+      "step": 1060
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.4516129032258067e-06,
+      "loss": 0.3739,
+      "step": 1070
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.4838709677419353e-06,
+      "loss": 0.4137,
+      "step": 1080
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9522982835769653,
+      "eval_runtime": 0.8607,
+      "eval_samples_per_second": 233.524,
+      "eval_steps_per_second": 8.133,
+      "step": 1085
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.516129032258065e-06,
+      "loss": 0.4227,
+      "step": 1090
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.5483870967741936e-06,
+      "loss": 0.3894,
+      "step": 1100
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.5806451612903225e-06,
+      "loss": 0.3273,
+      "step": 1110
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.612903225806452e-06,
+      "loss": 0.386,
+      "step": 1120
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.645161290322581e-06,
+      "loss": 0.3879,
+      "step": 1130
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.67741935483871e-06,
+      "loss": 0.3911,
+      "step": 1140
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.709677419354839e-06,
+      "loss": 0.4133,
+      "step": 1150
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.741935483870968e-06,
+      "loss": 0.3704,
+      "step": 1160
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.774193548387097e-06,
+      "loss": 0.3424,
+      "step": 1170
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.8064516129032257e-06,
+      "loss": 0.3193,
+      "step": 1180
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.838709677419355e-06,
+      "loss": 0.3229,
+      "step": 1190
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.870967741935484e-06,
+      "loss": 0.3421,
+      "step": 1200
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.903225806451613e-06,
+      "loss": 0.3697,
+      "step": 1210
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.935483870967742e-06,
+      "loss": 0.4259,
+      "step": 1220
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.967741935483871e-06,
+      "loss": 0.3597,
+      "step": 1230
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.3525,
+      "step": 1240
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9394405484199524,
+      "eval_runtime": 0.8363,
+      "eval_samples_per_second": 240.333,
+      "eval_steps_per_second": 8.37,
+      "step": 1240
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.032258064516129e-06,
+      "loss": 0.3051,
+      "step": 1250
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.064516129032258e-06,
+      "loss": 0.4013,
+      "step": 1260
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 4.096774193548387e-06,
+      "loss": 0.3597,
+      "step": 1270
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.1290322580645165e-06,
+      "loss": 0.3322,
+      "step": 1280
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 4.161290322580645e-06,
+      "loss": 0.3643,
+      "step": 1290
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.193548387096774e-06,
+      "loss": 0.332,
+      "step": 1300
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.225806451612904e-06,
+      "loss": 0.3851,
+      "step": 1310
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.258064516129032e-06,
+      "loss": 0.3631,
+      "step": 1320
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.290322580645161e-06,
+      "loss": 0.3544,
+      "step": 1330
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.322580645161291e-06,
+      "loss": 0.4164,
+      "step": 1340
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 4.35483870967742e-06,
+      "loss": 0.3993,
+      "step": 1350
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 4.387096774193548e-06,
+      "loss": 0.361,
+      "step": 1360
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 4.419354838709678e-06,
+      "loss": 0.3252,
+      "step": 1370
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 4.451612903225807e-06,
+      "loss": 0.3392,
+      "step": 1380
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 4.4838709677419354e-06,
+      "loss": 0.4063,
+      "step": 1390
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9382967948913574,
+      "eval_runtime": 0.8611,
+      "eval_samples_per_second": 233.417,
+      "eval_steps_per_second": 8.129,
+      "step": 1395
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 4.516129032258065e-06,
+      "loss": 0.3193,
+      "step": 1400
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 4.548387096774194e-06,
+      "loss": 0.3485,
+      "step": 1410
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 4.580645161290323e-06,
+      "loss": 0.3669,
+      "step": 1420
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 4.612903225806452e-06,
+      "loss": 0.3371,
+      "step": 1430
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 4.6451612903225815e-06,
+      "loss": 0.3468,
+      "step": 1440
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 4.67741935483871e-06,
+      "loss": 0.345,
+      "step": 1450
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 4.7096774193548385e-06,
+      "loss": 0.3142,
+      "step": 1460
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 4.741935483870968e-06,
+      "loss": 0.319,
+      "step": 1470
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.774193548387097e-06,
+      "loss": 0.3528,
+      "step": 1480
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 4.806451612903226e-06,
+      "loss": 0.3384,
+      "step": 1490
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 4.838709677419355e-06,
+      "loss": 0.3653,
+      "step": 1500
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 4.870967741935484e-06,
+      "loss": 0.3679,
+      "step": 1510
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 4.903225806451613e-06,
+      "loss": 0.3588,
+      "step": 1520
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.9354838709677425e-06,
+      "loss": 0.362,
+      "step": 1530
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 4.967741935483871e-06,
+      "loss": 0.4236,
+      "step": 1540
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 5e-06,
+      "loss": 0.3399,
+      "step": 1550
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9325306415557861,
+      "eval_runtime": 0.8386,
+      "eval_samples_per_second": 239.688,
+      "eval_steps_per_second": 8.347,
+      "step": 1550
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 5.032258064516129e-06,
+      "loss": 0.4063,
+      "step": 1560
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 5.064516129032258e-06,
+      "loss": 0.3587,
+      "step": 1570
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 5.096774193548387e-06,
+      "loss": 0.368,
+      "step": 1580
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 5.129032258064516e-06,
+      "loss": 0.3673,
+      "step": 1590
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 5.161290322580646e-06,
+      "loss": 0.3324,
+      "step": 1600
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 5.193548387096774e-06,
+      "loss": 0.3494,
+      "step": 1610
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 5.2258064516129035e-06,
+      "loss": 0.3456,
+      "step": 1620
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 5.258064516129033e-06,
+      "loss": 0.381,
+      "step": 1630
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 5.2903225806451614e-06,
+      "loss": 0.3479,
+      "step": 1640
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 5.32258064516129e-06,
+      "loss": 0.3271,
+      "step": 1650
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 5.35483870967742e-06,
+      "loss": 0.3386,
+      "step": 1660
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 5.387096774193549e-06,
+      "loss": 0.3766,
+      "step": 1670
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 5.419354838709677e-06,
+      "loss": 0.3107,
+      "step": 1680
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 5.451612903225807e-06,
+      "loss": 0.3228,
+      "step": 1690
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 5.483870967741936e-06,
+      "loss": 0.3811,
+      "step": 1700
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9428999423980713,
+      "eval_runtime": 0.829,
+      "eval_samples_per_second": 242.469,
+      "eval_steps_per_second": 8.444,
+      "step": 1705
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 5.5161290322580645e-06,
+      "loss": 0.4347,
+      "step": 1710
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 5.548387096774194e-06,
+      "loss": 0.331,
+      "step": 1720
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 5.580645161290323e-06,
+      "loss": 0.3919,
+      "step": 1730
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 5.612903225806452e-06,
+      "loss": 0.3833,
+      "step": 1740
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 5.64516129032258e-06,
+      "loss": 0.3585,
+      "step": 1750
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 5.677419354838711e-06,
+      "loss": 0.3144,
+      "step": 1760
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 5.709677419354839e-06,
+      "loss": 0.3368,
+      "step": 1770
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 5.741935483870968e-06,
+      "loss": 0.3844,
+      "step": 1780
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 5.774193548387097e-06,
+      "loss": 0.3611,
+      "step": 1790
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 5.806451612903226e-06,
+      "loss": 0.3694,
+      "step": 1800
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 5.838709677419355e-06,
+      "loss": 0.3497,
+      "step": 1810
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 5.870967741935484e-06,
+      "loss": 0.4001,
+      "step": 1820
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 5.903225806451613e-06,
+      "loss": 0.3539,
+      "step": 1830
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 5.935483870967742e-06,
+      "loss": 0.3699,
+      "step": 1840
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 5.967741935483872e-06,
+      "loss": 0.3747,
+      "step": 1850
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 6e-06,
+      "loss": 0.2965,
+      "step": 1860
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9209400415420532,
+      "eval_runtime": 0.863,
+      "eval_samples_per_second": 232.911,
+      "eval_steps_per_second": 8.111,
+      "step": 1860
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 6.0322580645161295e-06,
+      "loss": 0.3912,
+      "step": 1870
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 6.064516129032258e-06,
+      "loss": 0.406,
+      "step": 1880
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 6.0967741935483874e-06,
+      "loss": 0.3304,
+      "step": 1890
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 6.129032258064516e-06,
+      "loss": 0.3943,
+      "step": 1900
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 6.161290322580645e-06,
+      "loss": 0.3152,
+      "step": 1910
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 6.193548387096775e-06,
+      "loss": 0.3745,
+      "step": 1920
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 6.225806451612903e-06,
+      "loss": 0.3761,
+      "step": 1930
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 6.258064516129032e-06,
+      "loss": 0.3933,
+      "step": 1940
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 6.290322580645161e-06,
+      "loss": 0.3376,
+      "step": 1950
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 6.322580645161291e-06,
+      "loss": 0.3558,
+      "step": 1960
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 6.35483870967742e-06,
+      "loss": 0.3388,
+      "step": 1970
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 6.387096774193549e-06,
+      "loss": 0.3425,
+      "step": 1980
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 6.419354838709678e-06,
+      "loss": 0.3189,
+      "step": 1990
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 6.451612903225806e-06,
+      "loss": 0.3294,
+      "step": 2000
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 6.483870967741936e-06,
+      "loss": 0.3418,
+      "step": 2010
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9341461062431335,
+      "eval_runtime": 0.846,
+      "eval_samples_per_second": 237.583,
+      "eval_steps_per_second": 8.274,
+      "step": 2015
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 6.516129032258064e-06,
+      "loss": 0.3615,
+      "step": 2020
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 6.5483870967741945e-06,
+      "loss": 0.386,
+      "step": 2030
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 6.580645161290323e-06,
+      "loss": 0.3471,
+      "step": 2040
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 6.6129032258064524e-06,
+      "loss": 0.3692,
+      "step": 2050
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 6.645161290322581e-06,
+      "loss": 0.4098,
+      "step": 2060
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 6.6774193548387095e-06,
+      "loss": 0.3171,
+      "step": 2070
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 6.709677419354839e-06,
+      "loss": 0.314,
+      "step": 2080
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 6.741935483870967e-06,
+      "loss": 0.3661,
+      "step": 2090
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 6.774193548387098e-06,
+      "loss": 0.3234,
+      "step": 2100
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 6.806451612903226e-06,
+      "loss": 0.3365,
+      "step": 2110
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 6.8387096774193555e-06,
+      "loss": 0.3531,
+      "step": 2120
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 6.870967741935484e-06,
+      "loss": 0.3714,
+      "step": 2130
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 6.9032258064516135e-06,
+      "loss": 0.3198,
+      "step": 2140
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 6.935483870967742e-06,
+      "loss": 0.316,
+      "step": 2150
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 6.9677419354838705e-06,
+      "loss": 0.3889,
+      "step": 2160
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 0.3375,
+      "step": 2170
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9031834602355957,
+      "eval_runtime": 0.8418,
+      "eval_samples_per_second": 238.782,
+      "eval_steps_per_second": 8.316,
+      "step": 2170
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 7.03225806451613e-06,
+      "loss": 0.3101,
+      "step": 2180
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 7.064516129032259e-06,
+      "loss": 0.3371,
+      "step": 2190
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 7.096774193548387e-06,
+      "loss": 0.3622,
+      "step": 2200
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 7.1290322580645166e-06,
+      "loss": 0.3822,
+      "step": 2210
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 7.161290322580645e-06,
+      "loss": 0.414,
+      "step": 2220
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 7.193548387096774e-06,
+      "loss": 0.3655,
+      "step": 2230
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 7.225806451612904e-06,
+      "loss": 0.3164,
+      "step": 2240
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 7.258064516129033e-06,
+      "loss": 0.3917,
+      "step": 2250
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 7.290322580645162e-06,
+      "loss": 0.371,
+      "step": 2260
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 7.322580645161291e-06,
+      "loss": 0.3977,
+      "step": 2270
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 7.35483870967742e-06,
+      "loss": 0.3368,
+      "step": 2280
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 7.387096774193548e-06,
+      "loss": 0.3858,
+      "step": 2290
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 7.419354838709678e-06,
+      "loss": 0.3688,
+      "step": 2300
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 7.451612903225806e-06,
+      "loss": 0.3465,
+      "step": 2310
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.483870967741936e-06,
+      "loss": 0.3322,
+      "step": 2320
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.919029176235199,
+      "eval_runtime": 0.8414,
+      "eval_samples_per_second": 238.881,
+      "eval_steps_per_second": 8.319,
+      "step": 2325
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.516129032258065e-06,
+      "loss": 0.3814,
+      "step": 2330
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.548387096774194e-06,
+      "loss": 0.3487,
+      "step": 2340
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 7.580645161290323e-06,
+      "loss": 0.3469,
+      "step": 2350
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.612903225806451e-06,
+      "loss": 0.3716,
+      "step": 2360
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.64516129032258e-06,
+      "loss": 0.3403,
+      "step": 2370
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 7.67741935483871e-06,
+      "loss": 0.3153,
+      "step": 2380
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.70967741935484e-06,
+      "loss": 0.3215,
+      "step": 2390
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 7.741935483870968e-06,
+      "loss": 0.3581,
+      "step": 2400
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 7.774193548387097e-06,
+      "loss": 0.3477,
+      "step": 2410
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 7.806451612903227e-06,
+      "loss": 0.3437,
+      "step": 2420
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 7.838709677419354e-06,
+      "loss": 0.3837,
+      "step": 2430
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 7.870967741935484e-06,
+      "loss": 0.3409,
+      "step": 2440
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 7.903225806451613e-06,
+      "loss": 0.3531,
+      "step": 2450
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 7.935483870967743e-06,
+      "loss": 0.3724,
+      "step": 2460
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 7.967741935483872e-06,
+      "loss": 0.3225,
+      "step": 2470
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.3482,
+      "step": 2480
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8908675312995911,
+      "eval_runtime": 0.8457,
+      "eval_samples_per_second": 237.675,
+      "eval_steps_per_second": 8.277,
+      "step": 2480
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 8.032258064516129e-06,
+      "loss": 0.4031,
+      "step": 2490
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 8.064516129032258e-06,
+      "loss": 0.3201,
+      "step": 2500
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 8.096774193548388e-06,
+      "loss": 0.3202,
+      "step": 2510
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 8.129032258064515e-06,
+      "loss": 0.3638,
+      "step": 2520
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 8.161290322580647e-06,
+      "loss": 0.3485,
+      "step": 2530
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 8.193548387096774e-06,
+      "loss": 0.3455,
+      "step": 2540
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.225806451612904e-06,
+      "loss": 0.36,
+      "step": 2550
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.258064516129033e-06,
+      "loss": 0.3969,
+      "step": 2560
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.29032258064516e-06,
+      "loss": 0.3288,
+      "step": 2570
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.32258064516129e-06,
+      "loss": 0.3214,
+      "step": 2580
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.35483870967742e-06,
+      "loss": 0.3148,
+      "step": 2590
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.387096774193549e-06,
+      "loss": 0.404,
+      "step": 2600
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 8.419354838709678e-06,
+      "loss": 0.3607,
+      "step": 2610
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 8.451612903225808e-06,
+      "loss": 0.3618,
+      "step": 2620
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 8.483870967741935e-06,
+      "loss": 0.3685,
+      "step": 2630
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9168297052383423,
+      "eval_runtime": 0.8425,
+      "eval_samples_per_second": 238.579,
+      "eval_steps_per_second": 8.309,
+      "step": 2635
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 8.516129032258065e-06,
+      "loss": 0.3095,
+      "step": 2640
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 8.548387096774194e-06,
+      "loss": 0.2926,
+      "step": 2650
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 8.580645161290322e-06,
+      "loss": 0.3219,
+      "step": 2660
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 8.612903225806453e-06,
+      "loss": 0.3686,
+      "step": 2670
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 8.645161290322582e-06,
+      "loss": 0.3548,
+      "step": 2680
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 8.67741935483871e-06,
+      "loss": 0.3454,
+      "step": 2690
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 8.70967741935484e-06,
+      "loss": 0.3511,
+      "step": 2700
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 8.741935483870969e-06,
+      "loss": 0.2993,
+      "step": 2710
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 8.774193548387096e-06,
+      "loss": 0.3482,
+      "step": 2720
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 8.806451612903226e-06,
+      "loss": 0.3623,
+      "step": 2730
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 8.838709677419357e-06,
+      "loss": 0.3717,
+      "step": 2740
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 8.870967741935484e-06,
+      "loss": 0.3292,
+      "step": 2750
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 8.903225806451614e-06,
+      "loss": 0.3425,
+      "step": 2760
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 8.935483870967743e-06,
+      "loss": 0.3703,
+      "step": 2770
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 8.967741935483871e-06,
+      "loss": 0.3081,
+      "step": 2780
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 9e-06,
+      "loss": 0.3687,
+      "step": 2790
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.938947856426239,
+      "eval_runtime": 0.8438,
+      "eval_samples_per_second": 238.203,
+      "eval_steps_per_second": 8.296,
+      "step": 2790
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 9.03225806451613e-06,
+      "loss": 0.3519,
+      "step": 2800
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 9.064516129032259e-06,
+      "loss": 0.3309,
+      "step": 2810
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 9.096774193548388e-06,
+      "loss": 0.4186,
+      "step": 2820
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 9.129032258064516e-06,
+      "loss": 0.3829,
+      "step": 2830
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 9.161290322580645e-06,
+      "loss": 0.458,
+      "step": 2840
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 9.193548387096775e-06,
+      "loss": 0.3946,
+      "step": 2850
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 9.225806451612904e-06,
+      "loss": 0.3001,
+      "step": 2860
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 9.258064516129032e-06,
+      "loss": 0.3687,
+      "step": 2870
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.290322580645163e-06,
+      "loss": 0.3536,
+      "step": 2880
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 9.32258064516129e-06,
+      "loss": 0.3073,
+      "step": 2890
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 9.35483870967742e-06,
+      "loss": 0.3706,
+      "step": 2900
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 9.38709677419355e-06,
+      "loss": 0.3277,
+      "step": 2910
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 9.419354838709677e-06,
+      "loss": 0.3384,
+      "step": 2920
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 9.451612903225806e-06,
+      "loss": 0.363,
+      "step": 2930
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 9.483870967741936e-06,
+      "loss": 0.3376,
+      "step": 2940
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9028215408325195,
+      "eval_runtime": 0.8455,
+      "eval_samples_per_second": 237.731,
+      "eval_steps_per_second": 8.279,
+      "step": 2945
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 9.516129032258064e-06,
+      "loss": 0.3655,
+      "step": 2950
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 9.548387096774195e-06,
+      "loss": 0.3243,
+      "step": 2960
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 9.580645161290324e-06,
+      "loss": 0.3268,
+      "step": 2970
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 9.612903225806452e-06,
+      "loss": 0.3482,
+      "step": 2980
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 9.645161290322581e-06,
+      "loss": 0.37,
+      "step": 2990
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 9.67741935483871e-06,
+      "loss": 0.342,
+      "step": 3000
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 9.709677419354838e-06,
+      "loss": 0.3389,
+      "step": 3010
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 9.741935483870967e-06,
+      "loss": 0.3894,
+      "step": 3020
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 9.774193548387099e-06,
+      "loss": 0.3554,
+      "step": 3030
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.806451612903226e-06,
+      "loss": 0.334,
+      "step": 3040
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 9.838709677419356e-06,
+      "loss": 0.3427,
+      "step": 3050
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 9.870967741935485e-06,
+      "loss": 0.32,
+      "step": 3060
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 9.903225806451613e-06,
+      "loss": 0.3653,
+      "step": 3070
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 9.935483870967742e-06,
+      "loss": 0.3118,
+      "step": 3080
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 9.967741935483871e-06,
+      "loss": 0.3013,
+      "step": 3090
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1e-05,
+      "loss": 0.3604,
+      "step": 3100
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9377031326293945,
+      "eval_runtime": 0.8392,
+      "eval_samples_per_second": 239.51,
+      "eval_steps_per_second": 8.341,
+      "step": 3100
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.003225806451613e-05,
+      "loss": 0.3295,
+      "step": 3110
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.0064516129032258e-05,
+      "loss": 0.351,
+      "step": 3120
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.0096774193548387e-05,
+      "loss": 0.3381,
+      "step": 3130
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.0129032258064517e-05,
+      "loss": 0.3238,
+      "step": 3140
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.0161290322580646e-05,
+      "loss": 0.3551,
+      "step": 3150
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.0193548387096774e-05,
+      "loss": 0.2933,
+      "step": 3160
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.0225806451612905e-05,
+      "loss": 0.3754,
+      "step": 3170
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.0258064516129032e-05,
+      "loss": 0.3297,
+      "step": 3180
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.0290322580645162e-05,
+      "loss": 0.2778,
+      "step": 3190
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.0322580645161291e-05,
+      "loss": 0.3395,
+      "step": 3200
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.0354838709677419e-05,
+      "loss": 0.3328,
+      "step": 3210
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.0387096774193548e-05,
+      "loss": 0.3573,
+      "step": 3220
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.0419354838709678e-05,
+      "loss": 0.3562,
+      "step": 3230
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.0451612903225807e-05,
+      "loss": 0.3404,
+      "step": 3240
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.0483870967741936e-05,
+      "loss": 0.3733,
+      "step": 3250
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8831413388252258,
+      "eval_runtime": 0.8806,
+      "eval_samples_per_second": 228.241,
+      "eval_steps_per_second": 7.949,
+      "step": 3255
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.0516129032258066e-05,
+      "loss": 0.3407,
+      "step": 3260
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.0548387096774193e-05,
+      "loss": 0.3967,
+      "step": 3270
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.0580645161290323e-05,
+      "loss": 0.3449,
+      "step": 3280
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.0612903225806452e-05,
+      "loss": 0.3947,
+      "step": 3290
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.064516129032258e-05,
+      "loss": 0.3706,
+      "step": 3300
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.0677419354838711e-05,
+      "loss": 0.3339,
+      "step": 3310
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.070967741935484e-05,
+      "loss": 0.3485,
+      "step": 3320
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.0741935483870968e-05,
+      "loss": 0.3876,
+      "step": 3330
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.0774193548387097e-05,
+      "loss": 0.3665,
+      "step": 3340
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.0806451612903227e-05,
+      "loss": 0.3759,
+      "step": 3350
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.0838709677419355e-05,
+      "loss": 0.3261,
+      "step": 3360
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.0870967741935484e-05,
+      "loss": 0.2801,
+      "step": 3370
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.0903225806451613e-05,
+      "loss": 0.3503,
+      "step": 3380
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0935483870967743e-05,
+      "loss": 0.3561,
+      "step": 3390
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0967741935483872e-05,
+      "loss": 0.3077,
+      "step": 3400
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.3521,
+      "step": 3410
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9040021896362305,
+      "eval_runtime": 0.8552,
+      "eval_samples_per_second": 235.02,
+      "eval_steps_per_second": 8.185,
+      "step": 3410
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.1032258064516129e-05,
+      "loss": 0.383,
+      "step": 3420
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.1064516129032258e-05,
+      "loss": 0.3693,
+      "step": 3430
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.1096774193548388e-05,
+      "loss": 0.3694,
+      "step": 3440
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.1129032258064517e-05,
+      "loss": 0.351,
+      "step": 3450
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.1161290322580647e-05,
+      "loss": 0.3348,
+      "step": 3460
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.1193548387096774e-05,
+      "loss": 0.3394,
+      "step": 3470
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.1225806451612904e-05,
+      "loss": 0.2785,
+      "step": 3480
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.1258064516129033e-05,
+      "loss": 0.3418,
+      "step": 3490
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.129032258064516e-05,
+      "loss": 0.3314,
+      "step": 3500
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.132258064516129e-05,
+      "loss": 0.3426,
+      "step": 3510
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.1354838709677421e-05,
+      "loss": 0.31,
+      "step": 3520
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.1387096774193549e-05,
+      "loss": 0.3017,
+      "step": 3530
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.1419354838709678e-05,
+      "loss": 0.3703,
+      "step": 3540
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.1451612903225808e-05,
+      "loss": 0.3361,
+      "step": 3550
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.1483870967741935e-05,
+      "loss": 0.3066,
+      "step": 3560
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9158064126968384,
+      "eval_runtime": 0.8647,
+      "eval_samples_per_second": 232.453,
+      "eval_steps_per_second": 8.095,
+      "step": 3565
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.1516129032258065e-05,
+      "loss": 0.3578,
+      "step": 3570
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.1548387096774194e-05,
+      "loss": 0.3618,
+      "step": 3580
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.1580645161290322e-05,
+      "loss": 0.4059,
+      "step": 3590
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.1612903225806453e-05,
+      "loss": 0.3253,
+      "step": 3600
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.1645161290322582e-05,
+      "loss": 0.3596,
+      "step": 3610
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.167741935483871e-05,
+      "loss": 0.3736,
+      "step": 3620
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.170967741935484e-05,
+      "loss": 0.3047,
+      "step": 3630
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.1741935483870969e-05,
+      "loss": 0.3434,
+      "step": 3640
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.1774193548387096e-05,
+      "loss": 0.3427,
+      "step": 3650
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.1806451612903226e-05,
+      "loss": 0.3412,
+      "step": 3660
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.1838709677419355e-05,
+      "loss": 0.3685,
+      "step": 3670
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.1870967741935484e-05,
+      "loss": 0.4098,
+      "step": 3680
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.1903225806451614e-05,
+      "loss": 0.3184,
+      "step": 3690
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.1935483870967743e-05,
+      "loss": 0.3204,
+      "step": 3700
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.1967741935483871e-05,
+      "loss": 0.3192,
+      "step": 3710
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.2e-05,
+      "loss": 0.3253,
+      "step": 3720
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9216181635856628,
+      "eval_runtime": 0.8278,
+      "eval_samples_per_second": 242.802,
+      "eval_steps_per_second": 8.456,
+      "step": 3720
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.203225806451613e-05,
+      "loss": 0.3432,
+      "step": 3730
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.2064516129032259e-05,
+      "loss": 0.3452,
+      "step": 3740
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.2096774193548388e-05,
+      "loss": 0.3825,
+      "step": 3750
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.2129032258064516e-05,
+      "loss": 0.3986,
+      "step": 3760
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.2161290322580646e-05,
+      "loss": 0.3624,
+      "step": 3770
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.2193548387096775e-05,
+      "loss": 0.3439,
+      "step": 3780
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.2225806451612903e-05,
+      "loss": 0.3497,
+      "step": 3790
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.2258064516129032e-05,
+      "loss": 0.2928,
+      "step": 3800
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.2290322580645163e-05,
+      "loss": 0.2865,
+      "step": 3810
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.232258064516129e-05,
+      "loss": 0.3283,
+      "step": 3820
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.235483870967742e-05,
+      "loss": 0.3192,
+      "step": 3830
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.238709677419355e-05,
+      "loss": 0.3804,
+      "step": 3840
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.2419354838709677e-05,
+      "loss": 0.347,
+      "step": 3850
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.2451612903225807e-05,
+      "loss": 0.3735,
+      "step": 3860
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.2483870967741936e-05,
+      "loss": 0.349,
+      "step": 3870
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9289106726646423,
+      "eval_runtime": 0.8399,
+      "eval_samples_per_second": 239.317,
+      "eval_steps_per_second": 8.334,
+      "step": 3875
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.2516129032258064e-05,
+      "loss": 0.4138,
+      "step": 3880
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.2548387096774195e-05,
+      "loss": 0.3182,
+      "step": 3890
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.2580645161290322e-05,
+      "loss": 0.3402,
+      "step": 3900
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.2612903225806452e-05,
+      "loss": 0.3188,
+      "step": 3910
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.2645161290322583e-05,
+      "loss": 0.3525,
+      "step": 3920
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.267741935483871e-05,
+      "loss": 0.3274,
+      "step": 3930
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.270967741935484e-05,
+      "loss": 0.3342,
+      "step": 3940
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.2741935483870968e-05,
+      "loss": 0.3695,
+      "step": 3950
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.2774193548387099e-05,
+      "loss": 0.3347,
+      "step": 3960
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.2806451612903225e-05,
+      "loss": 0.3155,
+      "step": 3970
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.2838709677419356e-05,
+      "loss": 0.3231,
+      "step": 3980
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.2870967741935485e-05,
+      "loss": 0.3341,
+      "step": 3990
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.2903225806451613e-05,
+      "loss": 0.3279,
+      "step": 4000
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.2935483870967744e-05,
+      "loss": 0.3801,
+      "step": 4010
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.2967741935483872e-05,
+      "loss": 0.3493,
+      "step": 4020
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.3308,
+      "step": 4030
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9479464292526245,
+      "eval_runtime": 0.8592,
+      "eval_samples_per_second": 233.934,
+      "eval_steps_per_second": 8.147,
+      "step": 4030
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.3032258064516129e-05,
+      "loss": 0.3416,
+      "step": 4040
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.3064516129032258e-05,
+      "loss": 0.3294,
+      "step": 4050
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.3096774193548389e-05,
+      "loss": 0.3599,
+      "step": 4060
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.3129032258064517e-05,
+      "loss": 0.3548,
+      "step": 4070
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.3161290322580646e-05,
+      "loss": 0.3776,
+      "step": 4080
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.3193548387096774e-05,
+      "loss": 0.3507,
+      "step": 4090
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.3225806451612905e-05,
+      "loss": 0.3652,
+      "step": 4100
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.3258064516129033e-05,
+      "loss": 0.3016,
+      "step": 4110
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.3290322580645162e-05,
+      "loss": 0.3619,
+      "step": 4120
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.3322580645161293e-05,
+      "loss": 0.3571,
+      "step": 4130
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.3354838709677419e-05,
+      "loss": 0.3735,
+      "step": 4140
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.338709677419355e-05,
+      "loss": 0.385,
+      "step": 4150
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.3419354838709678e-05,
+      "loss": 0.3072,
+      "step": 4160
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.3451612903225807e-05,
+      "loss": 0.3397,
+      "step": 4170
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.3483870967741935e-05,
+      "loss": 0.3078,
+      "step": 4180
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8556873202323914,
+      "eval_runtime": 0.8448,
+      "eval_samples_per_second": 237.917,
+      "eval_steps_per_second": 8.286,
+      "step": 4185
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.3516129032258066e-05,
+      "loss": 0.4126,
+      "step": 4190
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.3548387096774195e-05,
+      "loss": 0.3549,
+      "step": 4200
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.3580645161290323e-05,
+      "loss": 0.3399,
+      "step": 4210
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.3612903225806452e-05,
+      "loss": 0.3183,
+      "step": 4220
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.364516129032258e-05,
+      "loss": 0.3201,
+      "step": 4230
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3677419354838711e-05,
+      "loss": 0.379,
+      "step": 4240
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.3709677419354839e-05,
+      "loss": 0.348,
+      "step": 4250
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.3741935483870968e-05,
+      "loss": 0.3143,
+      "step": 4260
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.37741935483871e-05,
+      "loss": 0.2707,
+      "step": 4270
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.3806451612903227e-05,
+      "loss": 0.2997,
+      "step": 4280
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.3838709677419356e-05,
+      "loss": 0.3204,
+      "step": 4290
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.3870967741935484e-05,
+      "loss": 0.2963,
+      "step": 4300
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3903225806451613e-05,
+      "loss": 0.3233,
+      "step": 4310
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.3935483870967741e-05,
+      "loss": 0.3712,
+      "step": 4320
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3967741935483872e-05,
+      "loss": 0.329,
+      "step": 4330
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.305,
+      "step": 4340
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.940169095993042,
+      "eval_runtime": 0.8423,
+      "eval_samples_per_second": 238.641,
+      "eval_steps_per_second": 8.311,
+      "step": 4340
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.403225806451613e-05,
+      "loss": 0.3336,
+      "step": 4350
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.406451612903226e-05,
+      "loss": 0.3042,
+      "step": 4360
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.4096774193548388e-05,
+      "loss": 0.3808,
+      "step": 4370
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.4129032258064517e-05,
+      "loss": 0.3318,
+      "step": 4380
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.4161290322580645e-05,
+      "loss": 0.2884,
+      "step": 4390
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.4193548387096774e-05,
+      "loss": 0.3422,
+      "step": 4400
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.4225806451612905e-05,
+      "loss": 0.3661,
+      "step": 4410
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.4258064516129033e-05,
+      "loss": 0.367,
+      "step": 4420
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.4290322580645163e-05,
+      "loss": 0.3577,
+      "step": 4430
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.432258064516129e-05,
+      "loss": 0.3313,
+      "step": 4440
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.4354838709677421e-05,
+      "loss": 0.3438,
+      "step": 4450
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.4387096774193547e-05,
+      "loss": 0.3667,
+      "step": 4460
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.4419354838709678e-05,
+      "loss": 0.3548,
+      "step": 4470
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.4451612903225808e-05,
+      "loss": 0.3542,
+      "step": 4480
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.4483870967741935e-05,
+      "loss": 0.3044,
+      "step": 4490
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9229084253311157,
+      "eval_runtime": 0.8386,
+      "eval_samples_per_second": 239.674,
+      "eval_steps_per_second": 8.347,
+      "step": 4495
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.4516129032258066e-05,
+      "loss": 0.3743,
+      "step": 4500
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.4548387096774194e-05,
+      "loss": 0.333,
+      "step": 4510
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.4580645161290324e-05,
+      "loss": 0.427,
+      "step": 4520
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.4612903225806451e-05,
+      "loss": 0.3886,
+      "step": 4530
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.4645161290322582e-05,
+      "loss": 0.3089,
+      "step": 4540
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.4677419354838708e-05,
+      "loss": 0.3413,
+      "step": 4550
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.470967741935484e-05,
+      "loss": 0.3575,
+      "step": 4560
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.4741935483870969e-05,
+      "loss": 0.3081,
+      "step": 4570
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.4774193548387096e-05,
+      "loss": 0.3277,
+      "step": 4580
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.4806451612903227e-05,
+      "loss": 0.3379,
+      "step": 4590
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.4838709677419355e-05,
+      "loss": 0.2833,
+      "step": 4600
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.4870967741935485e-05,
+      "loss": 0.4066,
+      "step": 4610
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.4903225806451612e-05,
+      "loss": 0.3161,
+      "step": 4620
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.4935483870967743e-05,
+      "loss": 0.3079,
+      "step": 4630
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.4967741935483873e-05,
+      "loss": 0.3349,
+      "step": 4640
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.5e-05,
+      "loss": 0.3194,
+      "step": 4650
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.945794939994812,
+      "eval_runtime": 0.8526,
+      "eval_samples_per_second": 235.749,
+      "eval_steps_per_second": 8.21,
+      "step": 4650
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.503225806451613e-05,
+      "loss": 0.3634,
+      "step": 4660
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.5064516129032257e-05,
+      "loss": 0.3063,
+      "step": 4670
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.5096774193548389e-05,
+      "loss": 0.3335,
+      "step": 4680
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.5129032258064516e-05,
+      "loss": 0.3171,
+      "step": 4690
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.5161290322580646e-05,
+      "loss": 0.3932,
+      "step": 4700
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.5193548387096777e-05,
+      "loss": 0.2977,
+      "step": 4710
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.5225806451612903e-05,
+      "loss": 0.3729,
+      "step": 4720
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.5258064516129034e-05,
+      "loss": 0.405,
+      "step": 4730
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.529032258064516e-05,
+      "loss": 0.3506,
+      "step": 4740
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.5322580645161292e-05,
+      "loss": 0.3567,
+      "step": 4750
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.535483870967742e-05,
+      "loss": 0.3355,
+      "step": 4760
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.5387096774193548e-05,
+      "loss": 0.3199,
+      "step": 4770
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.541935483870968e-05,
+      "loss": 0.3043,
+      "step": 4780
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.5451612903225807e-05,
+      "loss": 0.3926,
+      "step": 4790
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.5483870967741936e-05,
+      "loss": 0.3436,
+      "step": 4800
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9381190538406372,
+      "eval_runtime": 0.837,
+      "eval_samples_per_second": 240.148,
+      "eval_steps_per_second": 8.363,
+      "step": 4805
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.5516129032258065e-05,
+      "loss": 0.3539,
+      "step": 4810
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.5548387096774195e-05,
+      "loss": 0.378,
+      "step": 4820
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.558064516129032e-05,
+      "loss": 0.36,
+      "step": 4830
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.5612903225806454e-05,
+      "loss": 0.3248,
+      "step": 4840
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.5645161290322583e-05,
+      "loss": 0.3304,
+      "step": 4850
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.567741935483871e-05,
+      "loss": 0.359,
+      "step": 4860
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.570967741935484e-05,
+      "loss": 0.3508,
+      "step": 4870
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.5741935483870968e-05,
+      "loss": 0.3525,
+      "step": 4880
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.5774193548387097e-05,
+      "loss": 0.3774,
+      "step": 4890
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.5806451612903226e-05,
+      "loss": 0.3039,
+      "step": 4900
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.5838709677419356e-05,
+      "loss": 0.4139,
+      "step": 4910
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.5870967741935485e-05,
+      "loss": 0.3134,
+      "step": 4920
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.5903225806451615e-05,
+      "loss": 0.3812,
+      "step": 4930
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.5935483870967744e-05,
+      "loss": 0.3573,
+      "step": 4940
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.596774193548387e-05,
+      "loss": 0.3527,
+      "step": 4950
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.3395,
+      "step": 4960
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9341874122619629,
+      "eval_runtime": 0.8375,
+      "eval_samples_per_second": 240.005,
+      "eval_steps_per_second": 8.358,
+      "step": 4960
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.603225806451613e-05,
+      "loss": 0.3788,
+      "step": 4970
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.6064516129032258e-05,
+      "loss": 0.3311,
+      "step": 4980
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.6096774193548387e-05,
+      "loss": 0.2967,
+      "step": 4990
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 0.3275,
+      "step": 5000
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.6161290322580646e-05,
+      "loss": 0.2934,
+      "step": 5010
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.6193548387096776e-05,
+      "loss": 0.3937,
+      "step": 5020
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.6225806451612905e-05,
+      "loss": 0.3154,
+      "step": 5030
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.625806451612903e-05,
+      "loss": 0.3644,
+      "step": 5040
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.6290322580645164e-05,
+      "loss": 0.3307,
+      "step": 5050
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.6322580645161293e-05,
+      "loss": 0.3854,
+      "step": 5060
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.635483870967742e-05,
+      "loss": 0.3667,
+      "step": 5070
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.638709677419355e-05,
+      "loss": 0.3334,
+      "step": 5080
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.6419354838709678e-05,
+      "loss": 0.3759,
+      "step": 5090
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.6451612903225807e-05,
+      "loss": 0.3439,
+      "step": 5100
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.6483870967741937e-05,
+      "loss": 0.3243,
+      "step": 5110
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9332064390182495,
+      "eval_runtime": 0.8448,
+      "eval_samples_per_second": 237.932,
+      "eval_steps_per_second": 8.286,
+      "step": 5115
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.6516129032258066e-05,
+      "loss": 0.3108,
+      "step": 5120
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.6548387096774195e-05,
+      "loss": 0.3124,
+      "step": 5130
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.658064516129032e-05,
+      "loss": 0.3114,
+      "step": 5140
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.6612903225806454e-05,
+      "loss": 0.3884,
+      "step": 5150
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.664516129032258e-05,
+      "loss": 0.2713,
+      "step": 5160
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.667741935483871e-05,
+      "loss": 0.3003,
+      "step": 5170
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 1.670967741935484e-05,
+      "loss": 0.35,
+      "step": 5180
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 1.6741935483870968e-05,
+      "loss": 0.2975,
+      "step": 5190
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 1.6774193548387098e-05,
+      "loss": 0.3632,
+      "step": 5200
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 1.6806451612903227e-05,
+      "loss": 0.3656,
+      "step": 5210
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 1.6838709677419356e-05,
+      "loss": 0.3301,
+      "step": 5220
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 1.6870967741935482e-05,
+      "loss": 0.3005,
+      "step": 5230
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 1.6903225806451615e-05,
+      "loss": 0.3415,
+      "step": 5240
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 1.693548387096774e-05,
+      "loss": 0.344,
+      "step": 5250
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 1.696774193548387e-05,
+      "loss": 0.3098,
+      "step": 5260
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.3557,
+      "step": 5270
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9473739266395569,
+      "eval_runtime": 0.8283,
+      "eval_samples_per_second": 242.68,
+      "eval_steps_per_second": 8.452,
+      "step": 5270
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 1.703225806451613e-05,
+      "loss": 0.3489,
+      "step": 5280
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 1.706451612903226e-05,
+      "loss": 0.3122,
+      "step": 5290
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 1.7096774193548388e-05,
+      "loss": 0.3439,
+      "step": 5300
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 1.7129032258064517e-05,
+      "loss": 0.3626,
+      "step": 5310
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 1.7161290322580643e-05,
+      "loss": 0.3145,
+      "step": 5320
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 1.7193548387096776e-05,
+      "loss": 0.3112,
+      "step": 5330
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 1.7225806451612906e-05,
+      "loss": 0.3329,
+      "step": 5340
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 1.725806451612903e-05,
+      "loss": 0.3189,
+      "step": 5350
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 1.7290322580645164e-05,
+      "loss": 0.3526,
+      "step": 5360
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 1.732258064516129e-05,
+      "loss": 0.3689,
+      "step": 5370
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 1.735483870967742e-05,
+      "loss": 0.2996,
+      "step": 5380
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 1.738709677419355e-05,
+      "loss": 0.3012,
+      "step": 5390
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 1.741935483870968e-05,
+      "loss": 0.3023,
+      "step": 5400
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 1.7451612903225808e-05,
+      "loss": 0.3037,
+      "step": 5410
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 1.7483870967741937e-05,
+      "loss": 0.3647,
+      "step": 5420
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9527412056922913,
+      "eval_runtime": 0.8523,
+      "eval_samples_per_second": 235.825,
+      "eval_steps_per_second": 8.213,
+      "step": 5425
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 1.7516129032258067e-05,
+      "loss": 0.3842,
+      "step": 5430
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 1.7548387096774193e-05,
+      "loss": 0.4339,
+      "step": 5440
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 1.7580645161290325e-05,
+      "loss": 0.2974,
+      "step": 5450
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 1.761290322580645e-05,
+      "loss": 0.3496,
+      "step": 5460
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 1.764516129032258e-05,
+      "loss": 0.3387,
+      "step": 5470
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 1.7677419354838713e-05,
+      "loss": 0.3835,
+      "step": 5480
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 1.770967741935484e-05,
+      "loss": 0.3295,
+      "step": 5490
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 1.774193548387097e-05,
+      "loss": 0.3258,
+      "step": 5500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 1.7774193548387098e-05,
+      "loss": 0.2835,
+      "step": 5510
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 1.7806451612903228e-05,
+      "loss": 0.3664,
+      "step": 5520
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 1.7838709677419354e-05,
+      "loss": 0.3784,
+      "step": 5530
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 1.7870967741935486e-05,
+      "loss": 0.301,
+      "step": 5540
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 1.7903225806451616e-05,
+      "loss": 0.3592,
+      "step": 5550
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 1.7935483870967742e-05,
+      "loss": 0.3389,
+      "step": 5560
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 1.796774193548387e-05,
+      "loss": 0.3387,
+      "step": 5570
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 1.8e-05,
+      "loss": 0.3239,
+      "step": 5580
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9291701316833496,
+      "eval_runtime": 0.8497,
+      "eval_samples_per_second": 236.564,
+      "eval_steps_per_second": 8.239,
+      "step": 5580
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 1.803225806451613e-05,
+      "loss": 0.3755,
+      "step": 5590
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 1.806451612903226e-05,
+      "loss": 0.3568,
+      "step": 5600
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 1.809677419354839e-05,
+      "loss": 0.374,
+      "step": 5610
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 1.8129032258064518e-05,
+      "loss": 0.3016,
+      "step": 5620
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 1.8161290322580647e-05,
+      "loss": 0.3601,
+      "step": 5630
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 1.8193548387096777e-05,
+      "loss": 0.3318,
+      "step": 5640
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 1.8225806451612903e-05,
+      "loss": 0.3451,
+      "step": 5650
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 1.8258064516129032e-05,
+      "loss": 0.3773,
+      "step": 5660
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 1.829032258064516e-05,
+      "loss": 0.3063,
+      "step": 5670
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 1.832258064516129e-05,
+      "loss": 0.3328,
+      "step": 5680
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 1.835483870967742e-05,
+      "loss": 0.3383,
+      "step": 5690
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 1.838709677419355e-05,
+      "loss": 0.3273,
+      "step": 5700
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 1.841935483870968e-05,
+      "loss": 0.4109,
+      "step": 5710
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 1.845161290322581e-05,
+      "loss": 0.3311,
+      "step": 5720
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 1.8483870967741938e-05,
+      "loss": 0.3631,
+      "step": 5730
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9420697689056396,
+      "eval_runtime": 0.8727,
+      "eval_samples_per_second": 230.323,
+      "eval_steps_per_second": 8.021,
+      "step": 5735
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 1.8516129032258064e-05,
+      "loss": 0.3645,
+      "step": 5740
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 1.8548387096774193e-05,
+      "loss": 0.3608,
+      "step": 5750
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 1.8580645161290326e-05,
+      "loss": 0.3687,
+      "step": 5760
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 1.8612903225806452e-05,
+      "loss": 0.2953,
+      "step": 5770
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 1.864516129032258e-05,
+      "loss": 0.3696,
+      "step": 5780
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 1.867741935483871e-05,
+      "loss": 0.3185,
+      "step": 5790
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 1.870967741935484e-05,
+      "loss": 0.334,
+      "step": 5800
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 1.8741935483870966e-05,
+      "loss": 0.3494,
+      "step": 5810
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 1.87741935483871e-05,
+      "loss": 0.3264,
+      "step": 5820
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 1.8806451612903225e-05,
+      "loss": 0.3374,
+      "step": 5830
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 1.8838709677419354e-05,
+      "loss": 0.3193,
+      "step": 5840
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 1.8870967741935487e-05,
+      "loss": 0.2975,
+      "step": 5850
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 1.8903225806451613e-05,
+      "loss": 0.3771,
+      "step": 5860
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 1.8935483870967742e-05,
+      "loss": 0.348,
+      "step": 5870
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 1.896774193548387e-05,
+      "loss": 0.3391,
+      "step": 5880
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 1.9e-05,
+      "loss": 0.3451,
+      "step": 5890
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9444811344146729,
+      "eval_runtime": 0.8534,
+      "eval_samples_per_second": 235.53,
+      "eval_steps_per_second": 8.203,
+      "step": 5890
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 1.9032258064516127e-05,
+      "loss": 0.335,
+      "step": 5900
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 1.906451612903226e-05,
+      "loss": 0.3026,
+      "step": 5910
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 1.909677419354839e-05,
+      "loss": 0.3932,
+      "step": 5920
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 1.9129032258064515e-05,
+      "loss": 0.3117,
+      "step": 5930
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 1.9161290322580648e-05,
+      "loss": 0.3354,
+      "step": 5940
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 1.9193548387096774e-05,
+      "loss": 0.3759,
+      "step": 5950
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 1.9225806451612903e-05,
+      "loss": 0.3809,
+      "step": 5960
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 1.9258064516129033e-05,
+      "loss": 0.3199,
+      "step": 5970
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 1.9290322580645162e-05,
+      "loss": 0.3592,
+      "step": 5980
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 1.932258064516129e-05,
+      "loss": 0.3076,
+      "step": 5990
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 1.935483870967742e-05,
+      "loss": 0.3442,
+      "step": 6000
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 1.938709677419355e-05,
+      "loss": 0.3592,
+      "step": 6010
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 1.9419354838709676e-05,
+      "loss": 0.2611,
+      "step": 6020
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 1.945161290322581e-05,
+      "loss": 0.3261,
+      "step": 6030
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 1.9483870967741935e-05,
+      "loss": 0.3127,
+      "step": 6040
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9261710047721863,
+      "eval_runtime": 0.8614,
+      "eval_samples_per_second": 233.35,
+      "eval_steps_per_second": 8.127,
+      "step": 6045
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 1.9516129032258064e-05,
+      "loss": 0.3681,
+      "step": 6050
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 1.9548387096774197e-05,
+      "loss": 0.3446,
+      "step": 6060
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 1.9580645161290323e-05,
+      "loss": 0.272,
+      "step": 6070
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 1.9612903225806452e-05,
+      "loss": 0.308,
+      "step": 6080
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 1.9645161290322582e-05,
+      "loss": 0.3195,
+      "step": 6090
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 1.967741935483871e-05,
+      "loss": 0.32,
+      "step": 6100
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 1.9709677419354837e-05,
+      "loss": 0.3261,
+      "step": 6110
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 1.974193548387097e-05,
+      "loss": 0.3753,
+      "step": 6120
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 1.97741935483871e-05,
+      "loss": 0.4015,
+      "step": 6130
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 1.9806451612903225e-05,
+      "loss": 0.3745,
+      "step": 6140
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 1.9838709677419358e-05,
+      "loss": 0.2977,
+      "step": 6150
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 1.9870967741935484e-05,
+      "loss": 0.2991,
+      "step": 6160
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 1.9903225806451613e-05,
+      "loss": 0.3264,
+      "step": 6170
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 1.9935483870967743e-05,
+      "loss": 0.2906,
+      "step": 6180
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 1.9967741935483872e-05,
+      "loss": 0.4166,
+      "step": 6190
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 2e-05,
+      "loss": 0.3382,
+      "step": 6200
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9197480082511902,
+      "eval_runtime": 0.8631,
+      "eval_samples_per_second": 232.875,
+      "eval_steps_per_second": 8.11,
+      "step": 6200
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 2.003225806451613e-05,
+      "loss": 0.3376,
+      "step": 6210
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 2.006451612903226e-05,
+      "loss": 0.3569,
+      "step": 6220
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 2.0096774193548386e-05,
+      "loss": 0.3699,
+      "step": 6230
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 2.0129032258064516e-05,
+      "loss": 0.35,
+      "step": 6240
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 2.0161290322580645e-05,
+      "loss": 0.3295,
+      "step": 6250
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 2.0193548387096775e-05,
+      "loss": 0.3463,
+      "step": 6260
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 2.0225806451612904e-05,
+      "loss": 0.3864,
+      "step": 6270
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 2.0258064516129033e-05,
+      "loss": 0.3672,
+      "step": 6280
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 2.0290322580645163e-05,
+      "loss": 0.3616,
+      "step": 6290
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 2.0322580645161292e-05,
+      "loss": 0.2948,
+      "step": 6300
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 2.035483870967742e-05,
+      "loss": 0.2914,
+      "step": 6310
+    },
+    {
+      "epoch": 40.77,
+      "learning_rate": 2.0387096774193547e-05,
+      "loss": 0.4073,
+      "step": 6320
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 2.0419354838709677e-05,
+      "loss": 0.3391,
+      "step": 6330
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 2.045161290322581e-05,
+      "loss": 0.3793,
+      "step": 6340
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 2.0483870967741936e-05,
+      "loss": 0.3157,
+      "step": 6350
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9437835812568665,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.824,
+      "eval_steps_per_second": 8.073,
+      "step": 6355
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 2.0516129032258065e-05,
+      "loss": 0.3196,
+      "step": 6360
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 2.0548387096774194e-05,
+      "loss": 0.3199,
+      "step": 6370
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 2.0580645161290324e-05,
+      "loss": 0.3663,
+      "step": 6380
+    },
+    {
+      "epoch": 41.23,
+      "learning_rate": 2.0612903225806453e-05,
+      "loss": 0.3623,
+      "step": 6390
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 2.0645161290322582e-05,
+      "loss": 0.3436,
+      "step": 6400
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 2.0677419354838712e-05,
+      "loss": 0.3197,
+      "step": 6410
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 2.0709677419354838e-05,
+      "loss": 0.3123,
+      "step": 6420
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 2.074193548387097e-05,
+      "loss": 0.2964,
+      "step": 6430
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 2.0774193548387097e-05,
+      "loss": 0.3492,
+      "step": 6440
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 2.0806451612903226e-05,
+      "loss": 0.3294,
+      "step": 6450
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 2.0838709677419355e-05,
+      "loss": 0.396,
+      "step": 6460
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 2.0870967741935485e-05,
+      "loss": 0.3479,
+      "step": 6470
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 2.0903225806451614e-05,
+      "loss": 0.3481,
+      "step": 6480
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 2.0935483870967743e-05,
+      "loss": 0.3047,
+      "step": 6490
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 2.0967741935483873e-05,
+      "loss": 0.3525,
+      "step": 6500
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 2.1e-05,
+      "loss": 0.3424,
+      "step": 6510
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.975294291973114,
+      "eval_runtime": 0.847,
+      "eval_samples_per_second": 237.322,
+      "eval_steps_per_second": 8.265,
+      "step": 6510
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 2.103225806451613e-05,
+      "loss": 0.3257,
+      "step": 6520
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 2.1064516129032258e-05,
+      "loss": 0.3587,
+      "step": 6530
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 2.1096774193548387e-05,
+      "loss": 0.3015,
+      "step": 6540
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 2.112903225806452e-05,
+      "loss": 0.3513,
+      "step": 6550
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 2.1161290322580646e-05,
+      "loss": 0.3261,
+      "step": 6560
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 2.1193548387096775e-05,
+      "loss": 0.2889,
+      "step": 6570
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 2.1225806451612904e-05,
+      "loss": 0.3425,
+      "step": 6580
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 2.1258064516129034e-05,
+      "loss": 0.3893,
+      "step": 6590
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 2.129032258064516e-05,
+      "loss": 0.3112,
+      "step": 6600
+    },
+    {
+      "epoch": 42.64,
+      "learning_rate": 2.1322580645161293e-05,
+      "loss": 0.351,
+      "step": 6610
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 2.1354838709677422e-05,
+      "loss": 0.3138,
+      "step": 6620
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 2.1387096774193548e-05,
+      "loss": 0.3506,
+      "step": 6630
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 2.141935483870968e-05,
+      "loss": 0.3292,
+      "step": 6640
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 2.1451612903225807e-05,
+      "loss": 0.2935,
+      "step": 6650
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 2.1483870967741936e-05,
+      "loss": 0.276,
+      "step": 6660
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9298369288444519,
+      "eval_runtime": 0.8696,
+      "eval_samples_per_second": 231.148,
+      "eval_steps_per_second": 8.05,
+      "step": 6665
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 2.1516129032258066e-05,
+      "loss": 0.342,
+      "step": 6670
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 2.1548387096774195e-05,
+      "loss": 0.289,
+      "step": 6680
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 2.1580645161290324e-05,
+      "loss": 0.3246,
+      "step": 6690
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 2.1612903225806454e-05,
+      "loss": 0.3877,
+      "step": 6700
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 2.1645161290322583e-05,
+      "loss": 0.3327,
+      "step": 6710
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 2.167741935483871e-05,
+      "loss": 0.3714,
+      "step": 6720
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 2.1709677419354842e-05,
+      "loss": 0.2788,
+      "step": 6730
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 2.1741935483870968e-05,
+      "loss": 0.3555,
+      "step": 6740
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 2.1774193548387097e-05,
+      "loss": 0.3167,
+      "step": 6750
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 2.1806451612903227e-05,
+      "loss": 0.3201,
+      "step": 6760
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 2.1838709677419356e-05,
+      "loss": 0.3684,
+      "step": 6770
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 2.1870967741935485e-05,
+      "loss": 0.3331,
+      "step": 6780
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 2.1903225806451615e-05,
+      "loss": 0.2962,
+      "step": 6790
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 2.1935483870967744e-05,
+      "loss": 0.3707,
+      "step": 6800
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 2.196774193548387e-05,
+      "loss": 0.3699,
+      "step": 6810
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.3626,
+      "step": 6820
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9185047149658203,
+      "eval_runtime": 0.8934,
+      "eval_samples_per_second": 224.989,
+      "eval_steps_per_second": 7.835,
+      "step": 6820
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 2.2032258064516132e-05,
+      "loss": 0.3763,
+      "step": 6830
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 2.2064516129032258e-05,
+      "loss": 0.3277,
+      "step": 6840
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 2.2096774193548388e-05,
+      "loss": 0.329,
+      "step": 6850
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 2.2129032258064517e-05,
+      "loss": 0.2858,
+      "step": 6860
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 2.2161290322580646e-05,
+      "loss": 0.3824,
+      "step": 6870
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 2.2193548387096776e-05,
+      "loss": 0.3814,
+      "step": 6880
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 2.2225806451612905e-05,
+      "loss": 0.3248,
+      "step": 6890
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 2.2258064516129034e-05,
+      "loss": 0.3534,
+      "step": 6900
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 2.229032258064516e-05,
+      "loss": 0.2758,
+      "step": 6910
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 2.2322580645161293e-05,
+      "loss": 0.3089,
+      "step": 6920
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 2.235483870967742e-05,
+      "loss": 0.3251,
+      "step": 6930
+    },
+    {
+      "epoch": 44.77,
+      "learning_rate": 2.238709677419355e-05,
+      "loss": 0.313,
+      "step": 6940
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 2.2419354838709678e-05,
+      "loss": 0.3536,
+      "step": 6950
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 2.2451612903225807e-05,
+      "loss": 0.3648,
+      "step": 6960
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 2.2483870967741937e-05,
+      "loss": 0.3086,
+      "step": 6970
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9207830429077148,
+      "eval_runtime": 0.8618,
+      "eval_samples_per_second": 233.232,
+      "eval_steps_per_second": 8.123,
+      "step": 6975
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 2.2516129032258066e-05,
+      "loss": 0.3467,
+      "step": 6980
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 2.2548387096774195e-05,
+      "loss": 0.3481,
+      "step": 6990
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 2.258064516129032e-05,
+      "loss": 0.3297,
+      "step": 7000
+    },
+    {
+      "epoch": 45.23,
+      "learning_rate": 2.2612903225806454e-05,
+      "loss": 0.3718,
+      "step": 7010
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 2.264516129032258e-05,
+      "loss": 0.3268,
+      "step": 7020
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 2.267741935483871e-05,
+      "loss": 0.3097,
+      "step": 7030
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 2.2709677419354842e-05,
+      "loss": 0.3236,
+      "step": 7040
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 2.274193548387097e-05,
+      "loss": 0.3513,
+      "step": 7050
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 2.2774193548387098e-05,
+      "loss": 0.3745,
+      "step": 7060
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 2.2806451612903227e-05,
+      "loss": 0.3452,
+      "step": 7070
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 2.2838709677419357e-05,
+      "loss": 0.347,
+      "step": 7080
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 2.2870967741935483e-05,
+      "loss": 0.368,
+      "step": 7090
+    },
+    {
+      "epoch": 45.8,
+      "learning_rate": 2.2903225806451615e-05,
+      "loss": 0.3739,
+      "step": 7100
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 2.293548387096774e-05,
+      "loss": 0.3612,
+      "step": 7110
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 2.296774193548387e-05,
+      "loss": 0.3139,
+      "step": 7120
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.3077,
+      "step": 7130
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8818454742431641,
+      "eval_runtime": 0.851,
+      "eval_samples_per_second": 236.204,
+      "eval_steps_per_second": 8.226,
+      "step": 7130
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 2.303225806451613e-05,
+      "loss": 0.343,
+      "step": 7140
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 2.306451612903226e-05,
+      "loss": 0.2897,
+      "step": 7150
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 2.3096774193548388e-05,
+      "loss": 0.3379,
+      "step": 7160
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 2.3129032258064518e-05,
+      "loss": 0.3607,
+      "step": 7170
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 2.3161290322580644e-05,
+      "loss": 0.2824,
+      "step": 7180
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 2.3193548387096776e-05,
+      "loss": 0.3729,
+      "step": 7190
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 2.3225806451612906e-05,
+      "loss": 0.3759,
+      "step": 7200
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 2.325806451612903e-05,
+      "loss": 0.4088,
+      "step": 7210
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 2.3290322580645164e-05,
+      "loss": 0.31,
+      "step": 7220
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 2.332258064516129e-05,
+      "loss": 0.3909,
+      "step": 7230
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 2.335483870967742e-05,
+      "loss": 0.3609,
+      "step": 7240
+    },
+    {
+      "epoch": 46.77,
+      "learning_rate": 2.338709677419355e-05,
+      "loss": 0.3719,
+      "step": 7250
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 2.341935483870968e-05,
+      "loss": 0.3399,
+      "step": 7260
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 2.3451612903225808e-05,
+      "loss": 0.3604,
+      "step": 7270
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 2.3483870967741937e-05,
+      "loss": 0.3368,
+      "step": 7280
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9004615545272827,
+      "eval_runtime": 0.8389,
+      "eval_samples_per_second": 239.587,
+      "eval_steps_per_second": 8.344,
+      "step": 7285
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 2.3516129032258067e-05,
+      "loss": 0.3654,
+      "step": 7290
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 2.3548387096774193e-05,
+      "loss": 0.3704,
+      "step": 7300
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 2.3580645161290325e-05,
+      "loss": 0.3357,
+      "step": 7310
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 2.361290322580645e-05,
+      "loss": 0.3139,
+      "step": 7320
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 2.364516129032258e-05,
+      "loss": 0.337,
+      "step": 7330
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 2.367741935483871e-05,
+      "loss": 0.32,
+      "step": 7340
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 2.370967741935484e-05,
+      "loss": 0.3354,
+      "step": 7350
+    },
+    {
+      "epoch": 47.48,
+      "learning_rate": 2.374193548387097e-05,
+      "loss": 0.3483,
+      "step": 7360
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 2.37741935483871e-05,
+      "loss": 0.3618,
+      "step": 7370
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 2.3806451612903228e-05,
+      "loss": 0.3683,
+      "step": 7380
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 2.3838709677419354e-05,
+      "loss": 0.341,
+      "step": 7390
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 2.3870967741935486e-05,
+      "loss": 0.3228,
+      "step": 7400
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 2.3903225806451616e-05,
+      "loss": 0.4166,
+      "step": 7410
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 2.3935483870967742e-05,
+      "loss": 0.336,
+      "step": 7420
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 2.396774193548387e-05,
+      "loss": 0.3708,
+      "step": 7430
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 2.4e-05,
+      "loss": 0.4088,
+      "step": 7440
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.912268340587616,
+      "eval_runtime": 0.8565,
+      "eval_samples_per_second": 234.663,
+      "eval_steps_per_second": 8.172,
+      "step": 7440
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 2.403225806451613e-05,
+      "loss": 0.3554,
+      "step": 7450
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 2.406451612903226e-05,
+      "loss": 0.3295,
+      "step": 7460
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 2.409677419354839e-05,
+      "loss": 0.353,
+      "step": 7470
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 2.4129032258064518e-05,
+      "loss": 0.3363,
+      "step": 7480
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 2.4161290322580648e-05,
+      "loss": 0.3311,
+      "step": 7490
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 2.4193548387096777e-05,
+      "loss": 0.2785,
+      "step": 7500
+    },
+    {
+      "epoch": 48.45,
+      "learning_rate": 2.4225806451612903e-05,
+      "loss": 0.3117,
+      "step": 7510
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 2.4258064516129032e-05,
+      "loss": 0.3237,
+      "step": 7520
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 2.429032258064516e-05,
+      "loss": 0.3184,
+      "step": 7530
+    },
+    {
+      "epoch": 48.64,
+      "learning_rate": 2.432258064516129e-05,
+      "loss": 0.3554,
+      "step": 7540
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 2.435483870967742e-05,
+      "loss": 0.3542,
+      "step": 7550
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 2.438709677419355e-05,
+      "loss": 0.3759,
+      "step": 7560
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 2.441935483870968e-05,
+      "loss": 0.3112,
+      "step": 7570
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 2.4451612903225805e-05,
+      "loss": 0.3168,
+      "step": 7580
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 2.4483870967741938e-05,
+      "loss": 0.3819,
+      "step": 7590
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9064768552780151,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.368,
+      "eval_steps_per_second": 8.058,
+      "step": 7595
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 2.4516129032258064e-05,
+      "loss": 0.3646,
+      "step": 7600
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 2.4548387096774193e-05,
+      "loss": 0.3866,
+      "step": 7610
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 2.4580645161290326e-05,
+      "loss": 0.3068,
+      "step": 7620
+    },
+    {
+      "epoch": 49.23,
+      "learning_rate": 2.4612903225806452e-05,
+      "loss": 0.3483,
+      "step": 7630
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 2.464516129032258e-05,
+      "loss": 0.3537,
+      "step": 7640
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 2.467741935483871e-05,
+      "loss": 0.3478,
+      "step": 7650
+    },
+    {
+      "epoch": 49.42,
+      "learning_rate": 2.470967741935484e-05,
+      "loss": 0.3721,
+      "step": 7660
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 2.4741935483870966e-05,
+      "loss": 0.3337,
+      "step": 7670
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 2.47741935483871e-05,
+      "loss": 0.3219,
+      "step": 7680
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 2.480645161290323e-05,
+      "loss": 0.3339,
+      "step": 7690
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 2.4838709677419354e-05,
+      "loss": 0.3906,
+      "step": 7700
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 2.4870967741935487e-05,
+      "loss": 0.3672,
+      "step": 7710
+    },
+    {
+      "epoch": 49.8,
+      "learning_rate": 2.4903225806451613e-05,
+      "loss": 0.3345,
+      "step": 7720
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 2.4935483870967742e-05,
+      "loss": 0.3368,
+      "step": 7730
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 2.4967741935483872e-05,
+      "loss": 0.3479,
+      "step": 7740
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 2.5e-05,
+      "loss": 0.3541,
+      "step": 7750
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8852261304855347,
+      "eval_runtime": 0.8805,
+      "eval_samples_per_second": 228.274,
+      "eval_steps_per_second": 7.95,
+      "step": 7750
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 2.5032258064516127e-05,
+      "loss": 0.35,
+      "step": 7760
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 2.5064516129032263e-05,
+      "loss": 0.3152,
+      "step": 7770
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 2.509677419354839e-05,
+      "loss": 0.3353,
+      "step": 7780
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 2.5129032258064515e-05,
+      "loss": 0.3452,
+      "step": 7790
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 2.5161290322580645e-05,
+      "loss": 0.3533,
+      "step": 7800
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 2.5193548387096777e-05,
+      "loss": 0.3547,
+      "step": 7810
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 2.5225806451612903e-05,
+      "loss": 0.3733,
+      "step": 7820
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 2.5258064516129033e-05,
+      "loss": 0.3644,
+      "step": 7830
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 2.5290322580645166e-05,
+      "loss": 0.3588,
+      "step": 7840
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 2.532258064516129e-05,
+      "loss": 0.3446,
+      "step": 7850
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 2.535483870967742e-05,
+      "loss": 0.3936,
+      "step": 7860
+    },
+    {
+      "epoch": 50.77,
+      "learning_rate": 2.5387096774193547e-05,
+      "loss": 0.3375,
+      "step": 7870
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 2.541935483870968e-05,
+      "loss": 0.3093,
+      "step": 7880
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 2.545161290322581e-05,
+      "loss": 0.3225,
+      "step": 7890
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 2.5483870967741935e-05,
+      "loss": 0.3652,
+      "step": 7900
+    },
+    {
+      "epoch": 51.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9167368412017822,
+      "eval_runtime": 0.8498,
+      "eval_samples_per_second": 236.522,
+      "eval_steps_per_second": 8.237,
+      "step": 7905
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 2.5516129032258068e-05,
+      "loss": 0.4288,
+      "step": 7910
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 2.5548387096774197e-05,
+      "loss": 0.3493,
+      "step": 7920
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 2.5580645161290323e-05,
+      "loss": 0.3808,
+      "step": 7930
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 2.561290322580645e-05,
+      "loss": 0.3397,
+      "step": 7940
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 2.5645161290322582e-05,
+      "loss": 0.3654,
+      "step": 7950
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 2.567741935483871e-05,
+      "loss": 0.352,
+      "step": 7960
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 2.5709677419354837e-05,
+      "loss": 0.3364,
+      "step": 7970
+    },
+    {
+      "epoch": 51.48,
+      "learning_rate": 2.574193548387097e-05,
+      "loss": 0.3348,
+      "step": 7980
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 2.57741935483871e-05,
+      "loss": 0.376,
+      "step": 7990
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 2.5806451612903226e-05,
+      "loss": 0.306,
+      "step": 8000
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 2.5838709677419355e-05,
+      "loss": 0.4043,
+      "step": 8010
+    },
+    {
+      "epoch": 51.74,
+      "learning_rate": 2.5870967741935488e-05,
+      "loss": 0.3333,
+      "step": 8020
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 2.5903225806451614e-05,
+      "loss": 0.365,
+      "step": 8030
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 2.5935483870967743e-05,
+      "loss": 0.3481,
+      "step": 8040
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 2.5967741935483876e-05,
+      "loss": 0.293,
+      "step": 8050
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.3267,
+      "step": 8060
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9290481209754944,
+      "eval_runtime": 0.843,
+      "eval_samples_per_second": 238.434,
+      "eval_steps_per_second": 8.304,
+      "step": 8060
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 2.603225806451613e-05,
+      "loss": 0.3858,
+      "step": 8070
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 2.6064516129032257e-05,
+      "loss": 0.3276,
+      "step": 8080
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 2.609677419354839e-05,
+      "loss": 0.3098,
+      "step": 8090
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 2.6129032258064516e-05,
+      "loss": 0.302,
+      "step": 8100
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 2.6161290322580645e-05,
+      "loss": 0.3394,
+      "step": 8110
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 2.6193548387096778e-05,
+      "loss": 0.2617,
+      "step": 8120
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 2.6225806451612904e-05,
+      "loss": 0.2985,
+      "step": 8130
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 2.6258064516129033e-05,
+      "loss": 0.3685,
+      "step": 8140
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 2.629032258064516e-05,
+      "loss": 0.3289,
+      "step": 8150
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 2.6322580645161292e-05,
+      "loss": 0.3268,
+      "step": 8160
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 2.635483870967742e-05,
+      "loss": 0.3726,
+      "step": 8170
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 2.6387096774193548e-05,
+      "loss": 0.3504,
+      "step": 8180
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 2.641935483870968e-05,
+      "loss": 0.2718,
+      "step": 8190
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 2.645161290322581e-05,
+      "loss": 0.3301,
+      "step": 8200
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 2.6483870967741936e-05,
+      "loss": 0.3057,
+      "step": 8210
+    },
+    {
+      "epoch": 53.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8862131834030151,
+      "eval_runtime": 0.8578,
+      "eval_samples_per_second": 234.31,
+      "eval_steps_per_second": 8.16,
+      "step": 8215
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 2.6516129032258065e-05,
+      "loss": 0.323,
+      "step": 8220
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 2.6548387096774198e-05,
+      "loss": 0.3412,
+      "step": 8230
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 2.6580645161290324e-05,
+      "loss": 0.3577,
+      "step": 8240
+    },
+    {
+      "epoch": 53.23,
+      "learning_rate": 2.661290322580645e-05,
+      "loss": 0.3025,
+      "step": 8250
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 2.6645161290322586e-05,
+      "loss": 0.2918,
+      "step": 8260
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 2.6677419354838712e-05,
+      "loss": 0.336,
+      "step": 8270
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 2.6709677419354838e-05,
+      "loss": 0.3668,
+      "step": 8280
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 2.6741935483870967e-05,
+      "loss": 0.3078,
+      "step": 8290
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 2.67741935483871e-05,
+      "loss": 0.3593,
+      "step": 8300
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 2.6806451612903226e-05,
+      "loss": 0.3017,
+      "step": 8310
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 2.6838709677419355e-05,
+      "loss": 0.3766,
+      "step": 8320
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 2.6870967741935488e-05,
+      "loss": 0.3655,
+      "step": 8330
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 2.6903225806451614e-05,
+      "loss": 0.3553,
+      "step": 8340
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 2.6935483870967744e-05,
+      "loss": 0.304,
+      "step": 8350
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 2.696774193548387e-05,
+      "loss": 0.3611,
+      "step": 8360
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.3169,
+      "step": 8370
+    },
+    {
+      "epoch": 54.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8655603528022766,
+      "eval_runtime": 0.842,
+      "eval_samples_per_second": 238.708,
+      "eval_steps_per_second": 8.313,
+      "step": 8370
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 2.7032258064516132e-05,
+      "loss": 0.3765,
+      "step": 8380
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 2.7064516129032258e-05,
+      "loss": 0.2981,
+      "step": 8390
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 2.709677419354839e-05,
+      "loss": 0.317,
+      "step": 8400
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 2.712903225806452e-05,
+      "loss": 0.3346,
+      "step": 8410
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 2.7161290322580646e-05,
+      "loss": 0.3626,
+      "step": 8420
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 2.7193548387096772e-05,
+      "loss": 0.3227,
+      "step": 8430
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 2.7225806451612905e-05,
+      "loss": 0.3248,
+      "step": 8440
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 2.7258064516129034e-05,
+      "loss": 0.3599,
+      "step": 8450
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 2.729032258064516e-05,
+      "loss": 0.3112,
+      "step": 8460
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 2.7322580645161293e-05,
+      "loss": 0.327,
+      "step": 8470
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 2.7354838709677422e-05,
+      "loss": 0.3565,
+      "step": 8480
+    },
+    {
+      "epoch": 54.77,
+      "learning_rate": 2.7387096774193548e-05,
+      "loss": 0.3197,
+      "step": 8490
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 2.7419354838709678e-05,
+      "loss": 0.3198,
+      "step": 8500
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 2.745161290322581e-05,
+      "loss": 0.3418,
+      "step": 8510
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 2.7483870967741936e-05,
+      "loss": 0.3715,
+      "step": 8520
+    },
+    {
+      "epoch": 55.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9334109425544739,
+      "eval_runtime": 0.8467,
+      "eval_samples_per_second": 237.394,
+      "eval_steps_per_second": 8.267,
+      "step": 8525
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 2.7516129032258066e-05,
+      "loss": 0.3678,
+      "step": 8530
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 2.75483870967742e-05,
+      "loss": 0.3526,
+      "step": 8540
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 2.7580645161290324e-05,
+      "loss": 0.3181,
+      "step": 8550
+    },
+    {
+      "epoch": 55.23,
+      "learning_rate": 2.7612903225806454e-05,
+      "loss": 0.313,
+      "step": 8560
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 2.764516129032258e-05,
+      "loss": 0.3636,
+      "step": 8570
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 2.7677419354838713e-05,
+      "loss": 0.326,
+      "step": 8580
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 2.7709677419354842e-05,
+      "loss": 0.3477,
+      "step": 8590
+    },
+    {
+      "epoch": 55.48,
+      "learning_rate": 2.7741935483870968e-05,
+      "loss": 0.3534,
+      "step": 8600
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 2.77741935483871e-05,
+      "loss": 0.3427,
+      "step": 8610
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 2.7806451612903227e-05,
+      "loss": 0.3573,
+      "step": 8620
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 2.7838709677419356e-05,
+      "loss": 0.3233,
+      "step": 8630
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 2.7870967741935482e-05,
+      "loss": 0.3753,
+      "step": 8640
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 2.7903225806451615e-05,
+      "loss": 0.3578,
+      "step": 8650
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 2.7935483870967744e-05,
+      "loss": 0.3135,
+      "step": 8660
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 2.796774193548387e-05,
+      "loss": 0.3358,
+      "step": 8670
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.3594,
+      "step": 8680
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8903777599334717,
+      "eval_runtime": 0.8874,
+      "eval_samples_per_second": 226.497,
+      "eval_steps_per_second": 7.888,
+      "step": 8680
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 2.8032258064516132e-05,
+      "loss": 0.3519,
+      "step": 8690
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 2.806451612903226e-05,
+      "loss": 0.3233,
+      "step": 8700
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 2.8096774193548388e-05,
+      "loss": 0.3271,
+      "step": 8710
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 2.812903225806452e-05,
+      "loss": 0.3263,
+      "step": 8720
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 2.8161290322580646e-05,
+      "loss": 0.3215,
+      "step": 8730
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 2.8193548387096776e-05,
+      "loss": 0.3587,
+      "step": 8740
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 2.822580645161291e-05,
+      "loss": 0.3645,
+      "step": 8750
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 2.8258064516129035e-05,
+      "loss": 0.4102,
+      "step": 8760
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 2.829032258064516e-05,
+      "loss": 0.35,
+      "step": 8770
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 2.832258064516129e-05,
+      "loss": 0.3633,
+      "step": 8780
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 2.8354838709677423e-05,
+      "loss": 0.4014,
+      "step": 8790
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 2.838709677419355e-05,
+      "loss": 0.3949,
+      "step": 8800
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 2.8419354838709678e-05,
+      "loss": 0.379,
+      "step": 8810
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 2.845161290322581e-05,
+      "loss": 0.2877,
+      "step": 8820
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 2.8483870967741937e-05,
+      "loss": 0.3266,
+      "step": 8830
+    },
+    {
+      "epoch": 57.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8890269994735718,
+      "eval_runtime": 0.8396,
+      "eval_samples_per_second": 239.396,
+      "eval_steps_per_second": 8.337,
+      "step": 8835
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 2.8516129032258066e-05,
+      "loss": 0.3358,
+      "step": 8840
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 2.8548387096774192e-05,
+      "loss": 0.3391,
+      "step": 8850
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 2.8580645161290325e-05,
+      "loss": 0.3493,
+      "step": 8860
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 2.8612903225806454e-05,
+      "loss": 0.3013,
+      "step": 8870
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 2.864516129032258e-05,
+      "loss": 0.3319,
+      "step": 8880
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 2.8677419354838713e-05,
+      "loss": 0.3186,
+      "step": 8890
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 2.8709677419354843e-05,
+      "loss": 0.3447,
+      "step": 8900
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 2.874193548387097e-05,
+      "loss": 0.3379,
+      "step": 8910
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 2.8774193548387095e-05,
+      "loss": 0.3234,
+      "step": 8920
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 2.880645161290323e-05,
+      "loss": 0.3784,
+      "step": 8930
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 2.8838709677419357e-05,
+      "loss": 0.3886,
+      "step": 8940
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 2.8870967741935483e-05,
+      "loss": 0.3257,
+      "step": 8950
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 2.8903225806451615e-05,
+      "loss": 0.3916,
+      "step": 8960
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 2.8935483870967745e-05,
+      "loss": 0.3694,
+      "step": 8970
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 2.896774193548387e-05,
+      "loss": 0.3396,
+      "step": 8980
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 2.9e-05,
+      "loss": 0.3393,
+      "step": 8990
+    },
+    {
+      "epoch": 58.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9295005798339844,
+      "eval_runtime": 0.8553,
+      "eval_samples_per_second": 235.012,
+      "eval_steps_per_second": 8.184,
+      "step": 8990
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 2.9032258064516133e-05,
+      "loss": 0.3477,
+      "step": 9000
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 2.906451612903226e-05,
+      "loss": 0.2807,
+      "step": 9010
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 2.909677419354839e-05,
+      "loss": 0.3342,
+      "step": 9020
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 2.912903225806452e-05,
+      "loss": 0.3446,
+      "step": 9030
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 2.9161290322580647e-05,
+      "loss": 0.3475,
+      "step": 9040
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 2.9193548387096776e-05,
+      "loss": 0.2771,
+      "step": 9050
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 2.9225806451612902e-05,
+      "loss": 0.3629,
+      "step": 9060
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 2.9258064516129035e-05,
+      "loss": 0.3691,
+      "step": 9070
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 2.9290322580645165e-05,
+      "loss": 0.3751,
+      "step": 9080
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 2.932258064516129e-05,
+      "loss": 0.3481,
+      "step": 9090
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 2.9354838709677417e-05,
+      "loss": 0.3716,
+      "step": 9100
+    },
+    {
+      "epoch": 58.77,
+      "learning_rate": 2.938709677419355e-05,
+      "loss": 0.3319,
+      "step": 9110
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 2.941935483870968e-05,
+      "loss": 0.3354,
+      "step": 9120
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 2.9451612903225805e-05,
+      "loss": 0.3193,
+      "step": 9130
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 2.9483870967741937e-05,
+      "loss": 0.3088,
+      "step": 9140
+    },
+    {
+      "epoch": 59.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.937752366065979,
+      "eval_runtime": 0.8615,
+      "eval_samples_per_second": 233.321,
+      "eval_steps_per_second": 8.126,
+      "step": 9145
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 2.9516129032258067e-05,
+      "loss": 0.3535,
+      "step": 9150
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 2.9548387096774193e-05,
+      "loss": 0.3531,
+      "step": 9160
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 2.9580645161290322e-05,
+      "loss": 0.3041,
+      "step": 9170
+    },
+    {
+      "epoch": 59.23,
+      "learning_rate": 2.9612903225806455e-05,
+      "loss": 0.3069,
+      "step": 9180
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 2.964516129032258e-05,
+      "loss": 0.307,
+      "step": 9190
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 2.967741935483871e-05,
+      "loss": 0.332,
+      "step": 9200
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 2.9709677419354843e-05,
+      "loss": 0.3255,
+      "step": 9210
+    },
+    {
+      "epoch": 59.48,
+      "learning_rate": 2.974193548387097e-05,
+      "loss": 0.364,
+      "step": 9220
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 2.97741935483871e-05,
+      "loss": 0.3625,
+      "step": 9230
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 2.9806451612903224e-05,
+      "loss": 0.3096,
+      "step": 9240
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 2.9838709677419357e-05,
+      "loss": 0.3553,
+      "step": 9250
+    },
+    {
+      "epoch": 59.74,
+      "learning_rate": 2.9870967741935487e-05,
+      "loss": 0.3065,
+      "step": 9260
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 2.9903225806451613e-05,
+      "loss": 0.3193,
+      "step": 9270
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 2.9935483870967745e-05,
+      "loss": 0.4069,
+      "step": 9280
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 2.996774193548387e-05,
+      "loss": 0.3818,
+      "step": 9290
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 3e-05,
+      "loss": 0.3108,
+      "step": 9300
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8675196766853333,
+      "eval_runtime": 0.9345,
+      "eval_samples_per_second": 215.093,
+      "eval_steps_per_second": 7.491,
+      "step": 9300
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 3.0032258064516127e-05,
+      "loss": 0.3618,
+      "step": 9310
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 3.006451612903226e-05,
+      "loss": 0.3068,
+      "step": 9320
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 3.009677419354839e-05,
+      "loss": 0.373,
+      "step": 9330
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 3.0129032258064515e-05,
+      "loss": 0.332,
+      "step": 9340
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 3.0161290322580648e-05,
+      "loss": 0.3406,
+      "step": 9350
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 3.0193548387096777e-05,
+      "loss": 0.3554,
+      "step": 9360
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 3.0225806451612903e-05,
+      "loss": 0.3901,
+      "step": 9370
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 3.0258064516129032e-05,
+      "loss": 0.3605,
+      "step": 9380
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 3.0290322580645165e-05,
+      "loss": 0.322,
+      "step": 9390
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 3.032258064516129e-05,
+      "loss": 0.3663,
+      "step": 9400
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 3.035483870967742e-05,
+      "loss": 0.3082,
+      "step": 9410
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 3.0387096774193553e-05,
+      "loss": 0.3329,
+      "step": 9420
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 3.041935483870968e-05,
+      "loss": 0.3291,
+      "step": 9430
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 3.0451612903225805e-05,
+      "loss": 0.3653,
+      "step": 9440
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 3.0483870967741935e-05,
+      "loss": 0.3742,
+      "step": 9450
+    },
+    {
+      "epoch": 61.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9092706441879272,
+      "eval_runtime": 0.845,
+      "eval_samples_per_second": 237.866,
+      "eval_steps_per_second": 8.284,
+      "step": 9455
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 3.0516129032258067e-05,
+      "loss": 0.3733,
+      "step": 9460
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 3.05483870967742e-05,
+      "loss": 0.3888,
+      "step": 9470
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 3.058064516129032e-05,
+      "loss": 0.3125,
+      "step": 9480
+    },
+    {
+      "epoch": 61.23,
+      "learning_rate": 3.0612903225806456e-05,
+      "loss": 0.327,
+      "step": 9490
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 3.0645161290322585e-05,
+      "loss": 0.3228,
+      "step": 9500
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 3.067741935483871e-05,
+      "loss": 0.3811,
+      "step": 9510
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 3.070967741935484e-05,
+      "loss": 0.3184,
+      "step": 9520
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 3.074193548387097e-05,
+      "loss": 0.3864,
+      "step": 9530
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 3.0774193548387096e-05,
+      "loss": 0.3868,
+      "step": 9540
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 3.0806451612903225e-05,
+      "loss": 0.3875,
+      "step": 9550
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 3.083870967741936e-05,
+      "loss": 0.3272,
+      "step": 9560
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 3.0870967741935484e-05,
+      "loss": 0.3142,
+      "step": 9570
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 3.090322580645161e-05,
+      "loss": 0.39,
+      "step": 9580
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 3.093548387096774e-05,
+      "loss": 0.342,
+      "step": 9590
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 3.096774193548387e-05,
+      "loss": 0.3153,
+      "step": 9600
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 3.1e-05,
+      "loss": 0.3394,
+      "step": 9610
+    },
+    {
+      "epoch": 62.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0552465915679932,
+      "eval_runtime": 0.8668,
+      "eval_samples_per_second": 231.899,
+      "eval_steps_per_second": 8.076,
+      "step": 9610
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 3.103225806451613e-05,
+      "loss": 0.3388,
+      "step": 9620
+    },
+    {
+      "epoch": 62.13,
+      "learning_rate": 3.106451612903226e-05,
+      "loss": 0.3255,
+      "step": 9630
+    },
+    {
+      "epoch": 62.19,
+      "learning_rate": 3.109677419354839e-05,
+      "loss": 0.3568,
+      "step": 9640
+    },
+    {
+      "epoch": 62.26,
+      "learning_rate": 3.112903225806452e-05,
+      "loss": 0.3066,
+      "step": 9650
+    },
+    {
+      "epoch": 62.32,
+      "learning_rate": 3.116129032258064e-05,
+      "loss": 0.3348,
+      "step": 9660
+    },
+    {
+      "epoch": 62.39,
+      "learning_rate": 3.119354838709678e-05,
+      "loss": 0.3216,
+      "step": 9670
+    },
+    {
+      "epoch": 62.45,
+      "learning_rate": 3.122580645161291e-05,
+      "loss": 0.3446,
+      "step": 9680
+    },
+    {
+      "epoch": 62.51,
+      "learning_rate": 3.125806451612903e-05,
+      "loss": 0.3923,
+      "step": 9690
+    },
+    {
+      "epoch": 62.58,
+      "learning_rate": 3.1290322580645166e-05,
+      "loss": 0.321,
+      "step": 9700
+    },
+    {
+      "epoch": 62.64,
+      "learning_rate": 3.1322580645161295e-05,
+      "loss": 0.323,
+      "step": 9710
+    },
+    {
+      "epoch": 62.71,
+      "learning_rate": 3.135483870967742e-05,
+      "loss": 0.3886,
+      "step": 9720
+    },
+    {
+      "epoch": 62.77,
+      "learning_rate": 3.138709677419355e-05,
+      "loss": 0.3185,
+      "step": 9730
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 3.141935483870968e-05,
+      "loss": 0.2977,
+      "step": 9740
+    },
+    {
+      "epoch": 62.9,
+      "learning_rate": 3.1451612903225806e-05,
+      "loss": 0.3388,
+      "step": 9750
+    },
+    {
+      "epoch": 62.96,
+      "learning_rate": 3.1483870967741935e-05,
+      "loss": 0.3437,
+      "step": 9760
+    },
+    {
+      "epoch": 63.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9263607859611511,
+      "eval_runtime": 0.8992,
+      "eval_samples_per_second": 223.536,
+      "eval_steps_per_second": 7.785,
+      "step": 9765
+    },
+    {
+      "epoch": 63.03,
+      "learning_rate": 3.151612903225807e-05,
+      "loss": 0.2904,
+      "step": 9770
+    },
+    {
+      "epoch": 63.1,
+      "learning_rate": 3.1548387096774194e-05,
+      "loss": 0.3341,
+      "step": 9780
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 3.1580645161290323e-05,
+      "loss": 0.3235,
+      "step": 9790
+    },
+    {
+      "epoch": 63.23,
+      "learning_rate": 3.161290322580645e-05,
+      "loss": 0.3433,
+      "step": 9800
+    },
+    {
+      "epoch": 63.29,
+      "learning_rate": 3.164516129032258e-05,
+      "loss": 0.3613,
+      "step": 9810
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 3.167741935483871e-05,
+      "loss": 0.4026,
+      "step": 9820
+    },
+    {
+      "epoch": 63.42,
+      "learning_rate": 3.170967741935484e-05,
+      "loss": 0.3424,
+      "step": 9830
+    },
+    {
+      "epoch": 63.48,
+      "learning_rate": 3.174193548387097e-05,
+      "loss": 0.3218,
+      "step": 9840
+    },
+    {
+      "epoch": 63.55,
+      "learning_rate": 3.17741935483871e-05,
+      "loss": 0.3373,
+      "step": 9850
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 3.180645161290323e-05,
+      "loss": 0.3928,
+      "step": 9860
+    },
+    {
+      "epoch": 63.68,
+      "learning_rate": 3.183870967741935e-05,
+      "loss": 0.3471,
+      "step": 9870
+    },
+    {
+      "epoch": 63.74,
+      "learning_rate": 3.187096774193549e-05,
+      "loss": 0.3167,
+      "step": 9880
+    },
+    {
+      "epoch": 63.8,
+      "learning_rate": 3.190322580645162e-05,
+      "loss": 0.3551,
+      "step": 9890
+    },
+    {
+      "epoch": 63.87,
+      "learning_rate": 3.193548387096774e-05,
+      "loss": 0.3139,
+      "step": 9900
+    },
+    {
+      "epoch": 63.93,
+      "learning_rate": 3.1967741935483876e-05,
+      "loss": 0.3159,
+      "step": 9910
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.323,
+      "step": 9920
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9444639086723328,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.469,
+      "eval_steps_per_second": 8.026,
+      "step": 9920
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 3.203225806451613e-05,
+      "loss": 0.3744,
+      "step": 9930
+    },
+    {
+      "epoch": 64.13,
+      "learning_rate": 3.206451612903226e-05,
+      "loss": 0.3579,
+      "step": 9940
+    },
+    {
+      "epoch": 64.19,
+      "learning_rate": 3.2096774193548393e-05,
+      "loss": 0.3275,
+      "step": 9950
+    },
+    {
+      "epoch": 64.26,
+      "learning_rate": 3.2129032258064516e-05,
+      "loss": 0.4097,
+      "step": 9960
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 3.2161290322580645e-05,
+      "loss": 0.3888,
+      "step": 9970
+    },
+    {
+      "epoch": 64.39,
+      "learning_rate": 3.2193548387096775e-05,
+      "loss": 0.3398,
+      "step": 9980
+    },
+    {
+      "epoch": 64.45,
+      "learning_rate": 3.2225806451612904e-05,
+      "loss": 0.4031,
+      "step": 9990
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 0.3371,
+      "step": 10000
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 3.229032258064516e-05,
+      "loss": 0.3425,
+      "step": 10010
+    },
+    {
+      "epoch": 64.64,
+      "learning_rate": 3.232258064516129e-05,
+      "loss": 0.3598,
+      "step": 10020
+    },
+    {
+      "epoch": 64.71,
+      "learning_rate": 3.235483870967742e-05,
+      "loss": 0.3119,
+      "step": 10030
+    },
+    {
+      "epoch": 64.77,
+      "learning_rate": 3.238709677419355e-05,
+      "loss": 0.3164,
+      "step": 10040
+    },
+    {
+      "epoch": 64.84,
+      "learning_rate": 3.241935483870968e-05,
+      "loss": 0.3117,
+      "step": 10050
+    },
+    {
+      "epoch": 64.9,
+      "learning_rate": 3.245161290322581e-05,
+      "loss": 0.3444,
+      "step": 10060
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 3.248387096774194e-05,
+      "loss": 0.3707,
+      "step": 10070
+    },
+    {
+      "epoch": 65.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9190543293952942,
+      "eval_runtime": 0.8625,
+      "eval_samples_per_second": 233.033,
+      "eval_steps_per_second": 8.116,
+      "step": 10075
+    },
+    {
+      "epoch": 65.03,
+      "learning_rate": 3.251612903225806e-05,
+      "loss": 0.3656,
+      "step": 10080
+    },
+    {
+      "epoch": 65.1,
+      "learning_rate": 3.25483870967742e-05,
+      "loss": 0.3401,
+      "step": 10090
+    },
+    {
+      "epoch": 65.16,
+      "learning_rate": 3.258064516129033e-05,
+      "loss": 0.3625,
+      "step": 10100
+    },
+    {
+      "epoch": 65.23,
+      "learning_rate": 3.261290322580645e-05,
+      "loss": 0.3279,
+      "step": 10110
+    },
+    {
+      "epoch": 65.29,
+      "learning_rate": 3.2645161290322586e-05,
+      "loss": 0.3878,
+      "step": 10120
+    },
+    {
+      "epoch": 65.35,
+      "learning_rate": 3.267741935483871e-05,
+      "loss": 0.3059,
+      "step": 10130
+    },
+    {
+      "epoch": 65.42,
+      "learning_rate": 3.270967741935484e-05,
+      "loss": 0.376,
+      "step": 10140
+    },
+    {
+      "epoch": 65.48,
+      "learning_rate": 3.274193548387097e-05,
+      "loss": 0.2943,
+      "step": 10150
+    },
+    {
+      "epoch": 65.55,
+      "learning_rate": 3.27741935483871e-05,
+      "loss": 0.3715,
+      "step": 10160
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 3.2806451612903226e-05,
+      "loss": 0.3452,
+      "step": 10170
+    },
+    {
+      "epoch": 65.68,
+      "learning_rate": 3.2838709677419356e-05,
+      "loss": 0.3792,
+      "step": 10180
+    },
+    {
+      "epoch": 65.74,
+      "learning_rate": 3.2870967741935485e-05,
+      "loss": 0.3524,
+      "step": 10190
+    },
+    {
+      "epoch": 65.8,
+      "learning_rate": 3.2903225806451614e-05,
+      "loss": 0.3257,
+      "step": 10200
+    },
+    {
+      "epoch": 65.87,
+      "learning_rate": 3.2935483870967744e-05,
+      "loss": 0.3752,
+      "step": 10210
+    },
+    {
+      "epoch": 65.93,
+      "learning_rate": 3.296774193548387e-05,
+      "loss": 0.3595,
+      "step": 10220
+    },
+    {
+      "epoch": 66.0,
+      "learning_rate": 3.3e-05,
+      "loss": 0.3119,
+      "step": 10230
+    },
+    {
+      "epoch": 66.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9501498341560364,
+      "eval_runtime": 0.8621,
+      "eval_samples_per_second": 233.154,
+      "eval_steps_per_second": 8.12,
+      "step": 10230
+    },
+    {
+      "epoch": 66.06,
+      "learning_rate": 3.303225806451613e-05,
+      "loss": 0.3192,
+      "step": 10240
+    },
+    {
+      "epoch": 66.13,
+      "learning_rate": 3.306451612903226e-05,
+      "loss": 0.2844,
+      "step": 10250
+    },
+    {
+      "epoch": 66.19,
+      "learning_rate": 3.309677419354839e-05,
+      "loss": 0.4106,
+      "step": 10260
+    },
+    {
+      "epoch": 66.26,
+      "learning_rate": 3.312903225806452e-05,
+      "loss": 0.3098,
+      "step": 10270
+    },
+    {
+      "epoch": 66.32,
+      "learning_rate": 3.316129032258064e-05,
+      "loss": 0.3962,
+      "step": 10280
+    },
+    {
+      "epoch": 66.39,
+      "learning_rate": 3.319354838709677e-05,
+      "loss": 0.3338,
+      "step": 10290
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 3.322580645161291e-05,
+      "loss": 0.383,
+      "step": 10300
+    },
+    {
+      "epoch": 66.51,
+      "learning_rate": 3.325806451612903e-05,
+      "loss": 0.3334,
+      "step": 10310
+    },
+    {
+      "epoch": 66.58,
+      "learning_rate": 3.329032258064516e-05,
+      "loss": 0.3063,
+      "step": 10320
+    },
+    {
+      "epoch": 66.64,
+      "learning_rate": 3.3322580645161296e-05,
+      "loss": 0.3624,
+      "step": 10330
+    },
+    {
+      "epoch": 66.71,
+      "learning_rate": 3.335483870967742e-05,
+      "loss": 0.3281,
+      "step": 10340
+    },
+    {
+      "epoch": 66.77,
+      "learning_rate": 3.338709677419355e-05,
+      "loss": 0.35,
+      "step": 10350
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 3.341935483870968e-05,
+      "loss": 0.3034,
+      "step": 10360
+    },
+    {
+      "epoch": 66.9,
+      "learning_rate": 3.345161290322581e-05,
+      "loss": 0.3499,
+      "step": 10370
+    },
+    {
+      "epoch": 66.96,
+      "learning_rate": 3.3483870967741936e-05,
+      "loss": 0.3623,
+      "step": 10380
+    },
+    {
+      "epoch": 67.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9688825607299805,
+      "eval_runtime": 0.8637,
+      "eval_samples_per_second": 232.716,
+      "eval_steps_per_second": 8.105,
+      "step": 10385
+    },
+    {
+      "epoch": 67.03,
+      "learning_rate": 3.3516129032258066e-05,
+      "loss": 0.3477,
+      "step": 10390
+    },
+    {
+      "epoch": 67.1,
+      "learning_rate": 3.3548387096774195e-05,
+      "loss": 0.3401,
+      "step": 10400
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 3.3580645161290325e-05,
+      "loss": 0.3474,
+      "step": 10410
+    },
+    {
+      "epoch": 67.23,
+      "learning_rate": 3.3612903225806454e-05,
+      "loss": 0.38,
+      "step": 10420
+    },
+    {
+      "epoch": 67.29,
+      "learning_rate": 3.3645161290322577e-05,
+      "loss": 0.3803,
+      "step": 10430
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 3.367741935483871e-05,
+      "loss": 0.4011,
+      "step": 10440
+    },
+    {
+      "epoch": 67.42,
+      "learning_rate": 3.370967741935484e-05,
+      "loss": 0.3471,
+      "step": 10450
+    },
+    {
+      "epoch": 67.48,
+      "learning_rate": 3.3741935483870965e-05,
+      "loss": 0.3901,
+      "step": 10460
+    },
+    {
+      "epoch": 67.55,
+      "learning_rate": 3.37741935483871e-05,
+      "loss": 0.3196,
+      "step": 10470
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 3.380645161290323e-05,
+      "loss": 0.3289,
+      "step": 10480
+    },
+    {
+      "epoch": 67.68,
+      "learning_rate": 3.383870967741935e-05,
+      "loss": 0.3363,
+      "step": 10490
+    },
+    {
+      "epoch": 67.74,
+      "learning_rate": 3.387096774193548e-05,
+      "loss": 0.4135,
+      "step": 10500
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 3.390322580645162e-05,
+      "loss": 0.3797,
+      "step": 10510
+    },
+    {
+      "epoch": 67.87,
+      "learning_rate": 3.393548387096774e-05,
+      "loss": 0.332,
+      "step": 10520
+    },
+    {
+      "epoch": 67.93,
+      "learning_rate": 3.396774193548387e-05,
+      "loss": 0.4143,
+      "step": 10530
+    },
+    {
+      "epoch": 68.0,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.3794,
+      "step": 10540
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.1084413528442383,
+      "eval_runtime": 0.8536,
+      "eval_samples_per_second": 235.486,
+      "eval_steps_per_second": 8.201,
+      "step": 10540
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 3.403225806451613e-05,
+      "loss": 0.3738,
+      "step": 10550
+    },
+    {
+      "epoch": 68.13,
+      "learning_rate": 3.406451612903226e-05,
+      "loss": 0.327,
+      "step": 10560
+    },
+    {
+      "epoch": 68.19,
+      "learning_rate": 3.409677419354839e-05,
+      "loss": 0.3104,
+      "step": 10570
+    },
+    {
+      "epoch": 68.26,
+      "learning_rate": 3.412903225806452e-05,
+      "loss": 0.3238,
+      "step": 10580
+    },
+    {
+      "epoch": 68.32,
+      "learning_rate": 3.416129032258065e-05,
+      "loss": 0.3298,
+      "step": 10590
+    },
+    {
+      "epoch": 68.39,
+      "learning_rate": 3.4193548387096776e-05,
+      "loss": 0.3475,
+      "step": 10600
+    },
+    {
+      "epoch": 68.45,
+      "learning_rate": 3.4225806451612905e-05,
+      "loss": 0.3155,
+      "step": 10610
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 3.4258064516129035e-05,
+      "loss": 0.3638,
+      "step": 10620
+    },
+    {
+      "epoch": 68.58,
+      "learning_rate": 3.4290322580645164e-05,
+      "loss": 0.403,
+      "step": 10630
+    },
+    {
+      "epoch": 68.64,
+      "learning_rate": 3.432258064516129e-05,
+      "loss": 0.3025,
+      "step": 10640
+    },
+    {
+      "epoch": 68.71,
+      "learning_rate": 3.435483870967742e-05,
+      "loss": 0.3247,
+      "step": 10650
+    },
+    {
+      "epoch": 68.77,
+      "learning_rate": 3.438709677419355e-05,
+      "loss": 0.325,
+      "step": 10660
+    },
+    {
+      "epoch": 68.84,
+      "learning_rate": 3.4419354838709675e-05,
+      "loss": 0.3739,
+      "step": 10670
+    },
+    {
+      "epoch": 68.9,
+      "learning_rate": 3.445161290322581e-05,
+      "loss": 0.3559,
+      "step": 10680
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 3.448387096774194e-05,
+      "loss": 0.374,
+      "step": 10690
+    },
+    {
+      "epoch": 69.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0028343200683594,
+      "eval_runtime": 0.8686,
+      "eval_samples_per_second": 231.399,
+      "eval_steps_per_second": 8.059,
+      "step": 10695
+    },
+    {
+      "epoch": 69.03,
+      "learning_rate": 3.451612903225806e-05,
+      "loss": 0.4091,
+      "step": 10700
+    },
+    {
+      "epoch": 69.1,
+      "learning_rate": 3.454838709677419e-05,
+      "loss": 0.3946,
+      "step": 10710
+    },
+    {
+      "epoch": 69.16,
+      "learning_rate": 3.458064516129033e-05,
+      "loss": 0.3062,
+      "step": 10720
+    },
+    {
+      "epoch": 69.23,
+      "learning_rate": 3.461290322580645e-05,
+      "loss": 0.3973,
+      "step": 10730
+    },
+    {
+      "epoch": 69.29,
+      "learning_rate": 3.464516129032258e-05,
+      "loss": 0.3296,
+      "step": 10740
+    },
+    {
+      "epoch": 69.35,
+      "learning_rate": 3.467741935483872e-05,
+      "loss": 0.3281,
+      "step": 10750
+    },
+    {
+      "epoch": 69.42,
+      "learning_rate": 3.470967741935484e-05,
+      "loss": 0.3424,
+      "step": 10760
+    },
+    {
+      "epoch": 69.48,
+      "learning_rate": 3.474193548387097e-05,
+      "loss": 0.33,
+      "step": 10770
+    },
+    {
+      "epoch": 69.55,
+      "learning_rate": 3.47741935483871e-05,
+      "loss": 0.3382,
+      "step": 10780
+    },
+    {
+      "epoch": 69.61,
+      "learning_rate": 3.480645161290323e-05,
+      "loss": 0.3744,
+      "step": 10790
+    },
+    {
+      "epoch": 69.68,
+      "learning_rate": 3.483870967741936e-05,
+      "loss": 0.3557,
+      "step": 10800
+    },
+    {
+      "epoch": 69.74,
+      "learning_rate": 3.4870967741935486e-05,
+      "loss": 0.347,
+      "step": 10810
+    },
+    {
+      "epoch": 69.8,
+      "learning_rate": 3.4903225806451616e-05,
+      "loss": 0.3479,
+      "step": 10820
+    },
+    {
+      "epoch": 69.87,
+      "learning_rate": 3.4935483870967745e-05,
+      "loss": 0.3578,
+      "step": 10830
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 3.4967741935483874e-05,
+      "loss": 0.3633,
+      "step": 10840
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 3.5e-05,
+      "loss": 0.4189,
+      "step": 10850
+    },
+    {
+      "epoch": 70.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0393540859222412,
+      "eval_runtime": 0.8898,
+      "eval_samples_per_second": 225.905,
+      "eval_steps_per_second": 7.867,
+      "step": 10850
+    },
+    {
+      "epoch": 70.06,
+      "learning_rate": 3.503225806451613e-05,
+      "loss": 0.3551,
+      "step": 10860
+    },
+    {
+      "epoch": 70.13,
+      "learning_rate": 3.506451612903226e-05,
+      "loss": 0.3187,
+      "step": 10870
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 3.5096774193548385e-05,
+      "loss": 0.3891,
+      "step": 10880
+    },
+    {
+      "epoch": 70.26,
+      "learning_rate": 3.512903225806452e-05,
+      "loss": 0.3453,
+      "step": 10890
+    },
+    {
+      "epoch": 70.32,
+      "learning_rate": 3.516129032258065e-05,
+      "loss": 0.3172,
+      "step": 10900
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 3.519354838709677e-05,
+      "loss": 0.3728,
+      "step": 10910
+    },
+    {
+      "epoch": 70.45,
+      "learning_rate": 3.52258064516129e-05,
+      "loss": 0.2762,
+      "step": 10920
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 3.525806451612904e-05,
+      "loss": 0.3418,
+      "step": 10930
+    },
+    {
+      "epoch": 70.58,
+      "learning_rate": 3.529032258064516e-05,
+      "loss": 0.3754,
+      "step": 10940
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 3.532258064516129e-05,
+      "loss": 0.352,
+      "step": 10950
+    },
+    {
+      "epoch": 70.71,
+      "learning_rate": 3.535483870967743e-05,
+      "loss": 0.2711,
+      "step": 10960
+    },
+    {
+      "epoch": 70.77,
+      "learning_rate": 3.538709677419355e-05,
+      "loss": 0.361,
+      "step": 10970
+    },
+    {
+      "epoch": 70.84,
+      "learning_rate": 3.541935483870968e-05,
+      "loss": 0.3201,
+      "step": 10980
+    },
+    {
+      "epoch": 70.9,
+      "learning_rate": 3.545161290322581e-05,
+      "loss": 0.3215,
+      "step": 10990
+    },
+    {
+      "epoch": 70.96,
+      "learning_rate": 3.548387096774194e-05,
+      "loss": 0.3492,
+      "step": 11000
+    },
+    {
+      "epoch": 71.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.039563775062561,
+      "eval_runtime": 0.8558,
+      "eval_samples_per_second": 234.877,
+      "eval_steps_per_second": 8.18,
+      "step": 11005
+    },
+    {
+      "epoch": 71.03,
+      "learning_rate": 3.551612903225807e-05,
+      "loss": 0.3457,
+      "step": 11010
+    },
+    {
+      "epoch": 71.1,
+      "learning_rate": 3.5548387096774196e-05,
+      "loss": 0.3454,
+      "step": 11020
+    },
+    {
+      "epoch": 71.16,
+      "learning_rate": 3.5580645161290326e-05,
+      "loss": 0.3616,
+      "step": 11030
+    },
+    {
+      "epoch": 71.23,
+      "learning_rate": 3.5612903225806455e-05,
+      "loss": 0.333,
+      "step": 11040
+    },
+    {
+      "epoch": 71.29,
+      "learning_rate": 3.5645161290322585e-05,
+      "loss": 0.3609,
+      "step": 11050
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 3.567741935483871e-05,
+      "loss": 0.3734,
+      "step": 11060
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 3.570967741935484e-05,
+      "loss": 0.28,
+      "step": 11070
+    },
+    {
+      "epoch": 71.48,
+      "learning_rate": 3.574193548387097e-05,
+      "loss": 0.4307,
+      "step": 11080
+    },
+    {
+      "epoch": 71.55,
+      "learning_rate": 3.5774193548387095e-05,
+      "loss": 0.3627,
+      "step": 11090
+    },
+    {
+      "epoch": 71.61,
+      "learning_rate": 3.580645161290323e-05,
+      "loss": 0.3371,
+      "step": 11100
+    },
+    {
+      "epoch": 71.68,
+      "learning_rate": 3.583870967741936e-05,
+      "loss": 0.2982,
+      "step": 11110
+    },
+    {
+      "epoch": 71.74,
+      "learning_rate": 3.5870967741935483e-05,
+      "loss": 0.3671,
+      "step": 11120
+    },
+    {
+      "epoch": 71.8,
+      "learning_rate": 3.590322580645161e-05,
+      "loss": 0.3858,
+      "step": 11130
+    },
+    {
+      "epoch": 71.87,
+      "learning_rate": 3.593548387096774e-05,
+      "loss": 0.3651,
+      "step": 11140
+    },
+    {
+      "epoch": 71.93,
+      "learning_rate": 3.596774193548387e-05,
+      "loss": 0.3286,
+      "step": 11150
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 3.6e-05,
+      "loss": 0.3695,
+      "step": 11160
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0077959299087524,
+      "eval_runtime": 0.8474,
+      "eval_samples_per_second": 237.197,
+      "eval_steps_per_second": 8.261,
+      "step": 11160
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 3.603225806451613e-05,
+      "loss": 0.3894,
+      "step": 11170
+    },
+    {
+      "epoch": 72.13,
+      "learning_rate": 3.606451612903226e-05,
+      "loss": 0.3541,
+      "step": 11180
+    },
+    {
+      "epoch": 72.19,
+      "learning_rate": 3.609677419354839e-05,
+      "loss": 0.338,
+      "step": 11190
+    },
+    {
+      "epoch": 72.26,
+      "learning_rate": 3.612903225806452e-05,
+      "loss": 0.3655,
+      "step": 11200
+    },
+    {
+      "epoch": 72.32,
+      "learning_rate": 3.616129032258065e-05,
+      "loss": 0.2724,
+      "step": 11210
+    },
+    {
+      "epoch": 72.39,
+      "learning_rate": 3.619354838709678e-05,
+      "loss": 0.3372,
+      "step": 11220
+    },
+    {
+      "epoch": 72.45,
+      "learning_rate": 3.6225806451612907e-05,
+      "loss": 0.3527,
+      "step": 11230
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 3.6258064516129036e-05,
+      "loss": 0.3356,
+      "step": 11240
+    },
+    {
+      "epoch": 72.58,
+      "learning_rate": 3.6290322580645165e-05,
+      "loss": 0.3838,
+      "step": 11250
+    },
+    {
+      "epoch": 72.64,
+      "learning_rate": 3.6322580645161295e-05,
+      "loss": 0.3069,
+      "step": 11260
+    },
+    {
+      "epoch": 72.71,
+      "learning_rate": 3.635483870967742e-05,
+      "loss": 0.316,
+      "step": 11270
+    },
+    {
+      "epoch": 72.77,
+      "learning_rate": 3.6387096774193553e-05,
+      "loss": 0.3801,
+      "step": 11280
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 3.641935483870968e-05,
+      "loss": 0.3375,
+      "step": 11290
+    },
+    {
+      "epoch": 72.9,
+      "learning_rate": 3.6451612903225805e-05,
+      "loss": 0.3182,
+      "step": 11300
+    },
+    {
+      "epoch": 72.96,
+      "learning_rate": 3.648387096774194e-05,
+      "loss": 0.3506,
+      "step": 11310
+    },
+    {
+      "epoch": 73.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0237681865692139,
+      "eval_runtime": 0.8597,
+      "eval_samples_per_second": 233.797,
+      "eval_steps_per_second": 8.142,
+      "step": 11315
+    },
+    {
+      "epoch": 73.03,
+      "learning_rate": 3.6516129032258064e-05,
+      "loss": 0.3592,
+      "step": 11320
+    },
+    {
+      "epoch": 73.1,
+      "learning_rate": 3.6548387096774194e-05,
+      "loss": 0.3652,
+      "step": 11330
+    },
+    {
+      "epoch": 73.16,
+      "learning_rate": 3.658064516129032e-05,
+      "loss": 0.441,
+      "step": 11340
+    },
+    {
+      "epoch": 73.23,
+      "learning_rate": 3.661290322580645e-05,
+      "loss": 0.3251,
+      "step": 11350
+    },
+    {
+      "epoch": 73.29,
+      "learning_rate": 3.664516129032258e-05,
+      "loss": 0.3709,
+      "step": 11360
+    },
+    {
+      "epoch": 73.35,
+      "learning_rate": 3.667741935483871e-05,
+      "loss": 0.3862,
+      "step": 11370
+    },
+    {
+      "epoch": 73.42,
+      "learning_rate": 3.670967741935484e-05,
+      "loss": 0.2961,
+      "step": 11380
+    },
+    {
+      "epoch": 73.48,
+      "learning_rate": 3.674193548387097e-05,
+      "loss": 0.3382,
+      "step": 11390
+    },
+    {
+      "epoch": 73.55,
+      "learning_rate": 3.67741935483871e-05,
+      "loss": 0.3104,
+      "step": 11400
+    },
+    {
+      "epoch": 73.61,
+      "learning_rate": 3.680645161290323e-05,
+      "loss": 0.3672,
+      "step": 11410
+    },
+    {
+      "epoch": 73.68,
+      "learning_rate": 3.683870967741936e-05,
+      "loss": 0.3453,
+      "step": 11420
+    },
+    {
+      "epoch": 73.74,
+      "learning_rate": 3.687096774193549e-05,
+      "loss": 0.3494,
+      "step": 11430
+    },
+    {
+      "epoch": 73.8,
+      "learning_rate": 3.690322580645162e-05,
+      "loss": 0.4021,
+      "step": 11440
+    },
+    {
+      "epoch": 73.87,
+      "learning_rate": 3.6935483870967746e-05,
+      "loss": 0.326,
+      "step": 11450
+    },
+    {
+      "epoch": 73.93,
+      "learning_rate": 3.6967741935483876e-05,
+      "loss": 0.3339,
+      "step": 11460
+    },
+    {
+      "epoch": 74.0,
+      "learning_rate": 3.7e-05,
+      "loss": 0.352,
+      "step": 11470
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9790213704109192,
+      "eval_runtime": 0.8409,
+      "eval_samples_per_second": 239.033,
+      "eval_steps_per_second": 8.325,
+      "step": 11470
+    },
+    {
+      "epoch": 74.06,
+      "learning_rate": 3.703225806451613e-05,
+      "loss": 0.3425,
+      "step": 11480
+    },
+    {
+      "epoch": 74.13,
+      "learning_rate": 3.7064516129032264e-05,
+      "loss": 0.3608,
+      "step": 11490
+    },
+    {
+      "epoch": 74.19,
+      "learning_rate": 3.7096774193548386e-05,
+      "loss": 0.3555,
+      "step": 11500
+    },
+    {
+      "epoch": 74.26,
+      "learning_rate": 3.7129032258064516e-05,
+      "loss": 0.342,
+      "step": 11510
+    },
+    {
+      "epoch": 74.32,
+      "learning_rate": 3.716129032258065e-05,
+      "loss": 0.3452,
+      "step": 11520
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 3.7193548387096774e-05,
+      "loss": 0.4109,
+      "step": 11530
+    },
+    {
+      "epoch": 74.45,
+      "learning_rate": 3.7225806451612904e-05,
+      "loss": 0.3122,
+      "step": 11540
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 3.725806451612903e-05,
+      "loss": 0.3606,
+      "step": 11550
+    },
+    {
+      "epoch": 74.58,
+      "learning_rate": 3.729032258064516e-05,
+      "loss": 0.3325,
+      "step": 11560
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 3.732258064516129e-05,
+      "loss": 0.3645,
+      "step": 11570
+    },
+    {
+      "epoch": 74.71,
+      "learning_rate": 3.735483870967742e-05,
+      "loss": 0.381,
+      "step": 11580
+    },
+    {
+      "epoch": 74.77,
+      "learning_rate": 3.738709677419355e-05,
+      "loss": 0.3878,
+      "step": 11590
+    },
+    {
+      "epoch": 74.84,
+      "learning_rate": 3.741935483870968e-05,
+      "loss": 0.3665,
+      "step": 11600
+    },
+    {
+      "epoch": 74.9,
+      "learning_rate": 3.745161290322581e-05,
+      "loss": 0.3309,
+      "step": 11610
+    },
+    {
+      "epoch": 74.96,
+      "learning_rate": 3.748387096774193e-05,
+      "loss": 0.3727,
+      "step": 11620
+    },
+    {
+      "epoch": 75.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9489021897315979,
+      "eval_runtime": 0.8493,
+      "eval_samples_per_second": 236.665,
+      "eval_steps_per_second": 8.242,
+      "step": 11625
+    },
+    {
+      "epoch": 75.03,
+      "learning_rate": 3.751612903225807e-05,
+      "loss": 0.3954,
+      "step": 11630
+    },
+    {
+      "epoch": 75.1,
+      "learning_rate": 3.75483870967742e-05,
+      "loss": 0.3463,
+      "step": 11640
+    },
+    {
+      "epoch": 75.16,
+      "learning_rate": 3.758064516129032e-05,
+      "loss": 0.291,
+      "step": 11650
+    },
+    {
+      "epoch": 75.23,
+      "learning_rate": 3.761290322580645e-05,
+      "loss": 0.3486,
+      "step": 11660
+    },
+    {
+      "epoch": 75.29,
+      "learning_rate": 3.7645161290322586e-05,
+      "loss": 0.322,
+      "step": 11670
+    },
+    {
+      "epoch": 75.35,
+      "learning_rate": 3.767741935483871e-05,
+      "loss": 0.3835,
+      "step": 11680
+    },
+    {
+      "epoch": 75.42,
+      "learning_rate": 3.770967741935484e-05,
+      "loss": 0.3722,
+      "step": 11690
+    },
+    {
+      "epoch": 75.48,
+      "learning_rate": 3.7741935483870974e-05,
+      "loss": 0.331,
+      "step": 11700
+    },
+    {
+      "epoch": 75.55,
+      "learning_rate": 3.7774193548387096e-05,
+      "loss": 0.4023,
+      "step": 11710
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 3.7806451612903226e-05,
+      "loss": 0.3474,
+      "step": 11720
+    },
+    {
+      "epoch": 75.68,
+      "learning_rate": 3.7838709677419355e-05,
+      "loss": 0.33,
+      "step": 11730
+    },
+    {
+      "epoch": 75.74,
+      "learning_rate": 3.7870967741935485e-05,
+      "loss": 0.3996,
+      "step": 11740
+    },
+    {
+      "epoch": 75.8,
+      "learning_rate": 3.7903225806451614e-05,
+      "loss": 0.3407,
+      "step": 11750
+    },
+    {
+      "epoch": 75.87,
+      "learning_rate": 3.793548387096774e-05,
+      "loss": 0.4119,
+      "step": 11760
+    },
+    {
+      "epoch": 75.93,
+      "learning_rate": 3.796774193548387e-05,
+      "loss": 0.3293,
+      "step": 11770
+    },
+    {
+      "epoch": 76.0,
+      "learning_rate": 3.8e-05,
+      "loss": 0.3802,
+      "step": 11780
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9710496068000793,
+      "eval_runtime": 0.8386,
+      "eval_samples_per_second": 239.691,
+      "eval_steps_per_second": 8.347,
+      "step": 11780
+    },
+    {
+      "epoch": 76.06,
+      "learning_rate": 3.803225806451613e-05,
+      "loss": 0.3724,
+      "step": 11790
+    },
+    {
+      "epoch": 76.13,
+      "learning_rate": 3.8064516129032254e-05,
+      "loss": 0.3452,
+      "step": 11800
+    },
+    {
+      "epoch": 76.19,
+      "learning_rate": 3.809677419354839e-05,
+      "loss": 0.3937,
+      "step": 11810
+    },
+    {
+      "epoch": 76.26,
+      "learning_rate": 3.812903225806452e-05,
+      "loss": 0.4232,
+      "step": 11820
+    },
+    {
+      "epoch": 76.32,
+      "learning_rate": 3.816129032258064e-05,
+      "loss": 0.4028,
+      "step": 11830
+    },
+    {
+      "epoch": 76.39,
+      "learning_rate": 3.819354838709678e-05,
+      "loss": 0.3361,
+      "step": 11840
+    },
+    {
+      "epoch": 76.45,
+      "learning_rate": 3.822580645161291e-05,
+      "loss": 0.3162,
+      "step": 11850
+    },
+    {
+      "epoch": 76.51,
+      "learning_rate": 3.825806451612903e-05,
+      "loss": 0.3306,
+      "step": 11860
+    },
+    {
+      "epoch": 76.58,
+      "learning_rate": 3.829032258064516e-05,
+      "loss": 0.3209,
+      "step": 11870
+    },
+    {
+      "epoch": 76.64,
+      "learning_rate": 3.8322580645161296e-05,
+      "loss": 0.3635,
+      "step": 11880
+    },
+    {
+      "epoch": 76.71,
+      "learning_rate": 3.835483870967742e-05,
+      "loss": 0.3173,
+      "step": 11890
+    },
+    {
+      "epoch": 76.77,
+      "learning_rate": 3.838709677419355e-05,
+      "loss": 0.3808,
+      "step": 11900
+    },
+    {
+      "epoch": 76.84,
+      "learning_rate": 3.8419354838709684e-05,
+      "loss": 0.4051,
+      "step": 11910
+    },
+    {
+      "epoch": 76.9,
+      "learning_rate": 3.845161290322581e-05,
+      "loss": 0.3547,
+      "step": 11920
+    },
+    {
+      "epoch": 76.96,
+      "learning_rate": 3.8483870967741936e-05,
+      "loss": 0.3788,
+      "step": 11930
+    },
+    {
+      "epoch": 77.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0000590085983276,
+      "eval_runtime": 0.8577,
+      "eval_samples_per_second": 234.354,
+      "eval_steps_per_second": 8.162,
+      "step": 11935
+    },
+    {
+      "epoch": 77.03,
+      "learning_rate": 3.8516129032258065e-05,
+      "loss": 0.3211,
+      "step": 11940
+    },
+    {
+      "epoch": 77.1,
+      "learning_rate": 3.8548387096774195e-05,
+      "loss": 0.362,
+      "step": 11950
+    },
+    {
+      "epoch": 77.16,
+      "learning_rate": 3.8580645161290324e-05,
+      "loss": 0.34,
+      "step": 11960
+    },
+    {
+      "epoch": 77.23,
+      "learning_rate": 3.8612903225806454e-05,
+      "loss": 0.3016,
+      "step": 11970
+    },
+    {
+      "epoch": 77.29,
+      "learning_rate": 3.864516129032258e-05,
+      "loss": 0.3919,
+      "step": 11980
+    },
+    {
+      "epoch": 77.35,
+      "learning_rate": 3.867741935483871e-05,
+      "loss": 0.3746,
+      "step": 11990
+    },
+    {
+      "epoch": 77.42,
+      "learning_rate": 3.870967741935484e-05,
+      "loss": 0.3697,
+      "step": 12000
+    },
+    {
+      "epoch": 77.48,
+      "learning_rate": 3.8741935483870964e-05,
+      "loss": 0.2719,
+      "step": 12010
+    },
+    {
+      "epoch": 77.55,
+      "learning_rate": 3.87741935483871e-05,
+      "loss": 0.3356,
+      "step": 12020
+    },
+    {
+      "epoch": 77.61,
+      "learning_rate": 3.880645161290323e-05,
+      "loss": 0.3498,
+      "step": 12030
+    },
+    {
+      "epoch": 77.68,
+      "learning_rate": 3.883870967741935e-05,
+      "loss": 0.3831,
+      "step": 12040
+    },
+    {
+      "epoch": 77.74,
+      "learning_rate": 3.887096774193549e-05,
+      "loss": 0.3656,
+      "step": 12050
+    },
+    {
+      "epoch": 77.8,
+      "learning_rate": 3.890322580645162e-05,
+      "loss": 0.3428,
+      "step": 12060
+    },
+    {
+      "epoch": 77.87,
+      "learning_rate": 3.893548387096774e-05,
+      "loss": 0.4146,
+      "step": 12070
+    },
+    {
+      "epoch": 77.93,
+      "learning_rate": 3.896774193548387e-05,
+      "loss": 0.3593,
+      "step": 12080
+    },
+    {
+      "epoch": 78.0,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.3882,
+      "step": 12090
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy": 0.7313432835820896,
+      "eval_loss": 1.0876059532165527,
+      "eval_runtime": 0.8611,
+      "eval_samples_per_second": 233.427,
+      "eval_steps_per_second": 8.129,
+      "step": 12090
+    },
+    {
+      "epoch": 78.06,
+      "learning_rate": 3.903225806451613e-05,
+      "loss": 0.3483,
+      "step": 12100
+    },
+    {
+      "epoch": 78.13,
+      "learning_rate": 3.906451612903226e-05,
+      "loss": 0.3379,
+      "step": 12110
+    },
+    {
+      "epoch": 78.19,
+      "learning_rate": 3.9096774193548394e-05,
+      "loss": 0.3472,
+      "step": 12120
+    },
+    {
+      "epoch": 78.26,
+      "learning_rate": 3.912903225806452e-05,
+      "loss": 0.3432,
+      "step": 12130
+    },
+    {
+      "epoch": 78.32,
+      "learning_rate": 3.9161290322580646e-05,
+      "loss": 0.3453,
+      "step": 12140
+    },
+    {
+      "epoch": 78.39,
+      "learning_rate": 3.9193548387096776e-05,
+      "loss": 0.3683,
+      "step": 12150
+    },
+    {
+      "epoch": 78.45,
+      "learning_rate": 3.9225806451612905e-05,
+      "loss": 0.3166,
+      "step": 12160
+    },
+    {
+      "epoch": 78.51,
+      "learning_rate": 3.9258064516129034e-05,
+      "loss": 0.4051,
+      "step": 12170
+    },
+    {
+      "epoch": 78.58,
+      "learning_rate": 3.9290322580645164e-05,
+      "loss": 0.3457,
+      "step": 12180
+    },
+    {
+      "epoch": 78.64,
+      "learning_rate": 3.932258064516129e-05,
+      "loss": 0.3851,
+      "step": 12190
+    },
+    {
+      "epoch": 78.71,
+      "learning_rate": 3.935483870967742e-05,
+      "loss": 0.274,
+      "step": 12200
+    },
+    {
+      "epoch": 78.77,
+      "learning_rate": 3.938709677419355e-05,
+      "loss": 0.3241,
+      "step": 12210
+    },
+    {
+      "epoch": 78.84,
+      "learning_rate": 3.9419354838709674e-05,
+      "loss": 0.361,
+      "step": 12220
+    },
+    {
+      "epoch": 78.9,
+      "learning_rate": 3.945161290322581e-05,
+      "loss": 0.3436,
+      "step": 12230
+    },
+    {
+      "epoch": 78.96,
+      "learning_rate": 3.948387096774194e-05,
+      "loss": 0.3226,
+      "step": 12240
+    },
+    {
+      "epoch": 79.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0137066841125488,
+      "eval_runtime": 0.8675,
+      "eval_samples_per_second": 231.69,
+      "eval_steps_per_second": 8.069,
+      "step": 12245
+    },
+    {
+      "epoch": 79.03,
+      "learning_rate": 3.951612903225806e-05,
+      "loss": 0.3456,
+      "step": 12250
+    },
+    {
+      "epoch": 79.1,
+      "learning_rate": 3.95483870967742e-05,
+      "loss": 0.3234,
+      "step": 12260
+    },
+    {
+      "epoch": 79.16,
+      "learning_rate": 3.958064516129033e-05,
+      "loss": 0.3376,
+      "step": 12270
+    },
+    {
+      "epoch": 79.23,
+      "learning_rate": 3.961290322580645e-05,
+      "loss": 0.3336,
+      "step": 12280
+    },
+    {
+      "epoch": 79.29,
+      "learning_rate": 3.964516129032258e-05,
+      "loss": 0.3336,
+      "step": 12290
+    },
+    {
+      "epoch": 79.35,
+      "learning_rate": 3.9677419354838716e-05,
+      "loss": 0.3317,
+      "step": 12300
+    },
+    {
+      "epoch": 79.42,
+      "learning_rate": 3.970967741935484e-05,
+      "loss": 0.393,
+      "step": 12310
+    },
+    {
+      "epoch": 79.48,
+      "learning_rate": 3.974193548387097e-05,
+      "loss": 0.3051,
+      "step": 12320
+    },
+    {
+      "epoch": 79.55,
+      "learning_rate": 3.97741935483871e-05,
+      "loss": 0.3875,
+      "step": 12330
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 3.980645161290323e-05,
+      "loss": 0.349,
+      "step": 12340
+    },
+    {
+      "epoch": 79.68,
+      "learning_rate": 3.9838709677419356e-05,
+      "loss": 0.3823,
+      "step": 12350
+    },
+    {
+      "epoch": 79.74,
+      "learning_rate": 3.9870967741935486e-05,
+      "loss": 0.3586,
+      "step": 12360
+    },
+    {
+      "epoch": 79.8,
+      "learning_rate": 3.9903225806451615e-05,
+      "loss": 0.332,
+      "step": 12370
+    },
+    {
+      "epoch": 79.87,
+      "learning_rate": 3.9935483870967745e-05,
+      "loss": 0.3366,
+      "step": 12380
+    },
+    {
+      "epoch": 79.93,
+      "learning_rate": 3.9967741935483874e-05,
+      "loss": 0.3505,
+      "step": 12390
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 4e-05,
+      "loss": 0.4021,
+      "step": 12400
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.7263681592039801,
+      "eval_loss": 1.1070302724838257,
+      "eval_runtime": 0.8736,
+      "eval_samples_per_second": 230.09,
+      "eval_steps_per_second": 8.013,
+      "step": 12400
+    },
+    {
+      "epoch": 80.06,
+      "learning_rate": 4.003225806451613e-05,
+      "loss": 0.3494,
+      "step": 12410
+    },
+    {
+      "epoch": 80.13,
+      "learning_rate": 4.006451612903226e-05,
+      "loss": 0.3459,
+      "step": 12420
+    },
+    {
+      "epoch": 80.19,
+      "learning_rate": 4.0096774193548385e-05,
+      "loss": 0.3931,
+      "step": 12430
+    },
+    {
+      "epoch": 80.26,
+      "learning_rate": 4.012903225806452e-05,
+      "loss": 0.336,
+      "step": 12440
+    },
+    {
+      "epoch": 80.32,
+      "learning_rate": 4.016129032258065e-05,
+      "loss": 0.3636,
+      "step": 12450
+    },
+    {
+      "epoch": 80.39,
+      "learning_rate": 4.019354838709677e-05,
+      "loss": 0.3557,
+      "step": 12460
+    },
+    {
+      "epoch": 80.45,
+      "learning_rate": 4.022580645161291e-05,
+      "loss": 0.3293,
+      "step": 12470
+    },
+    {
+      "epoch": 80.51,
+      "learning_rate": 4.025806451612903e-05,
+      "loss": 0.329,
+      "step": 12480
+    },
+    {
+      "epoch": 80.58,
+      "learning_rate": 4.029032258064516e-05,
+      "loss": 0.3455,
+      "step": 12490
+    },
+    {
+      "epoch": 80.64,
+      "learning_rate": 4.032258064516129e-05,
+      "loss": 0.3919,
+      "step": 12500
+    },
+    {
+      "epoch": 80.71,
+      "learning_rate": 4.035483870967742e-05,
+      "loss": 0.4229,
+      "step": 12510
+    },
+    {
+      "epoch": 80.77,
+      "learning_rate": 4.038709677419355e-05,
+      "loss": 0.3888,
+      "step": 12520
+    },
+    {
+      "epoch": 80.84,
+      "learning_rate": 4.041935483870968e-05,
+      "loss": 0.3065,
+      "step": 12530
+    },
+    {
+      "epoch": 80.9,
+      "learning_rate": 4.045161290322581e-05,
+      "loss": 0.3433,
+      "step": 12540
+    },
+    {
+      "epoch": 80.96,
+      "learning_rate": 4.048387096774194e-05,
+      "loss": 0.3755,
+      "step": 12550
+    },
+    {
+      "epoch": 81.0,
+      "eval_accuracy": 0.7263681592039801,
+      "eval_loss": 1.112458348274231,
+      "eval_runtime": 0.8705,
+      "eval_samples_per_second": 230.898,
+      "eval_steps_per_second": 8.041,
+      "step": 12555
+    },
+    {
+      "epoch": 81.03,
+      "learning_rate": 4.0516129032258067e-05,
+      "loss": 0.3918,
+      "step": 12560
+    },
+    {
+      "epoch": 81.1,
+      "learning_rate": 4.0548387096774196e-05,
+      "loss": 0.3743,
+      "step": 12570
+    },
+    {
+      "epoch": 81.16,
+      "learning_rate": 4.0580645161290325e-05,
+      "loss": 0.4113,
+      "step": 12580
+    },
+    {
+      "epoch": 81.23,
+      "learning_rate": 4.0612903225806455e-05,
+      "loss": 0.2856,
+      "step": 12590
+    },
+    {
+      "epoch": 81.29,
+      "learning_rate": 4.0645161290322584e-05,
+      "loss": 0.3901,
+      "step": 12600
+    },
+    {
+      "epoch": 81.35,
+      "learning_rate": 4.0677419354838713e-05,
+      "loss": 0.3149,
+      "step": 12610
+    },
+    {
+      "epoch": 81.42,
+      "learning_rate": 4.070967741935484e-05,
+      "loss": 0.3391,
+      "step": 12620
+    },
+    {
+      "epoch": 81.48,
+      "learning_rate": 4.0741935483870965e-05,
+      "loss": 0.3502,
+      "step": 12630
+    },
+    {
+      "epoch": 81.55,
+      "learning_rate": 4.0774193548387095e-05,
+      "loss": 0.3728,
+      "step": 12640
+    },
+    {
+      "epoch": 81.61,
+      "learning_rate": 4.080645161290323e-05,
+      "loss": 0.3643,
+      "step": 12650
+    },
+    {
+      "epoch": 81.68,
+      "learning_rate": 4.0838709677419354e-05,
+      "loss": 0.3975,
+      "step": 12660
+    },
+    {
+      "epoch": 81.74,
+      "learning_rate": 4.087096774193548e-05,
+      "loss": 0.3036,
+      "step": 12670
+    },
+    {
+      "epoch": 81.8,
+      "learning_rate": 4.090322580645162e-05,
+      "loss": 0.3415,
+      "step": 12680
+    },
+    {
+      "epoch": 81.87,
+      "learning_rate": 4.093548387096774e-05,
+      "loss": 0.3781,
+      "step": 12690
+    },
+    {
+      "epoch": 81.93,
+      "learning_rate": 4.096774193548387e-05,
+      "loss": 0.3748,
+      "step": 12700
+    },
+    {
+      "epoch": 82.0,
+      "learning_rate": 4.1e-05,
+      "loss": 0.368,
+      "step": 12710
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0504125356674194,
+      "eval_runtime": 0.8708,
+      "eval_samples_per_second": 230.823,
+      "eval_steps_per_second": 8.039,
+      "step": 12710
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 4.103225806451613e-05,
+      "loss": 0.3762,
+      "step": 12720
+    },
+    {
+      "epoch": 82.13,
+      "learning_rate": 4.106451612903226e-05,
+      "loss": 0.333,
+      "step": 12730
+    },
+    {
+      "epoch": 82.19,
+      "learning_rate": 4.109677419354839e-05,
+      "loss": 0.4295,
+      "step": 12740
+    },
+    {
+      "epoch": 82.26,
+      "learning_rate": 4.112903225806452e-05,
+      "loss": 0.3731,
+      "step": 12750
+    },
+    {
+      "epoch": 82.32,
+      "learning_rate": 4.116129032258065e-05,
+      "loss": 0.3224,
+      "step": 12760
+    },
+    {
+      "epoch": 82.39,
+      "learning_rate": 4.119354838709678e-05,
+      "loss": 0.3756,
+      "step": 12770
+    },
+    {
+      "epoch": 82.45,
+      "learning_rate": 4.1225806451612906e-05,
+      "loss": 0.3151,
+      "step": 12780
+    },
+    {
+      "epoch": 82.51,
+      "learning_rate": 4.1258064516129036e-05,
+      "loss": 0.3175,
+      "step": 12790
+    },
+    {
+      "epoch": 82.58,
+      "learning_rate": 4.1290322580645165e-05,
+      "loss": 0.3503,
+      "step": 12800
+    },
+    {
+      "epoch": 82.64,
+      "learning_rate": 4.132258064516129e-05,
+      "loss": 0.3819,
+      "step": 12810
+    },
+    {
+      "epoch": 82.71,
+      "learning_rate": 4.1354838709677424e-05,
+      "loss": 0.3824,
+      "step": 12820
+    },
+    {
+      "epoch": 82.77,
+      "learning_rate": 4.138709677419355e-05,
+      "loss": 0.3282,
+      "step": 12830
+    },
+    {
+      "epoch": 82.84,
+      "learning_rate": 4.1419354838709676e-05,
+      "loss": 0.3464,
+      "step": 12840
+    },
+    {
+      "epoch": 82.9,
+      "learning_rate": 4.1451612903225805e-05,
+      "loss": 0.4278,
+      "step": 12850
+    },
+    {
+      "epoch": 82.96,
+      "learning_rate": 4.148387096774194e-05,
+      "loss": 0.3586,
+      "step": 12860
+    },
+    {
+      "epoch": 83.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0221822261810303,
+      "eval_runtime": 0.8477,
+      "eval_samples_per_second": 237.119,
+      "eval_steps_per_second": 8.258,
+      "step": 12865
+    },
+    {
+      "epoch": 83.03,
+      "learning_rate": 4.1516129032258064e-05,
+      "loss": 0.4008,
+      "step": 12870
+    },
+    {
+      "epoch": 83.1,
+      "learning_rate": 4.154838709677419e-05,
+      "loss": 0.3575,
+      "step": 12880
+    },
+    {
+      "epoch": 83.16,
+      "learning_rate": 4.158064516129033e-05,
+      "loss": 0.3427,
+      "step": 12890
+    },
+    {
+      "epoch": 83.23,
+      "learning_rate": 4.161290322580645e-05,
+      "loss": 0.3701,
+      "step": 12900
+    },
+    {
+      "epoch": 83.29,
+      "learning_rate": 4.164516129032258e-05,
+      "loss": 0.3186,
+      "step": 12910
+    },
+    {
+      "epoch": 83.35,
+      "learning_rate": 4.167741935483871e-05,
+      "loss": 0.4221,
+      "step": 12920
+    },
+    {
+      "epoch": 83.42,
+      "learning_rate": 4.170967741935484e-05,
+      "loss": 0.345,
+      "step": 12930
+    },
+    {
+      "epoch": 83.48,
+      "learning_rate": 4.174193548387097e-05,
+      "loss": 0.3711,
+      "step": 12940
+    },
+    {
+      "epoch": 83.55,
+      "learning_rate": 4.17741935483871e-05,
+      "loss": 0.349,
+      "step": 12950
+    },
+    {
+      "epoch": 83.61,
+      "learning_rate": 4.180645161290323e-05,
+      "loss": 0.3028,
+      "step": 12960
+    },
+    {
+      "epoch": 83.68,
+      "learning_rate": 4.183870967741936e-05,
+      "loss": 0.3421,
+      "step": 12970
+    },
+    {
+      "epoch": 83.74,
+      "learning_rate": 4.187096774193549e-05,
+      "loss": 0.3625,
+      "step": 12980
+    },
+    {
+      "epoch": 83.8,
+      "learning_rate": 4.190322580645161e-05,
+      "loss": 0.3203,
+      "step": 12990
+    },
+    {
+      "epoch": 83.87,
+      "learning_rate": 4.1935483870967746e-05,
+      "loss": 0.3309,
+      "step": 13000
+    },
+    {
+      "epoch": 83.93,
+      "learning_rate": 4.1967741935483875e-05,
+      "loss": 0.361,
+      "step": 13010
+    },
+    {
+      "epoch": 84.0,
+      "learning_rate": 4.2e-05,
+      "loss": 0.3566,
+      "step": 13020
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9528164863586426,
+      "eval_runtime": 0.8674,
+      "eval_samples_per_second": 231.723,
+      "eval_steps_per_second": 8.07,
+      "step": 13020
+    },
+    {
+      "epoch": 84.06,
+      "learning_rate": 4.2032258064516134e-05,
+      "loss": 0.3768,
+      "step": 13030
+    },
+    {
+      "epoch": 84.13,
+      "learning_rate": 4.206451612903226e-05,
+      "loss": 0.3238,
+      "step": 13040
+    },
+    {
+      "epoch": 84.19,
+      "learning_rate": 4.2096774193548386e-05,
+      "loss": 0.3913,
+      "step": 13050
+    },
+    {
+      "epoch": 84.26,
+      "learning_rate": 4.2129032258064515e-05,
+      "loss": 0.415,
+      "step": 13060
+    },
+    {
+      "epoch": 84.32,
+      "learning_rate": 4.216129032258065e-05,
+      "loss": 0.3353,
+      "step": 13070
+    },
+    {
+      "epoch": 84.39,
+      "learning_rate": 4.2193548387096774e-05,
+      "loss": 0.3486,
+      "step": 13080
+    },
+    {
+      "epoch": 84.45,
+      "learning_rate": 4.22258064516129e-05,
+      "loss": 0.3455,
+      "step": 13090
+    },
+    {
+      "epoch": 84.51,
+      "learning_rate": 4.225806451612904e-05,
+      "loss": 0.3078,
+      "step": 13100
+    },
+    {
+      "epoch": 84.58,
+      "learning_rate": 4.229032258064516e-05,
+      "loss": 0.3398,
+      "step": 13110
+    },
+    {
+      "epoch": 84.64,
+      "learning_rate": 4.232258064516129e-05,
+      "loss": 0.2944,
+      "step": 13120
+    },
+    {
+      "epoch": 84.71,
+      "learning_rate": 4.235483870967742e-05,
+      "loss": 0.342,
+      "step": 13130
+    },
+    {
+      "epoch": 84.77,
+      "learning_rate": 4.238709677419355e-05,
+      "loss": 0.3756,
+      "step": 13140
+    },
+    {
+      "epoch": 84.84,
+      "learning_rate": 4.241935483870968e-05,
+      "loss": 0.3804,
+      "step": 13150
+    },
+    {
+      "epoch": 84.9,
+      "learning_rate": 4.245161290322581e-05,
+      "loss": 0.3909,
+      "step": 13160
+    },
+    {
+      "epoch": 84.96,
+      "learning_rate": 4.248387096774194e-05,
+      "loss": 0.3637,
+      "step": 13170
+    },
+    {
+      "epoch": 85.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9612098932266235,
+      "eval_runtime": 0.8617,
+      "eval_samples_per_second": 233.252,
+      "eval_steps_per_second": 8.123,
+      "step": 13175
+    },
+    {
+      "epoch": 85.03,
+      "learning_rate": 4.251612903225807e-05,
+      "loss": 0.3196,
+      "step": 13180
+    },
+    {
+      "epoch": 85.1,
+      "learning_rate": 4.25483870967742e-05,
+      "loss": 0.322,
+      "step": 13190
+    },
+    {
+      "epoch": 85.16,
+      "learning_rate": 4.258064516129032e-05,
+      "loss": 0.3515,
+      "step": 13200
+    },
+    {
+      "epoch": 85.23,
+      "learning_rate": 4.2612903225806456e-05,
+      "loss": 0.3309,
+      "step": 13210
+    },
+    {
+      "epoch": 85.29,
+      "learning_rate": 4.2645161290322585e-05,
+      "loss": 0.3996,
+      "step": 13220
+    },
+    {
+      "epoch": 85.35,
+      "learning_rate": 4.267741935483871e-05,
+      "loss": 0.3378,
+      "step": 13230
+    },
+    {
+      "epoch": 85.42,
+      "learning_rate": 4.2709677419354844e-05,
+      "loss": 0.3121,
+      "step": 13240
+    },
+    {
+      "epoch": 85.48,
+      "learning_rate": 4.2741935483870973e-05,
+      "loss": 0.3302,
+      "step": 13250
+    },
+    {
+      "epoch": 85.55,
+      "learning_rate": 4.2774193548387096e-05,
+      "loss": 0.373,
+      "step": 13260
+    },
+    {
+      "epoch": 85.61,
+      "learning_rate": 4.2806451612903225e-05,
+      "loss": 0.3599,
+      "step": 13270
+    },
+    {
+      "epoch": 85.68,
+      "learning_rate": 4.283870967741936e-05,
+      "loss": 0.3727,
+      "step": 13280
+    },
+    {
+      "epoch": 85.74,
+      "learning_rate": 4.2870967741935484e-05,
+      "loss": 0.3854,
+      "step": 13290
+    },
+    {
+      "epoch": 85.8,
+      "learning_rate": 4.2903225806451614e-05,
+      "loss": 0.3386,
+      "step": 13300
+    },
+    {
+      "epoch": 85.87,
+      "learning_rate": 4.293548387096775e-05,
+      "loss": 0.351,
+      "step": 13310
+    },
+    {
+      "epoch": 85.93,
+      "learning_rate": 4.296774193548387e-05,
+      "loss": 0.3614,
+      "step": 13320
+    },
+    {
+      "epoch": 86.0,
+      "learning_rate": 4.3e-05,
+      "loss": 0.3243,
+      "step": 13330
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0125961303710938,
+      "eval_runtime": 0.8517,
+      "eval_samples_per_second": 235.987,
+      "eval_steps_per_second": 8.218,
+      "step": 13330
+    },
+    {
+      "epoch": 86.06,
+      "learning_rate": 4.303225806451613e-05,
+      "loss": 0.3936,
+      "step": 13340
+    },
+    {
+      "epoch": 86.13,
+      "learning_rate": 4.306451612903226e-05,
+      "loss": 0.3735,
+      "step": 13350
+    },
+    {
+      "epoch": 86.19,
+      "learning_rate": 4.309677419354839e-05,
+      "loss": 0.4335,
+      "step": 13360
+    },
+    {
+      "epoch": 86.26,
+      "learning_rate": 4.312903225806452e-05,
+      "loss": 0.3528,
+      "step": 13370
+    },
+    {
+      "epoch": 86.32,
+      "learning_rate": 4.316129032258065e-05,
+      "loss": 0.369,
+      "step": 13380
+    },
+    {
+      "epoch": 86.39,
+      "learning_rate": 4.319354838709678e-05,
+      "loss": 0.4021,
+      "step": 13390
+    },
+    {
+      "epoch": 86.45,
+      "learning_rate": 4.322580645161291e-05,
+      "loss": 0.3534,
+      "step": 13400
+    },
+    {
+      "epoch": 86.51,
+      "learning_rate": 4.325806451612903e-05,
+      "loss": 0.3374,
+      "step": 13410
+    },
+    {
+      "epoch": 86.58,
+      "learning_rate": 4.3290322580645166e-05,
+      "loss": 0.3489,
+      "step": 13420
+    },
+    {
+      "epoch": 86.64,
+      "learning_rate": 4.3322580645161295e-05,
+      "loss": 0.3499,
+      "step": 13430
+    },
+    {
+      "epoch": 86.71,
+      "learning_rate": 4.335483870967742e-05,
+      "loss": 0.363,
+      "step": 13440
+    },
+    {
+      "epoch": 86.77,
+      "learning_rate": 4.3387096774193554e-05,
+      "loss": 0.3899,
+      "step": 13450
+    },
+    {
+      "epoch": 86.84,
+      "learning_rate": 4.3419354838709684e-05,
+      "loss": 0.3697,
+      "step": 13460
+    },
+    {
+      "epoch": 86.9,
+      "learning_rate": 4.3451612903225806e-05,
+      "loss": 0.3616,
+      "step": 13470
+    },
+    {
+      "epoch": 86.96,
+      "learning_rate": 4.3483870967741936e-05,
+      "loss": 0.3575,
+      "step": 13480
+    },
+    {
+      "epoch": 87.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0440675020217896,
+      "eval_runtime": 0.8732,
+      "eval_samples_per_second": 230.191,
+      "eval_steps_per_second": 8.017,
+      "step": 13485
+    },
+    {
+      "epoch": 87.03,
+      "learning_rate": 4.351612903225807e-05,
+      "loss": 0.4083,
+      "step": 13490
+    },
+    {
+      "epoch": 87.1,
+      "learning_rate": 4.3548387096774194e-05,
+      "loss": 0.3125,
+      "step": 13500
+    },
+    {
+      "epoch": 87.16,
+      "learning_rate": 4.3580645161290324e-05,
+      "loss": 0.3944,
+      "step": 13510
+    },
+    {
+      "epoch": 87.23,
+      "learning_rate": 4.361290322580645e-05,
+      "loss": 0.3516,
+      "step": 13520
+    },
+    {
+      "epoch": 87.29,
+      "learning_rate": 4.364516129032258e-05,
+      "loss": 0.3559,
+      "step": 13530
+    },
+    {
+      "epoch": 87.35,
+      "learning_rate": 4.367741935483871e-05,
+      "loss": 0.3458,
+      "step": 13540
+    },
+    {
+      "epoch": 87.42,
+      "learning_rate": 4.370967741935484e-05,
+      "loss": 0.38,
+      "step": 13550
+    },
+    {
+      "epoch": 87.48,
+      "learning_rate": 4.374193548387097e-05,
+      "loss": 0.3346,
+      "step": 13560
+    },
+    {
+      "epoch": 87.55,
+      "learning_rate": 4.37741935483871e-05,
+      "loss": 0.3177,
+      "step": 13570
+    },
+    {
+      "epoch": 87.61,
+      "learning_rate": 4.380645161290323e-05,
+      "loss": 0.4357,
+      "step": 13580
+    },
+    {
+      "epoch": 87.68,
+      "learning_rate": 4.383870967741936e-05,
+      "loss": 0.3269,
+      "step": 13590
+    },
+    {
+      "epoch": 87.74,
+      "learning_rate": 4.387096774193549e-05,
+      "loss": 0.3971,
+      "step": 13600
+    },
+    {
+      "epoch": 87.8,
+      "learning_rate": 4.390322580645162e-05,
+      "loss": 0.3313,
+      "step": 13610
+    },
+    {
+      "epoch": 87.87,
+      "learning_rate": 4.393548387096774e-05,
+      "loss": 0.3242,
+      "step": 13620
+    },
+    {
+      "epoch": 87.93,
+      "learning_rate": 4.3967741935483876e-05,
+      "loss": 0.3152,
+      "step": 13630
+    },
+    {
+      "epoch": 88.0,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.3586,
+      "step": 13640
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0270942449569702,
+      "eval_runtime": 0.8719,
+      "eval_samples_per_second": 230.537,
+      "eval_steps_per_second": 8.029,
+      "step": 13640
+    },
+    {
+      "epoch": 88.06,
+      "learning_rate": 4.403225806451613e-05,
+      "loss": 0.336,
+      "step": 13650
+    },
+    {
+      "epoch": 88.13,
+      "learning_rate": 4.4064516129032264e-05,
+      "loss": 0.3088,
+      "step": 13660
+    },
+    {
+      "epoch": 88.19,
+      "learning_rate": 4.409677419354839e-05,
+      "loss": 0.3487,
+      "step": 13670
+    },
+    {
+      "epoch": 88.26,
+      "learning_rate": 4.4129032258064516e-05,
+      "loss": 0.3652,
+      "step": 13680
+    },
+    {
+      "epoch": 88.32,
+      "learning_rate": 4.4161290322580646e-05,
+      "loss": 0.4336,
+      "step": 13690
+    },
+    {
+      "epoch": 88.39,
+      "learning_rate": 4.4193548387096775e-05,
+      "loss": 0.3482,
+      "step": 13700
+    },
+    {
+      "epoch": 88.45,
+      "learning_rate": 4.4225806451612905e-05,
+      "loss": 0.3209,
+      "step": 13710
+    },
+    {
+      "epoch": 88.51,
+      "learning_rate": 4.4258064516129034e-05,
+      "loss": 0.3689,
+      "step": 13720
+    },
+    {
+      "epoch": 88.58,
+      "learning_rate": 4.429032258064516e-05,
+      "loss": 0.3982,
+      "step": 13730
+    },
+    {
+      "epoch": 88.64,
+      "learning_rate": 4.432258064516129e-05,
+      "loss": 0.3388,
+      "step": 13740
+    },
+    {
+      "epoch": 88.71,
+      "learning_rate": 4.435483870967742e-05,
+      "loss": 0.3962,
+      "step": 13750
+    },
+    {
+      "epoch": 88.77,
+      "learning_rate": 4.438709677419355e-05,
+      "loss": 0.3796,
+      "step": 13760
+    },
+    {
+      "epoch": 88.84,
+      "learning_rate": 4.441935483870968e-05,
+      "loss": 0.2861,
+      "step": 13770
+    },
+    {
+      "epoch": 88.9,
+      "learning_rate": 4.445161290322581e-05,
+      "loss": 0.3692,
+      "step": 13780
+    },
+    {
+      "epoch": 88.96,
+      "learning_rate": 4.448387096774194e-05,
+      "loss": 0.3414,
+      "step": 13790
+    },
+    {
+      "epoch": 89.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9725929498672485,
+      "eval_runtime": 0.8558,
+      "eval_samples_per_second": 234.856,
+      "eval_steps_per_second": 8.179,
+      "step": 13795
+    },
+    {
+      "epoch": 89.03,
+      "learning_rate": 4.451612903225807e-05,
+      "loss": 0.3929,
+      "step": 13800
+    },
+    {
+      "epoch": 89.1,
+      "learning_rate": 4.45483870967742e-05,
+      "loss": 0.3106,
+      "step": 13810
+    },
+    {
+      "epoch": 89.16,
+      "learning_rate": 4.458064516129032e-05,
+      "loss": 0.4012,
+      "step": 13820
+    },
+    {
+      "epoch": 89.23,
+      "learning_rate": 4.461290322580645e-05,
+      "loss": 0.3569,
+      "step": 13830
+    },
+    {
+      "epoch": 89.29,
+      "learning_rate": 4.4645161290322586e-05,
+      "loss": 0.3562,
+      "step": 13840
+    },
+    {
+      "epoch": 89.35,
+      "learning_rate": 4.467741935483871e-05,
+      "loss": 0.3507,
+      "step": 13850
+    },
+    {
+      "epoch": 89.42,
+      "learning_rate": 4.470967741935484e-05,
+      "loss": 0.3701,
+      "step": 13860
+    },
+    {
+      "epoch": 89.48,
+      "learning_rate": 4.4741935483870975e-05,
+      "loss": 0.3355,
+      "step": 13870
+    },
+    {
+      "epoch": 89.55,
+      "learning_rate": 4.47741935483871e-05,
+      "loss": 0.3575,
+      "step": 13880
+    },
+    {
+      "epoch": 89.61,
+      "learning_rate": 4.4806451612903227e-05,
+      "loss": 0.3658,
+      "step": 13890
+    },
+    {
+      "epoch": 89.68,
+      "learning_rate": 4.4838709677419356e-05,
+      "loss": 0.3935,
+      "step": 13900
+    },
+    {
+      "epoch": 89.74,
+      "learning_rate": 4.4870967741935485e-05,
+      "loss": 0.3098,
+      "step": 13910
+    },
+    {
+      "epoch": 89.8,
+      "learning_rate": 4.4903225806451615e-05,
+      "loss": 0.3995,
+      "step": 13920
+    },
+    {
+      "epoch": 89.87,
+      "learning_rate": 4.4935483870967744e-05,
+      "loss": 0.3461,
+      "step": 13930
+    },
+    {
+      "epoch": 89.93,
+      "learning_rate": 4.4967741935483873e-05,
+      "loss": 0.3746,
+      "step": 13940
+    },
+    {
+      "epoch": 90.0,
+      "learning_rate": 4.5e-05,
+      "loss": 0.357,
+      "step": 13950
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9348400235176086,
+      "eval_runtime": 0.8492,
+      "eval_samples_per_second": 236.689,
+      "eval_steps_per_second": 8.243,
+      "step": 13950
+    },
+    {
+      "epoch": 90.06,
+      "learning_rate": 4.503225806451613e-05,
+      "loss": 0.3163,
+      "step": 13960
+    },
+    {
+      "epoch": 90.13,
+      "learning_rate": 4.5064516129032255e-05,
+      "loss": 0.4242,
+      "step": 13970
+    },
+    {
+      "epoch": 90.19,
+      "learning_rate": 4.509677419354839e-05,
+      "loss": 0.345,
+      "step": 13980
+    },
+    {
+      "epoch": 90.26,
+      "learning_rate": 4.512903225806452e-05,
+      "loss": 0.3884,
+      "step": 13990
+    },
+    {
+      "epoch": 90.32,
+      "learning_rate": 4.516129032258064e-05,
+      "loss": 0.3673,
+      "step": 14000
+    },
+    {
+      "epoch": 90.39,
+      "learning_rate": 4.519354838709678e-05,
+      "loss": 0.3587,
+      "step": 14010
+    },
+    {
+      "epoch": 90.45,
+      "learning_rate": 4.522580645161291e-05,
+      "loss": 0.3638,
+      "step": 14020
+    },
+    {
+      "epoch": 90.51,
+      "learning_rate": 4.525806451612903e-05,
+      "loss": 0.3798,
+      "step": 14030
+    },
+    {
+      "epoch": 90.58,
+      "learning_rate": 4.529032258064516e-05,
+      "loss": 0.3646,
+      "step": 14040
+    },
+    {
+      "epoch": 90.64,
+      "learning_rate": 4.53225806451613e-05,
+      "loss": 0.3516,
+      "step": 14050
+    },
+    {
+      "epoch": 90.71,
+      "learning_rate": 4.535483870967742e-05,
+      "loss": 0.4118,
+      "step": 14060
+    },
+    {
+      "epoch": 90.77,
+      "learning_rate": 4.538709677419355e-05,
+      "loss": 0.3555,
+      "step": 14070
+    },
+    {
+      "epoch": 90.84,
+      "learning_rate": 4.5419354838709685e-05,
+      "loss": 0.3753,
+      "step": 14080
+    },
+    {
+      "epoch": 90.9,
+      "learning_rate": 4.545161290322581e-05,
+      "loss": 0.3584,
+      "step": 14090
+    },
+    {
+      "epoch": 90.96,
+      "learning_rate": 4.548387096774194e-05,
+      "loss": 0.3762,
+      "step": 14100
+    },
+    {
+      "epoch": 91.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9526757597923279,
+      "eval_runtime": 0.8506,
+      "eval_samples_per_second": 236.312,
+      "eval_steps_per_second": 8.23,
+      "step": 14105
+    },
+    {
+      "epoch": 91.03,
+      "learning_rate": 4.5516129032258066e-05,
+      "loss": 0.3782,
+      "step": 14110
+    },
+    {
+      "epoch": 91.1,
+      "learning_rate": 4.5548387096774196e-05,
+      "loss": 0.3652,
+      "step": 14120
+    },
+    {
+      "epoch": 91.16,
+      "learning_rate": 4.5580645161290325e-05,
+      "loss": 0.3598,
+      "step": 14130
+    },
+    {
+      "epoch": 91.23,
+      "learning_rate": 4.5612903225806454e-05,
+      "loss": 0.3805,
+      "step": 14140
+    },
+    {
+      "epoch": 91.29,
+      "learning_rate": 4.5645161290322584e-05,
+      "loss": 0.3363,
+      "step": 14150
+    },
+    {
+      "epoch": 91.35,
+      "learning_rate": 4.567741935483871e-05,
+      "loss": 0.3683,
+      "step": 14160
+    },
+    {
+      "epoch": 91.42,
+      "learning_rate": 4.570967741935484e-05,
+      "loss": 0.3797,
+      "step": 14170
+    },
+    {
+      "epoch": 91.48,
+      "learning_rate": 4.5741935483870965e-05,
+      "loss": 0.3636,
+      "step": 14180
+    },
+    {
+      "epoch": 91.55,
+      "learning_rate": 4.57741935483871e-05,
+      "loss": 0.3704,
+      "step": 14190
+    },
+    {
+      "epoch": 91.61,
+      "learning_rate": 4.580645161290323e-05,
+      "loss": 0.4015,
+      "step": 14200
+    },
+    {
+      "epoch": 91.68,
+      "learning_rate": 4.583870967741935e-05,
+      "loss": 0.3636,
+      "step": 14210
+    },
+    {
+      "epoch": 91.74,
+      "learning_rate": 4.587096774193548e-05,
+      "loss": 0.3902,
+      "step": 14220
+    },
+    {
+      "epoch": 91.8,
+      "learning_rate": 4.590322580645162e-05,
+      "loss": 0.3621,
+      "step": 14230
+    },
+    {
+      "epoch": 91.87,
+      "learning_rate": 4.593548387096774e-05,
+      "loss": 0.3475,
+      "step": 14240
+    },
+    {
+      "epoch": 91.93,
+      "learning_rate": 4.596774193548387e-05,
+      "loss": 0.4021,
+      "step": 14250
+    },
+    {
+      "epoch": 92.0,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.3729,
+      "step": 14260
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9189078211784363,
+      "eval_runtime": 0.8434,
+      "eval_samples_per_second": 238.333,
+      "eval_steps_per_second": 8.3,
+      "step": 14260
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 4.603225806451613e-05,
+      "loss": 0.3796,
+      "step": 14270
+    },
+    {
+      "epoch": 92.13,
+      "learning_rate": 4.606451612903226e-05,
+      "loss": 0.3258,
+      "step": 14280
+    },
+    {
+      "epoch": 92.19,
+      "learning_rate": 4.609677419354839e-05,
+      "loss": 0.3961,
+      "step": 14290
+    },
+    {
+      "epoch": 92.26,
+      "learning_rate": 4.612903225806452e-05,
+      "loss": 0.3794,
+      "step": 14300
+    },
+    {
+      "epoch": 92.32,
+      "learning_rate": 4.616129032258065e-05,
+      "loss": 0.3831,
+      "step": 14310
+    },
+    {
+      "epoch": 92.39,
+      "learning_rate": 4.6193548387096776e-05,
+      "loss": 0.3228,
+      "step": 14320
+    },
+    {
+      "epoch": 92.45,
+      "learning_rate": 4.6225806451612906e-05,
+      "loss": 0.3648,
+      "step": 14330
+    },
+    {
+      "epoch": 92.51,
+      "learning_rate": 4.6258064516129035e-05,
+      "loss": 0.3778,
+      "step": 14340
+    },
+    {
+      "epoch": 92.58,
+      "learning_rate": 4.6290322580645164e-05,
+      "loss": 0.3743,
+      "step": 14350
+    },
+    {
+      "epoch": 92.64,
+      "learning_rate": 4.632258064516129e-05,
+      "loss": 0.3706,
+      "step": 14360
+    },
+    {
+      "epoch": 92.71,
+      "learning_rate": 4.635483870967742e-05,
+      "loss": 0.3798,
+      "step": 14370
+    },
+    {
+      "epoch": 92.77,
+      "learning_rate": 4.638709677419355e-05,
+      "loss": 0.3741,
+      "step": 14380
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 4.6419354838709675e-05,
+      "loss": 0.3148,
+      "step": 14390
+    },
+    {
+      "epoch": 92.9,
+      "learning_rate": 4.645161290322581e-05,
+      "loss": 0.3719,
+      "step": 14400
+    },
+    {
+      "epoch": 92.96,
+      "learning_rate": 4.648387096774194e-05,
+      "loss": 0.4242,
+      "step": 14410
+    },
+    {
+      "epoch": 93.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9253067374229431,
+      "eval_runtime": 0.8838,
+      "eval_samples_per_second": 227.417,
+      "eval_steps_per_second": 7.92,
+      "step": 14415
+    },
+    {
+      "epoch": 93.03,
+      "learning_rate": 4.651612903225806e-05,
+      "loss": 0.3926,
+      "step": 14420
+    },
+    {
+      "epoch": 93.1,
+      "learning_rate": 4.654838709677419e-05,
+      "loss": 0.3171,
+      "step": 14430
+    },
+    {
+      "epoch": 93.16,
+      "learning_rate": 4.658064516129033e-05,
+      "loss": 0.3432,
+      "step": 14440
+    },
+    {
+      "epoch": 93.23,
+      "learning_rate": 4.661290322580645e-05,
+      "loss": 0.4105,
+      "step": 14450
+    },
+    {
+      "epoch": 93.29,
+      "learning_rate": 4.664516129032258e-05,
+      "loss": 0.347,
+      "step": 14460
+    },
+    {
+      "epoch": 93.35,
+      "learning_rate": 4.667741935483872e-05,
+      "loss": 0.3932,
+      "step": 14470
+    },
+    {
+      "epoch": 93.42,
+      "learning_rate": 4.670967741935484e-05,
+      "loss": 0.3361,
+      "step": 14480
+    },
+    {
+      "epoch": 93.48,
+      "learning_rate": 4.674193548387097e-05,
+      "loss": 0.3673,
+      "step": 14490
+    },
+    {
+      "epoch": 93.55,
+      "learning_rate": 4.67741935483871e-05,
+      "loss": 0.3742,
+      "step": 14500
+    },
+    {
+      "epoch": 93.61,
+      "learning_rate": 4.680645161290323e-05,
+      "loss": 0.3758,
+      "step": 14510
+    },
+    {
+      "epoch": 93.68,
+      "learning_rate": 4.683870967741936e-05,
+      "loss": 0.3257,
+      "step": 14520
+    },
+    {
+      "epoch": 93.74,
+      "learning_rate": 4.6870967741935486e-05,
+      "loss": 0.3739,
+      "step": 14530
+    },
+    {
+      "epoch": 93.8,
+      "learning_rate": 4.6903225806451616e-05,
+      "loss": 0.4113,
+      "step": 14540
+    },
+    {
+      "epoch": 93.87,
+      "learning_rate": 4.6935483870967745e-05,
+      "loss": 0.4127,
+      "step": 14550
+    },
+    {
+      "epoch": 93.93,
+      "learning_rate": 4.6967741935483875e-05,
+      "loss": 0.3947,
+      "step": 14560
+    },
+    {
+      "epoch": 94.0,
+      "learning_rate": 4.7e-05,
+      "loss": 0.3004,
+      "step": 14570
+    },
+    {
+      "epoch": 94.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9500876665115356,
+      "eval_runtime": 0.8445,
+      "eval_samples_per_second": 238.015,
+      "eval_steps_per_second": 8.289,
+      "step": 14570
+    },
+    {
+      "epoch": 94.06,
+      "learning_rate": 4.7032258064516133e-05,
+      "loss": 0.3152,
+      "step": 14580
+    },
+    {
+      "epoch": 94.13,
+      "learning_rate": 4.706451612903226e-05,
+      "loss": 0.3173,
+      "step": 14590
+    },
+    {
+      "epoch": 94.19,
+      "learning_rate": 4.7096774193548385e-05,
+      "loss": 0.379,
+      "step": 14600
+    },
+    {
+      "epoch": 94.26,
+      "learning_rate": 4.712903225806452e-05,
+      "loss": 0.3597,
+      "step": 14610
+    },
+    {
+      "epoch": 94.32,
+      "learning_rate": 4.716129032258065e-05,
+      "loss": 0.3671,
+      "step": 14620
+    },
+    {
+      "epoch": 94.39,
+      "learning_rate": 4.7193548387096774e-05,
+      "loss": 0.4464,
+      "step": 14630
+    },
+    {
+      "epoch": 94.45,
+      "learning_rate": 4.72258064516129e-05,
+      "loss": 0.4043,
+      "step": 14640
+    },
+    {
+      "epoch": 94.51,
+      "learning_rate": 4.725806451612904e-05,
+      "loss": 0.338,
+      "step": 14650
+    },
+    {
+      "epoch": 94.58,
+      "learning_rate": 4.729032258064516e-05,
+      "loss": 0.3525,
+      "step": 14660
+    },
+    {
+      "epoch": 94.64,
+      "learning_rate": 4.732258064516129e-05,
+      "loss": 0.3923,
+      "step": 14670
+    },
+    {
+      "epoch": 94.71,
+      "learning_rate": 4.735483870967742e-05,
+      "loss": 0.3438,
+      "step": 14680
+    },
+    {
+      "epoch": 94.77,
+      "learning_rate": 4.738709677419355e-05,
+      "loss": 0.3231,
+      "step": 14690
+    },
+    {
+      "epoch": 94.84,
+      "learning_rate": 4.741935483870968e-05,
+      "loss": 0.3588,
+      "step": 14700
+    },
+    {
+      "epoch": 94.9,
+      "learning_rate": 4.745161290322581e-05,
+      "loss": 0.3263,
+      "step": 14710
+    },
+    {
+      "epoch": 94.96,
+      "learning_rate": 4.748387096774194e-05,
+      "loss": 0.3884,
+      "step": 14720
+    },
+    {
+      "epoch": 95.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.9513483047485352,
+      "eval_runtime": 0.8464,
+      "eval_samples_per_second": 237.472,
+      "eval_steps_per_second": 8.27,
+      "step": 14725
+    },
+    {
+      "epoch": 95.03,
+      "learning_rate": 4.751612903225807e-05,
+      "loss": 0.4073,
+      "step": 14730
+    },
+    {
+      "epoch": 95.1,
+      "learning_rate": 4.75483870967742e-05,
+      "loss": 0.4089,
+      "step": 14740
+    },
+    {
+      "epoch": 95.16,
+      "learning_rate": 4.7580645161290326e-05,
+      "loss": 0.4172,
+      "step": 14750
+    },
+    {
+      "epoch": 95.23,
+      "learning_rate": 4.7612903225806455e-05,
+      "loss": 0.3039,
+      "step": 14760
+    },
+    {
+      "epoch": 95.29,
+      "learning_rate": 4.7645161290322585e-05,
+      "loss": 0.412,
+      "step": 14770
+    },
+    {
+      "epoch": 95.35,
+      "learning_rate": 4.767741935483871e-05,
+      "loss": 0.3296,
+      "step": 14780
+    },
+    {
+      "epoch": 95.42,
+      "learning_rate": 4.7709677419354844e-05,
+      "loss": 0.3381,
+      "step": 14790
+    },
+    {
+      "epoch": 95.48,
+      "learning_rate": 4.774193548387097e-05,
+      "loss": 0.333,
+      "step": 14800
+    },
+    {
+      "epoch": 95.55,
+      "learning_rate": 4.7774193548387096e-05,
+      "loss": 0.3924,
+      "step": 14810
+    },
+    {
+      "epoch": 95.61,
+      "learning_rate": 4.780645161290323e-05,
+      "loss": 0.2816,
+      "step": 14820
+    },
+    {
+      "epoch": 95.68,
+      "learning_rate": 4.783870967741936e-05,
+      "loss": 0.364,
+      "step": 14830
+    },
+    {
+      "epoch": 95.74,
+      "learning_rate": 4.7870967741935484e-05,
+      "loss": 0.3487,
+      "step": 14840
+    },
+    {
+      "epoch": 95.8,
+      "learning_rate": 4.790322580645161e-05,
+      "loss": 0.322,
+      "step": 14850
+    },
+    {
+      "epoch": 95.87,
+      "learning_rate": 4.793548387096774e-05,
+      "loss": 0.3789,
+      "step": 14860
+    },
+    {
+      "epoch": 95.93,
+      "learning_rate": 4.796774193548387e-05,
+      "loss": 0.341,
+      "step": 14870
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 4.8e-05,
+      "loss": 0.3609,
+      "step": 14880
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9719735383987427,
+      "eval_runtime": 0.8415,
+      "eval_samples_per_second": 238.855,
+      "eval_steps_per_second": 8.318,
+      "step": 14880
+    },
+    {
+      "epoch": 96.06,
+      "learning_rate": 4.803225806451613e-05,
+      "loss": 0.4142,
+      "step": 14890
+    },
+    {
+      "epoch": 96.13,
+      "learning_rate": 4.806451612903226e-05,
+      "loss": 0.3119,
+      "step": 14900
+    },
+    {
+      "epoch": 96.19,
+      "learning_rate": 4.809677419354839e-05,
+      "loss": 0.3851,
+      "step": 14910
+    },
+    {
+      "epoch": 96.26,
+      "learning_rate": 4.812903225806452e-05,
+      "loss": 0.2969,
+      "step": 14920
+    },
+    {
+      "epoch": 96.32,
+      "learning_rate": 4.816129032258065e-05,
+      "loss": 0.3722,
+      "step": 14930
+    },
+    {
+      "epoch": 96.39,
+      "learning_rate": 4.819354838709678e-05,
+      "loss": 0.365,
+      "step": 14940
+    },
+    {
+      "epoch": 96.45,
+      "learning_rate": 4.822580645161291e-05,
+      "loss": 0.332,
+      "step": 14950
+    },
+    {
+      "epoch": 96.51,
+      "learning_rate": 4.8258064516129036e-05,
+      "loss": 0.3437,
+      "step": 14960
+    },
+    {
+      "epoch": 96.58,
+      "learning_rate": 4.8290322580645166e-05,
+      "loss": 0.3847,
+      "step": 14970
+    },
+    {
+      "epoch": 96.64,
+      "learning_rate": 4.8322580645161295e-05,
+      "loss": 0.3087,
+      "step": 14980
+    },
+    {
+      "epoch": 96.71,
+      "learning_rate": 4.835483870967742e-05,
+      "loss": 0.3769,
+      "step": 14990
+    },
+    {
+      "epoch": 96.77,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 0.3504,
+      "step": 15000
+    },
+    {
+      "epoch": 96.84,
+      "learning_rate": 4.8419354838709676e-05,
+      "loss": 0.3381,
+      "step": 15010
+    },
+    {
+      "epoch": 96.9,
+      "learning_rate": 4.8451612903225806e-05,
+      "loss": 0.2908,
+      "step": 15020
+    },
+    {
+      "epoch": 96.96,
+      "learning_rate": 4.848387096774194e-05,
+      "loss": 0.3036,
+      "step": 15030
+    },
+    {
+      "epoch": 97.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9717298150062561,
+      "eval_runtime": 0.8815,
+      "eval_samples_per_second": 228.014,
+      "eval_steps_per_second": 7.941,
+      "step": 15035
+    },
+    {
+      "epoch": 97.03,
+      "learning_rate": 4.8516129032258065e-05,
+      "loss": 0.4189,
+      "step": 15040
+    },
+    {
+      "epoch": 97.1,
+      "learning_rate": 4.8548387096774194e-05,
+      "loss": 0.3834,
+      "step": 15050
+    },
+    {
+      "epoch": 97.16,
+      "learning_rate": 4.858064516129032e-05,
+      "loss": 0.3578,
+      "step": 15060
+    },
+    {
+      "epoch": 97.23,
+      "learning_rate": 4.861290322580645e-05,
+      "loss": 0.3518,
+      "step": 15070
+    },
+    {
+      "epoch": 97.29,
+      "learning_rate": 4.864516129032258e-05,
+      "loss": 0.3052,
+      "step": 15080
+    },
+    {
+      "epoch": 97.35,
+      "learning_rate": 4.867741935483871e-05,
+      "loss": 0.3494,
+      "step": 15090
+    },
+    {
+      "epoch": 97.42,
+      "learning_rate": 4.870967741935484e-05,
+      "loss": 0.3428,
+      "step": 15100
+    },
+    {
+      "epoch": 97.48,
+      "learning_rate": 4.874193548387097e-05,
+      "loss": 0.3646,
+      "step": 15110
+    },
+    {
+      "epoch": 97.55,
+      "learning_rate": 4.87741935483871e-05,
+      "loss": 0.3566,
+      "step": 15120
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 4.880645161290323e-05,
+      "loss": 0.4361,
+      "step": 15130
+    },
+    {
+      "epoch": 97.68,
+      "learning_rate": 4.883870967741936e-05,
+      "loss": 0.3915,
+      "step": 15140
+    },
+    {
+      "epoch": 97.74,
+      "learning_rate": 4.887096774193549e-05,
+      "loss": 0.3379,
+      "step": 15150
+    },
+    {
+      "epoch": 97.8,
+      "learning_rate": 4.890322580645161e-05,
+      "loss": 0.3846,
+      "step": 15160
+    },
+    {
+      "epoch": 97.87,
+      "learning_rate": 4.8935483870967746e-05,
+      "loss": 0.3563,
+      "step": 15170
+    },
+    {
+      "epoch": 97.93,
+      "learning_rate": 4.8967741935483876e-05,
+      "loss": 0.3878,
+      "step": 15180
+    },
+    {
+      "epoch": 98.0,
+      "learning_rate": 4.9e-05,
+      "loss": 0.3643,
+      "step": 15190
+    },
+    {
+      "epoch": 98.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0178523063659668,
+      "eval_runtime": 0.8377,
+      "eval_samples_per_second": 239.939,
+      "eval_steps_per_second": 8.356,
+      "step": 15190
+    },
+    {
+      "epoch": 98.06,
+      "learning_rate": 4.903225806451613e-05,
+      "loss": 0.414,
+      "step": 15200
+    },
+    {
+      "epoch": 98.13,
+      "learning_rate": 4.9064516129032264e-05,
+      "loss": 0.3465,
+      "step": 15210
+    },
+    {
+      "epoch": 98.19,
+      "learning_rate": 4.9096774193548387e-05,
+      "loss": 0.3177,
+      "step": 15220
+    },
+    {
+      "epoch": 98.26,
+      "learning_rate": 4.9129032258064516e-05,
+      "loss": 0.37,
+      "step": 15230
+    },
+    {
+      "epoch": 98.32,
+      "learning_rate": 4.916129032258065e-05,
+      "loss": 0.4192,
+      "step": 15240
+    },
+    {
+      "epoch": 98.39,
+      "learning_rate": 4.9193548387096775e-05,
+      "loss": 0.3952,
+      "step": 15250
+    },
+    {
+      "epoch": 98.45,
+      "learning_rate": 4.9225806451612904e-05,
+      "loss": 0.3246,
+      "step": 15260
+    },
+    {
+      "epoch": 98.51,
+      "learning_rate": 4.9258064516129033e-05,
+      "loss": 0.3856,
+      "step": 15270
+    },
+    {
+      "epoch": 98.58,
+      "learning_rate": 4.929032258064516e-05,
+      "loss": 0.3842,
+      "step": 15280
+    },
+    {
+      "epoch": 98.64,
+      "learning_rate": 4.932258064516129e-05,
+      "loss": 0.3491,
+      "step": 15290
+    },
+    {
+      "epoch": 98.71,
+      "learning_rate": 4.935483870967742e-05,
+      "loss": 0.3567,
+      "step": 15300
+    },
+    {
+      "epoch": 98.77,
+      "learning_rate": 4.938709677419355e-05,
+      "loss": 0.4346,
+      "step": 15310
+    },
+    {
+      "epoch": 98.84,
+      "learning_rate": 4.941935483870968e-05,
+      "loss": 0.4138,
+      "step": 15320
+    },
+    {
+      "epoch": 98.9,
+      "learning_rate": 4.945161290322581e-05,
+      "loss": 0.3704,
+      "step": 15330
+    },
+    {
+      "epoch": 98.96,
+      "learning_rate": 4.948387096774193e-05,
+      "loss": 0.303,
+      "step": 15340
+    },
+    {
+      "epoch": 99.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9921143651008606,
+      "eval_runtime": 0.8576,
+      "eval_samples_per_second": 234.37,
+      "eval_steps_per_second": 8.162,
+      "step": 15345
+    },
+    {
+      "epoch": 99.03,
+      "learning_rate": 4.951612903225807e-05,
+      "loss": 0.3924,
+      "step": 15350
+    },
+    {
+      "epoch": 99.1,
+      "learning_rate": 4.95483870967742e-05,
+      "loss": 0.335,
+      "step": 15360
+    },
+    {
+      "epoch": 99.16,
+      "learning_rate": 4.958064516129032e-05,
+      "loss": 0.35,
+      "step": 15370
+    },
+    {
+      "epoch": 99.23,
+      "learning_rate": 4.961290322580646e-05,
+      "loss": 0.3363,
+      "step": 15380
+    },
+    {
+      "epoch": 99.29,
+      "learning_rate": 4.9645161290322586e-05,
+      "loss": 0.4206,
+      "step": 15390
+    },
+    {
+      "epoch": 99.35,
+      "learning_rate": 4.967741935483871e-05,
+      "loss": 0.3412,
+      "step": 15400
+    },
+    {
+      "epoch": 99.42,
+      "learning_rate": 4.970967741935484e-05,
+      "loss": 0.3235,
+      "step": 15410
+    },
+    {
+      "epoch": 99.48,
+      "learning_rate": 4.9741935483870974e-05,
+      "loss": 0.3592,
+      "step": 15420
+    },
+    {
+      "epoch": 99.55,
+      "learning_rate": 4.97741935483871e-05,
+      "loss": 0.3618,
+      "step": 15430
+    },
+    {
+      "epoch": 99.61,
+      "learning_rate": 4.9806451612903226e-05,
+      "loss": 0.3884,
+      "step": 15440
+    },
+    {
+      "epoch": 99.68,
+      "learning_rate": 4.983870967741936e-05,
+      "loss": 0.3651,
+      "step": 15450
+    },
+    {
+      "epoch": 99.74,
+      "learning_rate": 4.9870967741935485e-05,
+      "loss": 0.387,
+      "step": 15460
+    },
+    {
+      "epoch": 99.8,
+      "learning_rate": 4.9903225806451614e-05,
+      "loss": 0.393,
+      "step": 15470
+    },
+    {
+      "epoch": 99.87,
+      "learning_rate": 4.9935483870967744e-05,
+      "loss": 0.334,
+      "step": 15480
+    },
+    {
+      "epoch": 99.93,
+      "learning_rate": 4.996774193548387e-05,
+      "loss": 0.3553,
+      "step": 15490
+    },
+    {
+      "epoch": 100.0,
+      "learning_rate": 5e-05,
+      "loss": 0.3579,
+      "step": 15500
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.953705370426178,
+      "eval_runtime": 0.8621,
+      "eval_samples_per_second": 233.141,
+      "eval_steps_per_second": 8.119,
+      "step": 15500
+    },
+    {
+      "epoch": 100.06,
+      "learning_rate": 4.999641577060932e-05,
+      "loss": 0.4032,
+      "step": 15510
+    },
+    {
+      "epoch": 100.13,
+      "learning_rate": 4.999283154121864e-05,
+      "loss": 0.4123,
+      "step": 15520
+    },
+    {
+      "epoch": 100.19,
+      "learning_rate": 4.998924731182796e-05,
+      "loss": 0.4147,
+      "step": 15530
+    },
+    {
+      "epoch": 100.26,
+      "learning_rate": 4.9985663082437275e-05,
+      "loss": 0.3558,
+      "step": 15540
+    },
+    {
+      "epoch": 100.32,
+      "learning_rate": 4.99820788530466e-05,
+      "loss": 0.4363,
+      "step": 15550
+    },
+    {
+      "epoch": 100.39,
+      "learning_rate": 4.997849462365591e-05,
+      "loss": 0.3473,
+      "step": 15560
+    },
+    {
+      "epoch": 100.45,
+      "learning_rate": 4.9974910394265237e-05,
+      "loss": 0.3736,
+      "step": 15570
+    },
+    {
+      "epoch": 100.51,
+      "learning_rate": 4.9971326164874555e-05,
+      "loss": 0.3261,
+      "step": 15580
+    },
+    {
+      "epoch": 100.58,
+      "learning_rate": 4.996774193548387e-05,
+      "loss": 0.3916,
+      "step": 15590
+    },
+    {
+      "epoch": 100.64,
+      "learning_rate": 4.996415770609319e-05,
+      "loss": 0.3597,
+      "step": 15600
+    },
+    {
+      "epoch": 100.71,
+      "learning_rate": 4.996057347670251e-05,
+      "loss": 0.3669,
+      "step": 15610
+    },
+    {
+      "epoch": 100.77,
+      "learning_rate": 4.995698924731183e-05,
+      "loss": 0.3716,
+      "step": 15620
+    },
+    {
+      "epoch": 100.84,
+      "learning_rate": 4.995340501792115e-05,
+      "loss": 0.4159,
+      "step": 15630
+    },
+    {
+      "epoch": 100.9,
+      "learning_rate": 4.9949820788530464e-05,
+      "loss": 0.3893,
+      "step": 15640
+    },
+    {
+      "epoch": 100.96,
+      "learning_rate": 4.994623655913979e-05,
+      "loss": 0.4025,
+      "step": 15650
+    },
+    {
+      "epoch": 101.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.959656298160553,
+      "eval_runtime": 0.8527,
+      "eval_samples_per_second": 235.722,
+      "eval_steps_per_second": 8.209,
+      "step": 15655
+    },
+    {
+      "epoch": 101.03,
+      "learning_rate": 4.994265232974911e-05,
+      "loss": 0.3283,
+      "step": 15660
+    },
+    {
+      "epoch": 101.1,
+      "learning_rate": 4.9939068100358425e-05,
+      "loss": 0.3563,
+      "step": 15670
+    },
+    {
+      "epoch": 101.16,
+      "learning_rate": 4.9935483870967744e-05,
+      "loss": 0.3664,
+      "step": 15680
+    },
+    {
+      "epoch": 101.23,
+      "learning_rate": 4.993189964157706e-05,
+      "loss": 0.3672,
+      "step": 15690
+    },
+    {
+      "epoch": 101.29,
+      "learning_rate": 4.992831541218638e-05,
+      "loss": 0.3586,
+      "step": 15700
+    },
+    {
+      "epoch": 101.35,
+      "learning_rate": 4.9924731182795705e-05,
+      "loss": 0.3521,
+      "step": 15710
+    },
+    {
+      "epoch": 101.42,
+      "learning_rate": 4.9921146953405016e-05,
+      "loss": 0.3666,
+      "step": 15720
+    },
+    {
+      "epoch": 101.48,
+      "learning_rate": 4.991756272401434e-05,
+      "loss": 0.3704,
+      "step": 15730
+    },
+    {
+      "epoch": 101.55,
+      "learning_rate": 4.991397849462366e-05,
+      "loss": 0.3176,
+      "step": 15740
+    },
+    {
+      "epoch": 101.61,
+      "learning_rate": 4.991039426523298e-05,
+      "loss": 0.2962,
+      "step": 15750
+    },
+    {
+      "epoch": 101.68,
+      "learning_rate": 4.9906810035842296e-05,
+      "loss": 0.3225,
+      "step": 15760
+    },
+    {
+      "epoch": 101.74,
+      "learning_rate": 4.9903225806451614e-05,
+      "loss": 0.3355,
+      "step": 15770
+    },
+    {
+      "epoch": 101.8,
+      "learning_rate": 4.989964157706093e-05,
+      "loss": 0.4375,
+      "step": 15780
+    },
+    {
+      "epoch": 101.87,
+      "learning_rate": 4.989605734767026e-05,
+      "loss": 0.3576,
+      "step": 15790
+    },
+    {
+      "epoch": 101.93,
+      "learning_rate": 4.989247311827957e-05,
+      "loss": 0.3925,
+      "step": 15800
+    },
+    {
+      "epoch": 102.0,
+      "learning_rate": 4.9888888888888894e-05,
+      "loss": 0.3798,
+      "step": 15810
+    },
+    {
+      "epoch": 102.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.921752393245697,
+      "eval_runtime": 0.8603,
+      "eval_samples_per_second": 233.647,
+      "eval_steps_per_second": 8.137,
+      "step": 15810
+    },
+    {
+      "epoch": 102.06,
+      "learning_rate": 4.988530465949821e-05,
+      "loss": 0.3845,
+      "step": 15820
+    },
+    {
+      "epoch": 102.13,
+      "learning_rate": 4.988172043010753e-05,
+      "loss": 0.346,
+      "step": 15830
+    },
+    {
+      "epoch": 102.19,
+      "learning_rate": 4.987813620071685e-05,
+      "loss": 0.3717,
+      "step": 15840
+    },
+    {
+      "epoch": 102.26,
+      "learning_rate": 4.987455197132617e-05,
+      "loss": 0.3679,
+      "step": 15850
+    },
+    {
+      "epoch": 102.32,
+      "learning_rate": 4.9870967741935485e-05,
+      "loss": 0.3224,
+      "step": 15860
+    },
+    {
+      "epoch": 102.39,
+      "learning_rate": 4.986738351254481e-05,
+      "loss": 0.4019,
+      "step": 15870
+    },
+    {
+      "epoch": 102.45,
+      "learning_rate": 4.986379928315412e-05,
+      "loss": 0.363,
+      "step": 15880
+    },
+    {
+      "epoch": 102.51,
+      "learning_rate": 4.986021505376344e-05,
+      "loss": 0.2912,
+      "step": 15890
+    },
+    {
+      "epoch": 102.58,
+      "learning_rate": 4.9856630824372764e-05,
+      "loss": 0.3258,
+      "step": 15900
+    },
+    {
+      "epoch": 102.64,
+      "learning_rate": 4.985304659498208e-05,
+      "loss": 0.3831,
+      "step": 15910
+    },
+    {
+      "epoch": 102.71,
+      "learning_rate": 4.98494623655914e-05,
+      "loss": 0.374,
+      "step": 15920
+    },
+    {
+      "epoch": 102.77,
+      "learning_rate": 4.984587813620072e-05,
+      "loss": 0.3668,
+      "step": 15930
+    },
+    {
+      "epoch": 102.84,
+      "learning_rate": 4.984229390681004e-05,
+      "loss": 0.3037,
+      "step": 15940
+    },
+    {
+      "epoch": 102.9,
+      "learning_rate": 4.983870967741936e-05,
+      "loss": 0.394,
+      "step": 15950
+    },
+    {
+      "epoch": 102.96,
+      "learning_rate": 4.9835125448028674e-05,
+      "loss": 0.3589,
+      "step": 15960
+    },
+    {
+      "epoch": 103.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.013307809829712,
+      "eval_runtime": 0.8798,
+      "eval_samples_per_second": 228.462,
+      "eval_steps_per_second": 7.956,
+      "step": 15965
+    },
+    {
+      "epoch": 103.03,
+      "learning_rate": 4.983154121863799e-05,
+      "loss": 0.3913,
+      "step": 15970
+    },
+    {
+      "epoch": 103.1,
+      "learning_rate": 4.982795698924732e-05,
+      "loss": 0.351,
+      "step": 15980
+    },
+    {
+      "epoch": 103.16,
+      "learning_rate": 4.9824372759856635e-05,
+      "loss": 0.3403,
+      "step": 15990
+    },
+    {
+      "epoch": 103.23,
+      "learning_rate": 4.982078853046595e-05,
+      "loss": 0.3438,
+      "step": 16000
+    },
+    {
+      "epoch": 103.29,
+      "learning_rate": 4.981720430107527e-05,
+      "loss": 0.3483,
+      "step": 16010
+    },
+    {
+      "epoch": 103.35,
+      "learning_rate": 4.981362007168459e-05,
+      "loss": 0.346,
+      "step": 16020
+    },
+    {
+      "epoch": 103.42,
+      "learning_rate": 4.9810035842293915e-05,
+      "loss": 0.4418,
+      "step": 16030
+    },
+    {
+      "epoch": 103.48,
+      "learning_rate": 4.9806451612903226e-05,
+      "loss": 0.3334,
+      "step": 16040
+    },
+    {
+      "epoch": 103.55,
+      "learning_rate": 4.9802867383512544e-05,
+      "loss": 0.3785,
+      "step": 16050
+    },
+    {
+      "epoch": 103.61,
+      "learning_rate": 4.979928315412187e-05,
+      "loss": 0.3315,
+      "step": 16060
+    },
+    {
+      "epoch": 103.68,
+      "learning_rate": 4.979569892473119e-05,
+      "loss": 0.3828,
+      "step": 16070
+    },
+    {
+      "epoch": 103.74,
+      "learning_rate": 4.9792114695340506e-05,
+      "loss": 0.325,
+      "step": 16080
+    },
+    {
+      "epoch": 103.8,
+      "learning_rate": 4.9788530465949824e-05,
+      "loss": 0.3869,
+      "step": 16090
+    },
+    {
+      "epoch": 103.87,
+      "learning_rate": 4.978494623655914e-05,
+      "loss": 0.3066,
+      "step": 16100
+    },
+    {
+      "epoch": 103.93,
+      "learning_rate": 4.978136200716847e-05,
+      "loss": 0.3752,
+      "step": 16110
+    },
+    {
+      "epoch": 104.0,
+      "learning_rate": 4.977777777777778e-05,
+      "loss": 0.3139,
+      "step": 16120
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.043976902961731,
+      "eval_runtime": 0.8589,
+      "eval_samples_per_second": 234.027,
+      "eval_steps_per_second": 8.15,
+      "step": 16120
+    },
+    {
+      "epoch": 104.06,
+      "learning_rate": 4.97741935483871e-05,
+      "loss": 0.3802,
+      "step": 16130
+    },
+    {
+      "epoch": 104.13,
+      "learning_rate": 4.977060931899642e-05,
+      "loss": 0.4086,
+      "step": 16140
+    },
+    {
+      "epoch": 104.19,
+      "learning_rate": 4.976702508960573e-05,
+      "loss": 0.3359,
+      "step": 16150
+    },
+    {
+      "epoch": 104.26,
+      "learning_rate": 4.976344086021506e-05,
+      "loss": 0.3005,
+      "step": 16160
+    },
+    {
+      "epoch": 104.32,
+      "learning_rate": 4.9759856630824376e-05,
+      "loss": 0.357,
+      "step": 16170
+    },
+    {
+      "epoch": 104.39,
+      "learning_rate": 4.9756272401433695e-05,
+      "loss": 0.356,
+      "step": 16180
+    },
+    {
+      "epoch": 104.45,
+      "learning_rate": 4.975268817204301e-05,
+      "loss": 0.3324,
+      "step": 16190
+    },
+    {
+      "epoch": 104.51,
+      "learning_rate": 4.974910394265233e-05,
+      "loss": 0.3136,
+      "step": 16200
+    },
+    {
+      "epoch": 104.58,
+      "learning_rate": 4.974551971326165e-05,
+      "loss": 0.3639,
+      "step": 16210
+    },
+    {
+      "epoch": 104.64,
+      "learning_rate": 4.9741935483870974e-05,
+      "loss": 0.3837,
+      "step": 16220
+    },
+    {
+      "epoch": 104.71,
+      "learning_rate": 4.9738351254480286e-05,
+      "loss": 0.3261,
+      "step": 16230
+    },
+    {
+      "epoch": 104.77,
+      "learning_rate": 4.9734767025089604e-05,
+      "loss": 0.3876,
+      "step": 16240
+    },
+    {
+      "epoch": 104.84,
+      "learning_rate": 4.973118279569893e-05,
+      "loss": 0.4264,
+      "step": 16250
+    },
+    {
+      "epoch": 104.9,
+      "learning_rate": 4.972759856630825e-05,
+      "loss": 0.3463,
+      "step": 16260
+    },
+    {
+      "epoch": 104.96,
+      "learning_rate": 4.9724014336917565e-05,
+      "loss": 0.3452,
+      "step": 16270
+    },
+    {
+      "epoch": 105.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.013669490814209,
+      "eval_runtime": 0.8647,
+      "eval_samples_per_second": 232.455,
+      "eval_steps_per_second": 8.095,
+      "step": 16275
+    },
+    {
+      "epoch": 105.03,
+      "learning_rate": 4.972043010752688e-05,
+      "loss": 0.3729,
+      "step": 16280
+    },
+    {
+      "epoch": 105.1,
+      "learning_rate": 4.97168458781362e-05,
+      "loss": 0.3812,
+      "step": 16290
+    },
+    {
+      "epoch": 105.16,
+      "learning_rate": 4.9713261648745527e-05,
+      "loss": 0.3782,
+      "step": 16300
+    },
+    {
+      "epoch": 105.23,
+      "learning_rate": 4.970967741935484e-05,
+      "loss": 0.3074,
+      "step": 16310
+    },
+    {
+      "epoch": 105.29,
+      "learning_rate": 4.9706093189964156e-05,
+      "loss": 0.3036,
+      "step": 16320
+    },
+    {
+      "epoch": 105.35,
+      "learning_rate": 4.970250896057348e-05,
+      "loss": 0.3173,
+      "step": 16330
+    },
+    {
+      "epoch": 105.42,
+      "learning_rate": 4.96989247311828e-05,
+      "loss": 0.3867,
+      "step": 16340
+    },
+    {
+      "epoch": 105.48,
+      "learning_rate": 4.969534050179212e-05,
+      "loss": 0.3727,
+      "step": 16350
+    },
+    {
+      "epoch": 105.55,
+      "learning_rate": 4.9691756272401436e-05,
+      "loss": 0.4325,
+      "step": 16360
+    },
+    {
+      "epoch": 105.61,
+      "learning_rate": 4.9688172043010754e-05,
+      "loss": 0.3694,
+      "step": 16370
+    },
+    {
+      "epoch": 105.68,
+      "learning_rate": 4.968458781362008e-05,
+      "loss": 0.37,
+      "step": 16380
+    },
+    {
+      "epoch": 105.74,
+      "learning_rate": 4.968100358422939e-05,
+      "loss": 0.3614,
+      "step": 16390
+    },
+    {
+      "epoch": 105.8,
+      "learning_rate": 4.967741935483871e-05,
+      "loss": 0.3419,
+      "step": 16400
+    },
+    {
+      "epoch": 105.87,
+      "learning_rate": 4.9673835125448034e-05,
+      "loss": 0.4037,
+      "step": 16410
+    },
+    {
+      "epoch": 105.93,
+      "learning_rate": 4.967025089605735e-05,
+      "loss": 0.3676,
+      "step": 16420
+    },
+    {
+      "epoch": 106.0,
+      "learning_rate": 4.966666666666667e-05,
+      "loss": 0.399,
+      "step": 16430
+    },
+    {
+      "epoch": 106.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.1307978630065918,
+      "eval_runtime": 0.8411,
+      "eval_samples_per_second": 238.983,
+      "eval_steps_per_second": 8.323,
+      "step": 16430
+    },
+    {
+      "epoch": 106.06,
+      "learning_rate": 4.966308243727599e-05,
+      "loss": 0.4203,
+      "step": 16440
+    },
+    {
+      "epoch": 106.13,
+      "learning_rate": 4.9659498207885306e-05,
+      "loss": 0.3742,
+      "step": 16450
+    },
+    {
+      "epoch": 106.19,
+      "learning_rate": 4.965591397849463e-05,
+      "loss": 0.3958,
+      "step": 16460
+    },
+    {
+      "epoch": 106.26,
+      "learning_rate": 4.965232974910394e-05,
+      "loss": 0.3241,
+      "step": 16470
+    },
+    {
+      "epoch": 106.32,
+      "learning_rate": 4.964874551971326e-05,
+      "loss": 0.3833,
+      "step": 16480
+    },
+    {
+      "epoch": 106.39,
+      "learning_rate": 4.9645161290322586e-05,
+      "loss": 0.3287,
+      "step": 16490
+    },
+    {
+      "epoch": 106.45,
+      "learning_rate": 4.9641577060931904e-05,
+      "loss": 0.3268,
+      "step": 16500
+    },
+    {
+      "epoch": 106.51,
+      "learning_rate": 4.963799283154122e-05,
+      "loss": 0.3316,
+      "step": 16510
+    },
+    {
+      "epoch": 106.58,
+      "learning_rate": 4.963440860215054e-05,
+      "loss": 0.311,
+      "step": 16520
+    },
+    {
+      "epoch": 106.64,
+      "learning_rate": 4.963082437275986e-05,
+      "loss": 0.3613,
+      "step": 16530
+    },
+    {
+      "epoch": 106.71,
+      "learning_rate": 4.962724014336918e-05,
+      "loss": 0.4229,
+      "step": 16540
+    },
+    {
+      "epoch": 106.77,
+      "learning_rate": 4.9623655913978495e-05,
+      "loss": 0.354,
+      "step": 16550
+    },
+    {
+      "epoch": 106.84,
+      "learning_rate": 4.9620071684587813e-05,
+      "loss": 0.3645,
+      "step": 16560
+    },
+    {
+      "epoch": 106.9,
+      "learning_rate": 4.961648745519714e-05,
+      "loss": 0.3642,
+      "step": 16570
+    },
+    {
+      "epoch": 106.96,
+      "learning_rate": 4.961290322580646e-05,
+      "loss": 0.3626,
+      "step": 16580
+    },
+    {
+      "epoch": 107.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9933374524116516,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 230.723,
+      "eval_steps_per_second": 8.035,
+      "step": 16585
+    },
+    {
+      "epoch": 107.03,
+      "learning_rate": 4.960931899641577e-05,
+      "loss": 0.3981,
+      "step": 16590
+    },
+    {
+      "epoch": 107.1,
+      "learning_rate": 4.960573476702509e-05,
+      "loss": 0.4129,
+      "step": 16600
+    },
+    {
+      "epoch": 107.16,
+      "learning_rate": 4.960215053763441e-05,
+      "loss": 0.3599,
+      "step": 16610
+    },
+    {
+      "epoch": 107.23,
+      "learning_rate": 4.959856630824373e-05,
+      "loss": 0.3585,
+      "step": 16620
+    },
+    {
+      "epoch": 107.29,
+      "learning_rate": 4.959498207885305e-05,
+      "loss": 0.3678,
+      "step": 16630
+    },
+    {
+      "epoch": 107.35,
+      "learning_rate": 4.9591397849462366e-05,
+      "loss": 0.349,
+      "step": 16640
+    },
+    {
+      "epoch": 107.42,
+      "learning_rate": 4.958781362007169e-05,
+      "loss": 0.3162,
+      "step": 16650
+    },
+    {
+      "epoch": 107.48,
+      "learning_rate": 4.958422939068101e-05,
+      "loss": 0.3487,
+      "step": 16660
+    },
+    {
+      "epoch": 107.55,
+      "learning_rate": 4.958064516129032e-05,
+      "loss": 0.3779,
+      "step": 16670
+    },
+    {
+      "epoch": 107.61,
+      "learning_rate": 4.9577060931899645e-05,
+      "loss": 0.4308,
+      "step": 16680
+    },
+    {
+      "epoch": 107.68,
+      "learning_rate": 4.9573476702508964e-05,
+      "loss": 0.3969,
+      "step": 16690
+    },
+    {
+      "epoch": 107.74,
+      "learning_rate": 4.956989247311828e-05,
+      "loss": 0.3977,
+      "step": 16700
+    },
+    {
+      "epoch": 107.8,
+      "learning_rate": 4.95663082437276e-05,
+      "loss": 0.3491,
+      "step": 16710
+    },
+    {
+      "epoch": 107.87,
+      "learning_rate": 4.956272401433692e-05,
+      "loss": 0.41,
+      "step": 16720
+    },
+    {
+      "epoch": 107.93,
+      "learning_rate": 4.955913978494624e-05,
+      "loss": 0.4062,
+      "step": 16730
+    },
+    {
+      "epoch": 108.0,
+      "learning_rate": 4.955555555555556e-05,
+      "loss": 0.3083,
+      "step": 16740
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9568660855293274,
+      "eval_runtime": 0.9303,
+      "eval_samples_per_second": 216.05,
+      "eval_steps_per_second": 7.524,
+      "step": 16740
+    },
+    {
+      "epoch": 108.06,
+      "learning_rate": 4.955197132616487e-05,
+      "loss": 0.3707,
+      "step": 16750
+    },
+    {
+      "epoch": 108.13,
+      "learning_rate": 4.95483870967742e-05,
+      "loss": 0.3956,
+      "step": 16760
+    },
+    {
+      "epoch": 108.19,
+      "learning_rate": 4.9544802867383516e-05,
+      "loss": 0.3235,
+      "step": 16770
+    },
+    {
+      "epoch": 108.26,
+      "learning_rate": 4.9541218637992834e-05,
+      "loss": 0.3478,
+      "step": 16780
+    },
+    {
+      "epoch": 108.32,
+      "learning_rate": 4.953763440860215e-05,
+      "loss": 0.3773,
+      "step": 16790
+    },
+    {
+      "epoch": 108.39,
+      "learning_rate": 4.953405017921147e-05,
+      "loss": 0.3353,
+      "step": 16800
+    },
+    {
+      "epoch": 108.45,
+      "learning_rate": 4.9530465949820796e-05,
+      "loss": 0.3635,
+      "step": 16810
+    },
+    {
+      "epoch": 108.51,
+      "learning_rate": 4.952688172043011e-05,
+      "loss": 0.3674,
+      "step": 16820
+    },
+    {
+      "epoch": 108.58,
+      "learning_rate": 4.9523297491039425e-05,
+      "loss": 0.3942,
+      "step": 16830
+    },
+    {
+      "epoch": 108.64,
+      "learning_rate": 4.951971326164875e-05,
+      "loss": 0.3632,
+      "step": 16840
+    },
+    {
+      "epoch": 108.71,
+      "learning_rate": 4.951612903225807e-05,
+      "loss": 0.3643,
+      "step": 16850
+    },
+    {
+      "epoch": 108.77,
+      "learning_rate": 4.951254480286738e-05,
+      "loss": 0.2618,
+      "step": 16860
+    },
+    {
+      "epoch": 108.84,
+      "learning_rate": 4.9508960573476705e-05,
+      "loss": 0.3613,
+      "step": 16870
+    },
+    {
+      "epoch": 108.9,
+      "learning_rate": 4.950537634408602e-05,
+      "loss": 0.4217,
+      "step": 16880
+    },
+    {
+      "epoch": 108.96,
+      "learning_rate": 4.950179211469534e-05,
+      "loss": 0.3661,
+      "step": 16890
+    },
+    {
+      "epoch": 109.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8978585600852966,
+      "eval_runtime": 0.8462,
+      "eval_samples_per_second": 237.542,
+      "eval_steps_per_second": 8.273,
+      "step": 16895
+    },
+    {
+      "epoch": 109.03,
+      "learning_rate": 4.949820788530466e-05,
+      "loss": 0.4153,
+      "step": 16900
+    },
+    {
+      "epoch": 109.1,
+      "learning_rate": 4.949462365591398e-05,
+      "loss": 0.349,
+      "step": 16910
+    },
+    {
+      "epoch": 109.16,
+      "learning_rate": 4.94910394265233e-05,
+      "loss": 0.3515,
+      "step": 16920
+    },
+    {
+      "epoch": 109.23,
+      "learning_rate": 4.948745519713262e-05,
+      "loss": 0.3107,
+      "step": 16930
+    },
+    {
+      "epoch": 109.29,
+      "learning_rate": 4.948387096774193e-05,
+      "loss": 0.4052,
+      "step": 16940
+    },
+    {
+      "epoch": 109.35,
+      "learning_rate": 4.948028673835126e-05,
+      "loss": 0.3243,
+      "step": 16950
+    },
+    {
+      "epoch": 109.42,
+      "learning_rate": 4.9476702508960576e-05,
+      "loss": 0.381,
+      "step": 16960
+    },
+    {
+      "epoch": 109.48,
+      "learning_rate": 4.9473118279569894e-05,
+      "loss": 0.3245,
+      "step": 16970
+    },
+    {
+      "epoch": 109.55,
+      "learning_rate": 4.946953405017921e-05,
+      "loss": 0.3773,
+      "step": 16980
+    },
+    {
+      "epoch": 109.61,
+      "learning_rate": 4.946594982078853e-05,
+      "loss": 0.3508,
+      "step": 16990
+    },
+    {
+      "epoch": 109.68,
+      "learning_rate": 4.9462365591397855e-05,
+      "loss": 0.4047,
+      "step": 17000
+    },
+    {
+      "epoch": 109.74,
+      "learning_rate": 4.945878136200717e-05,
+      "loss": 0.3456,
+      "step": 17010
+    },
+    {
+      "epoch": 109.8,
+      "learning_rate": 4.9455197132616485e-05,
+      "loss": 0.4434,
+      "step": 17020
+    },
+    {
+      "epoch": 109.87,
+      "learning_rate": 4.945161290322581e-05,
+      "loss": 0.3078,
+      "step": 17030
+    },
+    {
+      "epoch": 109.93,
+      "learning_rate": 4.944802867383513e-05,
+      "loss": 0.3371,
+      "step": 17040
+    },
+    {
+      "epoch": 110.0,
+      "learning_rate": 4.9444444444444446e-05,
+      "loss": 0.4002,
+      "step": 17050
+    },
+    {
+      "epoch": 110.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9831101298332214,
+      "eval_runtime": 0.8449,
+      "eval_samples_per_second": 237.908,
+      "eval_steps_per_second": 8.285,
+      "step": 17050
+    },
+    {
+      "epoch": 110.06,
+      "learning_rate": 4.9440860215053764e-05,
+      "loss": 0.3394,
+      "step": 17060
+    },
+    {
+      "epoch": 110.13,
+      "learning_rate": 4.943727598566308e-05,
+      "loss": 0.3547,
+      "step": 17070
+    },
+    {
+      "epoch": 110.19,
+      "learning_rate": 4.943369175627241e-05,
+      "loss": 0.416,
+      "step": 17080
+    },
+    {
+      "epoch": 110.26,
+      "learning_rate": 4.9430107526881726e-05,
+      "loss": 0.3219,
+      "step": 17090
+    },
+    {
+      "epoch": 110.32,
+      "learning_rate": 4.942652329749104e-05,
+      "loss": 0.3856,
+      "step": 17100
+    },
+    {
+      "epoch": 110.39,
+      "learning_rate": 4.942293906810036e-05,
+      "loss": 0.3848,
+      "step": 17110
+    },
+    {
+      "epoch": 110.45,
+      "learning_rate": 4.941935483870968e-05,
+      "loss": 0.3574,
+      "step": 17120
+    },
+    {
+      "epoch": 110.51,
+      "learning_rate": 4.9415770609319e-05,
+      "loss": 0.3767,
+      "step": 17130
+    },
+    {
+      "epoch": 110.58,
+      "learning_rate": 4.941218637992832e-05,
+      "loss": 0.3444,
+      "step": 17140
+    },
+    {
+      "epoch": 110.64,
+      "learning_rate": 4.9408602150537635e-05,
+      "loss": 0.3775,
+      "step": 17150
+    },
+    {
+      "epoch": 110.71,
+      "learning_rate": 4.940501792114696e-05,
+      "loss": 0.3392,
+      "step": 17160
+    },
+    {
+      "epoch": 110.77,
+      "learning_rate": 4.940143369175628e-05,
+      "loss": 0.353,
+      "step": 17170
+    },
+    {
+      "epoch": 110.84,
+      "learning_rate": 4.939784946236559e-05,
+      "loss": 0.3373,
+      "step": 17180
+    },
+    {
+      "epoch": 110.9,
+      "learning_rate": 4.9394265232974915e-05,
+      "loss": 0.3431,
+      "step": 17190
+    },
+    {
+      "epoch": 110.96,
+      "learning_rate": 4.939068100358423e-05,
+      "loss": 0.3274,
+      "step": 17200
+    },
+    {
+      "epoch": 111.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.001416563987732,
+      "eval_runtime": 0.8794,
+      "eval_samples_per_second": 228.572,
+      "eval_steps_per_second": 7.96,
+      "step": 17205
+    },
+    {
+      "epoch": 111.03,
+      "learning_rate": 4.938709677419355e-05,
+      "loss": 0.4155,
+      "step": 17210
+    },
+    {
+      "epoch": 111.1,
+      "learning_rate": 4.938351254480287e-05,
+      "loss": 0.3058,
+      "step": 17220
+    },
+    {
+      "epoch": 111.16,
+      "learning_rate": 4.937992831541219e-05,
+      "loss": 0.4206,
+      "step": 17230
+    },
+    {
+      "epoch": 111.23,
+      "learning_rate": 4.9376344086021506e-05,
+      "loss": 0.406,
+      "step": 17240
+    },
+    {
+      "epoch": 111.29,
+      "learning_rate": 4.937275985663083e-05,
+      "loss": 0.3236,
+      "step": 17250
+    },
+    {
+      "epoch": 111.35,
+      "learning_rate": 4.936917562724014e-05,
+      "loss": 0.4001,
+      "step": 17260
+    },
+    {
+      "epoch": 111.42,
+      "learning_rate": 4.936559139784947e-05,
+      "loss": 0.367,
+      "step": 17270
+    },
+    {
+      "epoch": 111.48,
+      "learning_rate": 4.9362007168458785e-05,
+      "loss": 0.436,
+      "step": 17280
+    },
+    {
+      "epoch": 111.55,
+      "learning_rate": 4.93584229390681e-05,
+      "loss": 0.3708,
+      "step": 17290
+    },
+    {
+      "epoch": 111.61,
+      "learning_rate": 4.935483870967742e-05,
+      "loss": 0.372,
+      "step": 17300
+    },
+    {
+      "epoch": 111.68,
+      "learning_rate": 4.935125448028674e-05,
+      "loss": 0.3461,
+      "step": 17310
+    },
+    {
+      "epoch": 111.74,
+      "learning_rate": 4.934767025089606e-05,
+      "loss": 0.3404,
+      "step": 17320
+    },
+    {
+      "epoch": 111.8,
+      "learning_rate": 4.934408602150538e-05,
+      "loss": 0.4164,
+      "step": 17330
+    },
+    {
+      "epoch": 111.87,
+      "learning_rate": 4.9340501792114694e-05,
+      "loss": 0.4147,
+      "step": 17340
+    },
+    {
+      "epoch": 111.93,
+      "learning_rate": 4.933691756272402e-05,
+      "loss": 0.3391,
+      "step": 17350
+    },
+    {
+      "epoch": 112.0,
+      "learning_rate": 4.933333333333334e-05,
+      "loss": 0.3686,
+      "step": 17360
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9588882923126221,
+      "eval_runtime": 0.8591,
+      "eval_samples_per_second": 233.97,
+      "eval_steps_per_second": 8.148,
+      "step": 17360
+    },
+    {
+      "epoch": 112.06,
+      "learning_rate": 4.9329749103942656e-05,
+      "loss": 0.3291,
+      "step": 17370
+    },
+    {
+      "epoch": 112.13,
+      "learning_rate": 4.9326164874551974e-05,
+      "loss": 0.3211,
+      "step": 17380
+    },
+    {
+      "epoch": 112.19,
+      "learning_rate": 4.932258064516129e-05,
+      "loss": 0.3613,
+      "step": 17390
+    },
+    {
+      "epoch": 112.26,
+      "learning_rate": 4.931899641577061e-05,
+      "loss": 0.3486,
+      "step": 17400
+    },
+    {
+      "epoch": 112.32,
+      "learning_rate": 4.9315412186379935e-05,
+      "loss": 0.3776,
+      "step": 17410
+    },
+    {
+      "epoch": 112.39,
+      "learning_rate": 4.931182795698925e-05,
+      "loss": 0.2763,
+      "step": 17420
+    },
+    {
+      "epoch": 112.45,
+      "learning_rate": 4.930824372759857e-05,
+      "loss": 0.3601,
+      "step": 17430
+    },
+    {
+      "epoch": 112.51,
+      "learning_rate": 4.930465949820789e-05,
+      "loss": 0.3464,
+      "step": 17440
+    },
+    {
+      "epoch": 112.58,
+      "learning_rate": 4.93010752688172e-05,
+      "loss": 0.4006,
+      "step": 17450
+    },
+    {
+      "epoch": 112.64,
+      "learning_rate": 4.9297491039426526e-05,
+      "loss": 0.3633,
+      "step": 17460
+    },
+    {
+      "epoch": 112.71,
+      "learning_rate": 4.9293906810035845e-05,
+      "loss": 0.378,
+      "step": 17470
+    },
+    {
+      "epoch": 112.77,
+      "learning_rate": 4.929032258064516e-05,
+      "loss": 0.3494,
+      "step": 17480
+    },
+    {
+      "epoch": 112.84,
+      "learning_rate": 4.928673835125448e-05,
+      "loss": 0.3484,
+      "step": 17490
+    },
+    {
+      "epoch": 112.9,
+      "learning_rate": 4.92831541218638e-05,
+      "loss": 0.3472,
+      "step": 17500
+    },
+    {
+      "epoch": 112.96,
+      "learning_rate": 4.9279569892473124e-05,
+      "loss": 0.3856,
+      "step": 17510
+    },
+    {
+      "epoch": 113.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9257412552833557,
+      "eval_runtime": 0.8544,
+      "eval_samples_per_second": 235.241,
+      "eval_steps_per_second": 8.192,
+      "step": 17515
+    },
+    {
+      "epoch": 113.03,
+      "learning_rate": 4.927598566308244e-05,
+      "loss": 0.4114,
+      "step": 17520
+    },
+    {
+      "epoch": 113.1,
+      "learning_rate": 4.9272401433691754e-05,
+      "loss": 0.3662,
+      "step": 17530
+    },
+    {
+      "epoch": 113.16,
+      "learning_rate": 4.926881720430108e-05,
+      "loss": 0.3352,
+      "step": 17540
+    },
+    {
+      "epoch": 113.23,
+      "learning_rate": 4.92652329749104e-05,
+      "loss": 0.3262,
+      "step": 17550
+    },
+    {
+      "epoch": 113.29,
+      "learning_rate": 4.9261648745519715e-05,
+      "loss": 0.3768,
+      "step": 17560
+    },
+    {
+      "epoch": 113.35,
+      "learning_rate": 4.9258064516129033e-05,
+      "loss": 0.3674,
+      "step": 17570
+    },
+    {
+      "epoch": 113.42,
+      "learning_rate": 4.925448028673835e-05,
+      "loss": 0.3859,
+      "step": 17580
+    },
+    {
+      "epoch": 113.48,
+      "learning_rate": 4.925089605734767e-05,
+      "loss": 0.3462,
+      "step": 17590
+    },
+    {
+      "epoch": 113.55,
+      "learning_rate": 4.9247311827956995e-05,
+      "loss": 0.3285,
+      "step": 17600
+    },
+    {
+      "epoch": 113.61,
+      "learning_rate": 4.9243727598566306e-05,
+      "loss": 0.3227,
+      "step": 17610
+    },
+    {
+      "epoch": 113.68,
+      "learning_rate": 4.924014336917563e-05,
+      "loss": 0.3353,
+      "step": 17620
+    },
+    {
+      "epoch": 113.74,
+      "learning_rate": 4.923655913978495e-05,
+      "loss": 0.4073,
+      "step": 17630
+    },
+    {
+      "epoch": 113.8,
+      "learning_rate": 4.923297491039427e-05,
+      "loss": 0.3446,
+      "step": 17640
+    },
+    {
+      "epoch": 113.87,
+      "learning_rate": 4.9229390681003586e-05,
+      "loss": 0.4031,
+      "step": 17650
+    },
+    {
+      "epoch": 113.93,
+      "learning_rate": 4.9225806451612904e-05,
+      "loss": 0.3733,
+      "step": 17660
+    },
+    {
+      "epoch": 114.0,
+      "learning_rate": 4.922222222222222e-05,
+      "loss": 0.2763,
+      "step": 17670
+    },
+    {
+      "epoch": 114.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9009706377983093,
+      "eval_runtime": 0.8525,
+      "eval_samples_per_second": 235.777,
+      "eval_steps_per_second": 8.211,
+      "step": 17670
+    },
+    {
+      "epoch": 114.06,
+      "learning_rate": 4.921863799283155e-05,
+      "loss": 0.3512,
+      "step": 17680
+    },
+    {
+      "epoch": 114.13,
+      "learning_rate": 4.921505376344086e-05,
+      "loss": 0.3555,
+      "step": 17690
+    },
+    {
+      "epoch": 114.19,
+      "learning_rate": 4.9211469534050184e-05,
+      "loss": 0.3335,
+      "step": 17700
+    },
+    {
+      "epoch": 114.26,
+      "learning_rate": 4.92078853046595e-05,
+      "loss": 0.3609,
+      "step": 17710
+    },
+    {
+      "epoch": 114.32,
+      "learning_rate": 4.920430107526882e-05,
+      "loss": 0.3318,
+      "step": 17720
+    },
+    {
+      "epoch": 114.39,
+      "learning_rate": 4.920071684587814e-05,
+      "loss": 0.3313,
+      "step": 17730
+    },
+    {
+      "epoch": 114.45,
+      "learning_rate": 4.9197132616487456e-05,
+      "loss": 0.3961,
+      "step": 17740
+    },
+    {
+      "epoch": 114.51,
+      "learning_rate": 4.9193548387096775e-05,
+      "loss": 0.3433,
+      "step": 17750
+    },
+    {
+      "epoch": 114.58,
+      "learning_rate": 4.91899641577061e-05,
+      "loss": 0.3435,
+      "step": 17760
+    },
+    {
+      "epoch": 114.64,
+      "learning_rate": 4.918637992831541e-05,
+      "loss": 0.4021,
+      "step": 17770
+    },
+    {
+      "epoch": 114.71,
+      "learning_rate": 4.9182795698924736e-05,
+      "loss": 0.3987,
+      "step": 17780
+    },
+    {
+      "epoch": 114.77,
+      "learning_rate": 4.9179211469534054e-05,
+      "loss": 0.3918,
+      "step": 17790
+    },
+    {
+      "epoch": 114.84,
+      "learning_rate": 4.917562724014337e-05,
+      "loss": 0.3314,
+      "step": 17800
+    },
+    {
+      "epoch": 114.9,
+      "learning_rate": 4.917204301075269e-05,
+      "loss": 0.3425,
+      "step": 17810
+    },
+    {
+      "epoch": 114.96,
+      "learning_rate": 4.916845878136201e-05,
+      "loss": 0.3064,
+      "step": 17820
+    },
+    {
+      "epoch": 115.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9248042106628418,
+      "eval_runtime": 0.85,
+      "eval_samples_per_second": 236.464,
+      "eval_steps_per_second": 8.235,
+      "step": 17825
+    },
+    {
+      "epoch": 115.03,
+      "learning_rate": 4.916487455197133e-05,
+      "loss": 0.3472,
+      "step": 17830
+    },
+    {
+      "epoch": 115.1,
+      "learning_rate": 4.916129032258065e-05,
+      "loss": 0.3129,
+      "step": 17840
+    },
+    {
+      "epoch": 115.16,
+      "learning_rate": 4.9157706093189964e-05,
+      "loss": 0.3497,
+      "step": 17850
+    },
+    {
+      "epoch": 115.23,
+      "learning_rate": 4.915412186379928e-05,
+      "loss": 0.3741,
+      "step": 17860
+    },
+    {
+      "epoch": 115.29,
+      "learning_rate": 4.915053763440861e-05,
+      "loss": 0.3871,
+      "step": 17870
+    },
+    {
+      "epoch": 115.35,
+      "learning_rate": 4.9146953405017925e-05,
+      "loss": 0.3465,
+      "step": 17880
+    },
+    {
+      "epoch": 115.42,
+      "learning_rate": 4.914336917562724e-05,
+      "loss": 0.3189,
+      "step": 17890
+    },
+    {
+      "epoch": 115.48,
+      "learning_rate": 4.913978494623656e-05,
+      "loss": 0.3618,
+      "step": 17900
+    },
+    {
+      "epoch": 115.55,
+      "learning_rate": 4.913620071684588e-05,
+      "loss": 0.3445,
+      "step": 17910
+    },
+    {
+      "epoch": 115.61,
+      "learning_rate": 4.9132616487455205e-05,
+      "loss": 0.3015,
+      "step": 17920
+    },
+    {
+      "epoch": 115.68,
+      "learning_rate": 4.9129032258064516e-05,
+      "loss": 0.2834,
+      "step": 17930
+    },
+    {
+      "epoch": 115.74,
+      "learning_rate": 4.9125448028673834e-05,
+      "loss": 0.3166,
+      "step": 17940
+    },
+    {
+      "epoch": 115.8,
+      "learning_rate": 4.912186379928316e-05,
+      "loss": 0.3663,
+      "step": 17950
+    },
+    {
+      "epoch": 115.87,
+      "learning_rate": 4.911827956989248e-05,
+      "loss": 0.3518,
+      "step": 17960
+    },
+    {
+      "epoch": 115.93,
+      "learning_rate": 4.9114695340501796e-05,
+      "loss": 0.4029,
+      "step": 17970
+    },
+    {
+      "epoch": 116.0,
+      "learning_rate": 4.9111111111111114e-05,
+      "loss": 0.4289,
+      "step": 17980
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9713186621665955,
+      "eval_runtime": 0.8653,
+      "eval_samples_per_second": 232.298,
+      "eval_steps_per_second": 8.09,
+      "step": 17980
+    },
+    {
+      "epoch": 116.06,
+      "learning_rate": 4.910752688172043e-05,
+      "loss": 0.4417,
+      "step": 17990
+    },
+    {
+      "epoch": 116.13,
+      "learning_rate": 4.910394265232976e-05,
+      "loss": 0.3164,
+      "step": 18000
+    },
+    {
+      "epoch": 116.19,
+      "learning_rate": 4.910035842293907e-05,
+      "loss": 0.3486,
+      "step": 18010
+    },
+    {
+      "epoch": 116.26,
+      "learning_rate": 4.9096774193548387e-05,
+      "loss": 0.3534,
+      "step": 18020
+    },
+    {
+      "epoch": 116.32,
+      "learning_rate": 4.909318996415771e-05,
+      "loss": 0.3265,
+      "step": 18030
+    },
+    {
+      "epoch": 116.39,
+      "learning_rate": 4.908960573476703e-05,
+      "loss": 0.325,
+      "step": 18040
+    },
+    {
+      "epoch": 116.45,
+      "learning_rate": 4.908602150537635e-05,
+      "loss": 0.3842,
+      "step": 18050
+    },
+    {
+      "epoch": 116.51,
+      "learning_rate": 4.9082437275985666e-05,
+      "loss": 0.375,
+      "step": 18060
+    },
+    {
+      "epoch": 116.58,
+      "learning_rate": 4.9078853046594984e-05,
+      "loss": 0.4008,
+      "step": 18070
+    },
+    {
+      "epoch": 116.64,
+      "learning_rate": 4.90752688172043e-05,
+      "loss": 0.3657,
+      "step": 18080
+    },
+    {
+      "epoch": 116.71,
+      "learning_rate": 4.907168458781362e-05,
+      "loss": 0.3763,
+      "step": 18090
+    },
+    {
+      "epoch": 116.77,
+      "learning_rate": 4.906810035842294e-05,
+      "loss": 0.3947,
+      "step": 18100
+    },
+    {
+      "epoch": 116.84,
+      "learning_rate": 4.9064516129032264e-05,
+      "loss": 0.3604,
+      "step": 18110
+    },
+    {
+      "epoch": 116.9,
+      "learning_rate": 4.9060931899641575e-05,
+      "loss": 0.2975,
+      "step": 18120
+    },
+    {
+      "epoch": 116.96,
+      "learning_rate": 4.90573476702509e-05,
+      "loss": 0.3103,
+      "step": 18130
+    },
+    {
+      "epoch": 117.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9784727096557617,
+      "eval_runtime": 0.8523,
+      "eval_samples_per_second": 235.828,
+      "eval_steps_per_second": 8.213,
+      "step": 18135
+    },
+    {
+      "epoch": 117.03,
+      "learning_rate": 4.905376344086022e-05,
+      "loss": 0.3431,
+      "step": 18140
+    },
+    {
+      "epoch": 117.1,
+      "learning_rate": 4.905017921146954e-05,
+      "loss": 0.3604,
+      "step": 18150
+    },
+    {
+      "epoch": 117.16,
+      "learning_rate": 4.9046594982078855e-05,
+      "loss": 0.426,
+      "step": 18160
+    },
+    {
+      "epoch": 117.23,
+      "learning_rate": 4.904301075268817e-05,
+      "loss": 0.3589,
+      "step": 18170
+    },
+    {
+      "epoch": 117.29,
+      "learning_rate": 4.903942652329749e-05,
+      "loss": 0.3954,
+      "step": 18180
+    },
+    {
+      "epoch": 117.35,
+      "learning_rate": 4.9035842293906816e-05,
+      "loss": 0.3801,
+      "step": 18190
+    },
+    {
+      "epoch": 117.42,
+      "learning_rate": 4.903225806451613e-05,
+      "loss": 0.3694,
+      "step": 18200
+    },
+    {
+      "epoch": 117.48,
+      "learning_rate": 4.9028673835125446e-05,
+      "loss": 0.2953,
+      "step": 18210
+    },
+    {
+      "epoch": 117.55,
+      "learning_rate": 4.902508960573477e-05,
+      "loss": 0.4135,
+      "step": 18220
+    },
+    {
+      "epoch": 117.61,
+      "learning_rate": 4.902150537634409e-05,
+      "loss": 0.3128,
+      "step": 18230
+    },
+    {
+      "epoch": 117.68,
+      "learning_rate": 4.901792114695341e-05,
+      "loss": 0.3545,
+      "step": 18240
+    },
+    {
+      "epoch": 117.74,
+      "learning_rate": 4.9014336917562726e-05,
+      "loss": 0.3055,
+      "step": 18250
+    },
+    {
+      "epoch": 117.8,
+      "learning_rate": 4.9010752688172044e-05,
+      "loss": 0.332,
+      "step": 18260
+    },
+    {
+      "epoch": 117.87,
+      "learning_rate": 4.900716845878137e-05,
+      "loss": 0.352,
+      "step": 18270
+    },
+    {
+      "epoch": 117.93,
+      "learning_rate": 4.900358422939068e-05,
+      "loss": 0.3399,
+      "step": 18280
+    },
+    {
+      "epoch": 118.0,
+      "learning_rate": 4.9e-05,
+      "loss": 0.408,
+      "step": 18290
+    },
+    {
+      "epoch": 118.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0508743524551392,
+      "eval_runtime": 0.8462,
+      "eval_samples_per_second": 237.53,
+      "eval_steps_per_second": 8.272,
+      "step": 18290
+    },
+    {
+      "epoch": 118.06,
+      "learning_rate": 4.8996415770609323e-05,
+      "loss": 0.3797,
+      "step": 18300
+    },
+    {
+      "epoch": 118.13,
+      "learning_rate": 4.899283154121864e-05,
+      "loss": 0.3508,
+      "step": 18310
+    },
+    {
+      "epoch": 118.19,
+      "learning_rate": 4.898924731182796e-05,
+      "loss": 0.3166,
+      "step": 18320
+    },
+    {
+      "epoch": 118.26,
+      "learning_rate": 4.898566308243728e-05,
+      "loss": 0.392,
+      "step": 18330
+    },
+    {
+      "epoch": 118.32,
+      "learning_rate": 4.8982078853046596e-05,
+      "loss": 0.3745,
+      "step": 18340
+    },
+    {
+      "epoch": 118.39,
+      "learning_rate": 4.897849462365592e-05,
+      "loss": 0.3718,
+      "step": 18350
+    },
+    {
+      "epoch": 118.45,
+      "learning_rate": 4.897491039426523e-05,
+      "loss": 0.3796,
+      "step": 18360
+    },
+    {
+      "epoch": 118.51,
+      "learning_rate": 4.897132616487455e-05,
+      "loss": 0.3896,
+      "step": 18370
+    },
+    {
+      "epoch": 118.58,
+      "learning_rate": 4.8967741935483876e-05,
+      "loss": 0.3556,
+      "step": 18380
+    },
+    {
+      "epoch": 118.64,
+      "learning_rate": 4.8964157706093194e-05,
+      "loss": 0.312,
+      "step": 18390
+    },
+    {
+      "epoch": 118.71,
+      "learning_rate": 4.896057347670251e-05,
+      "loss": 0.3552,
+      "step": 18400
+    },
+    {
+      "epoch": 118.77,
+      "learning_rate": 4.895698924731183e-05,
+      "loss": 0.3534,
+      "step": 18410
+    },
+    {
+      "epoch": 118.84,
+      "learning_rate": 4.895340501792115e-05,
+      "loss": 0.3915,
+      "step": 18420
+    },
+    {
+      "epoch": 118.9,
+      "learning_rate": 4.8949820788530474e-05,
+      "loss": 0.3845,
+      "step": 18430
+    },
+    {
+      "epoch": 118.96,
+      "learning_rate": 4.8946236559139785e-05,
+      "loss": 0.3526,
+      "step": 18440
+    },
+    {
+      "epoch": 119.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9923292398452759,
+      "eval_runtime": 0.8404,
+      "eval_samples_per_second": 239.174,
+      "eval_steps_per_second": 8.329,
+      "step": 18445
+    },
+    {
+      "epoch": 119.03,
+      "learning_rate": 4.89426523297491e-05,
+      "loss": 0.3458,
+      "step": 18450
+    },
+    {
+      "epoch": 119.1,
+      "learning_rate": 4.893906810035843e-05,
+      "loss": 0.349,
+      "step": 18460
+    },
+    {
+      "epoch": 119.16,
+      "learning_rate": 4.8935483870967746e-05,
+      "loss": 0.335,
+      "step": 18470
+    },
+    {
+      "epoch": 119.23,
+      "learning_rate": 4.8931899641577065e-05,
+      "loss": 0.3535,
+      "step": 18480
+    },
+    {
+      "epoch": 119.29,
+      "learning_rate": 4.892831541218638e-05,
+      "loss": 0.3814,
+      "step": 18490
+    },
+    {
+      "epoch": 119.35,
+      "learning_rate": 4.89247311827957e-05,
+      "loss": 0.2976,
+      "step": 18500
+    },
+    {
+      "epoch": 119.42,
+      "learning_rate": 4.892114695340502e-05,
+      "loss": 0.3319,
+      "step": 18510
+    },
+    {
+      "epoch": 119.48,
+      "learning_rate": 4.891756272401434e-05,
+      "loss": 0.3966,
+      "step": 18520
+    },
+    {
+      "epoch": 119.55,
+      "learning_rate": 4.8913978494623656e-05,
+      "loss": 0.3432,
+      "step": 18530
+    },
+    {
+      "epoch": 119.61,
+      "learning_rate": 4.891039426523298e-05,
+      "loss": 0.3638,
+      "step": 18540
+    },
+    {
+      "epoch": 119.68,
+      "learning_rate": 4.89068100358423e-05,
+      "loss": 0.3589,
+      "step": 18550
+    },
+    {
+      "epoch": 119.74,
+      "learning_rate": 4.890322580645161e-05,
+      "loss": 0.3231,
+      "step": 18560
+    },
+    {
+      "epoch": 119.8,
+      "learning_rate": 4.8899641577060935e-05,
+      "loss": 0.3628,
+      "step": 18570
+    },
+    {
+      "epoch": 119.87,
+      "learning_rate": 4.8896057347670253e-05,
+      "loss": 0.359,
+      "step": 18580
+    },
+    {
+      "epoch": 119.93,
+      "learning_rate": 4.889247311827957e-05,
+      "loss": 0.3364,
+      "step": 18590
+    },
+    {
+      "epoch": 120.0,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 0.319,
+      "step": 18600
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.010260820388794,
+      "eval_runtime": 0.8393,
+      "eval_samples_per_second": 239.491,
+      "eval_steps_per_second": 8.34,
+      "step": 18600
+    },
+    {
+      "epoch": 120.06,
+      "learning_rate": 4.888530465949821e-05,
+      "loss": 0.4486,
+      "step": 18610
+    },
+    {
+      "epoch": 120.13,
+      "learning_rate": 4.888172043010753e-05,
+      "loss": 0.3376,
+      "step": 18620
+    },
+    {
+      "epoch": 120.19,
+      "learning_rate": 4.887813620071685e-05,
+      "loss": 0.3239,
+      "step": 18630
+    },
+    {
+      "epoch": 120.26,
+      "learning_rate": 4.887455197132616e-05,
+      "loss": 0.3723,
+      "step": 18640
+    },
+    {
+      "epoch": 120.32,
+      "learning_rate": 4.887096774193549e-05,
+      "loss": 0.3875,
+      "step": 18650
+    },
+    {
+      "epoch": 120.39,
+      "learning_rate": 4.8867383512544806e-05,
+      "loss": 0.356,
+      "step": 18660
+    },
+    {
+      "epoch": 120.45,
+      "learning_rate": 4.8863799283154124e-05,
+      "loss": 0.3814,
+      "step": 18670
+    },
+    {
+      "epoch": 120.51,
+      "learning_rate": 4.886021505376344e-05,
+      "loss": 0.316,
+      "step": 18680
+    },
+    {
+      "epoch": 120.58,
+      "learning_rate": 4.885663082437276e-05,
+      "loss": 0.3511,
+      "step": 18690
+    },
+    {
+      "epoch": 120.64,
+      "learning_rate": 4.8853046594982085e-05,
+      "loss": 0.3924,
+      "step": 18700
+    },
+    {
+      "epoch": 120.71,
+      "learning_rate": 4.88494623655914e-05,
+      "loss": 0.3582,
+      "step": 18710
+    },
+    {
+      "epoch": 120.77,
+      "learning_rate": 4.8845878136200715e-05,
+      "loss": 0.3178,
+      "step": 18720
+    },
+    {
+      "epoch": 120.84,
+      "learning_rate": 4.884229390681004e-05,
+      "loss": 0.3531,
+      "step": 18730
+    },
+    {
+      "epoch": 120.9,
+      "learning_rate": 4.883870967741936e-05,
+      "loss": 0.3422,
+      "step": 18740
+    },
+    {
+      "epoch": 120.96,
+      "learning_rate": 4.8835125448028677e-05,
+      "loss": 0.3761,
+      "step": 18750
+    },
+    {
+      "epoch": 121.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9890701770782471,
+      "eval_runtime": 0.8484,
+      "eval_samples_per_second": 236.906,
+      "eval_steps_per_second": 8.25,
+      "step": 18755
+    },
+    {
+      "epoch": 121.03,
+      "learning_rate": 4.8831541218637995e-05,
+      "loss": 0.3516,
+      "step": 18760
+    },
+    {
+      "epoch": 121.1,
+      "learning_rate": 4.882795698924731e-05,
+      "loss": 0.2861,
+      "step": 18770
+    },
+    {
+      "epoch": 121.16,
+      "learning_rate": 4.882437275985664e-05,
+      "loss": 0.3082,
+      "step": 18780
+    },
+    {
+      "epoch": 121.23,
+      "learning_rate": 4.882078853046595e-05,
+      "loss": 0.393,
+      "step": 18790
+    },
+    {
+      "epoch": 121.29,
+      "learning_rate": 4.881720430107527e-05,
+      "loss": 0.3471,
+      "step": 18800
+    },
+    {
+      "epoch": 121.35,
+      "learning_rate": 4.881362007168459e-05,
+      "loss": 0.3825,
+      "step": 18810
+    },
+    {
+      "epoch": 121.42,
+      "learning_rate": 4.881003584229391e-05,
+      "loss": 0.3126,
+      "step": 18820
+    },
+    {
+      "epoch": 121.48,
+      "learning_rate": 4.880645161290323e-05,
+      "loss": 0.3495,
+      "step": 18830
+    },
+    {
+      "epoch": 121.55,
+      "learning_rate": 4.880286738351255e-05,
+      "loss": 0.3338,
+      "step": 18840
+    },
+    {
+      "epoch": 121.61,
+      "learning_rate": 4.8799283154121865e-05,
+      "loss": 0.3456,
+      "step": 18850
+    },
+    {
+      "epoch": 121.68,
+      "learning_rate": 4.8795698924731184e-05,
+      "loss": 0.331,
+      "step": 18860
+    },
+    {
+      "epoch": 121.74,
+      "learning_rate": 4.87921146953405e-05,
+      "loss": 0.3726,
+      "step": 18870
+    },
+    {
+      "epoch": 121.8,
+      "learning_rate": 4.878853046594982e-05,
+      "loss": 0.3544,
+      "step": 18880
+    },
+    {
+      "epoch": 121.87,
+      "learning_rate": 4.8784946236559145e-05,
+      "loss": 0.4035,
+      "step": 18890
+    },
+    {
+      "epoch": 121.93,
+      "learning_rate": 4.878136200716846e-05,
+      "loss": 0.3533,
+      "step": 18900
+    },
+    {
+      "epoch": 122.0,
+      "learning_rate": 4.8777777777777775e-05,
+      "loss": 0.3705,
+      "step": 18910
+    },
+    {
+      "epoch": 122.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9804939031600952,
+      "eval_runtime": 0.8549,
+      "eval_samples_per_second": 235.108,
+      "eval_steps_per_second": 8.188,
+      "step": 18910
+    },
+    {
+      "epoch": 122.06,
+      "learning_rate": 4.87741935483871e-05,
+      "loss": 0.3365,
+      "step": 18920
+    },
+    {
+      "epoch": 122.13,
+      "learning_rate": 4.877060931899642e-05,
+      "loss": 0.3224,
+      "step": 18930
+    },
+    {
+      "epoch": 122.19,
+      "learning_rate": 4.8767025089605736e-05,
+      "loss": 0.3233,
+      "step": 18940
+    },
+    {
+      "epoch": 122.26,
+      "learning_rate": 4.8763440860215054e-05,
+      "loss": 0.3404,
+      "step": 18950
+    },
+    {
+      "epoch": 122.32,
+      "learning_rate": 4.875985663082437e-05,
+      "loss": 0.3109,
+      "step": 18960
+    },
+    {
+      "epoch": 122.39,
+      "learning_rate": 4.87562724014337e-05,
+      "loss": 0.3231,
+      "step": 18970
+    },
+    {
+      "epoch": 122.45,
+      "learning_rate": 4.8752688172043016e-05,
+      "loss": 0.4158,
+      "step": 18980
+    },
+    {
+      "epoch": 122.51,
+      "learning_rate": 4.874910394265233e-05,
+      "loss": 0.3324,
+      "step": 18990
+    },
+    {
+      "epoch": 122.58,
+      "learning_rate": 4.874551971326165e-05,
+      "loss": 0.3448,
+      "step": 19000
+    },
+    {
+      "epoch": 122.64,
+      "learning_rate": 4.874193548387097e-05,
+      "loss": 0.3915,
+      "step": 19010
+    },
+    {
+      "epoch": 122.71,
+      "learning_rate": 4.873835125448029e-05,
+      "loss": 0.3435,
+      "step": 19020
+    },
+    {
+      "epoch": 122.77,
+      "learning_rate": 4.8734767025089607e-05,
+      "loss": 0.3383,
+      "step": 19030
+    },
+    {
+      "epoch": 122.84,
+      "learning_rate": 4.8731182795698925e-05,
+      "loss": 0.376,
+      "step": 19040
+    },
+    {
+      "epoch": 122.9,
+      "learning_rate": 4.872759856630825e-05,
+      "loss": 0.3504,
+      "step": 19050
+    },
+    {
+      "epoch": 122.96,
+      "learning_rate": 4.872401433691757e-05,
+      "loss": 0.3734,
+      "step": 19060
+    },
+    {
+      "epoch": 123.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9798120260238647,
+      "eval_runtime": 0.8435,
+      "eval_samples_per_second": 238.307,
+      "eval_steps_per_second": 8.299,
+      "step": 19065
+    },
+    {
+      "epoch": 123.03,
+      "learning_rate": 4.872043010752688e-05,
+      "loss": 0.3615,
+      "step": 19070
+    },
+    {
+      "epoch": 123.1,
+      "learning_rate": 4.8716845878136204e-05,
+      "loss": 0.3687,
+      "step": 19080
+    },
+    {
+      "epoch": 123.16,
+      "learning_rate": 4.871326164874552e-05,
+      "loss": 0.3255,
+      "step": 19090
+    },
+    {
+      "epoch": 123.23,
+      "learning_rate": 4.870967741935484e-05,
+      "loss": 0.3608,
+      "step": 19100
+    },
+    {
+      "epoch": 123.29,
+      "learning_rate": 4.870609318996416e-05,
+      "loss": 0.3268,
+      "step": 19110
+    },
+    {
+      "epoch": 123.35,
+      "learning_rate": 4.870250896057348e-05,
+      "loss": 0.3324,
+      "step": 19120
+    },
+    {
+      "epoch": 123.42,
+      "learning_rate": 4.86989247311828e-05,
+      "loss": 0.4141,
+      "step": 19130
+    },
+    {
+      "epoch": 123.48,
+      "learning_rate": 4.869534050179212e-05,
+      "loss": 0.3581,
+      "step": 19140
+    },
+    {
+      "epoch": 123.55,
+      "learning_rate": 4.869175627240143e-05,
+      "loss": 0.3323,
+      "step": 19150
+    },
+    {
+      "epoch": 123.61,
+      "learning_rate": 4.868817204301076e-05,
+      "loss": 0.3629,
+      "step": 19160
+    },
+    {
+      "epoch": 123.68,
+      "learning_rate": 4.8684587813620075e-05,
+      "loss": 0.4002,
+      "step": 19170
+    },
+    {
+      "epoch": 123.74,
+      "learning_rate": 4.868100358422939e-05,
+      "loss": 0.3925,
+      "step": 19180
+    },
+    {
+      "epoch": 123.8,
+      "learning_rate": 4.867741935483871e-05,
+      "loss": 0.4041,
+      "step": 19190
+    },
+    {
+      "epoch": 123.87,
+      "learning_rate": 4.867383512544803e-05,
+      "loss": 0.3213,
+      "step": 19200
+    },
+    {
+      "epoch": 123.93,
+      "learning_rate": 4.867025089605735e-05,
+      "loss": 0.359,
+      "step": 19210
+    },
+    {
+      "epoch": 124.0,
+      "learning_rate": 4.866666666666667e-05,
+      "loss": 0.357,
+      "step": 19220
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0030487775802612,
+      "eval_runtime": 0.8645,
+      "eval_samples_per_second": 232.506,
+      "eval_steps_per_second": 8.097,
+      "step": 19220
+    },
+    {
+      "epoch": 124.06,
+      "learning_rate": 4.8663082437275984e-05,
+      "loss": 0.3651,
+      "step": 19230
+    },
+    {
+      "epoch": 124.13,
+      "learning_rate": 4.865949820788531e-05,
+      "loss": 0.3308,
+      "step": 19240
+    },
+    {
+      "epoch": 124.19,
+      "learning_rate": 4.865591397849463e-05,
+      "loss": 0.3687,
+      "step": 19250
+    },
+    {
+      "epoch": 124.26,
+      "learning_rate": 4.8652329749103946e-05,
+      "loss": 0.3339,
+      "step": 19260
+    },
+    {
+      "epoch": 124.32,
+      "learning_rate": 4.8648745519713264e-05,
+      "loss": 0.3475,
+      "step": 19270
+    },
+    {
+      "epoch": 124.39,
+      "learning_rate": 4.864516129032258e-05,
+      "loss": 0.3412,
+      "step": 19280
+    },
+    {
+      "epoch": 124.45,
+      "learning_rate": 4.86415770609319e-05,
+      "loss": 0.339,
+      "step": 19290
+    },
+    {
+      "epoch": 124.51,
+      "learning_rate": 4.8637992831541225e-05,
+      "loss": 0.3118,
+      "step": 19300
+    },
+    {
+      "epoch": 124.58,
+      "learning_rate": 4.863440860215054e-05,
+      "loss": 0.3065,
+      "step": 19310
+    },
+    {
+      "epoch": 124.64,
+      "learning_rate": 4.863082437275986e-05,
+      "loss": 0.3915,
+      "step": 19320
+    },
+    {
+      "epoch": 124.71,
+      "learning_rate": 4.862724014336918e-05,
+      "loss": 0.3949,
+      "step": 19330
+    },
+    {
+      "epoch": 124.77,
+      "learning_rate": 4.862365591397849e-05,
+      "loss": 0.35,
+      "step": 19340
+    },
+    {
+      "epoch": 124.84,
+      "learning_rate": 4.8620071684587816e-05,
+      "loss": 0.3692,
+      "step": 19350
+    },
+    {
+      "epoch": 124.9,
+      "learning_rate": 4.8616487455197134e-05,
+      "loss": 0.3981,
+      "step": 19360
+    },
+    {
+      "epoch": 124.96,
+      "learning_rate": 4.861290322580645e-05,
+      "loss": 0.4097,
+      "step": 19370
+    },
+    {
+      "epoch": 125.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0437721014022827,
+      "eval_runtime": 0.8666,
+      "eval_samples_per_second": 231.944,
+      "eval_steps_per_second": 8.078,
+      "step": 19375
+    },
+    {
+      "epoch": 125.03,
+      "learning_rate": 4.860931899641577e-05,
+      "loss": 0.3419,
+      "step": 19380
+    },
+    {
+      "epoch": 125.1,
+      "learning_rate": 4.860573476702509e-05,
+      "loss": 0.3506,
+      "step": 19390
+    },
+    {
+      "epoch": 125.16,
+      "learning_rate": 4.8602150537634414e-05,
+      "loss": 0.3809,
+      "step": 19400
+    },
+    {
+      "epoch": 125.23,
+      "learning_rate": 4.859856630824373e-05,
+      "loss": 0.3845,
+      "step": 19410
+    },
+    {
+      "epoch": 125.29,
+      "learning_rate": 4.8594982078853044e-05,
+      "loss": 0.3416,
+      "step": 19420
+    },
+    {
+      "epoch": 125.35,
+      "learning_rate": 4.859139784946237e-05,
+      "loss": 0.35,
+      "step": 19430
+    },
+    {
+      "epoch": 125.42,
+      "learning_rate": 4.858781362007169e-05,
+      "loss": 0.3995,
+      "step": 19440
+    },
+    {
+      "epoch": 125.48,
+      "learning_rate": 4.8584229390681005e-05,
+      "loss": 0.2986,
+      "step": 19450
+    },
+    {
+      "epoch": 125.55,
+      "learning_rate": 4.858064516129032e-05,
+      "loss": 0.325,
+      "step": 19460
+    },
+    {
+      "epoch": 125.61,
+      "learning_rate": 4.857706093189964e-05,
+      "loss": 0.3829,
+      "step": 19470
+    },
+    {
+      "epoch": 125.68,
+      "learning_rate": 4.8573476702508966e-05,
+      "loss": 0.3023,
+      "step": 19480
+    },
+    {
+      "epoch": 125.74,
+      "learning_rate": 4.8569892473118285e-05,
+      "loss": 0.334,
+      "step": 19490
+    },
+    {
+      "epoch": 125.8,
+      "learning_rate": 4.8566308243727596e-05,
+      "loss": 0.4159,
+      "step": 19500
+    },
+    {
+      "epoch": 125.87,
+      "learning_rate": 4.856272401433692e-05,
+      "loss": 0.3443,
+      "step": 19510
+    },
+    {
+      "epoch": 125.93,
+      "learning_rate": 4.855913978494624e-05,
+      "loss": 0.3208,
+      "step": 19520
+    },
+    {
+      "epoch": 126.0,
+      "learning_rate": 4.855555555555556e-05,
+      "loss": 0.3525,
+      "step": 19530
+    },
+    {
+      "epoch": 126.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0122992992401123,
+      "eval_runtime": 0.8676,
+      "eval_samples_per_second": 231.667,
+      "eval_steps_per_second": 8.068,
+      "step": 19530
+    },
+    {
+      "epoch": 126.06,
+      "learning_rate": 4.8551971326164876e-05,
+      "loss": 0.3408,
+      "step": 19540
+    },
+    {
+      "epoch": 126.13,
+      "learning_rate": 4.8548387096774194e-05,
+      "loss": 0.3888,
+      "step": 19550
+    },
+    {
+      "epoch": 126.19,
+      "learning_rate": 4.854480286738351e-05,
+      "loss": 0.3713,
+      "step": 19560
+    },
+    {
+      "epoch": 126.26,
+      "learning_rate": 4.854121863799284e-05,
+      "loss": 0.3404,
+      "step": 19570
+    },
+    {
+      "epoch": 126.32,
+      "learning_rate": 4.853763440860215e-05,
+      "loss": 0.38,
+      "step": 19580
+    },
+    {
+      "epoch": 126.39,
+      "learning_rate": 4.8534050179211473e-05,
+      "loss": 0.3199,
+      "step": 19590
+    },
+    {
+      "epoch": 126.45,
+      "learning_rate": 4.853046594982079e-05,
+      "loss": 0.3671,
+      "step": 19600
+    },
+    {
+      "epoch": 126.51,
+      "learning_rate": 4.852688172043011e-05,
+      "loss": 0.2969,
+      "step": 19610
+    },
+    {
+      "epoch": 126.58,
+      "learning_rate": 4.852329749103943e-05,
+      "loss": 0.4109,
+      "step": 19620
+    },
+    {
+      "epoch": 126.64,
+      "learning_rate": 4.8519713261648746e-05,
+      "loss": 0.408,
+      "step": 19630
+    },
+    {
+      "epoch": 126.71,
+      "learning_rate": 4.8516129032258065e-05,
+      "loss": 0.3581,
+      "step": 19640
+    },
+    {
+      "epoch": 126.77,
+      "learning_rate": 4.851254480286739e-05,
+      "loss": 0.3105,
+      "step": 19650
+    },
+    {
+      "epoch": 126.84,
+      "learning_rate": 4.85089605734767e-05,
+      "loss": 0.4126,
+      "step": 19660
+    },
+    {
+      "epoch": 126.9,
+      "learning_rate": 4.8505376344086026e-05,
+      "loss": 0.3165,
+      "step": 19670
+    },
+    {
+      "epoch": 126.96,
+      "learning_rate": 4.8501792114695344e-05,
+      "loss": 0.3324,
+      "step": 19680
+    },
+    {
+      "epoch": 127.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.922193169593811,
+      "eval_runtime": 0.8376,
+      "eval_samples_per_second": 239.978,
+      "eval_steps_per_second": 8.357,
+      "step": 19685
+    },
+    {
+      "epoch": 127.03,
+      "learning_rate": 4.849820788530466e-05,
+      "loss": 0.3116,
+      "step": 19690
+    },
+    {
+      "epoch": 127.1,
+      "learning_rate": 4.849462365591398e-05,
+      "loss": 0.3799,
+      "step": 19700
+    },
+    {
+      "epoch": 127.16,
+      "learning_rate": 4.84910394265233e-05,
+      "loss": 0.3637,
+      "step": 19710
+    },
+    {
+      "epoch": 127.23,
+      "learning_rate": 4.848745519713262e-05,
+      "loss": 0.3557,
+      "step": 19720
+    },
+    {
+      "epoch": 127.29,
+      "learning_rate": 4.848387096774194e-05,
+      "loss": 0.3366,
+      "step": 19730
+    },
+    {
+      "epoch": 127.35,
+      "learning_rate": 4.848028673835125e-05,
+      "loss": 0.3646,
+      "step": 19740
+    },
+    {
+      "epoch": 127.42,
+      "learning_rate": 4.847670250896058e-05,
+      "loss": 0.337,
+      "step": 19750
+    },
+    {
+      "epoch": 127.48,
+      "learning_rate": 4.8473118279569897e-05,
+      "loss": 0.3552,
+      "step": 19760
+    },
+    {
+      "epoch": 127.55,
+      "learning_rate": 4.8469534050179215e-05,
+      "loss": 0.3936,
+      "step": 19770
+    },
+    {
+      "epoch": 127.61,
+      "learning_rate": 4.846594982078853e-05,
+      "loss": 0.3656,
+      "step": 19780
+    },
+    {
+      "epoch": 127.68,
+      "learning_rate": 4.846236559139785e-05,
+      "loss": 0.4082,
+      "step": 19790
+    },
+    {
+      "epoch": 127.74,
+      "learning_rate": 4.845878136200717e-05,
+      "loss": 0.3517,
+      "step": 19800
+    },
+    {
+      "epoch": 127.8,
+      "learning_rate": 4.8455197132616494e-05,
+      "loss": 0.334,
+      "step": 19810
+    },
+    {
+      "epoch": 127.87,
+      "learning_rate": 4.8451612903225806e-05,
+      "loss": 0.3197,
+      "step": 19820
+    },
+    {
+      "epoch": 127.93,
+      "learning_rate": 4.844802867383513e-05,
+      "loss": 0.3884,
+      "step": 19830
+    },
+    {
+      "epoch": 128.0,
+      "learning_rate": 4.844444444444445e-05,
+      "loss": 0.3752,
+      "step": 19840
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0103905200958252,
+      "eval_runtime": 0.8501,
+      "eval_samples_per_second": 236.456,
+      "eval_steps_per_second": 8.235,
+      "step": 19840
+    },
+    {
+      "epoch": 128.06,
+      "learning_rate": 4.844086021505377e-05,
+      "loss": 0.2798,
+      "step": 19850
+    },
+    {
+      "epoch": 128.13,
+      "learning_rate": 4.8437275985663085e-05,
+      "loss": 0.3619,
+      "step": 19860
+    },
+    {
+      "epoch": 128.19,
+      "learning_rate": 4.8433691756272404e-05,
+      "loss": 0.333,
+      "step": 19870
+    },
+    {
+      "epoch": 128.26,
+      "learning_rate": 4.843010752688172e-05,
+      "loss": 0.356,
+      "step": 19880
+    },
+    {
+      "epoch": 128.32,
+      "learning_rate": 4.842652329749105e-05,
+      "loss": 0.337,
+      "step": 19890
+    },
+    {
+      "epoch": 128.39,
+      "learning_rate": 4.842293906810036e-05,
+      "loss": 0.3534,
+      "step": 19900
+    },
+    {
+      "epoch": 128.45,
+      "learning_rate": 4.8419354838709676e-05,
+      "loss": 0.3767,
+      "step": 19910
+    },
+    {
+      "epoch": 128.51,
+      "learning_rate": 4.8415770609319e-05,
+      "loss": 0.3098,
+      "step": 19920
+    },
+    {
+      "epoch": 128.58,
+      "learning_rate": 4.841218637992832e-05,
+      "loss": 0.3427,
+      "step": 19930
+    },
+    {
+      "epoch": 128.64,
+      "learning_rate": 4.840860215053764e-05,
+      "loss": 0.3803,
+      "step": 19940
+    },
+    {
+      "epoch": 128.71,
+      "learning_rate": 4.8405017921146956e-05,
+      "loss": 0.3291,
+      "step": 19950
+    },
+    {
+      "epoch": 128.77,
+      "learning_rate": 4.8401433691756274e-05,
+      "loss": 0.3035,
+      "step": 19960
+    },
+    {
+      "epoch": 128.84,
+      "learning_rate": 4.83978494623656e-05,
+      "loss": 0.32,
+      "step": 19970
+    },
+    {
+      "epoch": 128.9,
+      "learning_rate": 4.839426523297491e-05,
+      "loss": 0.3536,
+      "step": 19980
+    },
+    {
+      "epoch": 128.96,
+      "learning_rate": 4.839068100358423e-05,
+      "loss": 0.3698,
+      "step": 19990
+    },
+    {
+      "epoch": 129.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9661576747894287,
+      "eval_runtime": 0.8841,
+      "eval_samples_per_second": 227.339,
+      "eval_steps_per_second": 7.917,
+      "step": 19995
+    },
+    {
+      "epoch": 129.03,
+      "learning_rate": 4.8387096774193554e-05,
+      "loss": 0.3718,
+      "step": 20000
+    },
+    {
+      "epoch": 129.1,
+      "learning_rate": 4.8383512544802865e-05,
+      "loss": 0.3912,
+      "step": 20010
+    },
+    {
+      "epoch": 129.16,
+      "learning_rate": 4.837992831541219e-05,
+      "loss": 0.3444,
+      "step": 20020
+    },
+    {
+      "epoch": 129.23,
+      "learning_rate": 4.837634408602151e-05,
+      "loss": 0.3285,
+      "step": 20030
+    },
+    {
+      "epoch": 129.29,
+      "learning_rate": 4.8372759856630827e-05,
+      "loss": 0.3614,
+      "step": 20040
+    },
+    {
+      "epoch": 129.35,
+      "learning_rate": 4.8369175627240145e-05,
+      "loss": 0.3193,
+      "step": 20050
+    },
+    {
+      "epoch": 129.42,
+      "learning_rate": 4.836559139784946e-05,
+      "loss": 0.3695,
+      "step": 20060
+    },
+    {
+      "epoch": 129.48,
+      "learning_rate": 4.836200716845878e-05,
+      "loss": 0.3692,
+      "step": 20070
+    },
+    {
+      "epoch": 129.55,
+      "learning_rate": 4.8358422939068106e-05,
+      "loss": 0.3684,
+      "step": 20080
+    },
+    {
+      "epoch": 129.61,
+      "learning_rate": 4.835483870967742e-05,
+      "loss": 0.3623,
+      "step": 20090
+    },
+    {
+      "epoch": 129.68,
+      "learning_rate": 4.835125448028674e-05,
+      "loss": 0.3133,
+      "step": 20100
+    },
+    {
+      "epoch": 129.74,
+      "learning_rate": 4.834767025089606e-05,
+      "loss": 0.3335,
+      "step": 20110
+    },
+    {
+      "epoch": 129.8,
+      "learning_rate": 4.834408602150538e-05,
+      "loss": 0.341,
+      "step": 20120
+    },
+    {
+      "epoch": 129.87,
+      "learning_rate": 4.83405017921147e-05,
+      "loss": 0.3286,
+      "step": 20130
+    },
+    {
+      "epoch": 129.93,
+      "learning_rate": 4.8336917562724015e-05,
+      "loss": 0.3261,
+      "step": 20140
+    },
+    {
+      "epoch": 130.0,
+      "learning_rate": 4.8333333333333334e-05,
+      "loss": 0.3647,
+      "step": 20150
+    },
+    {
+      "epoch": 130.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.945006251335144,
+      "eval_runtime": 0.861,
+      "eval_samples_per_second": 233.452,
+      "eval_steps_per_second": 8.13,
+      "step": 20150
+    },
+    {
+      "epoch": 130.06,
+      "learning_rate": 4.832974910394266e-05,
+      "loss": 0.3756,
+      "step": 20160
+    },
+    {
+      "epoch": 130.13,
+      "learning_rate": 4.832616487455197e-05,
+      "loss": 0.2802,
+      "step": 20170
+    },
+    {
+      "epoch": 130.19,
+      "learning_rate": 4.8322580645161295e-05,
+      "loss": 0.3667,
+      "step": 20180
+    },
+    {
+      "epoch": 130.26,
+      "learning_rate": 4.831899641577061e-05,
+      "loss": 0.3198,
+      "step": 20190
+    },
+    {
+      "epoch": 130.32,
+      "learning_rate": 4.831541218637993e-05,
+      "loss": 0.3573,
+      "step": 20200
+    },
+    {
+      "epoch": 130.39,
+      "learning_rate": 4.831182795698925e-05,
+      "loss": 0.4393,
+      "step": 20210
+    },
+    {
+      "epoch": 130.45,
+      "learning_rate": 4.830824372759857e-05,
+      "loss": 0.3508,
+      "step": 20220
+    },
+    {
+      "epoch": 130.51,
+      "learning_rate": 4.8304659498207886e-05,
+      "loss": 0.3677,
+      "step": 20230
+    },
+    {
+      "epoch": 130.58,
+      "learning_rate": 4.830107526881721e-05,
+      "loss": 0.3526,
+      "step": 20240
+    },
+    {
+      "epoch": 130.64,
+      "learning_rate": 4.829749103942652e-05,
+      "loss": 0.3851,
+      "step": 20250
+    },
+    {
+      "epoch": 130.71,
+      "learning_rate": 4.829390681003584e-05,
+      "loss": 0.338,
+      "step": 20260
+    },
+    {
+      "epoch": 130.77,
+      "learning_rate": 4.8290322580645166e-05,
+      "loss": 0.3726,
+      "step": 20270
+    },
+    {
+      "epoch": 130.84,
+      "learning_rate": 4.8286738351254484e-05,
+      "loss": 0.3762,
+      "step": 20280
+    },
+    {
+      "epoch": 130.9,
+      "learning_rate": 4.82831541218638e-05,
+      "loss": 0.3416,
+      "step": 20290
+    },
+    {
+      "epoch": 130.96,
+      "learning_rate": 4.827956989247312e-05,
+      "loss": 0.3773,
+      "step": 20300
+    },
+    {
+      "epoch": 131.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9183463454246521,
+      "eval_runtime": 0.8476,
+      "eval_samples_per_second": 237.129,
+      "eval_steps_per_second": 8.258,
+      "step": 20305
+    },
+    {
+      "epoch": 131.03,
+      "learning_rate": 4.827598566308244e-05,
+      "loss": 0.3683,
+      "step": 20310
+    },
+    {
+      "epoch": 131.1,
+      "learning_rate": 4.8272401433691763e-05,
+      "loss": 0.3472,
+      "step": 20320
+    },
+    {
+      "epoch": 131.16,
+      "learning_rate": 4.8268817204301075e-05,
+      "loss": 0.3077,
+      "step": 20330
+    },
+    {
+      "epoch": 131.23,
+      "learning_rate": 4.826523297491039e-05,
+      "loss": 0.3353,
+      "step": 20340
+    },
+    {
+      "epoch": 131.29,
+      "learning_rate": 4.826164874551972e-05,
+      "loss": 0.3598,
+      "step": 20350
+    },
+    {
+      "epoch": 131.35,
+      "learning_rate": 4.8258064516129036e-05,
+      "loss": 0.3358,
+      "step": 20360
+    },
+    {
+      "epoch": 131.42,
+      "learning_rate": 4.8254480286738354e-05,
+      "loss": 0.3502,
+      "step": 20370
+    },
+    {
+      "epoch": 131.48,
+      "learning_rate": 4.825089605734767e-05,
+      "loss": 0.3459,
+      "step": 20380
+    },
+    {
+      "epoch": 131.55,
+      "learning_rate": 4.824731182795699e-05,
+      "loss": 0.423,
+      "step": 20390
+    },
+    {
+      "epoch": 131.61,
+      "learning_rate": 4.8243727598566316e-05,
+      "loss": 0.3785,
+      "step": 20400
+    },
+    {
+      "epoch": 131.68,
+      "learning_rate": 4.824014336917563e-05,
+      "loss": 0.3059,
+      "step": 20410
+    },
+    {
+      "epoch": 131.74,
+      "learning_rate": 4.8236559139784945e-05,
+      "loss": 0.3666,
+      "step": 20420
+    },
+    {
+      "epoch": 131.8,
+      "learning_rate": 4.823297491039427e-05,
+      "loss": 0.2971,
+      "step": 20430
+    },
+    {
+      "epoch": 131.87,
+      "learning_rate": 4.822939068100359e-05,
+      "loss": 0.3678,
+      "step": 20440
+    },
+    {
+      "epoch": 131.93,
+      "learning_rate": 4.822580645161291e-05,
+      "loss": 0.4113,
+      "step": 20450
+    },
+    {
+      "epoch": 132.0,
+      "learning_rate": 4.8222222222222225e-05,
+      "loss": 0.3232,
+      "step": 20460
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.8992859125137329,
+      "eval_runtime": 0.8528,
+      "eval_samples_per_second": 235.698,
+      "eval_steps_per_second": 8.208,
+      "step": 20460
+    },
+    {
+      "epoch": 132.06,
+      "learning_rate": 4.821863799283154e-05,
+      "loss": 0.4176,
+      "step": 20470
+    },
+    {
+      "epoch": 132.13,
+      "learning_rate": 4.821505376344087e-05,
+      "loss": 0.3003,
+      "step": 20480
+    },
+    {
+      "epoch": 132.19,
+      "learning_rate": 4.821146953405018e-05,
+      "loss": 0.3569,
+      "step": 20490
+    },
+    {
+      "epoch": 132.26,
+      "learning_rate": 4.82078853046595e-05,
+      "loss": 0.4206,
+      "step": 20500
+    },
+    {
+      "epoch": 132.32,
+      "learning_rate": 4.820430107526882e-05,
+      "loss": 0.304,
+      "step": 20510
+    },
+    {
+      "epoch": 132.39,
+      "learning_rate": 4.820071684587814e-05,
+      "loss": 0.3046,
+      "step": 20520
+    },
+    {
+      "epoch": 132.45,
+      "learning_rate": 4.819713261648746e-05,
+      "loss": 0.4405,
+      "step": 20530
+    },
+    {
+      "epoch": 132.51,
+      "learning_rate": 4.819354838709678e-05,
+      "loss": 0.331,
+      "step": 20540
+    },
+    {
+      "epoch": 132.58,
+      "learning_rate": 4.8189964157706096e-05,
+      "loss": 0.3193,
+      "step": 20550
+    },
+    {
+      "epoch": 132.64,
+      "learning_rate": 4.8186379928315414e-05,
+      "loss": 0.3093,
+      "step": 20560
+    },
+    {
+      "epoch": 132.71,
+      "learning_rate": 4.818279569892473e-05,
+      "loss": 0.3262,
+      "step": 20570
+    },
+    {
+      "epoch": 132.77,
+      "learning_rate": 4.817921146953405e-05,
+      "loss": 0.2942,
+      "step": 20580
+    },
+    {
+      "epoch": 132.84,
+      "learning_rate": 4.8175627240143375e-05,
+      "loss": 0.3476,
+      "step": 20590
+    },
+    {
+      "epoch": 132.9,
+      "learning_rate": 4.8172043010752693e-05,
+      "loss": 0.3391,
+      "step": 20600
+    },
+    {
+      "epoch": 132.96,
+      "learning_rate": 4.8168458781362005e-05,
+      "loss": 0.3563,
+      "step": 20610
+    },
+    {
+      "epoch": 133.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.8847560882568359,
+      "eval_runtime": 0.8536,
+      "eval_samples_per_second": 235.484,
+      "eval_steps_per_second": 8.201,
+      "step": 20615
+    },
+    {
+      "epoch": 133.03,
+      "learning_rate": 4.816487455197133e-05,
+      "loss": 0.3498,
+      "step": 20620
+    },
+    {
+      "epoch": 133.1,
+      "learning_rate": 4.816129032258065e-05,
+      "loss": 0.3479,
+      "step": 20630
+    },
+    {
+      "epoch": 133.16,
+      "learning_rate": 4.8157706093189966e-05,
+      "loss": 0.4069,
+      "step": 20640
+    },
+    {
+      "epoch": 133.23,
+      "learning_rate": 4.8154121863799285e-05,
+      "loss": 0.3464,
+      "step": 20650
+    },
+    {
+      "epoch": 133.29,
+      "learning_rate": 4.81505376344086e-05,
+      "loss": 0.3658,
+      "step": 20660
+    },
+    {
+      "epoch": 133.35,
+      "learning_rate": 4.814695340501793e-05,
+      "loss": 0.3202,
+      "step": 20670
+    },
+    {
+      "epoch": 133.42,
+      "learning_rate": 4.814336917562724e-05,
+      "loss": 0.2982,
+      "step": 20680
+    },
+    {
+      "epoch": 133.48,
+      "learning_rate": 4.813978494623656e-05,
+      "loss": 0.3548,
+      "step": 20690
+    },
+    {
+      "epoch": 133.55,
+      "learning_rate": 4.813620071684588e-05,
+      "loss": 0.3201,
+      "step": 20700
+    },
+    {
+      "epoch": 133.61,
+      "learning_rate": 4.81326164874552e-05,
+      "loss": 0.3408,
+      "step": 20710
+    },
+    {
+      "epoch": 133.68,
+      "learning_rate": 4.812903225806452e-05,
+      "loss": 0.3956,
+      "step": 20720
+    },
+    {
+      "epoch": 133.74,
+      "learning_rate": 4.812544802867384e-05,
+      "loss": 0.2744,
+      "step": 20730
+    },
+    {
+      "epoch": 133.8,
+      "learning_rate": 4.8121863799283155e-05,
+      "loss": 0.3189,
+      "step": 20740
+    },
+    {
+      "epoch": 133.87,
+      "learning_rate": 4.811827956989248e-05,
+      "loss": 0.3673,
+      "step": 20750
+    },
+    {
+      "epoch": 133.93,
+      "learning_rate": 4.811469534050179e-05,
+      "loss": 0.3843,
+      "step": 20760
+    },
+    {
+      "epoch": 134.0,
+      "learning_rate": 4.811111111111111e-05,
+      "loss": 0.3412,
+      "step": 20770
+    },
+    {
+      "epoch": 134.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.005935549736023,
+      "eval_runtime": 0.8579,
+      "eval_samples_per_second": 234.305,
+      "eval_steps_per_second": 8.16,
+      "step": 20770
+    },
+    {
+      "epoch": 134.06,
+      "learning_rate": 4.8107526881720435e-05,
+      "loss": 0.3543,
+      "step": 20780
+    },
+    {
+      "epoch": 134.13,
+      "learning_rate": 4.810394265232975e-05,
+      "loss": 0.3197,
+      "step": 20790
+    },
+    {
+      "epoch": 134.19,
+      "learning_rate": 4.810035842293907e-05,
+      "loss": 0.3623,
+      "step": 20800
+    },
+    {
+      "epoch": 134.26,
+      "learning_rate": 4.809677419354839e-05,
+      "loss": 0.3695,
+      "step": 20810
+    },
+    {
+      "epoch": 134.32,
+      "learning_rate": 4.809318996415771e-05,
+      "loss": 0.3551,
+      "step": 20820
+    },
+    {
+      "epoch": 134.39,
+      "learning_rate": 4.808960573476703e-05,
+      "loss": 0.3411,
+      "step": 20830
+    },
+    {
+      "epoch": 134.45,
+      "learning_rate": 4.8086021505376344e-05,
+      "loss": 0.3673,
+      "step": 20840
+    },
+    {
+      "epoch": 134.51,
+      "learning_rate": 4.808243727598566e-05,
+      "loss": 0.3641,
+      "step": 20850
+    },
+    {
+      "epoch": 134.58,
+      "learning_rate": 4.807885304659499e-05,
+      "loss": 0.3566,
+      "step": 20860
+    },
+    {
+      "epoch": 134.64,
+      "learning_rate": 4.8075268817204305e-05,
+      "loss": 0.3139,
+      "step": 20870
+    },
+    {
+      "epoch": 134.71,
+      "learning_rate": 4.807168458781362e-05,
+      "loss": 0.3208,
+      "step": 20880
+    },
+    {
+      "epoch": 134.77,
+      "learning_rate": 4.806810035842294e-05,
+      "loss": 0.3848,
+      "step": 20890
+    },
+    {
+      "epoch": 134.84,
+      "learning_rate": 4.806451612903226e-05,
+      "loss": 0.3935,
+      "step": 20900
+    },
+    {
+      "epoch": 134.9,
+      "learning_rate": 4.806093189964158e-05,
+      "loss": 0.3555,
+      "step": 20910
+    },
+    {
+      "epoch": 134.96,
+      "learning_rate": 4.8057347670250896e-05,
+      "loss": 0.3258,
+      "step": 20920
+    },
+    {
+      "epoch": 135.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.0843429565429688,
+      "eval_runtime": 0.8361,
+      "eval_samples_per_second": 240.411,
+      "eval_steps_per_second": 8.373,
+      "step": 20925
+    },
+    {
+      "epoch": 135.03,
+      "learning_rate": 4.8053763440860215e-05,
+      "loss": 0.2928,
+      "step": 20930
+    },
+    {
+      "epoch": 135.1,
+      "learning_rate": 4.805017921146954e-05,
+      "loss": 0.3364,
+      "step": 20940
+    },
+    {
+      "epoch": 135.16,
+      "learning_rate": 4.804659498207886e-05,
+      "loss": 0.3408,
+      "step": 20950
+    },
+    {
+      "epoch": 135.23,
+      "learning_rate": 4.804301075268817e-05,
+      "loss": 0.3468,
+      "step": 20960
+    },
+    {
+      "epoch": 135.29,
+      "learning_rate": 4.8039426523297494e-05,
+      "loss": 0.3921,
+      "step": 20970
+    },
+    {
+      "epoch": 135.35,
+      "learning_rate": 4.803584229390681e-05,
+      "loss": 0.3377,
+      "step": 20980
+    },
+    {
+      "epoch": 135.42,
+      "learning_rate": 4.803225806451613e-05,
+      "loss": 0.3912,
+      "step": 20990
+    },
+    {
+      "epoch": 135.48,
+      "learning_rate": 4.802867383512545e-05,
+      "loss": 0.3972,
+      "step": 21000
+    },
+    {
+      "epoch": 135.55,
+      "learning_rate": 4.802508960573477e-05,
+      "loss": 0.3425,
+      "step": 21010
+    },
+    {
+      "epoch": 135.61,
+      "learning_rate": 4.802150537634409e-05,
+      "loss": 0.3419,
+      "step": 21020
+    },
+    {
+      "epoch": 135.68,
+      "learning_rate": 4.801792114695341e-05,
+      "loss": 0.3676,
+      "step": 21030
+    },
+    {
+      "epoch": 135.74,
+      "learning_rate": 4.801433691756272e-05,
+      "loss": 0.3478,
+      "step": 21040
+    },
+    {
+      "epoch": 135.8,
+      "learning_rate": 4.8010752688172047e-05,
+      "loss": 0.343,
+      "step": 21050
+    },
+    {
+      "epoch": 135.87,
+      "learning_rate": 4.8007168458781365e-05,
+      "loss": 0.3812,
+      "step": 21060
+    },
+    {
+      "epoch": 135.93,
+      "learning_rate": 4.800358422939068e-05,
+      "loss": 0.3926,
+      "step": 21070
+    },
+    {
+      "epoch": 136.0,
+      "learning_rate": 4.8e-05,
+      "loss": 0.3487,
+      "step": 21080
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 0.9971176981925964,
+      "eval_runtime": 0.8476,
+      "eval_samples_per_second": 237.145,
+      "eval_steps_per_second": 8.259,
+      "step": 21080
+    },
+    {
+      "epoch": 136.06,
+      "learning_rate": 4.799641577060932e-05,
+      "loss": 0.3277,
+      "step": 21090
+    },
+    {
+      "epoch": 136.13,
+      "learning_rate": 4.7992831541218644e-05,
+      "loss": 0.3301,
+      "step": 21100
+    },
+    {
+      "epoch": 136.19,
+      "learning_rate": 4.798924731182796e-05,
+      "loss": 0.3725,
+      "step": 21110
+    },
+    {
+      "epoch": 136.26,
+      "learning_rate": 4.7985663082437274e-05,
+      "loss": 0.4092,
+      "step": 21120
+    },
+    {
+      "epoch": 136.32,
+      "learning_rate": 4.79820788530466e-05,
+      "loss": 0.3506,
+      "step": 21130
+    },
+    {
+      "epoch": 136.39,
+      "learning_rate": 4.797849462365592e-05,
+      "loss": 0.3711,
+      "step": 21140
+    },
+    {
+      "epoch": 136.45,
+      "learning_rate": 4.7974910394265235e-05,
+      "loss": 0.3482,
+      "step": 21150
+    },
+    {
+      "epoch": 136.51,
+      "learning_rate": 4.7971326164874554e-05,
+      "loss": 0.3474,
+      "step": 21160
+    },
+    {
+      "epoch": 136.58,
+      "learning_rate": 4.796774193548387e-05,
+      "loss": 0.37,
+      "step": 21170
+    },
+    {
+      "epoch": 136.64,
+      "learning_rate": 4.79641577060932e-05,
+      "loss": 0.346,
+      "step": 21180
+    },
+    {
+      "epoch": 136.71,
+      "learning_rate": 4.7960573476702515e-05,
+      "loss": 0.3376,
+      "step": 21190
+    },
+    {
+      "epoch": 136.77,
+      "learning_rate": 4.7956989247311826e-05,
+      "loss": 0.3416,
+      "step": 21200
+    },
+    {
+      "epoch": 136.84,
+      "learning_rate": 4.795340501792115e-05,
+      "loss": 0.3002,
+      "step": 21210
+    },
+    {
+      "epoch": 136.9,
+      "learning_rate": 4.794982078853047e-05,
+      "loss": 0.301,
+      "step": 21220
+    },
+    {
+      "epoch": 136.96,
+      "learning_rate": 4.794623655913979e-05,
+      "loss": 0.3335,
+      "step": 21230
+    },
+    {
+      "epoch": 137.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9973180890083313,
+      "eval_runtime": 0.8393,
+      "eval_samples_per_second": 239.474,
+      "eval_steps_per_second": 8.34,
+      "step": 21235
+    },
+    {
+      "epoch": 137.03,
+      "learning_rate": 4.7942652329749106e-05,
+      "loss": 0.37,
+      "step": 21240
+    },
+    {
+      "epoch": 137.1,
+      "learning_rate": 4.7939068100358424e-05,
+      "loss": 0.3687,
+      "step": 21250
+    },
+    {
+      "epoch": 137.16,
+      "learning_rate": 4.793548387096774e-05,
+      "loss": 0.4037,
+      "step": 21260
+    },
+    {
+      "epoch": 137.23,
+      "learning_rate": 4.793189964157706e-05,
+      "loss": 0.3749,
+      "step": 21270
+    },
+    {
+      "epoch": 137.29,
+      "learning_rate": 4.792831541218638e-05,
+      "loss": 0.3737,
+      "step": 21280
+    },
+    {
+      "epoch": 137.35,
+      "learning_rate": 4.7924731182795704e-05,
+      "loss": 0.385,
+      "step": 21290
+    },
+    {
+      "epoch": 137.42,
+      "learning_rate": 4.792114695340502e-05,
+      "loss": 0.3063,
+      "step": 21300
+    },
+    {
+      "epoch": 137.48,
+      "learning_rate": 4.7917562724014333e-05,
+      "loss": 0.341,
+      "step": 21310
+    },
+    {
+      "epoch": 137.55,
+      "learning_rate": 4.791397849462366e-05,
+      "loss": 0.3275,
+      "step": 21320
+    },
+    {
+      "epoch": 137.61,
+      "learning_rate": 4.791039426523298e-05,
+      "loss": 0.3497,
+      "step": 21330
+    },
+    {
+      "epoch": 137.68,
+      "learning_rate": 4.7906810035842295e-05,
+      "loss": 0.38,
+      "step": 21340
+    },
+    {
+      "epoch": 137.74,
+      "learning_rate": 4.790322580645161e-05,
+      "loss": 0.3089,
+      "step": 21350
+    },
+    {
+      "epoch": 137.8,
+      "learning_rate": 4.789964157706093e-05,
+      "loss": 0.3495,
+      "step": 21360
+    },
+    {
+      "epoch": 137.87,
+      "learning_rate": 4.7896057347670256e-05,
+      "loss": 0.3193,
+      "step": 21370
+    },
+    {
+      "epoch": 137.93,
+      "learning_rate": 4.7892473118279574e-05,
+      "loss": 0.3621,
+      "step": 21380
+    },
+    {
+      "epoch": 138.0,
+      "learning_rate": 4.7888888888888886e-05,
+      "loss": 0.3784,
+      "step": 21390
+    },
+    {
+      "epoch": 138.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9534862041473389,
+      "eval_runtime": 0.8446,
+      "eval_samples_per_second": 237.978,
+      "eval_steps_per_second": 8.288,
+      "step": 21390
+    },
+    {
+      "epoch": 138.06,
+      "learning_rate": 4.788530465949821e-05,
+      "loss": 0.3318,
+      "step": 21400
+    },
+    {
+      "epoch": 138.13,
+      "learning_rate": 4.788172043010753e-05,
+      "loss": 0.3244,
+      "step": 21410
+    },
+    {
+      "epoch": 138.19,
+      "learning_rate": 4.787813620071685e-05,
+      "loss": 0.3458,
+      "step": 21420
+    },
+    {
+      "epoch": 138.26,
+      "learning_rate": 4.7874551971326165e-05,
+      "loss": 0.328,
+      "step": 21430
+    },
+    {
+      "epoch": 138.32,
+      "learning_rate": 4.7870967741935484e-05,
+      "loss": 0.363,
+      "step": 21440
+    },
+    {
+      "epoch": 138.39,
+      "learning_rate": 4.786738351254481e-05,
+      "loss": 0.366,
+      "step": 21450
+    },
+    {
+      "epoch": 138.45,
+      "learning_rate": 4.786379928315413e-05,
+      "loss": 0.3292,
+      "step": 21460
+    },
+    {
+      "epoch": 138.51,
+      "learning_rate": 4.786021505376344e-05,
+      "loss": 0.3457,
+      "step": 21470
+    },
+    {
+      "epoch": 138.58,
+      "learning_rate": 4.785663082437276e-05,
+      "loss": 0.3424,
+      "step": 21480
+    },
+    {
+      "epoch": 138.64,
+      "learning_rate": 4.785304659498208e-05,
+      "loss": 0.3243,
+      "step": 21490
+    },
+    {
+      "epoch": 138.71,
+      "learning_rate": 4.78494623655914e-05,
+      "loss": 0.3591,
+      "step": 21500
+    },
+    {
+      "epoch": 138.77,
+      "learning_rate": 4.784587813620072e-05,
+      "loss": 0.3293,
+      "step": 21510
+    },
+    {
+      "epoch": 138.84,
+      "learning_rate": 4.7842293906810036e-05,
+      "loss": 0.3066,
+      "step": 21520
+    },
+    {
+      "epoch": 138.9,
+      "learning_rate": 4.783870967741936e-05,
+      "loss": 0.3644,
+      "step": 21530
+    },
+    {
+      "epoch": 138.96,
+      "learning_rate": 4.783512544802868e-05,
+      "loss": 0.3556,
+      "step": 21540
+    },
+    {
+      "epoch": 139.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9415268301963806,
+      "eval_runtime": 0.8552,
+      "eval_samples_per_second": 235.023,
+      "eval_steps_per_second": 8.185,
+      "step": 21545
+    },
+    {
+      "epoch": 139.03,
+      "learning_rate": 4.783154121863799e-05,
+      "loss": 0.3864,
+      "step": 21550
+    },
+    {
+      "epoch": 139.1,
+      "learning_rate": 4.7827956989247316e-05,
+      "loss": 0.3518,
+      "step": 21560
+    },
+    {
+      "epoch": 139.16,
+      "learning_rate": 4.7824372759856634e-05,
+      "loss": 0.3324,
+      "step": 21570
+    },
+    {
+      "epoch": 139.23,
+      "learning_rate": 4.782078853046595e-05,
+      "loss": 0.3461,
+      "step": 21580
+    },
+    {
+      "epoch": 139.29,
+      "learning_rate": 4.781720430107527e-05,
+      "loss": 0.3791,
+      "step": 21590
+    },
+    {
+      "epoch": 139.35,
+      "learning_rate": 4.781362007168459e-05,
+      "loss": 0.3879,
+      "step": 21600
+    },
+    {
+      "epoch": 139.42,
+      "learning_rate": 4.781003584229391e-05,
+      "loss": 0.401,
+      "step": 21610
+    },
+    {
+      "epoch": 139.48,
+      "learning_rate": 4.780645161290323e-05,
+      "loss": 0.3721,
+      "step": 21620
+    },
+    {
+      "epoch": 139.55,
+      "learning_rate": 4.780286738351254e-05,
+      "loss": 0.3437,
+      "step": 21630
+    },
+    {
+      "epoch": 139.61,
+      "learning_rate": 4.779928315412187e-05,
+      "loss": 0.3188,
+      "step": 21640
+    },
+    {
+      "epoch": 139.68,
+      "learning_rate": 4.7795698924731186e-05,
+      "loss": 0.3575,
+      "step": 21650
+    },
+    {
+      "epoch": 139.74,
+      "learning_rate": 4.7792114695340505e-05,
+      "loss": 0.3043,
+      "step": 21660
+    },
+    {
+      "epoch": 139.8,
+      "learning_rate": 4.778853046594982e-05,
+      "loss": 0.3663,
+      "step": 21670
+    },
+    {
+      "epoch": 139.87,
+      "learning_rate": 4.778494623655914e-05,
+      "loss": 0.3649,
+      "step": 21680
+    },
+    {
+      "epoch": 139.93,
+      "learning_rate": 4.778136200716846e-05,
+      "loss": 0.3897,
+      "step": 21690
+    },
+    {
+      "epoch": 140.0,
+      "learning_rate": 4.7777777777777784e-05,
+      "loss": 0.3417,
+      "step": 21700
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9821661710739136,
+      "eval_runtime": 0.8718,
+      "eval_samples_per_second": 230.569,
+      "eval_steps_per_second": 8.03,
+      "step": 21700
+    },
+    {
+      "epoch": 140.06,
+      "learning_rate": 4.7774193548387096e-05,
+      "loss": 0.3567,
+      "step": 21710
+    },
+    {
+      "epoch": 140.13,
+      "learning_rate": 4.777060931899642e-05,
+      "loss": 0.3272,
+      "step": 21720
+    },
+    {
+      "epoch": 140.19,
+      "learning_rate": 4.776702508960574e-05,
+      "loss": 0.3718,
+      "step": 21730
+    },
+    {
+      "epoch": 140.26,
+      "learning_rate": 4.776344086021506e-05,
+      "loss": 0.3778,
+      "step": 21740
+    },
+    {
+      "epoch": 140.32,
+      "learning_rate": 4.7759856630824375e-05,
+      "loss": 0.3404,
+      "step": 21750
+    },
+    {
+      "epoch": 140.39,
+      "learning_rate": 4.775627240143369e-05,
+      "loss": 0.3124,
+      "step": 21760
+    },
+    {
+      "epoch": 140.45,
+      "learning_rate": 4.775268817204301e-05,
+      "loss": 0.3237,
+      "step": 21770
+    },
+    {
+      "epoch": 140.51,
+      "learning_rate": 4.7749103942652337e-05,
+      "loss": 0.3637,
+      "step": 21780
+    },
+    {
+      "epoch": 140.58,
+      "learning_rate": 4.774551971326165e-05,
+      "loss": 0.3829,
+      "step": 21790
+    },
+    {
+      "epoch": 140.64,
+      "learning_rate": 4.774193548387097e-05,
+      "loss": 0.3766,
+      "step": 21800
+    },
+    {
+      "epoch": 140.71,
+      "learning_rate": 4.773835125448029e-05,
+      "loss": 0.3476,
+      "step": 21810
+    },
+    {
+      "epoch": 140.77,
+      "learning_rate": 4.773476702508961e-05,
+      "loss": 0.3121,
+      "step": 21820
+    },
+    {
+      "epoch": 140.84,
+      "learning_rate": 4.773118279569893e-05,
+      "loss": 0.3472,
+      "step": 21830
+    },
+    {
+      "epoch": 140.9,
+      "learning_rate": 4.7727598566308246e-05,
+      "loss": 0.3353,
+      "step": 21840
+    },
+    {
+      "epoch": 140.96,
+      "learning_rate": 4.7724014336917564e-05,
+      "loss": 0.2932,
+      "step": 21850
+    },
+    {
+      "epoch": 141.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9863076210021973,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.123,
+      "eval_steps_per_second": 8.154,
+      "step": 21855
+    },
+    {
+      "epoch": 141.03,
+      "learning_rate": 4.772043010752689e-05,
+      "loss": 0.3528,
+      "step": 21860
+    },
+    {
+      "epoch": 141.1,
+      "learning_rate": 4.77168458781362e-05,
+      "loss": 0.3616,
+      "step": 21870
+    },
+    {
+      "epoch": 141.16,
+      "learning_rate": 4.771326164874552e-05,
+      "loss": 0.3531,
+      "step": 21880
+    },
+    {
+      "epoch": 141.23,
+      "learning_rate": 4.7709677419354844e-05,
+      "loss": 0.3567,
+      "step": 21890
+    },
+    {
+      "epoch": 141.29,
+      "learning_rate": 4.7706093189964155e-05,
+      "loss": 0.2474,
+      "step": 21900
+    },
+    {
+      "epoch": 141.35,
+      "learning_rate": 4.770250896057348e-05,
+      "loss": 0.3202,
+      "step": 21910
+    },
+    {
+      "epoch": 141.42,
+      "learning_rate": 4.76989247311828e-05,
+      "loss": 0.3628,
+      "step": 21920
+    },
+    {
+      "epoch": 141.48,
+      "learning_rate": 4.7695340501792116e-05,
+      "loss": 0.3483,
+      "step": 21930
+    },
+    {
+      "epoch": 141.55,
+      "learning_rate": 4.7691756272401435e-05,
+      "loss": 0.313,
+      "step": 21940
+    },
+    {
+      "epoch": 141.61,
+      "learning_rate": 4.768817204301075e-05,
+      "loss": 0.341,
+      "step": 21950
+    },
+    {
+      "epoch": 141.68,
+      "learning_rate": 4.768458781362007e-05,
+      "loss": 0.4271,
+      "step": 21960
+    },
+    {
+      "epoch": 141.74,
+      "learning_rate": 4.7681003584229396e-05,
+      "loss": 0.3739,
+      "step": 21970
+    },
+    {
+      "epoch": 141.8,
+      "learning_rate": 4.767741935483871e-05,
+      "loss": 0.3465,
+      "step": 21980
+    },
+    {
+      "epoch": 141.87,
+      "learning_rate": 4.767383512544803e-05,
+      "loss": 0.4164,
+      "step": 21990
+    },
+    {
+      "epoch": 141.93,
+      "learning_rate": 4.767025089605735e-05,
+      "loss": 0.343,
+      "step": 22000
+    },
+    {
+      "epoch": 142.0,
+      "learning_rate": 4.766666666666667e-05,
+      "loss": 0.4086,
+      "step": 22010
+    },
+    {
+      "epoch": 142.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9371060132980347,
+      "eval_runtime": 0.8588,
+      "eval_samples_per_second": 234.055,
+      "eval_steps_per_second": 8.151,
+      "step": 22010
+    },
+    {
+      "epoch": 142.06,
+      "learning_rate": 4.766308243727599e-05,
+      "loss": 0.358,
+      "step": 22020
+    },
+    {
+      "epoch": 142.13,
+      "learning_rate": 4.7659498207885305e-05,
+      "loss": 0.3595,
+      "step": 22030
+    },
+    {
+      "epoch": 142.19,
+      "learning_rate": 4.7655913978494623e-05,
+      "loss": 0.3461,
+      "step": 22040
+    },
+    {
+      "epoch": 142.26,
+      "learning_rate": 4.765232974910395e-05,
+      "loss": 0.3134,
+      "step": 22050
+    },
+    {
+      "epoch": 142.32,
+      "learning_rate": 4.764874551971326e-05,
+      "loss": 0.3777,
+      "step": 22060
+    },
+    {
+      "epoch": 142.39,
+      "learning_rate": 4.7645161290322585e-05,
+      "loss": 0.3661,
+      "step": 22070
+    },
+    {
+      "epoch": 142.45,
+      "learning_rate": 4.76415770609319e-05,
+      "loss": 0.3044,
+      "step": 22080
+    },
+    {
+      "epoch": 142.51,
+      "learning_rate": 4.763799283154122e-05,
+      "loss": 0.3873,
+      "step": 22090
+    },
+    {
+      "epoch": 142.58,
+      "learning_rate": 4.763440860215054e-05,
+      "loss": 0.3437,
+      "step": 22100
+    },
+    {
+      "epoch": 142.64,
+      "learning_rate": 4.763082437275986e-05,
+      "loss": 0.3512,
+      "step": 22110
+    },
+    {
+      "epoch": 142.71,
+      "learning_rate": 4.7627240143369176e-05,
+      "loss": 0.3275,
+      "step": 22120
+    },
+    {
+      "epoch": 142.77,
+      "learning_rate": 4.76236559139785e-05,
+      "loss": 0.3026,
+      "step": 22130
+    },
+    {
+      "epoch": 142.84,
+      "learning_rate": 4.762007168458781e-05,
+      "loss": 0.2943,
+      "step": 22140
+    },
+    {
+      "epoch": 142.9,
+      "learning_rate": 4.761648745519714e-05,
+      "loss": 0.2986,
+      "step": 22150
+    },
+    {
+      "epoch": 142.96,
+      "learning_rate": 4.7612903225806455e-05,
+      "loss": 0.3324,
+      "step": 22160
+    },
+    {
+      "epoch": 143.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9442649483680725,
+      "eval_runtime": 0.8486,
+      "eval_samples_per_second": 236.85,
+      "eval_steps_per_second": 8.249,
+      "step": 22165
+    },
+    {
+      "epoch": 143.03,
+      "learning_rate": 4.7609318996415774e-05,
+      "loss": 0.2935,
+      "step": 22170
+    },
+    {
+      "epoch": 143.1,
+      "learning_rate": 4.760573476702509e-05,
+      "loss": 0.3181,
+      "step": 22180
+    },
+    {
+      "epoch": 143.16,
+      "learning_rate": 4.760215053763441e-05,
+      "loss": 0.3453,
+      "step": 22190
+    },
+    {
+      "epoch": 143.23,
+      "learning_rate": 4.759856630824373e-05,
+      "loss": 0.3289,
+      "step": 22200
+    },
+    {
+      "epoch": 143.29,
+      "learning_rate": 4.759498207885305e-05,
+      "loss": 0.36,
+      "step": 22210
+    },
+    {
+      "epoch": 143.35,
+      "learning_rate": 4.7591397849462365e-05,
+      "loss": 0.2701,
+      "step": 22220
+    },
+    {
+      "epoch": 143.42,
+      "learning_rate": 4.758781362007168e-05,
+      "loss": 0.3619,
+      "step": 22230
+    },
+    {
+      "epoch": 143.48,
+      "learning_rate": 4.758422939068101e-05,
+      "loss": 0.296,
+      "step": 22240
+    },
+    {
+      "epoch": 143.55,
+      "learning_rate": 4.7580645161290326e-05,
+      "loss": 0.3335,
+      "step": 22250
+    },
+    {
+      "epoch": 143.61,
+      "learning_rate": 4.7577060931899644e-05,
+      "loss": 0.3815,
+      "step": 22260
+    },
+    {
+      "epoch": 143.68,
+      "learning_rate": 4.757347670250896e-05,
+      "loss": 0.3621,
+      "step": 22270
+    },
+    {
+      "epoch": 143.74,
+      "learning_rate": 4.756989247311828e-05,
+      "loss": 0.3511,
+      "step": 22280
+    },
+    {
+      "epoch": 143.8,
+      "learning_rate": 4.7566308243727606e-05,
+      "loss": 0.3488,
+      "step": 22290
+    },
+    {
+      "epoch": 143.87,
+      "learning_rate": 4.756272401433692e-05,
+      "loss": 0.3359,
+      "step": 22300
+    },
+    {
+      "epoch": 143.93,
+      "learning_rate": 4.7559139784946235e-05,
+      "loss": 0.3503,
+      "step": 22310
+    },
+    {
+      "epoch": 144.0,
+      "learning_rate": 4.755555555555556e-05,
+      "loss": 0.3205,
+      "step": 22320
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9946515560150146,
+      "eval_runtime": 0.8509,
+      "eval_samples_per_second": 236.21,
+      "eval_steps_per_second": 8.226,
+      "step": 22320
+    },
+    {
+      "epoch": 144.06,
+      "learning_rate": 4.755197132616488e-05,
+      "loss": 0.4478,
+      "step": 22330
+    },
+    {
+      "epoch": 144.13,
+      "learning_rate": 4.75483870967742e-05,
+      "loss": 0.303,
+      "step": 22340
+    },
+    {
+      "epoch": 144.19,
+      "learning_rate": 4.7544802867383515e-05,
+      "loss": 0.3254,
+      "step": 22350
+    },
+    {
+      "epoch": 144.26,
+      "learning_rate": 4.754121863799283e-05,
+      "loss": 0.3048,
+      "step": 22360
+    },
+    {
+      "epoch": 144.32,
+      "learning_rate": 4.753763440860216e-05,
+      "loss": 0.3471,
+      "step": 22370
+    },
+    {
+      "epoch": 144.39,
+      "learning_rate": 4.753405017921147e-05,
+      "loss": 0.3838,
+      "step": 22380
+    },
+    {
+      "epoch": 144.45,
+      "learning_rate": 4.753046594982079e-05,
+      "loss": 0.3219,
+      "step": 22390
+    },
+    {
+      "epoch": 144.51,
+      "learning_rate": 4.752688172043011e-05,
+      "loss": 0.3163,
+      "step": 22400
+    },
+    {
+      "epoch": 144.58,
+      "learning_rate": 4.752329749103943e-05,
+      "loss": 0.349,
+      "step": 22410
+    },
+    {
+      "epoch": 144.64,
+      "learning_rate": 4.751971326164875e-05,
+      "loss": 0.2982,
+      "step": 22420
+    },
+    {
+      "epoch": 144.71,
+      "learning_rate": 4.751612903225807e-05,
+      "loss": 0.356,
+      "step": 22430
+    },
+    {
+      "epoch": 144.77,
+      "learning_rate": 4.7512544802867386e-05,
+      "loss": 0.3346,
+      "step": 22440
+    },
+    {
+      "epoch": 144.84,
+      "learning_rate": 4.750896057347671e-05,
+      "loss": 0.3606,
+      "step": 22450
+    },
+    {
+      "epoch": 144.9,
+      "learning_rate": 4.750537634408602e-05,
+      "loss": 0.3445,
+      "step": 22460
+    },
+    {
+      "epoch": 144.96,
+      "learning_rate": 4.750179211469534e-05,
+      "loss": 0.3135,
+      "step": 22470
+    },
+    {
+      "epoch": 145.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9233006834983826,
+      "eval_runtime": 0.8586,
+      "eval_samples_per_second": 234.107,
+      "eval_steps_per_second": 8.153,
+      "step": 22475
+    },
+    {
+      "epoch": 145.03,
+      "learning_rate": 4.7498207885304665e-05,
+      "loss": 0.3403,
+      "step": 22480
+    },
+    {
+      "epoch": 145.1,
+      "learning_rate": 4.749462365591398e-05,
+      "loss": 0.265,
+      "step": 22490
+    },
+    {
+      "epoch": 145.16,
+      "learning_rate": 4.74910394265233e-05,
+      "loss": 0.3417,
+      "step": 22500
+    },
+    {
+      "epoch": 145.23,
+      "learning_rate": 4.748745519713262e-05,
+      "loss": 0.4045,
+      "step": 22510
+    },
+    {
+      "epoch": 145.29,
+      "learning_rate": 4.748387096774194e-05,
+      "loss": 0.3776,
+      "step": 22520
+    },
+    {
+      "epoch": 145.35,
+      "learning_rate": 4.7480286738351256e-05,
+      "loss": 0.3564,
+      "step": 22530
+    },
+    {
+      "epoch": 145.42,
+      "learning_rate": 4.7476702508960574e-05,
+      "loss": 0.3566,
+      "step": 22540
+    },
+    {
+      "epoch": 145.48,
+      "learning_rate": 4.747311827956989e-05,
+      "loss": 0.3368,
+      "step": 22550
+    },
+    {
+      "epoch": 145.55,
+      "learning_rate": 4.746953405017922e-05,
+      "loss": 0.3222,
+      "step": 22560
+    },
+    {
+      "epoch": 145.61,
+      "learning_rate": 4.746594982078853e-05,
+      "loss": 0.3843,
+      "step": 22570
+    },
+    {
+      "epoch": 145.68,
+      "learning_rate": 4.746236559139785e-05,
+      "loss": 0.3525,
+      "step": 22580
+    },
+    {
+      "epoch": 145.74,
+      "learning_rate": 4.745878136200717e-05,
+      "loss": 0.3579,
+      "step": 22590
+    },
+    {
+      "epoch": 145.8,
+      "learning_rate": 4.745519713261649e-05,
+      "loss": 0.3366,
+      "step": 22600
+    },
+    {
+      "epoch": 145.87,
+      "learning_rate": 4.745161290322581e-05,
+      "loss": 0.3559,
+      "step": 22610
+    },
+    {
+      "epoch": 145.93,
+      "learning_rate": 4.744802867383513e-05,
+      "loss": 0.3257,
+      "step": 22620
+    },
+    {
+      "epoch": 146.0,
+      "learning_rate": 4.7444444444444445e-05,
+      "loss": 0.3452,
+      "step": 22630
+    },
+    {
+      "epoch": 146.0,
+      "eval_accuracy": 0.7313432835820896,
+      "eval_loss": 0.993377685546875,
+      "eval_runtime": 0.856,
+      "eval_samples_per_second": 234.823,
+      "eval_steps_per_second": 8.178,
+      "step": 22630
+    },
+    {
+      "epoch": 146.06,
+      "learning_rate": 4.744086021505377e-05,
+      "loss": 0.3243,
+      "step": 22640
+    },
+    {
+      "epoch": 146.13,
+      "learning_rate": 4.743727598566308e-05,
+      "loss": 0.301,
+      "step": 22650
+    },
+    {
+      "epoch": 146.19,
+      "learning_rate": 4.74336917562724e-05,
+      "loss": 0.3553,
+      "step": 22660
+    },
+    {
+      "epoch": 146.26,
+      "learning_rate": 4.7430107526881725e-05,
+      "loss": 0.3324,
+      "step": 22670
+    },
+    {
+      "epoch": 146.32,
+      "learning_rate": 4.742652329749104e-05,
+      "loss": 0.3763,
+      "step": 22680
+    },
+    {
+      "epoch": 146.39,
+      "learning_rate": 4.742293906810036e-05,
+      "loss": 0.3653,
+      "step": 22690
+    },
+    {
+      "epoch": 146.45,
+      "learning_rate": 4.741935483870968e-05,
+      "loss": 0.35,
+      "step": 22700
+    },
+    {
+      "epoch": 146.51,
+      "learning_rate": 4.7415770609319e-05,
+      "loss": 0.3626,
+      "step": 22710
+    },
+    {
+      "epoch": 146.58,
+      "learning_rate": 4.741218637992832e-05,
+      "loss": 0.2664,
+      "step": 22720
+    },
+    {
+      "epoch": 146.64,
+      "learning_rate": 4.7408602150537634e-05,
+      "loss": 0.3531,
+      "step": 22730
+    },
+    {
+      "epoch": 146.71,
+      "learning_rate": 4.740501792114695e-05,
+      "loss": 0.3518,
+      "step": 22740
+    },
+    {
+      "epoch": 146.77,
+      "learning_rate": 4.740143369175628e-05,
+      "loss": 0.3172,
+      "step": 22750
+    },
+    {
+      "epoch": 146.84,
+      "learning_rate": 4.7397849462365595e-05,
+      "loss": 0.3005,
+      "step": 22760
+    },
+    {
+      "epoch": 146.9,
+      "learning_rate": 4.739426523297491e-05,
+      "loss": 0.3641,
+      "step": 22770
+    },
+    {
+      "epoch": 146.96,
+      "learning_rate": 4.739068100358423e-05,
+      "loss": 0.372,
+      "step": 22780
+    },
+    {
+      "epoch": 147.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9624363780021667,
+      "eval_runtime": 0.8824,
+      "eval_samples_per_second": 227.788,
+      "eval_steps_per_second": 7.933,
+      "step": 22785
+    },
+    {
+      "epoch": 147.03,
+      "learning_rate": 4.738709677419355e-05,
+      "loss": 0.3471,
+      "step": 22790
+    },
+    {
+      "epoch": 147.1,
+      "learning_rate": 4.7383512544802875e-05,
+      "loss": 0.3724,
+      "step": 22800
+    },
+    {
+      "epoch": 147.16,
+      "learning_rate": 4.7379928315412186e-05,
+      "loss": 0.3278,
+      "step": 22810
+    },
+    {
+      "epoch": 147.23,
+      "learning_rate": 4.7376344086021504e-05,
+      "loss": 0.3602,
+      "step": 22820
+    },
+    {
+      "epoch": 147.29,
+      "learning_rate": 4.737275985663083e-05,
+      "loss": 0.4329,
+      "step": 22830
+    },
+    {
+      "epoch": 147.35,
+      "learning_rate": 4.736917562724015e-05,
+      "loss": 0.3451,
+      "step": 22840
+    },
+    {
+      "epoch": 147.42,
+      "learning_rate": 4.7365591397849466e-05,
+      "loss": 0.3066,
+      "step": 22850
+    },
+    {
+      "epoch": 147.48,
+      "learning_rate": 4.7362007168458784e-05,
+      "loss": 0.3297,
+      "step": 22860
+    },
+    {
+      "epoch": 147.55,
+      "learning_rate": 4.73584229390681e-05,
+      "loss": 0.3433,
+      "step": 22870
+    },
+    {
+      "epoch": 147.61,
+      "learning_rate": 4.735483870967742e-05,
+      "loss": 0.302,
+      "step": 22880
+    },
+    {
+      "epoch": 147.68,
+      "learning_rate": 4.735125448028674e-05,
+      "loss": 0.3179,
+      "step": 22890
+    },
+    {
+      "epoch": 147.74,
+      "learning_rate": 4.734767025089606e-05,
+      "loss": 0.3255,
+      "step": 22900
+    },
+    {
+      "epoch": 147.8,
+      "learning_rate": 4.734408602150538e-05,
+      "loss": 0.3311,
+      "step": 22910
+    },
+    {
+      "epoch": 147.87,
+      "learning_rate": 4.73405017921147e-05,
+      "loss": 0.326,
+      "step": 22920
+    },
+    {
+      "epoch": 147.93,
+      "learning_rate": 4.733691756272401e-05,
+      "loss": 0.3622,
+      "step": 22930
+    },
+    {
+      "epoch": 148.0,
+      "learning_rate": 4.7333333333333336e-05,
+      "loss": 0.3473,
+      "step": 22940
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.940039336681366,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.367,
+      "eval_steps_per_second": 8.058,
+      "step": 22940
+    },
+    {
+      "epoch": 148.06,
+      "learning_rate": 4.7329749103942655e-05,
+      "loss": 0.3612,
+      "step": 22950
+    },
+    {
+      "epoch": 148.13,
+      "learning_rate": 4.732616487455197e-05,
+      "loss": 0.4366,
+      "step": 22960
+    },
+    {
+      "epoch": 148.19,
+      "learning_rate": 4.732258064516129e-05,
+      "loss": 0.3481,
+      "step": 22970
+    },
+    {
+      "epoch": 148.26,
+      "learning_rate": 4.731899641577061e-05,
+      "loss": 0.317,
+      "step": 22980
+    },
+    {
+      "epoch": 148.32,
+      "learning_rate": 4.7315412186379934e-05,
+      "loss": 0.305,
+      "step": 22990
+    },
+    {
+      "epoch": 148.39,
+      "learning_rate": 4.731182795698925e-05,
+      "loss": 0.3444,
+      "step": 23000
+    },
+    {
+      "epoch": 148.45,
+      "learning_rate": 4.7308243727598564e-05,
+      "loss": 0.3244,
+      "step": 23010
+    },
+    {
+      "epoch": 148.51,
+      "learning_rate": 4.730465949820789e-05,
+      "loss": 0.3677,
+      "step": 23020
+    },
+    {
+      "epoch": 148.58,
+      "learning_rate": 4.730107526881721e-05,
+      "loss": 0.3549,
+      "step": 23030
+    },
+    {
+      "epoch": 148.64,
+      "learning_rate": 4.7297491039426525e-05,
+      "loss": 0.2846,
+      "step": 23040
+    },
+    {
+      "epoch": 148.71,
+      "learning_rate": 4.7293906810035843e-05,
+      "loss": 0.3488,
+      "step": 23050
+    },
+    {
+      "epoch": 148.77,
+      "learning_rate": 4.729032258064516e-05,
+      "loss": 0.3823,
+      "step": 23060
+    },
+    {
+      "epoch": 148.84,
+      "learning_rate": 4.728673835125449e-05,
+      "loss": 0.3142,
+      "step": 23070
+    },
+    {
+      "epoch": 148.9,
+      "learning_rate": 4.7283154121863805e-05,
+      "loss": 0.3502,
+      "step": 23080
+    },
+    {
+      "epoch": 148.96,
+      "learning_rate": 4.7279569892473116e-05,
+      "loss": 0.3394,
+      "step": 23090
+    },
+    {
+      "epoch": 149.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.063481092453003,
+      "eval_runtime": 0.8809,
+      "eval_samples_per_second": 228.173,
+      "eval_steps_per_second": 7.946,
+      "step": 23095
+    },
+    {
+      "epoch": 149.03,
+      "learning_rate": 4.727598566308244e-05,
+      "loss": 0.3375,
+      "step": 23100
+    },
+    {
+      "epoch": 149.1,
+      "learning_rate": 4.727240143369176e-05,
+      "loss": 0.3044,
+      "step": 23110
+    },
+    {
+      "epoch": 149.16,
+      "learning_rate": 4.726881720430108e-05,
+      "loss": 0.372,
+      "step": 23120
+    },
+    {
+      "epoch": 149.23,
+      "learning_rate": 4.7265232974910396e-05,
+      "loss": 0.3442,
+      "step": 23130
+    },
+    {
+      "epoch": 149.29,
+      "learning_rate": 4.7261648745519714e-05,
+      "loss": 0.3623,
+      "step": 23140
+    },
+    {
+      "epoch": 149.35,
+      "learning_rate": 4.725806451612904e-05,
+      "loss": 0.3388,
+      "step": 23150
+    },
+    {
+      "epoch": 149.42,
+      "learning_rate": 4.725448028673836e-05,
+      "loss": 0.3291,
+      "step": 23160
+    },
+    {
+      "epoch": 149.48,
+      "learning_rate": 4.725089605734767e-05,
+      "loss": 0.3377,
+      "step": 23170
+    },
+    {
+      "epoch": 149.55,
+      "learning_rate": 4.7247311827956994e-05,
+      "loss": 0.3717,
+      "step": 23180
+    },
+    {
+      "epoch": 149.61,
+      "learning_rate": 4.724372759856631e-05,
+      "loss": 0.3759,
+      "step": 23190
+    },
+    {
+      "epoch": 149.68,
+      "learning_rate": 4.724014336917563e-05,
+      "loss": 0.321,
+      "step": 23200
+    },
+    {
+      "epoch": 149.74,
+      "learning_rate": 4.723655913978495e-05,
+      "loss": 0.35,
+      "step": 23210
+    },
+    {
+      "epoch": 149.8,
+      "learning_rate": 4.7232974910394266e-05,
+      "loss": 0.3384,
+      "step": 23220
+    },
+    {
+      "epoch": 149.87,
+      "learning_rate": 4.7229390681003585e-05,
+      "loss": 0.3371,
+      "step": 23230
+    },
+    {
+      "epoch": 149.93,
+      "learning_rate": 4.72258064516129e-05,
+      "loss": 0.3052,
+      "step": 23240
+    },
+    {
+      "epoch": 150.0,
+      "learning_rate": 4.722222222222222e-05,
+      "loss": 0.3264,
+      "step": 23250
+    },
+    {
+      "epoch": 150.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0731580257415771,
+      "eval_runtime": 0.8704,
+      "eval_samples_per_second": 230.928,
+      "eval_steps_per_second": 8.042,
+      "step": 23250
+    },
+    {
+      "epoch": 150.06,
+      "learning_rate": 4.7218637992831546e-05,
+      "loss": 0.3045,
+      "step": 23260
+    },
+    {
+      "epoch": 150.13,
+      "learning_rate": 4.7215053763440864e-05,
+      "loss": 0.3161,
+      "step": 23270
+    },
+    {
+      "epoch": 150.19,
+      "learning_rate": 4.7211469534050176e-05,
+      "loss": 0.3601,
+      "step": 23280
+    },
+    {
+      "epoch": 150.26,
+      "learning_rate": 4.72078853046595e-05,
+      "loss": 0.339,
+      "step": 23290
+    },
+    {
+      "epoch": 150.32,
+      "learning_rate": 4.720430107526882e-05,
+      "loss": 0.338,
+      "step": 23300
+    },
+    {
+      "epoch": 150.39,
+      "learning_rate": 4.720071684587814e-05,
+      "loss": 0.3002,
+      "step": 23310
+    },
+    {
+      "epoch": 150.45,
+      "learning_rate": 4.7197132616487455e-05,
+      "loss": 0.4131,
+      "step": 23320
+    },
+    {
+      "epoch": 150.51,
+      "learning_rate": 4.7193548387096774e-05,
+      "loss": 0.3878,
+      "step": 23330
+    },
+    {
+      "epoch": 150.58,
+      "learning_rate": 4.71899641577061e-05,
+      "loss": 0.3496,
+      "step": 23340
+    },
+    {
+      "epoch": 150.64,
+      "learning_rate": 4.718637992831542e-05,
+      "loss": 0.3603,
+      "step": 23350
+    },
+    {
+      "epoch": 150.71,
+      "learning_rate": 4.718279569892473e-05,
+      "loss": 0.2994,
+      "step": 23360
+    },
+    {
+      "epoch": 150.77,
+      "learning_rate": 4.717921146953405e-05,
+      "loss": 0.3814,
+      "step": 23370
+    },
+    {
+      "epoch": 150.84,
+      "learning_rate": 4.717562724014337e-05,
+      "loss": 0.3559,
+      "step": 23380
+    },
+    {
+      "epoch": 150.9,
+      "learning_rate": 4.717204301075269e-05,
+      "loss": 0.3411,
+      "step": 23390
+    },
+    {
+      "epoch": 150.96,
+      "learning_rate": 4.716845878136201e-05,
+      "loss": 0.3458,
+      "step": 23400
+    },
+    {
+      "epoch": 151.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0544376373291016,
+      "eval_runtime": 0.8507,
+      "eval_samples_per_second": 236.287,
+      "eval_steps_per_second": 8.229,
+      "step": 23405
+    },
+    {
+      "epoch": 151.03,
+      "learning_rate": 4.7164874551971326e-05,
+      "loss": 0.4012,
+      "step": 23410
+    },
+    {
+      "epoch": 151.1,
+      "learning_rate": 4.716129032258065e-05,
+      "loss": 0.3672,
+      "step": 23420
+    },
+    {
+      "epoch": 151.16,
+      "learning_rate": 4.715770609318997e-05,
+      "loss": 0.3027,
+      "step": 23430
+    },
+    {
+      "epoch": 151.23,
+      "learning_rate": 4.715412186379928e-05,
+      "loss": 0.2945,
+      "step": 23440
+    },
+    {
+      "epoch": 151.29,
+      "learning_rate": 4.7150537634408606e-05,
+      "loss": 0.2768,
+      "step": 23450
+    },
+    {
+      "epoch": 151.35,
+      "learning_rate": 4.7146953405017924e-05,
+      "loss": 0.3233,
+      "step": 23460
+    },
+    {
+      "epoch": 151.42,
+      "learning_rate": 4.714336917562724e-05,
+      "loss": 0.3694,
+      "step": 23470
+    },
+    {
+      "epoch": 151.48,
+      "learning_rate": 4.713978494623656e-05,
+      "loss": 0.2957,
+      "step": 23480
+    },
+    {
+      "epoch": 151.55,
+      "learning_rate": 4.713620071684588e-05,
+      "loss": 0.3954,
+      "step": 23490
+    },
+    {
+      "epoch": 151.61,
+      "learning_rate": 4.71326164874552e-05,
+      "loss": 0.3436,
+      "step": 23500
+    },
+    {
+      "epoch": 151.68,
+      "learning_rate": 4.712903225806452e-05,
+      "loss": 0.3224,
+      "step": 23510
+    },
+    {
+      "epoch": 151.74,
+      "learning_rate": 4.712544802867383e-05,
+      "loss": 0.3164,
+      "step": 23520
+    },
+    {
+      "epoch": 151.8,
+      "learning_rate": 4.712186379928316e-05,
+      "loss": 0.3438,
+      "step": 23530
+    },
+    {
+      "epoch": 151.87,
+      "learning_rate": 4.7118279569892476e-05,
+      "loss": 0.354,
+      "step": 23540
+    },
+    {
+      "epoch": 151.93,
+      "learning_rate": 4.7114695340501794e-05,
+      "loss": 0.3026,
+      "step": 23550
+    },
+    {
+      "epoch": 152.0,
+      "learning_rate": 4.711111111111111e-05,
+      "loss": 0.3362,
+      "step": 23560
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9986661672592163,
+      "eval_runtime": 0.8378,
+      "eval_samples_per_second": 239.903,
+      "eval_steps_per_second": 8.355,
+      "step": 23560
+    },
+    {
+      "epoch": 152.06,
+      "learning_rate": 4.710752688172043e-05,
+      "loss": 0.3338,
+      "step": 23570
+    },
+    {
+      "epoch": 152.13,
+      "learning_rate": 4.710394265232975e-05,
+      "loss": 0.3445,
+      "step": 23580
+    },
+    {
+      "epoch": 152.19,
+      "learning_rate": 4.7100358422939074e-05,
+      "loss": 0.3059,
+      "step": 23590
+    },
+    {
+      "epoch": 152.26,
+      "learning_rate": 4.7096774193548385e-05,
+      "loss": 0.3604,
+      "step": 23600
+    },
+    {
+      "epoch": 152.32,
+      "learning_rate": 4.709318996415771e-05,
+      "loss": 0.3577,
+      "step": 23610
+    },
+    {
+      "epoch": 152.39,
+      "learning_rate": 4.708960573476703e-05,
+      "loss": 0.3277,
+      "step": 23620
+    },
+    {
+      "epoch": 152.45,
+      "learning_rate": 4.708602150537635e-05,
+      "loss": 0.326,
+      "step": 23630
+    },
+    {
+      "epoch": 152.51,
+      "learning_rate": 4.7082437275985665e-05,
+      "loss": 0.3425,
+      "step": 23640
+    },
+    {
+      "epoch": 152.58,
+      "learning_rate": 4.707885304659498e-05,
+      "loss": 0.3048,
+      "step": 23650
+    },
+    {
+      "epoch": 152.64,
+      "learning_rate": 4.70752688172043e-05,
+      "loss": 0.4067,
+      "step": 23660
+    },
+    {
+      "epoch": 152.71,
+      "learning_rate": 4.7071684587813626e-05,
+      "loss": 0.3132,
+      "step": 23670
+    },
+    {
+      "epoch": 152.77,
+      "learning_rate": 4.706810035842294e-05,
+      "loss": 0.3452,
+      "step": 23680
+    },
+    {
+      "epoch": 152.84,
+      "learning_rate": 4.706451612903226e-05,
+      "loss": 0.3598,
+      "step": 23690
+    },
+    {
+      "epoch": 152.9,
+      "learning_rate": 4.706093189964158e-05,
+      "loss": 0.3777,
+      "step": 23700
+    },
+    {
+      "epoch": 152.96,
+      "learning_rate": 4.70573476702509e-05,
+      "loss": 0.3306,
+      "step": 23710
+    },
+    {
+      "epoch": 153.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9995443820953369,
+      "eval_runtime": 0.8748,
+      "eval_samples_per_second": 229.756,
+      "eval_steps_per_second": 8.001,
+      "step": 23715
+    },
+    {
+      "epoch": 153.03,
+      "learning_rate": 4.705376344086022e-05,
+      "loss": 0.3158,
+      "step": 23720
+    },
+    {
+      "epoch": 153.1,
+      "learning_rate": 4.7050179211469536e-05,
+      "loss": 0.3447,
+      "step": 23730
+    },
+    {
+      "epoch": 153.16,
+      "learning_rate": 4.7046594982078854e-05,
+      "loss": 0.3771,
+      "step": 23740
+    },
+    {
+      "epoch": 153.23,
+      "learning_rate": 4.704301075268818e-05,
+      "loss": 0.3242,
+      "step": 23750
+    },
+    {
+      "epoch": 153.29,
+      "learning_rate": 4.703942652329749e-05,
+      "loss": 0.32,
+      "step": 23760
+    },
+    {
+      "epoch": 153.35,
+      "learning_rate": 4.7035842293906815e-05,
+      "loss": 0.2856,
+      "step": 23770
+    },
+    {
+      "epoch": 153.42,
+      "learning_rate": 4.7032258064516133e-05,
+      "loss": 0.3686,
+      "step": 23780
+    },
+    {
+      "epoch": 153.48,
+      "learning_rate": 4.702867383512545e-05,
+      "loss": 0.3621,
+      "step": 23790
+    },
+    {
+      "epoch": 153.55,
+      "learning_rate": 4.702508960573477e-05,
+      "loss": 0.3645,
+      "step": 23800
+    },
+    {
+      "epoch": 153.61,
+      "learning_rate": 4.702150537634409e-05,
+      "loss": 0.2909,
+      "step": 23810
+    },
+    {
+      "epoch": 153.68,
+      "learning_rate": 4.7017921146953406e-05,
+      "loss": 0.2906,
+      "step": 23820
+    },
+    {
+      "epoch": 153.74,
+      "learning_rate": 4.7014336917562724e-05,
+      "loss": 0.3264,
+      "step": 23830
+    },
+    {
+      "epoch": 153.8,
+      "learning_rate": 4.701075268817204e-05,
+      "loss": 0.3601,
+      "step": 23840
+    },
+    {
+      "epoch": 153.87,
+      "learning_rate": 4.700716845878137e-05,
+      "loss": 0.314,
+      "step": 23850
+    },
+    {
+      "epoch": 153.93,
+      "learning_rate": 4.7003584229390686e-05,
+      "loss": 0.3455,
+      "step": 23860
+    },
+    {
+      "epoch": 154.0,
+      "learning_rate": 4.7e-05,
+      "loss": 0.3861,
+      "step": 23870
+    },
+    {
+      "epoch": 154.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0079283714294434,
+      "eval_runtime": 0.8583,
+      "eval_samples_per_second": 234.19,
+      "eval_steps_per_second": 8.156,
+      "step": 23870
+    },
+    {
+      "epoch": 154.06,
+      "learning_rate": 4.699641577060932e-05,
+      "loss": 0.3239,
+      "step": 23880
+    },
+    {
+      "epoch": 154.13,
+      "learning_rate": 4.699283154121864e-05,
+      "loss": 0.3146,
+      "step": 23890
+    },
+    {
+      "epoch": 154.19,
+      "learning_rate": 4.698924731182796e-05,
+      "loss": 0.3731,
+      "step": 23900
+    },
+    {
+      "epoch": 154.26,
+      "learning_rate": 4.698566308243728e-05,
+      "loss": 0.3058,
+      "step": 23910
+    },
+    {
+      "epoch": 154.32,
+      "learning_rate": 4.6982078853046595e-05,
+      "loss": 0.3451,
+      "step": 23920
+    },
+    {
+      "epoch": 154.39,
+      "learning_rate": 4.697849462365591e-05,
+      "loss": 0.3048,
+      "step": 23930
+    },
+    {
+      "epoch": 154.45,
+      "learning_rate": 4.697491039426524e-05,
+      "loss": 0.3376,
+      "step": 23940
+    },
+    {
+      "epoch": 154.51,
+      "learning_rate": 4.697132616487455e-05,
+      "loss": 0.321,
+      "step": 23950
+    },
+    {
+      "epoch": 154.58,
+      "learning_rate": 4.6967741935483875e-05,
+      "loss": 0.3415,
+      "step": 23960
+    },
+    {
+      "epoch": 154.64,
+      "learning_rate": 4.696415770609319e-05,
+      "loss": 0.3191,
+      "step": 23970
+    },
+    {
+      "epoch": 154.71,
+      "learning_rate": 4.696057347670251e-05,
+      "loss": 0.3408,
+      "step": 23980
+    },
+    {
+      "epoch": 154.77,
+      "learning_rate": 4.695698924731183e-05,
+      "loss": 0.3148,
+      "step": 23990
+    },
+    {
+      "epoch": 154.84,
+      "learning_rate": 4.695340501792115e-05,
+      "loss": 0.3246,
+      "step": 24000
+    },
+    {
+      "epoch": 154.9,
+      "learning_rate": 4.6949820788530466e-05,
+      "loss": 0.4117,
+      "step": 24010
+    },
+    {
+      "epoch": 154.96,
+      "learning_rate": 4.694623655913979e-05,
+      "loss": 0.3896,
+      "step": 24020
+    },
+    {
+      "epoch": 155.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9906116724014282,
+      "eval_runtime": 0.842,
+      "eval_samples_per_second": 238.713,
+      "eval_steps_per_second": 8.313,
+      "step": 24025
+    },
+    {
+      "epoch": 155.03,
+      "learning_rate": 4.69426523297491e-05,
+      "loss": 0.3894,
+      "step": 24030
+    },
+    {
+      "epoch": 155.1,
+      "learning_rate": 4.693906810035843e-05,
+      "loss": 0.3163,
+      "step": 24040
+    },
+    {
+      "epoch": 155.16,
+      "learning_rate": 4.6935483870967745e-05,
+      "loss": 0.3556,
+      "step": 24050
+    },
+    {
+      "epoch": 155.23,
+      "learning_rate": 4.6931899641577063e-05,
+      "loss": 0.2884,
+      "step": 24060
+    },
+    {
+      "epoch": 155.29,
+      "learning_rate": 4.692831541218638e-05,
+      "loss": 0.3685,
+      "step": 24070
+    },
+    {
+      "epoch": 155.35,
+      "learning_rate": 4.69247311827957e-05,
+      "loss": 0.3491,
+      "step": 24080
+    },
+    {
+      "epoch": 155.42,
+      "learning_rate": 4.692114695340502e-05,
+      "loss": 0.3567,
+      "step": 24090
+    },
+    {
+      "epoch": 155.48,
+      "learning_rate": 4.691756272401434e-05,
+      "loss": 0.318,
+      "step": 24100
+    },
+    {
+      "epoch": 155.55,
+      "learning_rate": 4.6913978494623654e-05,
+      "loss": 0.2977,
+      "step": 24110
+    },
+    {
+      "epoch": 155.61,
+      "learning_rate": 4.691039426523298e-05,
+      "loss": 0.2964,
+      "step": 24120
+    },
+    {
+      "epoch": 155.68,
+      "learning_rate": 4.69068100358423e-05,
+      "loss": 0.3091,
+      "step": 24130
+    },
+    {
+      "epoch": 155.74,
+      "learning_rate": 4.6903225806451616e-05,
+      "loss": 0.3042,
+      "step": 24140
+    },
+    {
+      "epoch": 155.8,
+      "learning_rate": 4.6899641577060934e-05,
+      "loss": 0.3083,
+      "step": 24150
+    },
+    {
+      "epoch": 155.87,
+      "learning_rate": 4.689605734767025e-05,
+      "loss": 0.3281,
+      "step": 24160
+    },
+    {
+      "epoch": 155.93,
+      "learning_rate": 4.689247311827957e-05,
+      "loss": 0.3724,
+      "step": 24170
+    },
+    {
+      "epoch": 156.0,
+      "learning_rate": 4.6888888888888895e-05,
+      "loss": 0.3218,
+      "step": 24180
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0362294912338257,
+      "eval_runtime": 0.8596,
+      "eval_samples_per_second": 233.824,
+      "eval_steps_per_second": 8.143,
+      "step": 24180
+    },
+    {
+      "epoch": 156.06,
+      "learning_rate": 4.688530465949821e-05,
+      "loss": 0.3703,
+      "step": 24190
+    },
+    {
+      "epoch": 156.13,
+      "learning_rate": 4.688172043010753e-05,
+      "loss": 0.2705,
+      "step": 24200
+    },
+    {
+      "epoch": 156.19,
+      "learning_rate": 4.687813620071685e-05,
+      "loss": 0.3618,
+      "step": 24210
+    },
+    {
+      "epoch": 156.26,
+      "learning_rate": 4.687455197132617e-05,
+      "loss": 0.3344,
+      "step": 24220
+    },
+    {
+      "epoch": 156.32,
+      "learning_rate": 4.6870967741935486e-05,
+      "loss": 0.3472,
+      "step": 24230
+    },
+    {
+      "epoch": 156.39,
+      "learning_rate": 4.6867383512544805e-05,
+      "loss": 0.2904,
+      "step": 24240
+    },
+    {
+      "epoch": 156.45,
+      "learning_rate": 4.686379928315412e-05,
+      "loss": 0.3402,
+      "step": 24250
+    },
+    {
+      "epoch": 156.51,
+      "learning_rate": 4.686021505376345e-05,
+      "loss": 0.3008,
+      "step": 24260
+    },
+    {
+      "epoch": 156.58,
+      "learning_rate": 4.685663082437276e-05,
+      "loss": 0.296,
+      "step": 24270
+    },
+    {
+      "epoch": 156.64,
+      "learning_rate": 4.685304659498208e-05,
+      "loss": 0.4295,
+      "step": 24280
+    },
+    {
+      "epoch": 156.71,
+      "learning_rate": 4.68494623655914e-05,
+      "loss": 0.3236,
+      "step": 24290
+    },
+    {
+      "epoch": 156.77,
+      "learning_rate": 4.684587813620072e-05,
+      "loss": 0.3377,
+      "step": 24300
+    },
+    {
+      "epoch": 156.84,
+      "learning_rate": 4.684229390681004e-05,
+      "loss": 0.3079,
+      "step": 24310
+    },
+    {
+      "epoch": 156.9,
+      "learning_rate": 4.683870967741936e-05,
+      "loss": 0.3237,
+      "step": 24320
+    },
+    {
+      "epoch": 156.96,
+      "learning_rate": 4.6835125448028675e-05,
+      "loss": 0.3523,
+      "step": 24330
+    },
+    {
+      "epoch": 157.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0087608098983765,
+      "eval_runtime": 0.891,
+      "eval_samples_per_second": 225.6,
+      "eval_steps_per_second": 7.857,
+      "step": 24335
+    },
+    {
+      "epoch": 157.03,
+      "learning_rate": 4.6831541218638e-05,
+      "loss": 0.3911,
+      "step": 24340
+    },
+    {
+      "epoch": 157.1,
+      "learning_rate": 4.682795698924731e-05,
+      "loss": 0.3063,
+      "step": 24350
+    },
+    {
+      "epoch": 157.16,
+      "learning_rate": 4.682437275985663e-05,
+      "loss": 0.317,
+      "step": 24360
+    },
+    {
+      "epoch": 157.23,
+      "learning_rate": 4.6820788530465955e-05,
+      "loss": 0.3494,
+      "step": 24370
+    },
+    {
+      "epoch": 157.29,
+      "learning_rate": 4.681720430107527e-05,
+      "loss": 0.2958,
+      "step": 24380
+    },
+    {
+      "epoch": 157.35,
+      "learning_rate": 4.681362007168459e-05,
+      "loss": 0.3154,
+      "step": 24390
+    },
+    {
+      "epoch": 157.42,
+      "learning_rate": 4.681003584229391e-05,
+      "loss": 0.3675,
+      "step": 24400
+    },
+    {
+      "epoch": 157.48,
+      "learning_rate": 4.680645161290323e-05,
+      "loss": 0.3462,
+      "step": 24410
+    },
+    {
+      "epoch": 157.55,
+      "learning_rate": 4.680286738351255e-05,
+      "loss": 0.3684,
+      "step": 24420
+    },
+    {
+      "epoch": 157.61,
+      "learning_rate": 4.6799283154121864e-05,
+      "loss": 0.3349,
+      "step": 24430
+    },
+    {
+      "epoch": 157.68,
+      "learning_rate": 4.679569892473118e-05,
+      "loss": 0.3669,
+      "step": 24440
+    },
+    {
+      "epoch": 157.74,
+      "learning_rate": 4.679211469534051e-05,
+      "loss": 0.348,
+      "step": 24450
+    },
+    {
+      "epoch": 157.8,
+      "learning_rate": 4.678853046594982e-05,
+      "loss": 0.3229,
+      "step": 24460
+    },
+    {
+      "epoch": 157.87,
+      "learning_rate": 4.6784946236559144e-05,
+      "loss": 0.3434,
+      "step": 24470
+    },
+    {
+      "epoch": 157.93,
+      "learning_rate": 4.678136200716846e-05,
+      "loss": 0.3117,
+      "step": 24480
+    },
+    {
+      "epoch": 158.0,
+      "learning_rate": 4.677777777777778e-05,
+      "loss": 0.3255,
+      "step": 24490
+    },
+    {
+      "epoch": 158.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0459743738174438,
+      "eval_runtime": 0.8628,
+      "eval_samples_per_second": 232.956,
+      "eval_steps_per_second": 8.113,
+      "step": 24490
+    },
+    {
+      "epoch": 158.06,
+      "learning_rate": 4.67741935483871e-05,
+      "loss": 0.3382,
+      "step": 24500
+    },
+    {
+      "epoch": 158.13,
+      "learning_rate": 4.6770609318996417e-05,
+      "loss": 0.3417,
+      "step": 24510
+    },
+    {
+      "epoch": 158.19,
+      "learning_rate": 4.6767025089605735e-05,
+      "loss": 0.2972,
+      "step": 24520
+    },
+    {
+      "epoch": 158.26,
+      "learning_rate": 4.676344086021506e-05,
+      "loss": 0.3067,
+      "step": 24530
+    },
+    {
+      "epoch": 158.32,
+      "learning_rate": 4.675985663082437e-05,
+      "loss": 0.3554,
+      "step": 24540
+    },
+    {
+      "epoch": 158.39,
+      "learning_rate": 4.6756272401433696e-05,
+      "loss": 0.3126,
+      "step": 24550
+    },
+    {
+      "epoch": 158.45,
+      "learning_rate": 4.6752688172043014e-05,
+      "loss": 0.3138,
+      "step": 24560
+    },
+    {
+      "epoch": 158.51,
+      "learning_rate": 4.674910394265233e-05,
+      "loss": 0.309,
+      "step": 24570
+    },
+    {
+      "epoch": 158.58,
+      "learning_rate": 4.674551971326165e-05,
+      "loss": 0.3524,
+      "step": 24580
+    },
+    {
+      "epoch": 158.64,
+      "learning_rate": 4.674193548387097e-05,
+      "loss": 0.3154,
+      "step": 24590
+    },
+    {
+      "epoch": 158.71,
+      "learning_rate": 4.673835125448029e-05,
+      "loss": 0.3178,
+      "step": 24600
+    },
+    {
+      "epoch": 158.77,
+      "learning_rate": 4.673476702508961e-05,
+      "loss": 0.2693,
+      "step": 24610
+    },
+    {
+      "epoch": 158.84,
+      "learning_rate": 4.6731182795698924e-05,
+      "loss": 0.3608,
+      "step": 24620
+    },
+    {
+      "epoch": 158.9,
+      "learning_rate": 4.672759856630824e-05,
+      "loss": 0.4185,
+      "step": 24630
+    },
+    {
+      "epoch": 158.96,
+      "learning_rate": 4.672401433691757e-05,
+      "loss": 0.3648,
+      "step": 24640
+    },
+    {
+      "epoch": 159.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9424301981925964,
+      "eval_runtime": 0.8657,
+      "eval_samples_per_second": 232.185,
+      "eval_steps_per_second": 8.086,
+      "step": 24645
+    },
+    {
+      "epoch": 159.03,
+      "learning_rate": 4.6720430107526885e-05,
+      "loss": 0.349,
+      "step": 24650
+    },
+    {
+      "epoch": 159.1,
+      "learning_rate": 4.67168458781362e-05,
+      "loss": 0.3987,
+      "step": 24660
+    },
+    {
+      "epoch": 159.16,
+      "learning_rate": 4.671326164874552e-05,
+      "loss": 0.3191,
+      "step": 24670
+    },
+    {
+      "epoch": 159.23,
+      "learning_rate": 4.670967741935484e-05,
+      "loss": 0.3001,
+      "step": 24680
+    },
+    {
+      "epoch": 159.29,
+      "learning_rate": 4.6706093189964165e-05,
+      "loss": 0.3061,
+      "step": 24690
+    },
+    {
+      "epoch": 159.35,
+      "learning_rate": 4.6702508960573476e-05,
+      "loss": 0.2737,
+      "step": 24700
+    },
+    {
+      "epoch": 159.42,
+      "learning_rate": 4.6698924731182794e-05,
+      "loss": 0.3506,
+      "step": 24710
+    },
+    {
+      "epoch": 159.48,
+      "learning_rate": 4.669534050179212e-05,
+      "loss": 0.3509,
+      "step": 24720
+    },
+    {
+      "epoch": 159.55,
+      "learning_rate": 4.669175627240144e-05,
+      "loss": 0.3358,
+      "step": 24730
+    },
+    {
+      "epoch": 159.61,
+      "learning_rate": 4.6688172043010756e-05,
+      "loss": 0.3065,
+      "step": 24740
+    },
+    {
+      "epoch": 159.68,
+      "learning_rate": 4.6684587813620074e-05,
+      "loss": 0.2822,
+      "step": 24750
+    },
+    {
+      "epoch": 159.74,
+      "learning_rate": 4.668100358422939e-05,
+      "loss": 0.3907,
+      "step": 24760
+    },
+    {
+      "epoch": 159.8,
+      "learning_rate": 4.667741935483872e-05,
+      "loss": 0.3082,
+      "step": 24770
+    },
+    {
+      "epoch": 159.87,
+      "learning_rate": 4.667383512544803e-05,
+      "loss": 0.3174,
+      "step": 24780
+    },
+    {
+      "epoch": 159.93,
+      "learning_rate": 4.667025089605735e-05,
+      "loss": 0.3498,
+      "step": 24790
+    },
+    {
+      "epoch": 160.0,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 0.4195,
+      "step": 24800
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9587200284004211,
+      "eval_runtime": 0.852,
+      "eval_samples_per_second": 235.912,
+      "eval_steps_per_second": 8.216,
+      "step": 24800
+    },
+    {
+      "epoch": 160.06,
+      "learning_rate": 4.666308243727599e-05,
+      "loss": 0.3362,
+      "step": 24810
+    },
+    {
+      "epoch": 160.13,
+      "learning_rate": 4.665949820788531e-05,
+      "loss": 0.3386,
+      "step": 24820
+    },
+    {
+      "epoch": 160.19,
+      "learning_rate": 4.6655913978494626e-05,
+      "loss": 0.3212,
+      "step": 24830
+    },
+    {
+      "epoch": 160.26,
+      "learning_rate": 4.6652329749103944e-05,
+      "loss": 0.3042,
+      "step": 24840
+    },
+    {
+      "epoch": 160.32,
+      "learning_rate": 4.664874551971327e-05,
+      "loss": 0.3195,
+      "step": 24850
+    },
+    {
+      "epoch": 160.39,
+      "learning_rate": 4.664516129032258e-05,
+      "loss": 0.3102,
+      "step": 24860
+    },
+    {
+      "epoch": 160.45,
+      "learning_rate": 4.66415770609319e-05,
+      "loss": 0.3751,
+      "step": 24870
+    },
+    {
+      "epoch": 160.51,
+      "learning_rate": 4.6637992831541224e-05,
+      "loss": 0.3726,
+      "step": 24880
+    },
+    {
+      "epoch": 160.58,
+      "learning_rate": 4.663440860215054e-05,
+      "loss": 0.3246,
+      "step": 24890
+    },
+    {
+      "epoch": 160.64,
+      "learning_rate": 4.6630824372759854e-05,
+      "loss": 0.4137,
+      "step": 24900
+    },
+    {
+      "epoch": 160.71,
+      "learning_rate": 4.662724014336918e-05,
+      "loss": 0.2803,
+      "step": 24910
+    },
+    {
+      "epoch": 160.77,
+      "learning_rate": 4.66236559139785e-05,
+      "loss": 0.3124,
+      "step": 24920
+    },
+    {
+      "epoch": 160.84,
+      "learning_rate": 4.6620071684587815e-05,
+      "loss": 0.3353,
+      "step": 24930
+    },
+    {
+      "epoch": 160.9,
+      "learning_rate": 4.661648745519713e-05,
+      "loss": 0.3491,
+      "step": 24940
+    },
+    {
+      "epoch": 160.96,
+      "learning_rate": 4.661290322580645e-05,
+      "loss": 0.3286,
+      "step": 24950
+    },
+    {
+      "epoch": 161.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.06204354763031,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.489,
+      "eval_steps_per_second": 8.027,
+      "step": 24955
+    },
+    {
+      "epoch": 161.03,
+      "learning_rate": 4.6609318996415776e-05,
+      "loss": 0.3554,
+      "step": 24960
+    },
+    {
+      "epoch": 161.1,
+      "learning_rate": 4.6605734767025095e-05,
+      "loss": 0.3334,
+      "step": 24970
+    },
+    {
+      "epoch": 161.16,
+      "learning_rate": 4.6602150537634406e-05,
+      "loss": 0.3681,
+      "step": 24980
+    },
+    {
+      "epoch": 161.23,
+      "learning_rate": 4.659856630824373e-05,
+      "loss": 0.3295,
+      "step": 24990
+    },
+    {
+      "epoch": 161.29,
+      "learning_rate": 4.659498207885305e-05,
+      "loss": 0.3188,
+      "step": 25000
+    },
+    {
+      "epoch": 161.35,
+      "learning_rate": 4.659139784946237e-05,
+      "loss": 0.3367,
+      "step": 25010
+    },
+    {
+      "epoch": 161.42,
+      "learning_rate": 4.6587813620071686e-05,
+      "loss": 0.3325,
+      "step": 25020
+    },
+    {
+      "epoch": 161.48,
+      "learning_rate": 4.6584229390681004e-05,
+      "loss": 0.3157,
+      "step": 25030
+    },
+    {
+      "epoch": 161.55,
+      "learning_rate": 4.658064516129033e-05,
+      "loss": 0.2971,
+      "step": 25040
+    },
+    {
+      "epoch": 161.61,
+      "learning_rate": 4.657706093189965e-05,
+      "loss": 0.2812,
+      "step": 25050
+    },
+    {
+      "epoch": 161.68,
+      "learning_rate": 4.657347670250896e-05,
+      "loss": 0.3285,
+      "step": 25060
+    },
+    {
+      "epoch": 161.74,
+      "learning_rate": 4.6569892473118283e-05,
+      "loss": 0.3209,
+      "step": 25070
+    },
+    {
+      "epoch": 161.8,
+      "learning_rate": 4.65663082437276e-05,
+      "loss": 0.3581,
+      "step": 25080
+    },
+    {
+      "epoch": 161.87,
+      "learning_rate": 4.656272401433692e-05,
+      "loss": 0.2685,
+      "step": 25090
+    },
+    {
+      "epoch": 161.93,
+      "learning_rate": 4.655913978494624e-05,
+      "loss": 0.2883,
+      "step": 25100
+    },
+    {
+      "epoch": 162.0,
+      "learning_rate": 4.6555555555555556e-05,
+      "loss": 0.3777,
+      "step": 25110
+    },
+    {
+      "epoch": 162.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0374493598937988,
+      "eval_runtime": 0.8806,
+      "eval_samples_per_second": 228.241,
+      "eval_steps_per_second": 7.949,
+      "step": 25110
+    },
+    {
+      "epoch": 162.06,
+      "learning_rate": 4.655197132616488e-05,
+      "loss": 0.3833,
+      "step": 25120
+    },
+    {
+      "epoch": 162.13,
+      "learning_rate": 4.654838709677419e-05,
+      "loss": 0.3011,
+      "step": 25130
+    },
+    {
+      "epoch": 162.19,
+      "learning_rate": 4.654480286738351e-05,
+      "loss": 0.3084,
+      "step": 25140
+    },
+    {
+      "epoch": 162.26,
+      "learning_rate": 4.6541218637992836e-05,
+      "loss": 0.2989,
+      "step": 25150
+    },
+    {
+      "epoch": 162.32,
+      "learning_rate": 4.6537634408602154e-05,
+      "loss": 0.2858,
+      "step": 25160
+    },
+    {
+      "epoch": 162.39,
+      "learning_rate": 4.653405017921147e-05,
+      "loss": 0.2897,
+      "step": 25170
+    },
+    {
+      "epoch": 162.45,
+      "learning_rate": 4.653046594982079e-05,
+      "loss": 0.3013,
+      "step": 25180
+    },
+    {
+      "epoch": 162.51,
+      "learning_rate": 4.652688172043011e-05,
+      "loss": 0.3083,
+      "step": 25190
+    },
+    {
+      "epoch": 162.58,
+      "learning_rate": 4.6523297491039434e-05,
+      "loss": 0.3511,
+      "step": 25200
+    },
+    {
+      "epoch": 162.64,
+      "learning_rate": 4.6519713261648745e-05,
+      "loss": 0.3644,
+      "step": 25210
+    },
+    {
+      "epoch": 162.71,
+      "learning_rate": 4.651612903225806e-05,
+      "loss": 0.3558,
+      "step": 25220
+    },
+    {
+      "epoch": 162.77,
+      "learning_rate": 4.651254480286739e-05,
+      "loss": 0.3558,
+      "step": 25230
+    },
+    {
+      "epoch": 162.84,
+      "learning_rate": 4.6508960573476707e-05,
+      "loss": 0.3215,
+      "step": 25240
+    },
+    {
+      "epoch": 162.9,
+      "learning_rate": 4.650537634408602e-05,
+      "loss": 0.3426,
+      "step": 25250
+    },
+    {
+      "epoch": 162.96,
+      "learning_rate": 4.650179211469534e-05,
+      "loss": 0.2927,
+      "step": 25260
+    },
+    {
+      "epoch": 163.0,
+      "eval_accuracy": 0.7313432835820896,
+      "eval_loss": 1.0674864053726196,
+      "eval_runtime": 0.8711,
+      "eval_samples_per_second": 230.737,
+      "eval_steps_per_second": 8.036,
+      "step": 25265
+    },
+    {
+      "epoch": 163.03,
+      "learning_rate": 4.649820788530466e-05,
+      "loss": 0.4235,
+      "step": 25270
+    },
+    {
+      "epoch": 163.1,
+      "learning_rate": 4.649462365591398e-05,
+      "loss": 0.36,
+      "step": 25280
+    },
+    {
+      "epoch": 163.16,
+      "learning_rate": 4.64910394265233e-05,
+      "loss": 0.3535,
+      "step": 25290
+    },
+    {
+      "epoch": 163.23,
+      "learning_rate": 4.6487455197132616e-05,
+      "loss": 0.3267,
+      "step": 25300
+    },
+    {
+      "epoch": 163.29,
+      "learning_rate": 4.648387096774194e-05,
+      "loss": 0.2941,
+      "step": 25310
+    },
+    {
+      "epoch": 163.35,
+      "learning_rate": 4.648028673835126e-05,
+      "loss": 0.3113,
+      "step": 25320
+    },
+    {
+      "epoch": 163.42,
+      "learning_rate": 4.647670250896057e-05,
+      "loss": 0.3131,
+      "step": 25330
+    },
+    {
+      "epoch": 163.48,
+      "learning_rate": 4.6473118279569895e-05,
+      "loss": 0.3263,
+      "step": 25340
+    },
+    {
+      "epoch": 163.55,
+      "learning_rate": 4.6469534050179214e-05,
+      "loss": 0.2959,
+      "step": 25350
+    },
+    {
+      "epoch": 163.61,
+      "learning_rate": 4.646594982078853e-05,
+      "loss": 0.3576,
+      "step": 25360
+    },
+    {
+      "epoch": 163.68,
+      "learning_rate": 4.646236559139785e-05,
+      "loss": 0.3314,
+      "step": 25370
+    },
+    {
+      "epoch": 163.74,
+      "learning_rate": 4.645878136200717e-05,
+      "loss": 0.304,
+      "step": 25380
+    },
+    {
+      "epoch": 163.8,
+      "learning_rate": 4.645519713261649e-05,
+      "loss": 0.3388,
+      "step": 25390
+    },
+    {
+      "epoch": 163.87,
+      "learning_rate": 4.645161290322581e-05,
+      "loss": 0.421,
+      "step": 25400
+    },
+    {
+      "epoch": 163.93,
+      "learning_rate": 4.644802867383512e-05,
+      "loss": 0.3323,
+      "step": 25410
+    },
+    {
+      "epoch": 164.0,
+      "learning_rate": 4.644444444444445e-05,
+      "loss": 0.3045,
+      "step": 25420
+    },
+    {
+      "epoch": 164.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0031135082244873,
+      "eval_runtime": 0.8528,
+      "eval_samples_per_second": 235.703,
+      "eval_steps_per_second": 8.209,
+      "step": 25420
+    },
+    {
+      "epoch": 164.06,
+      "learning_rate": 4.6440860215053766e-05,
+      "loss": 0.3893,
+      "step": 25430
+    },
+    {
+      "epoch": 164.13,
+      "learning_rate": 4.6437275985663084e-05,
+      "loss": 0.3281,
+      "step": 25440
+    },
+    {
+      "epoch": 164.19,
+      "learning_rate": 4.64336917562724e-05,
+      "loss": 0.295,
+      "step": 25450
+    },
+    {
+      "epoch": 164.26,
+      "learning_rate": 4.643010752688172e-05,
+      "loss": 0.3267,
+      "step": 25460
+    },
+    {
+      "epoch": 164.32,
+      "learning_rate": 4.6426523297491046e-05,
+      "loss": 0.373,
+      "step": 25470
+    },
+    {
+      "epoch": 164.39,
+      "learning_rate": 4.6422939068100364e-05,
+      "loss": 0.3099,
+      "step": 25480
+    },
+    {
+      "epoch": 164.45,
+      "learning_rate": 4.6419354838709675e-05,
+      "loss": 0.3439,
+      "step": 25490
+    },
+    {
+      "epoch": 164.51,
+      "learning_rate": 4.6415770609319e-05,
+      "loss": 0.2761,
+      "step": 25500
+    },
+    {
+      "epoch": 164.58,
+      "learning_rate": 4.641218637992832e-05,
+      "loss": 0.3422,
+      "step": 25510
+    },
+    {
+      "epoch": 164.64,
+      "learning_rate": 4.6408602150537637e-05,
+      "loss": 0.2779,
+      "step": 25520
+    },
+    {
+      "epoch": 164.71,
+      "learning_rate": 4.6405017921146955e-05,
+      "loss": 0.3549,
+      "step": 25530
+    },
+    {
+      "epoch": 164.77,
+      "learning_rate": 4.640143369175627e-05,
+      "loss": 0.3165,
+      "step": 25540
+    },
+    {
+      "epoch": 164.84,
+      "learning_rate": 4.63978494623656e-05,
+      "loss": 0.3358,
+      "step": 25550
+    },
+    {
+      "epoch": 164.9,
+      "learning_rate": 4.6394265232974916e-05,
+      "loss": 0.3355,
+      "step": 25560
+    },
+    {
+      "epoch": 164.96,
+      "learning_rate": 4.639068100358423e-05,
+      "loss": 0.3262,
+      "step": 25570
+    },
+    {
+      "epoch": 165.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0606225728988647,
+      "eval_runtime": 0.8703,
+      "eval_samples_per_second": 230.963,
+      "eval_steps_per_second": 8.044,
+      "step": 25575
+    },
+    {
+      "epoch": 165.03,
+      "learning_rate": 4.638709677419355e-05,
+      "loss": 0.2714,
+      "step": 25580
+    },
+    {
+      "epoch": 165.1,
+      "learning_rate": 4.638351254480287e-05,
+      "loss": 0.3131,
+      "step": 25590
+    },
+    {
+      "epoch": 165.16,
+      "learning_rate": 4.637992831541219e-05,
+      "loss": 0.3094,
+      "step": 25600
+    },
+    {
+      "epoch": 165.23,
+      "learning_rate": 4.637634408602151e-05,
+      "loss": 0.3265,
+      "step": 25610
+    },
+    {
+      "epoch": 165.29,
+      "learning_rate": 4.6372759856630825e-05,
+      "loss": 0.3259,
+      "step": 25620
+    },
+    {
+      "epoch": 165.35,
+      "learning_rate": 4.6369175627240144e-05,
+      "loss": 0.3149,
+      "step": 25630
+    },
+    {
+      "epoch": 165.42,
+      "learning_rate": 4.636559139784947e-05,
+      "loss": 0.3465,
+      "step": 25640
+    },
+    {
+      "epoch": 165.48,
+      "learning_rate": 4.636200716845878e-05,
+      "loss": 0.3565,
+      "step": 25650
+    },
+    {
+      "epoch": 165.55,
+      "learning_rate": 4.6358422939068105e-05,
+      "loss": 0.3097,
+      "step": 25660
+    },
+    {
+      "epoch": 165.61,
+      "learning_rate": 4.635483870967742e-05,
+      "loss": 0.3898,
+      "step": 25670
+    },
+    {
+      "epoch": 165.68,
+      "learning_rate": 4.635125448028674e-05,
+      "loss": 0.3299,
+      "step": 25680
+    },
+    {
+      "epoch": 165.74,
+      "learning_rate": 4.634767025089606e-05,
+      "loss": 0.3642,
+      "step": 25690
+    },
+    {
+      "epoch": 165.8,
+      "learning_rate": 4.634408602150538e-05,
+      "loss": 0.3377,
+      "step": 25700
+    },
+    {
+      "epoch": 165.87,
+      "learning_rate": 4.6340501792114696e-05,
+      "loss": 0.3787,
+      "step": 25710
+    },
+    {
+      "epoch": 165.93,
+      "learning_rate": 4.633691756272402e-05,
+      "loss": 0.3168,
+      "step": 25720
+    },
+    {
+      "epoch": 166.0,
+      "learning_rate": 4.633333333333333e-05,
+      "loss": 0.326,
+      "step": 25730
+    },
+    {
+      "epoch": 166.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9952878952026367,
+      "eval_runtime": 0.8455,
+      "eval_samples_per_second": 237.737,
+      "eval_steps_per_second": 8.279,
+      "step": 25730
+    },
+    {
+      "epoch": 166.06,
+      "learning_rate": 4.632974910394266e-05,
+      "loss": 0.3489,
+      "step": 25740
+    },
+    {
+      "epoch": 166.13,
+      "learning_rate": 4.6326164874551976e-05,
+      "loss": 0.3186,
+      "step": 25750
+    },
+    {
+      "epoch": 166.19,
+      "learning_rate": 4.632258064516129e-05,
+      "loss": 0.3561,
+      "step": 25760
+    },
+    {
+      "epoch": 166.26,
+      "learning_rate": 4.631899641577061e-05,
+      "loss": 0.3229,
+      "step": 25770
+    },
+    {
+      "epoch": 166.32,
+      "learning_rate": 4.631541218637993e-05,
+      "loss": 0.3147,
+      "step": 25780
+    },
+    {
+      "epoch": 166.39,
+      "learning_rate": 4.631182795698925e-05,
+      "loss": 0.3578,
+      "step": 25790
+    },
+    {
+      "epoch": 166.45,
+      "learning_rate": 4.630824372759857e-05,
+      "loss": 0.3386,
+      "step": 25800
+    },
+    {
+      "epoch": 166.51,
+      "learning_rate": 4.6304659498207885e-05,
+      "loss": 0.3758,
+      "step": 25810
+    },
+    {
+      "epoch": 166.58,
+      "learning_rate": 4.630107526881721e-05,
+      "loss": 0.2856,
+      "step": 25820
+    },
+    {
+      "epoch": 166.64,
+      "learning_rate": 4.629749103942653e-05,
+      "loss": 0.302,
+      "step": 25830
+    },
+    {
+      "epoch": 166.71,
+      "learning_rate": 4.629390681003584e-05,
+      "loss": 0.3189,
+      "step": 25840
+    },
+    {
+      "epoch": 166.77,
+      "learning_rate": 4.6290322580645164e-05,
+      "loss": 0.3442,
+      "step": 25850
+    },
+    {
+      "epoch": 166.84,
+      "learning_rate": 4.628673835125448e-05,
+      "loss": 0.3276,
+      "step": 25860
+    },
+    {
+      "epoch": 166.9,
+      "learning_rate": 4.62831541218638e-05,
+      "loss": 0.3791,
+      "step": 25870
+    },
+    {
+      "epoch": 166.96,
+      "learning_rate": 4.627956989247312e-05,
+      "loss": 0.3168,
+      "step": 25880
+    },
+    {
+      "epoch": 167.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0390187501907349,
+      "eval_runtime": 0.8926,
+      "eval_samples_per_second": 225.195,
+      "eval_steps_per_second": 7.843,
+      "step": 25885
+    },
+    {
+      "epoch": 167.03,
+      "learning_rate": 4.627598566308244e-05,
+      "loss": 0.3596,
+      "step": 25890
+    },
+    {
+      "epoch": 167.1,
+      "learning_rate": 4.6272401433691755e-05,
+      "loss": 0.3565,
+      "step": 25900
+    },
+    {
+      "epoch": 167.16,
+      "learning_rate": 4.626881720430108e-05,
+      "loss": 0.2744,
+      "step": 25910
+    },
+    {
+      "epoch": 167.23,
+      "learning_rate": 4.626523297491039e-05,
+      "loss": 0.3848,
+      "step": 25920
+    },
+    {
+      "epoch": 167.29,
+      "learning_rate": 4.626164874551972e-05,
+      "loss": 0.3069,
+      "step": 25930
+    },
+    {
+      "epoch": 167.35,
+      "learning_rate": 4.6258064516129035e-05,
+      "loss": 0.2454,
+      "step": 25940
+    },
+    {
+      "epoch": 167.42,
+      "learning_rate": 4.625448028673835e-05,
+      "loss": 0.275,
+      "step": 25950
+    },
+    {
+      "epoch": 167.48,
+      "learning_rate": 4.625089605734767e-05,
+      "loss": 0.3397,
+      "step": 25960
+    },
+    {
+      "epoch": 167.55,
+      "learning_rate": 4.624731182795699e-05,
+      "loss": 0.3253,
+      "step": 25970
+    },
+    {
+      "epoch": 167.61,
+      "learning_rate": 4.624372759856631e-05,
+      "loss": 0.362,
+      "step": 25980
+    },
+    {
+      "epoch": 167.68,
+      "learning_rate": 4.624014336917563e-05,
+      "loss": 0.3326,
+      "step": 25990
+    },
+    {
+      "epoch": 167.74,
+      "learning_rate": 4.6236559139784944e-05,
+      "loss": 0.3057,
+      "step": 26000
+    },
+    {
+      "epoch": 167.8,
+      "learning_rate": 4.623297491039427e-05,
+      "loss": 0.3267,
+      "step": 26010
+    },
+    {
+      "epoch": 167.87,
+      "learning_rate": 4.622939068100359e-05,
+      "loss": 0.3185,
+      "step": 26020
+    },
+    {
+      "epoch": 167.93,
+      "learning_rate": 4.6225806451612906e-05,
+      "loss": 0.3613,
+      "step": 26030
+    },
+    {
+      "epoch": 168.0,
+      "learning_rate": 4.6222222222222224e-05,
+      "loss": 0.2853,
+      "step": 26040
+    },
+    {
+      "epoch": 168.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0287158489227295,
+      "eval_runtime": 0.8759,
+      "eval_samples_per_second": 229.471,
+      "eval_steps_per_second": 7.992,
+      "step": 26040
+    },
+    {
+      "epoch": 168.06,
+      "learning_rate": 4.621863799283154e-05,
+      "loss": 0.3642,
+      "step": 26050
+    },
+    {
+      "epoch": 168.13,
+      "learning_rate": 4.621505376344086e-05,
+      "loss": 0.3494,
+      "step": 26060
+    },
+    {
+      "epoch": 168.19,
+      "learning_rate": 4.6211469534050185e-05,
+      "loss": 0.3019,
+      "step": 26070
+    },
+    {
+      "epoch": 168.26,
+      "learning_rate": 4.62078853046595e-05,
+      "loss": 0.3001,
+      "step": 26080
+    },
+    {
+      "epoch": 168.32,
+      "learning_rate": 4.620430107526882e-05,
+      "loss": 0.3121,
+      "step": 26090
+    },
+    {
+      "epoch": 168.39,
+      "learning_rate": 4.620071684587814e-05,
+      "loss": 0.3325,
+      "step": 26100
+    },
+    {
+      "epoch": 168.45,
+      "learning_rate": 4.619713261648746e-05,
+      "loss": 0.3141,
+      "step": 26110
+    },
+    {
+      "epoch": 168.51,
+      "learning_rate": 4.6193548387096776e-05,
+      "loss": 0.269,
+      "step": 26120
+    },
+    {
+      "epoch": 168.58,
+      "learning_rate": 4.6189964157706095e-05,
+      "loss": 0.3629,
+      "step": 26130
+    },
+    {
+      "epoch": 168.64,
+      "learning_rate": 4.618637992831541e-05,
+      "loss": 0.3731,
+      "step": 26140
+    },
+    {
+      "epoch": 168.71,
+      "learning_rate": 4.618279569892474e-05,
+      "loss": 0.3654,
+      "step": 26150
+    },
+    {
+      "epoch": 168.77,
+      "learning_rate": 4.617921146953405e-05,
+      "loss": 0.296,
+      "step": 26160
+    },
+    {
+      "epoch": 168.84,
+      "learning_rate": 4.6175627240143374e-05,
+      "loss": 0.3522,
+      "step": 26170
+    },
+    {
+      "epoch": 168.9,
+      "learning_rate": 4.617204301075269e-05,
+      "loss": 0.3087,
+      "step": 26180
+    },
+    {
+      "epoch": 168.96,
+      "learning_rate": 4.616845878136201e-05,
+      "loss": 0.3606,
+      "step": 26190
+    },
+    {
+      "epoch": 169.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.039176344871521,
+      "eval_runtime": 0.8584,
+      "eval_samples_per_second": 234.165,
+      "eval_steps_per_second": 8.155,
+      "step": 26195
+    },
+    {
+      "epoch": 169.03,
+      "learning_rate": 4.616487455197133e-05,
+      "loss": 0.323,
+      "step": 26200
+    },
+    {
+      "epoch": 169.1,
+      "learning_rate": 4.616129032258065e-05,
+      "loss": 0.3085,
+      "step": 26210
+    },
+    {
+      "epoch": 169.16,
+      "learning_rate": 4.6157706093189965e-05,
+      "loss": 0.2992,
+      "step": 26220
+    },
+    {
+      "epoch": 169.23,
+      "learning_rate": 4.615412186379929e-05,
+      "loss": 0.3251,
+      "step": 26230
+    },
+    {
+      "epoch": 169.29,
+      "learning_rate": 4.61505376344086e-05,
+      "loss": 0.3567,
+      "step": 26240
+    },
+    {
+      "epoch": 169.35,
+      "learning_rate": 4.614695340501792e-05,
+      "loss": 0.2895,
+      "step": 26250
+    },
+    {
+      "epoch": 169.42,
+      "learning_rate": 4.6143369175627245e-05,
+      "loss": 0.3264,
+      "step": 26260
+    },
+    {
+      "epoch": 169.48,
+      "learning_rate": 4.613978494623656e-05,
+      "loss": 0.3343,
+      "step": 26270
+    },
+    {
+      "epoch": 169.55,
+      "learning_rate": 4.613620071684588e-05,
+      "loss": 0.3118,
+      "step": 26280
+    },
+    {
+      "epoch": 169.61,
+      "learning_rate": 4.61326164874552e-05,
+      "loss": 0.3483,
+      "step": 26290
+    },
+    {
+      "epoch": 169.68,
+      "learning_rate": 4.612903225806452e-05,
+      "loss": 0.313,
+      "step": 26300
+    },
+    {
+      "epoch": 169.74,
+      "learning_rate": 4.612544802867384e-05,
+      "loss": 0.3511,
+      "step": 26310
+    },
+    {
+      "epoch": 169.8,
+      "learning_rate": 4.6121863799283154e-05,
+      "loss": 0.3475,
+      "step": 26320
+    },
+    {
+      "epoch": 169.87,
+      "learning_rate": 4.611827956989247e-05,
+      "loss": 0.2829,
+      "step": 26330
+    },
+    {
+      "epoch": 169.93,
+      "learning_rate": 4.61146953405018e-05,
+      "loss": 0.3231,
+      "step": 26340
+    },
+    {
+      "epoch": 170.0,
+      "learning_rate": 4.6111111111111115e-05,
+      "loss": 0.3417,
+      "step": 26350
+    },
+    {
+      "epoch": 170.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0167739391326904,
+      "eval_runtime": 0.8534,
+      "eval_samples_per_second": 235.52,
+      "eval_steps_per_second": 8.202,
+      "step": 26350
+    },
+    {
+      "epoch": 170.06,
+      "learning_rate": 4.6107526881720434e-05,
+      "loss": 0.3489,
+      "step": 26360
+    },
+    {
+      "epoch": 170.13,
+      "learning_rate": 4.610394265232975e-05,
+      "loss": 0.4138,
+      "step": 26370
+    },
+    {
+      "epoch": 170.19,
+      "learning_rate": 4.610035842293907e-05,
+      "loss": 0.3353,
+      "step": 26380
+    },
+    {
+      "epoch": 170.26,
+      "learning_rate": 4.609677419354839e-05,
+      "loss": 0.3341,
+      "step": 26390
+    },
+    {
+      "epoch": 170.32,
+      "learning_rate": 4.6093189964157706e-05,
+      "loss": 0.3342,
+      "step": 26400
+    },
+    {
+      "epoch": 170.39,
+      "learning_rate": 4.6089605734767025e-05,
+      "loss": 0.3236,
+      "step": 26410
+    },
+    {
+      "epoch": 170.45,
+      "learning_rate": 4.608602150537635e-05,
+      "loss": 0.3243,
+      "step": 26420
+    },
+    {
+      "epoch": 170.51,
+      "learning_rate": 4.608243727598566e-05,
+      "loss": 0.3177,
+      "step": 26430
+    },
+    {
+      "epoch": 170.58,
+      "learning_rate": 4.6078853046594986e-05,
+      "loss": 0.3269,
+      "step": 26440
+    },
+    {
+      "epoch": 170.64,
+      "learning_rate": 4.6075268817204304e-05,
+      "loss": 0.3432,
+      "step": 26450
+    },
+    {
+      "epoch": 170.71,
+      "learning_rate": 4.607168458781362e-05,
+      "loss": 0.3005,
+      "step": 26460
+    },
+    {
+      "epoch": 170.77,
+      "learning_rate": 4.606810035842294e-05,
+      "loss": 0.3145,
+      "step": 26470
+    },
+    {
+      "epoch": 170.84,
+      "learning_rate": 4.606451612903226e-05,
+      "loss": 0.2783,
+      "step": 26480
+    },
+    {
+      "epoch": 170.9,
+      "learning_rate": 4.606093189964158e-05,
+      "loss": 0.3948,
+      "step": 26490
+    },
+    {
+      "epoch": 170.96,
+      "learning_rate": 4.60573476702509e-05,
+      "loss": 0.327,
+      "step": 26500
+    },
+    {
+      "epoch": 171.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.09117591381073,
+      "eval_runtime": 0.8809,
+      "eval_samples_per_second": 228.163,
+      "eval_steps_per_second": 7.946,
+      "step": 26505
+    },
+    {
+      "epoch": 171.03,
+      "learning_rate": 4.6053763440860213e-05,
+      "loss": 0.4166,
+      "step": 26510
+    },
+    {
+      "epoch": 171.1,
+      "learning_rate": 4.605017921146954e-05,
+      "loss": 0.3045,
+      "step": 26520
+    },
+    {
+      "epoch": 171.16,
+      "learning_rate": 4.6046594982078857e-05,
+      "loss": 0.3742,
+      "step": 26530
+    },
+    {
+      "epoch": 171.23,
+      "learning_rate": 4.6043010752688175e-05,
+      "loss": 0.3572,
+      "step": 26540
+    },
+    {
+      "epoch": 171.29,
+      "learning_rate": 4.603942652329749e-05,
+      "loss": 0.3055,
+      "step": 26550
+    },
+    {
+      "epoch": 171.35,
+      "learning_rate": 4.603584229390681e-05,
+      "loss": 0.2934,
+      "step": 26560
+    },
+    {
+      "epoch": 171.42,
+      "learning_rate": 4.603225806451613e-05,
+      "loss": 0.3257,
+      "step": 26570
+    },
+    {
+      "epoch": 171.48,
+      "learning_rate": 4.6028673835125454e-05,
+      "loss": 0.2717,
+      "step": 26580
+    },
+    {
+      "epoch": 171.55,
+      "learning_rate": 4.6025089605734766e-05,
+      "loss": 0.337,
+      "step": 26590
+    },
+    {
+      "epoch": 171.61,
+      "learning_rate": 4.6021505376344084e-05,
+      "loss": 0.3011,
+      "step": 26600
+    },
+    {
+      "epoch": 171.68,
+      "learning_rate": 4.601792114695341e-05,
+      "loss": 0.3261,
+      "step": 26610
+    },
+    {
+      "epoch": 171.74,
+      "learning_rate": 4.601433691756273e-05,
+      "loss": 0.3218,
+      "step": 26620
+    },
+    {
+      "epoch": 171.8,
+      "learning_rate": 4.6010752688172045e-05,
+      "loss": 0.3012,
+      "step": 26630
+    },
+    {
+      "epoch": 171.87,
+      "learning_rate": 4.6007168458781364e-05,
+      "loss": 0.3159,
+      "step": 26640
+    },
+    {
+      "epoch": 171.93,
+      "learning_rate": 4.600358422939068e-05,
+      "loss": 0.3013,
+      "step": 26650
+    },
+    {
+      "epoch": 172.0,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.3562,
+      "step": 26660
+    },
+    {
+      "epoch": 172.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0509366989135742,
+      "eval_runtime": 0.8441,
+      "eval_samples_per_second": 238.114,
+      "eval_steps_per_second": 8.293,
+      "step": 26660
+    },
+    {
+      "epoch": 172.06,
+      "learning_rate": 4.599641577060932e-05,
+      "loss": 0.3242,
+      "step": 26670
+    },
+    {
+      "epoch": 172.13,
+      "learning_rate": 4.5992831541218636e-05,
+      "loss": 0.3672,
+      "step": 26680
+    },
+    {
+      "epoch": 172.19,
+      "learning_rate": 4.598924731182796e-05,
+      "loss": 0.3327,
+      "step": 26690
+    },
+    {
+      "epoch": 172.26,
+      "learning_rate": 4.598566308243728e-05,
+      "loss": 0.365,
+      "step": 26700
+    },
+    {
+      "epoch": 172.32,
+      "learning_rate": 4.59820788530466e-05,
+      "loss": 0.3315,
+      "step": 26710
+    },
+    {
+      "epoch": 172.39,
+      "learning_rate": 4.5978494623655916e-05,
+      "loss": 0.3175,
+      "step": 26720
+    },
+    {
+      "epoch": 172.45,
+      "learning_rate": 4.5974910394265234e-05,
+      "loss": 0.2895,
+      "step": 26730
+    },
+    {
+      "epoch": 172.51,
+      "learning_rate": 4.597132616487456e-05,
+      "loss": 0.3079,
+      "step": 26740
+    },
+    {
+      "epoch": 172.58,
+      "learning_rate": 4.596774193548387e-05,
+      "loss": 0.3184,
+      "step": 26750
+    },
+    {
+      "epoch": 172.64,
+      "learning_rate": 4.596415770609319e-05,
+      "loss": 0.3482,
+      "step": 26760
+    },
+    {
+      "epoch": 172.71,
+      "learning_rate": 4.5960573476702514e-05,
+      "loss": 0.3089,
+      "step": 26770
+    },
+    {
+      "epoch": 172.77,
+      "learning_rate": 4.595698924731183e-05,
+      "loss": 0.3358,
+      "step": 26780
+    },
+    {
+      "epoch": 172.84,
+      "learning_rate": 4.595340501792115e-05,
+      "loss": 0.2959,
+      "step": 26790
+    },
+    {
+      "epoch": 172.9,
+      "learning_rate": 4.594982078853047e-05,
+      "loss": 0.3801,
+      "step": 26800
+    },
+    {
+      "epoch": 172.96,
+      "learning_rate": 4.594623655913979e-05,
+      "loss": 0.3598,
+      "step": 26810
+    },
+    {
+      "epoch": 173.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0551728010177612,
+      "eval_runtime": 0.8879,
+      "eval_samples_per_second": 226.377,
+      "eval_steps_per_second": 7.884,
+      "step": 26815
+    },
+    {
+      "epoch": 173.03,
+      "learning_rate": 4.594265232974911e-05,
+      "loss": 0.3674,
+      "step": 26820
+    },
+    {
+      "epoch": 173.1,
+      "learning_rate": 4.593906810035842e-05,
+      "loss": 0.2952,
+      "step": 26830
+    },
+    {
+      "epoch": 173.16,
+      "learning_rate": 4.593548387096774e-05,
+      "loss": 0.2701,
+      "step": 26840
+    },
+    {
+      "epoch": 173.23,
+      "learning_rate": 4.5931899641577066e-05,
+      "loss": 0.349,
+      "step": 26850
+    },
+    {
+      "epoch": 173.29,
+      "learning_rate": 4.5928315412186384e-05,
+      "loss": 0.3254,
+      "step": 26860
+    },
+    {
+      "epoch": 173.35,
+      "learning_rate": 4.59247311827957e-05,
+      "loss": 0.3505,
+      "step": 26870
+    },
+    {
+      "epoch": 173.42,
+      "learning_rate": 4.592114695340502e-05,
+      "loss": 0.3527,
+      "step": 26880
+    },
+    {
+      "epoch": 173.48,
+      "learning_rate": 4.591756272401434e-05,
+      "loss": 0.3235,
+      "step": 26890
+    },
+    {
+      "epoch": 173.55,
+      "learning_rate": 4.591397849462366e-05,
+      "loss": 0.3308,
+      "step": 26900
+    },
+    {
+      "epoch": 173.61,
+      "learning_rate": 4.5910394265232975e-05,
+      "loss": 0.3454,
+      "step": 26910
+    },
+    {
+      "epoch": 173.68,
+      "learning_rate": 4.5906810035842294e-05,
+      "loss": 0.29,
+      "step": 26920
+    },
+    {
+      "epoch": 173.74,
+      "learning_rate": 4.590322580645162e-05,
+      "loss": 0.3313,
+      "step": 26930
+    },
+    {
+      "epoch": 173.8,
+      "learning_rate": 4.589964157706094e-05,
+      "loss": 0.3725,
+      "step": 26940
+    },
+    {
+      "epoch": 173.87,
+      "learning_rate": 4.589605734767025e-05,
+      "loss": 0.3303,
+      "step": 26950
+    },
+    {
+      "epoch": 173.93,
+      "learning_rate": 4.589247311827957e-05,
+      "loss": 0.341,
+      "step": 26960
+    },
+    {
+      "epoch": 174.0,
+      "learning_rate": 4.588888888888889e-05,
+      "loss": 0.2555,
+      "step": 26970
+    },
+    {
+      "epoch": 174.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0453872680664062,
+      "eval_runtime": 0.8481,
+      "eval_samples_per_second": 236.992,
+      "eval_steps_per_second": 8.253,
+      "step": 26970
+    },
+    {
+      "epoch": 174.06,
+      "learning_rate": 4.588530465949821e-05,
+      "loss": 0.2582,
+      "step": 26980
+    },
+    {
+      "epoch": 174.13,
+      "learning_rate": 4.588172043010753e-05,
+      "loss": 0.3449,
+      "step": 26990
+    },
+    {
+      "epoch": 174.19,
+      "learning_rate": 4.5878136200716846e-05,
+      "loss": 0.2914,
+      "step": 27000
+    },
+    {
+      "epoch": 174.26,
+      "learning_rate": 4.587455197132617e-05,
+      "loss": 0.3457,
+      "step": 27010
+    },
+    {
+      "epoch": 174.32,
+      "learning_rate": 4.587096774193548e-05,
+      "loss": 0.3484,
+      "step": 27020
+    },
+    {
+      "epoch": 174.39,
+      "learning_rate": 4.58673835125448e-05,
+      "loss": 0.2964,
+      "step": 27030
+    },
+    {
+      "epoch": 174.45,
+      "learning_rate": 4.5863799283154126e-05,
+      "loss": 0.31,
+      "step": 27040
+    },
+    {
+      "epoch": 174.51,
+      "learning_rate": 4.5860215053763444e-05,
+      "loss": 0.3009,
+      "step": 27050
+    },
+    {
+      "epoch": 174.58,
+      "learning_rate": 4.585663082437276e-05,
+      "loss": 0.3062,
+      "step": 27060
+    },
+    {
+      "epoch": 174.64,
+      "learning_rate": 4.585304659498208e-05,
+      "loss": 0.3677,
+      "step": 27070
+    },
+    {
+      "epoch": 174.71,
+      "learning_rate": 4.58494623655914e-05,
+      "loss": 0.3088,
+      "step": 27080
+    },
+    {
+      "epoch": 174.77,
+      "learning_rate": 4.5845878136200724e-05,
+      "loss": 0.3417,
+      "step": 27090
+    },
+    {
+      "epoch": 174.84,
+      "learning_rate": 4.5842293906810035e-05,
+      "loss": 0.2972,
+      "step": 27100
+    },
+    {
+      "epoch": 174.9,
+      "learning_rate": 4.583870967741935e-05,
+      "loss": 0.3844,
+      "step": 27110
+    },
+    {
+      "epoch": 174.96,
+      "learning_rate": 4.583512544802868e-05,
+      "loss": 0.3139,
+      "step": 27120
+    },
+    {
+      "epoch": 175.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0590002536773682,
+      "eval_runtime": 0.8904,
+      "eval_samples_per_second": 225.751,
+      "eval_steps_per_second": 7.862,
+      "step": 27125
+    },
+    {
+      "epoch": 175.03,
+      "learning_rate": 4.5831541218637996e-05,
+      "loss": 0.363,
+      "step": 27130
+    },
+    {
+      "epoch": 175.1,
+      "learning_rate": 4.5827956989247315e-05,
+      "loss": 0.3564,
+      "step": 27140
+    },
+    {
+      "epoch": 175.16,
+      "learning_rate": 4.582437275985663e-05,
+      "loss": 0.2948,
+      "step": 27150
+    },
+    {
+      "epoch": 175.23,
+      "learning_rate": 4.582078853046595e-05,
+      "loss": 0.3893,
+      "step": 27160
+    },
+    {
+      "epoch": 175.29,
+      "learning_rate": 4.5817204301075276e-05,
+      "loss": 0.367,
+      "step": 27170
+    },
+    {
+      "epoch": 175.35,
+      "learning_rate": 4.581362007168459e-05,
+      "loss": 0.3728,
+      "step": 27180
+    },
+    {
+      "epoch": 175.42,
+      "learning_rate": 4.5810035842293906e-05,
+      "loss": 0.3284,
+      "step": 27190
+    },
+    {
+      "epoch": 175.48,
+      "learning_rate": 4.580645161290323e-05,
+      "loss": 0.3002,
+      "step": 27200
+    },
+    {
+      "epoch": 175.55,
+      "learning_rate": 4.580286738351255e-05,
+      "loss": 0.3837,
+      "step": 27210
+    },
+    {
+      "epoch": 175.61,
+      "learning_rate": 4.579928315412187e-05,
+      "loss": 0.3564,
+      "step": 27220
+    },
+    {
+      "epoch": 175.68,
+      "learning_rate": 4.5795698924731185e-05,
+      "loss": 0.3349,
+      "step": 27230
+    },
+    {
+      "epoch": 175.74,
+      "learning_rate": 4.57921146953405e-05,
+      "loss": 0.2867,
+      "step": 27240
+    },
+    {
+      "epoch": 175.8,
+      "learning_rate": 4.578853046594982e-05,
+      "loss": 0.2637,
+      "step": 27250
+    },
+    {
+      "epoch": 175.87,
+      "learning_rate": 4.578494623655914e-05,
+      "loss": 0.3181,
+      "step": 27260
+    },
+    {
+      "epoch": 175.93,
+      "learning_rate": 4.578136200716846e-05,
+      "loss": 0.3721,
+      "step": 27270
+    },
+    {
+      "epoch": 176.0,
+      "learning_rate": 4.577777777777778e-05,
+      "loss": 0.3172,
+      "step": 27280
+    },
+    {
+      "epoch": 176.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.025319218635559,
+      "eval_runtime": 0.8877,
+      "eval_samples_per_second": 226.432,
+      "eval_steps_per_second": 7.886,
+      "step": 27280
+    },
+    {
+      "epoch": 176.06,
+      "learning_rate": 4.57741935483871e-05,
+      "loss": 0.3443,
+      "step": 27290
+    },
+    {
+      "epoch": 176.13,
+      "learning_rate": 4.577060931899641e-05,
+      "loss": 0.3427,
+      "step": 27300
+    },
+    {
+      "epoch": 176.19,
+      "learning_rate": 4.576702508960574e-05,
+      "loss": 0.2981,
+      "step": 27310
+    },
+    {
+      "epoch": 176.26,
+      "learning_rate": 4.5763440860215056e-05,
+      "loss": 0.3581,
+      "step": 27320
+    },
+    {
+      "epoch": 176.32,
+      "learning_rate": 4.5759856630824374e-05,
+      "loss": 0.2984,
+      "step": 27330
+    },
+    {
+      "epoch": 176.39,
+      "learning_rate": 4.575627240143369e-05,
+      "loss": 0.2851,
+      "step": 27340
+    },
+    {
+      "epoch": 176.45,
+      "learning_rate": 4.575268817204301e-05,
+      "loss": 0.337,
+      "step": 27350
+    },
+    {
+      "epoch": 176.51,
+      "learning_rate": 4.5749103942652335e-05,
+      "loss": 0.276,
+      "step": 27360
+    },
+    {
+      "epoch": 176.58,
+      "learning_rate": 4.5745519713261654e-05,
+      "loss": 0.3122,
+      "step": 27370
+    },
+    {
+      "epoch": 176.64,
+      "learning_rate": 4.5741935483870965e-05,
+      "loss": 0.3601,
+      "step": 27380
+    },
+    {
+      "epoch": 176.71,
+      "learning_rate": 4.573835125448029e-05,
+      "loss": 0.306,
+      "step": 27390
+    },
+    {
+      "epoch": 176.77,
+      "learning_rate": 4.573476702508961e-05,
+      "loss": 0.299,
+      "step": 27400
+    },
+    {
+      "epoch": 176.84,
+      "learning_rate": 4.5731182795698926e-05,
+      "loss": 0.3549,
+      "step": 27410
+    },
+    {
+      "epoch": 176.9,
+      "learning_rate": 4.5727598566308245e-05,
+      "loss": 0.2886,
+      "step": 27420
+    },
+    {
+      "epoch": 176.96,
+      "learning_rate": 4.572401433691756e-05,
+      "loss": 0.3057,
+      "step": 27430
+    },
+    {
+      "epoch": 177.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.068280577659607,
+      "eval_runtime": 0.8575,
+      "eval_samples_per_second": 234.403,
+      "eval_steps_per_second": 8.163,
+      "step": 27435
+    },
+    {
+      "epoch": 177.03,
+      "learning_rate": 4.572043010752689e-05,
+      "loss": 0.2968,
+      "step": 27440
+    },
+    {
+      "epoch": 177.1,
+      "learning_rate": 4.5716845878136206e-05,
+      "loss": 0.3315,
+      "step": 27450
+    },
+    {
+      "epoch": 177.16,
+      "learning_rate": 4.571326164874552e-05,
+      "loss": 0.3637,
+      "step": 27460
+    },
+    {
+      "epoch": 177.23,
+      "learning_rate": 4.570967741935484e-05,
+      "loss": 0.3006,
+      "step": 27470
+    },
+    {
+      "epoch": 177.29,
+      "learning_rate": 4.570609318996416e-05,
+      "loss": 0.2876,
+      "step": 27480
+    },
+    {
+      "epoch": 177.35,
+      "learning_rate": 4.570250896057348e-05,
+      "loss": 0.2955,
+      "step": 27490
+    },
+    {
+      "epoch": 177.42,
+      "learning_rate": 4.56989247311828e-05,
+      "loss": 0.3137,
+      "step": 27500
+    },
+    {
+      "epoch": 177.48,
+      "learning_rate": 4.5695340501792115e-05,
+      "loss": 0.3232,
+      "step": 27510
+    },
+    {
+      "epoch": 177.55,
+      "learning_rate": 4.569175627240144e-05,
+      "loss": 0.3289,
+      "step": 27520
+    },
+    {
+      "epoch": 177.61,
+      "learning_rate": 4.568817204301076e-05,
+      "loss": 0.2948,
+      "step": 27530
+    },
+    {
+      "epoch": 177.68,
+      "learning_rate": 4.568458781362007e-05,
+      "loss": 0.3163,
+      "step": 27540
+    },
+    {
+      "epoch": 177.74,
+      "learning_rate": 4.5681003584229395e-05,
+      "loss": 0.2686,
+      "step": 27550
+    },
+    {
+      "epoch": 177.8,
+      "learning_rate": 4.567741935483871e-05,
+      "loss": 0.3167,
+      "step": 27560
+    },
+    {
+      "epoch": 177.87,
+      "learning_rate": 4.567383512544803e-05,
+      "loss": 0.3583,
+      "step": 27570
+    },
+    {
+      "epoch": 177.93,
+      "learning_rate": 4.567025089605735e-05,
+      "loss": 0.3107,
+      "step": 27580
+    },
+    {
+      "epoch": 178.0,
+      "learning_rate": 4.566666666666667e-05,
+      "loss": 0.3709,
+      "step": 27590
+    },
+    {
+      "epoch": 178.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.109498381614685,
+      "eval_runtime": 0.852,
+      "eval_samples_per_second": 235.914,
+      "eval_steps_per_second": 8.216,
+      "step": 27590
+    },
+    {
+      "epoch": 178.06,
+      "learning_rate": 4.5663082437275986e-05,
+      "loss": 0.3718,
+      "step": 27600
+    },
+    {
+      "epoch": 178.13,
+      "learning_rate": 4.565949820788531e-05,
+      "loss": 0.3069,
+      "step": 27610
+    },
+    {
+      "epoch": 178.19,
+      "learning_rate": 4.565591397849462e-05,
+      "loss": 0.3199,
+      "step": 27620
+    },
+    {
+      "epoch": 178.26,
+      "learning_rate": 4.565232974910395e-05,
+      "loss": 0.3033,
+      "step": 27630
+    },
+    {
+      "epoch": 178.32,
+      "learning_rate": 4.5648745519713265e-05,
+      "loss": 0.3328,
+      "step": 27640
+    },
+    {
+      "epoch": 178.39,
+      "learning_rate": 4.5645161290322584e-05,
+      "loss": 0.3287,
+      "step": 27650
+    },
+    {
+      "epoch": 178.45,
+      "learning_rate": 4.56415770609319e-05,
+      "loss": 0.3321,
+      "step": 27660
+    },
+    {
+      "epoch": 178.51,
+      "learning_rate": 4.563799283154122e-05,
+      "loss": 0.4062,
+      "step": 27670
+    },
+    {
+      "epoch": 178.58,
+      "learning_rate": 4.563440860215054e-05,
+      "loss": 0.3003,
+      "step": 27680
+    },
+    {
+      "epoch": 178.64,
+      "learning_rate": 4.5630824372759856e-05,
+      "loss": 0.3843,
+      "step": 27690
+    },
+    {
+      "epoch": 178.71,
+      "learning_rate": 4.5627240143369175e-05,
+      "loss": 0.3253,
+      "step": 27700
+    },
+    {
+      "epoch": 178.77,
+      "learning_rate": 4.56236559139785e-05,
+      "loss": 0.2888,
+      "step": 27710
+    },
+    {
+      "epoch": 178.84,
+      "learning_rate": 4.562007168458782e-05,
+      "loss": 0.2826,
+      "step": 27720
+    },
+    {
+      "epoch": 178.9,
+      "learning_rate": 4.561648745519713e-05,
+      "loss": 0.309,
+      "step": 27730
+    },
+    {
+      "epoch": 178.96,
+      "learning_rate": 4.5612903225806454e-05,
+      "loss": 0.2748,
+      "step": 27740
+    },
+    {
+      "epoch": 179.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0317072868347168,
+      "eval_runtime": 0.8662,
+      "eval_samples_per_second": 232.052,
+      "eval_steps_per_second": 8.081,
+      "step": 27745
+    },
+    {
+      "epoch": 179.03,
+      "learning_rate": 4.560931899641577e-05,
+      "loss": 0.3668,
+      "step": 27750
+    },
+    {
+      "epoch": 179.1,
+      "learning_rate": 4.560573476702509e-05,
+      "loss": 0.3135,
+      "step": 27760
+    },
+    {
+      "epoch": 179.16,
+      "learning_rate": 4.560215053763441e-05,
+      "loss": 0.3622,
+      "step": 27770
+    },
+    {
+      "epoch": 179.23,
+      "learning_rate": 4.559856630824373e-05,
+      "loss": 0.2959,
+      "step": 27780
+    },
+    {
+      "epoch": 179.29,
+      "learning_rate": 4.559498207885305e-05,
+      "loss": 0.3051,
+      "step": 27790
+    },
+    {
+      "epoch": 179.35,
+      "learning_rate": 4.559139784946237e-05,
+      "loss": 0.338,
+      "step": 27800
+    },
+    {
+      "epoch": 179.42,
+      "learning_rate": 4.558781362007168e-05,
+      "loss": 0.3303,
+      "step": 27810
+    },
+    {
+      "epoch": 179.48,
+      "learning_rate": 4.558422939068101e-05,
+      "loss": 0.3019,
+      "step": 27820
+    },
+    {
+      "epoch": 179.55,
+      "learning_rate": 4.5580645161290325e-05,
+      "loss": 0.3344,
+      "step": 27830
+    },
+    {
+      "epoch": 179.61,
+      "learning_rate": 4.557706093189964e-05,
+      "loss": 0.2882,
+      "step": 27840
+    },
+    {
+      "epoch": 179.68,
+      "learning_rate": 4.557347670250896e-05,
+      "loss": 0.2932,
+      "step": 27850
+    },
+    {
+      "epoch": 179.74,
+      "learning_rate": 4.556989247311828e-05,
+      "loss": 0.3058,
+      "step": 27860
+    },
+    {
+      "epoch": 179.8,
+      "learning_rate": 4.5566308243727604e-05,
+      "loss": 0.2934,
+      "step": 27870
+    },
+    {
+      "epoch": 179.87,
+      "learning_rate": 4.556272401433692e-05,
+      "loss": 0.3269,
+      "step": 27880
+    },
+    {
+      "epoch": 179.93,
+      "learning_rate": 4.5559139784946234e-05,
+      "loss": 0.2993,
+      "step": 27890
+    },
+    {
+      "epoch": 180.0,
+      "learning_rate": 4.555555555555556e-05,
+      "loss": 0.3186,
+      "step": 27900
+    },
+    {
+      "epoch": 180.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0359281301498413,
+      "eval_runtime": 0.8539,
+      "eval_samples_per_second": 235.379,
+      "eval_steps_per_second": 8.197,
+      "step": 27900
+    },
+    {
+      "epoch": 180.06,
+      "learning_rate": 4.555197132616488e-05,
+      "loss": 0.3539,
+      "step": 27910
+    },
+    {
+      "epoch": 180.13,
+      "learning_rate": 4.5548387096774196e-05,
+      "loss": 0.2875,
+      "step": 27920
+    },
+    {
+      "epoch": 180.19,
+      "learning_rate": 4.5544802867383514e-05,
+      "loss": 0.2908,
+      "step": 27930
+    },
+    {
+      "epoch": 180.26,
+      "learning_rate": 4.554121863799283e-05,
+      "loss": 0.3345,
+      "step": 27940
+    },
+    {
+      "epoch": 180.32,
+      "learning_rate": 4.553763440860215e-05,
+      "loss": 0.3136,
+      "step": 27950
+    },
+    {
+      "epoch": 180.39,
+      "learning_rate": 4.5534050179211475e-05,
+      "loss": 0.2961,
+      "step": 27960
+    },
+    {
+      "epoch": 180.45,
+      "learning_rate": 4.5530465949820787e-05,
+      "loss": 0.3222,
+      "step": 27970
+    },
+    {
+      "epoch": 180.51,
+      "learning_rate": 4.552688172043011e-05,
+      "loss": 0.3581,
+      "step": 27980
+    },
+    {
+      "epoch": 180.58,
+      "learning_rate": 4.552329749103943e-05,
+      "loss": 0.287,
+      "step": 27990
+    },
+    {
+      "epoch": 180.64,
+      "learning_rate": 4.551971326164875e-05,
+      "loss": 0.2973,
+      "step": 28000
+    },
+    {
+      "epoch": 180.71,
+      "learning_rate": 4.5516129032258066e-05,
+      "loss": 0.3264,
+      "step": 28010
+    },
+    {
+      "epoch": 180.77,
+      "learning_rate": 4.5512544802867384e-05,
+      "loss": 0.2975,
+      "step": 28020
+    },
+    {
+      "epoch": 180.84,
+      "learning_rate": 4.55089605734767e-05,
+      "loss": 0.2887,
+      "step": 28030
+    },
+    {
+      "epoch": 180.9,
+      "learning_rate": 4.550537634408603e-05,
+      "loss": 0.3141,
+      "step": 28040
+    },
+    {
+      "epoch": 180.96,
+      "learning_rate": 4.550179211469534e-05,
+      "loss": 0.3178,
+      "step": 28050
+    },
+    {
+      "epoch": 181.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9878270626068115,
+      "eval_runtime": 0.8528,
+      "eval_samples_per_second": 235.697,
+      "eval_steps_per_second": 8.208,
+      "step": 28055
+    },
+    {
+      "epoch": 181.03,
+      "learning_rate": 4.5498207885304664e-05,
+      "loss": 0.3839,
+      "step": 28060
+    },
+    {
+      "epoch": 181.1,
+      "learning_rate": 4.549462365591398e-05,
+      "loss": 0.3253,
+      "step": 28070
+    },
+    {
+      "epoch": 181.16,
+      "learning_rate": 4.54910394265233e-05,
+      "loss": 0.298,
+      "step": 28080
+    },
+    {
+      "epoch": 181.23,
+      "learning_rate": 4.548745519713262e-05,
+      "loss": 0.3209,
+      "step": 28090
+    },
+    {
+      "epoch": 181.29,
+      "learning_rate": 4.548387096774194e-05,
+      "loss": 0.3713,
+      "step": 28100
+    },
+    {
+      "epoch": 181.35,
+      "learning_rate": 4.5480286738351255e-05,
+      "loss": 0.3435,
+      "step": 28110
+    },
+    {
+      "epoch": 181.42,
+      "learning_rate": 4.547670250896058e-05,
+      "loss": 0.3322,
+      "step": 28120
+    },
+    {
+      "epoch": 181.48,
+      "learning_rate": 4.547311827956989e-05,
+      "loss": 0.2818,
+      "step": 28130
+    },
+    {
+      "epoch": 181.55,
+      "learning_rate": 4.5469534050179216e-05,
+      "loss": 0.3557,
+      "step": 28140
+    },
+    {
+      "epoch": 181.61,
+      "learning_rate": 4.5465949820788535e-05,
+      "loss": 0.3098,
+      "step": 28150
+    },
+    {
+      "epoch": 181.68,
+      "learning_rate": 4.546236559139785e-05,
+      "loss": 0.3129,
+      "step": 28160
+    },
+    {
+      "epoch": 181.74,
+      "learning_rate": 4.545878136200717e-05,
+      "loss": 0.3206,
+      "step": 28170
+    },
+    {
+      "epoch": 181.8,
+      "learning_rate": 4.545519713261649e-05,
+      "loss": 0.3473,
+      "step": 28180
+    },
+    {
+      "epoch": 181.87,
+      "learning_rate": 4.545161290322581e-05,
+      "loss": 0.2721,
+      "step": 28190
+    },
+    {
+      "epoch": 181.93,
+      "learning_rate": 4.544802867383513e-05,
+      "loss": 0.307,
+      "step": 28200
+    },
+    {
+      "epoch": 182.0,
+      "learning_rate": 4.5444444444444444e-05,
+      "loss": 0.313,
+      "step": 28210
+    },
+    {
+      "epoch": 182.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9939073920249939,
+      "eval_runtime": 0.869,
+      "eval_samples_per_second": 231.313,
+      "eval_steps_per_second": 8.056,
+      "step": 28210
+    },
+    {
+      "epoch": 182.06,
+      "learning_rate": 4.544086021505377e-05,
+      "loss": 0.3532,
+      "step": 28220
+    },
+    {
+      "epoch": 182.13,
+      "learning_rate": 4.543727598566309e-05,
+      "loss": 0.3747,
+      "step": 28230
+    },
+    {
+      "epoch": 182.19,
+      "learning_rate": 4.5433691756272405e-05,
+      "loss": 0.3015,
+      "step": 28240
+    },
+    {
+      "epoch": 182.26,
+      "learning_rate": 4.543010752688172e-05,
+      "loss": 0.3879,
+      "step": 28250
+    },
+    {
+      "epoch": 182.32,
+      "learning_rate": 4.542652329749104e-05,
+      "loss": 0.318,
+      "step": 28260
+    },
+    {
+      "epoch": 182.39,
+      "learning_rate": 4.542293906810036e-05,
+      "loss": 0.292,
+      "step": 28270
+    },
+    {
+      "epoch": 182.45,
+      "learning_rate": 4.5419354838709685e-05,
+      "loss": 0.3103,
+      "step": 28280
+    },
+    {
+      "epoch": 182.51,
+      "learning_rate": 4.5415770609318996e-05,
+      "loss": 0.3585,
+      "step": 28290
+    },
+    {
+      "epoch": 182.58,
+      "learning_rate": 4.5412186379928314e-05,
+      "loss": 0.356,
+      "step": 28300
+    },
+    {
+      "epoch": 182.64,
+      "learning_rate": 4.540860215053764e-05,
+      "loss": 0.3239,
+      "step": 28310
+    },
+    {
+      "epoch": 182.71,
+      "learning_rate": 4.540501792114695e-05,
+      "loss": 0.2977,
+      "step": 28320
+    },
+    {
+      "epoch": 182.77,
+      "learning_rate": 4.5401433691756276e-05,
+      "loss": 0.3036,
+      "step": 28330
+    },
+    {
+      "epoch": 182.84,
+      "learning_rate": 4.5397849462365594e-05,
+      "loss": 0.3186,
+      "step": 28340
+    },
+    {
+      "epoch": 182.9,
+      "learning_rate": 4.539426523297491e-05,
+      "loss": 0.3046,
+      "step": 28350
+    },
+    {
+      "epoch": 182.96,
+      "learning_rate": 4.539068100358423e-05,
+      "loss": 0.3467,
+      "step": 28360
+    },
+    {
+      "epoch": 183.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.1165199279785156,
+      "eval_runtime": 0.8537,
+      "eval_samples_per_second": 235.435,
+      "eval_steps_per_second": 8.199,
+      "step": 28365
+    },
+    {
+      "epoch": 183.03,
+      "learning_rate": 4.538709677419355e-05,
+      "loss": 0.3327,
+      "step": 28370
+    },
+    {
+      "epoch": 183.1,
+      "learning_rate": 4.538351254480287e-05,
+      "loss": 0.318,
+      "step": 28380
+    },
+    {
+      "epoch": 183.16,
+      "learning_rate": 4.537992831541219e-05,
+      "loss": 0.3212,
+      "step": 28390
+    },
+    {
+      "epoch": 183.23,
+      "learning_rate": 4.53763440860215e-05,
+      "loss": 0.3574,
+      "step": 28400
+    },
+    {
+      "epoch": 183.29,
+      "learning_rate": 4.537275985663083e-05,
+      "loss": 0.2927,
+      "step": 28410
+    },
+    {
+      "epoch": 183.35,
+      "learning_rate": 4.5369175627240146e-05,
+      "loss": 0.3407,
+      "step": 28420
+    },
+    {
+      "epoch": 183.42,
+      "learning_rate": 4.5365591397849465e-05,
+      "loss": 0.3419,
+      "step": 28430
+    },
+    {
+      "epoch": 183.48,
+      "learning_rate": 4.536200716845878e-05,
+      "loss": 0.3859,
+      "step": 28440
+    },
+    {
+      "epoch": 183.55,
+      "learning_rate": 4.53584229390681e-05,
+      "loss": 0.2729,
+      "step": 28450
+    },
+    {
+      "epoch": 183.61,
+      "learning_rate": 4.535483870967742e-05,
+      "loss": 0.3199,
+      "step": 28460
+    },
+    {
+      "epoch": 183.68,
+      "learning_rate": 4.5351254480286744e-05,
+      "loss": 0.2782,
+      "step": 28470
+    },
+    {
+      "epoch": 183.74,
+      "learning_rate": 4.5347670250896056e-05,
+      "loss": 0.3019,
+      "step": 28480
+    },
+    {
+      "epoch": 183.8,
+      "learning_rate": 4.534408602150538e-05,
+      "loss": 0.2915,
+      "step": 28490
+    },
+    {
+      "epoch": 183.87,
+      "learning_rate": 4.53405017921147e-05,
+      "loss": 0.3005,
+      "step": 28500
+    },
+    {
+      "epoch": 183.93,
+      "learning_rate": 4.533691756272402e-05,
+      "loss": 0.3141,
+      "step": 28510
+    },
+    {
+      "epoch": 184.0,
+      "learning_rate": 4.5333333333333335e-05,
+      "loss": 0.347,
+      "step": 28520
+    },
+    {
+      "epoch": 184.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.056614637374878,
+      "eval_runtime": 0.8726,
+      "eval_samples_per_second": 230.342,
+      "eval_steps_per_second": 8.022,
+      "step": 28520
+    },
+    {
+      "epoch": 184.06,
+      "learning_rate": 4.5329749103942653e-05,
+      "loss": 0.3422,
+      "step": 28530
+    },
+    {
+      "epoch": 184.13,
+      "learning_rate": 4.532616487455197e-05,
+      "loss": 0.2948,
+      "step": 28540
+    },
+    {
+      "epoch": 184.19,
+      "learning_rate": 4.53225806451613e-05,
+      "loss": 0.2655,
+      "step": 28550
+    },
+    {
+      "epoch": 184.26,
+      "learning_rate": 4.531899641577061e-05,
+      "loss": 0.3719,
+      "step": 28560
+    },
+    {
+      "epoch": 184.32,
+      "learning_rate": 4.531541218637993e-05,
+      "loss": 0.31,
+      "step": 28570
+    },
+    {
+      "epoch": 184.39,
+      "learning_rate": 4.531182795698925e-05,
+      "loss": 0.2459,
+      "step": 28580
+    },
+    {
+      "epoch": 184.45,
+      "learning_rate": 4.530824372759857e-05,
+      "loss": 0.3645,
+      "step": 28590
+    },
+    {
+      "epoch": 184.51,
+      "learning_rate": 4.530465949820789e-05,
+      "loss": 0.2797,
+      "step": 28600
+    },
+    {
+      "epoch": 184.58,
+      "learning_rate": 4.5301075268817206e-05,
+      "loss": 0.3187,
+      "step": 28610
+    },
+    {
+      "epoch": 184.64,
+      "learning_rate": 4.5297491039426524e-05,
+      "loss": 0.329,
+      "step": 28620
+    },
+    {
+      "epoch": 184.71,
+      "learning_rate": 4.529390681003585e-05,
+      "loss": 0.3289,
+      "step": 28630
+    },
+    {
+      "epoch": 184.77,
+      "learning_rate": 4.529032258064516e-05,
+      "loss": 0.3354,
+      "step": 28640
+    },
+    {
+      "epoch": 184.84,
+      "learning_rate": 4.528673835125448e-05,
+      "loss": 0.2131,
+      "step": 28650
+    },
+    {
+      "epoch": 184.9,
+      "learning_rate": 4.5283154121863804e-05,
+      "loss": 0.3597,
+      "step": 28660
+    },
+    {
+      "epoch": 184.96,
+      "learning_rate": 4.527956989247312e-05,
+      "loss": 0.3031,
+      "step": 28670
+    },
+    {
+      "epoch": 185.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9843950271606445,
+      "eval_runtime": 0.8589,
+      "eval_samples_per_second": 234.017,
+      "eval_steps_per_second": 8.15,
+      "step": 28675
+    },
+    {
+      "epoch": 185.03,
+      "learning_rate": 4.527598566308244e-05,
+      "loss": 0.3265,
+      "step": 28680
+    },
+    {
+      "epoch": 185.1,
+      "learning_rate": 4.527240143369176e-05,
+      "loss": 0.3067,
+      "step": 28690
+    },
+    {
+      "epoch": 185.16,
+      "learning_rate": 4.5268817204301076e-05,
+      "loss": 0.3488,
+      "step": 28700
+    },
+    {
+      "epoch": 185.23,
+      "learning_rate": 4.52652329749104e-05,
+      "loss": 0.2794,
+      "step": 28710
+    },
+    {
+      "epoch": 185.29,
+      "learning_rate": 4.526164874551971e-05,
+      "loss": 0.2676,
+      "step": 28720
+    },
+    {
+      "epoch": 185.35,
+      "learning_rate": 4.525806451612903e-05,
+      "loss": 0.2813,
+      "step": 28730
+    },
+    {
+      "epoch": 185.42,
+      "learning_rate": 4.5254480286738356e-05,
+      "loss": 0.3371,
+      "step": 28740
+    },
+    {
+      "epoch": 185.48,
+      "learning_rate": 4.5250896057347674e-05,
+      "loss": 0.281,
+      "step": 28750
+    },
+    {
+      "epoch": 185.55,
+      "learning_rate": 4.524731182795699e-05,
+      "loss": 0.3087,
+      "step": 28760
+    },
+    {
+      "epoch": 185.61,
+      "learning_rate": 4.524372759856631e-05,
+      "loss": 0.2989,
+      "step": 28770
+    },
+    {
+      "epoch": 185.68,
+      "learning_rate": 4.524014336917563e-05,
+      "loss": 0.3072,
+      "step": 28780
+    },
+    {
+      "epoch": 185.74,
+      "learning_rate": 4.5236559139784954e-05,
+      "loss": 0.3132,
+      "step": 28790
+    },
+    {
+      "epoch": 185.8,
+      "learning_rate": 4.5232974910394265e-05,
+      "loss": 0.3306,
+      "step": 28800
+    },
+    {
+      "epoch": 185.87,
+      "learning_rate": 4.5229390681003584e-05,
+      "loss": 0.3191,
+      "step": 28810
+    },
+    {
+      "epoch": 185.93,
+      "learning_rate": 4.522580645161291e-05,
+      "loss": 0.2863,
+      "step": 28820
+    },
+    {
+      "epoch": 186.0,
+      "learning_rate": 4.522222222222223e-05,
+      "loss": 0.3644,
+      "step": 28830
+    },
+    {
+      "epoch": 186.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0274704694747925,
+      "eval_runtime": 0.8432,
+      "eval_samples_per_second": 238.366,
+      "eval_steps_per_second": 8.301,
+      "step": 28830
+    },
+    {
+      "epoch": 186.06,
+      "learning_rate": 4.5218637992831545e-05,
+      "loss": 0.2958,
+      "step": 28840
+    },
+    {
+      "epoch": 186.13,
+      "learning_rate": 4.521505376344086e-05,
+      "loss": 0.3382,
+      "step": 28850
+    },
+    {
+      "epoch": 186.19,
+      "learning_rate": 4.521146953405018e-05,
+      "loss": 0.3324,
+      "step": 28860
+    },
+    {
+      "epoch": 186.26,
+      "learning_rate": 4.5207885304659506e-05,
+      "loss": 0.2713,
+      "step": 28870
+    },
+    {
+      "epoch": 186.32,
+      "learning_rate": 4.520430107526882e-05,
+      "loss": 0.3726,
+      "step": 28880
+    },
+    {
+      "epoch": 186.39,
+      "learning_rate": 4.5200716845878136e-05,
+      "loss": 0.3552,
+      "step": 28890
+    },
+    {
+      "epoch": 186.45,
+      "learning_rate": 4.519713261648746e-05,
+      "loss": 0.3577,
+      "step": 28900
+    },
+    {
+      "epoch": 186.51,
+      "learning_rate": 4.519354838709678e-05,
+      "loss": 0.2936,
+      "step": 28910
+    },
+    {
+      "epoch": 186.58,
+      "learning_rate": 4.51899641577061e-05,
+      "loss": 0.3295,
+      "step": 28920
+    },
+    {
+      "epoch": 186.64,
+      "learning_rate": 4.5186379928315416e-05,
+      "loss": 0.352,
+      "step": 28930
+    },
+    {
+      "epoch": 186.71,
+      "learning_rate": 4.5182795698924734e-05,
+      "loss": 0.3125,
+      "step": 28940
+    },
+    {
+      "epoch": 186.77,
+      "learning_rate": 4.517921146953405e-05,
+      "loss": 0.2923,
+      "step": 28950
+    },
+    {
+      "epoch": 186.84,
+      "learning_rate": 4.517562724014337e-05,
+      "loss": 0.3365,
+      "step": 28960
+    },
+    {
+      "epoch": 186.9,
+      "learning_rate": 4.517204301075269e-05,
+      "loss": 0.2871,
+      "step": 28970
+    },
+    {
+      "epoch": 186.96,
+      "learning_rate": 4.516845878136201e-05,
+      "loss": 0.361,
+      "step": 28980
+    },
+    {
+      "epoch": 187.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9896905422210693,
+      "eval_runtime": 0.8706,
+      "eval_samples_per_second": 230.877,
+      "eval_steps_per_second": 8.041,
+      "step": 28985
+    },
+    {
+      "epoch": 187.03,
+      "learning_rate": 4.5164874551971325e-05,
+      "loss": 0.3346,
+      "step": 28990
+    },
+    {
+      "epoch": 187.1,
+      "learning_rate": 4.516129032258064e-05,
+      "loss": 0.2751,
+      "step": 29000
+    },
+    {
+      "epoch": 187.16,
+      "learning_rate": 4.515770609318997e-05,
+      "loss": 0.3205,
+      "step": 29010
+    },
+    {
+      "epoch": 187.23,
+      "learning_rate": 4.5154121863799286e-05,
+      "loss": 0.3116,
+      "step": 29020
+    },
+    {
+      "epoch": 187.29,
+      "learning_rate": 4.5150537634408604e-05,
+      "loss": 0.3369,
+      "step": 29030
+    },
+    {
+      "epoch": 187.35,
+      "learning_rate": 4.514695340501792e-05,
+      "loss": 0.3053,
+      "step": 29040
+    },
+    {
+      "epoch": 187.42,
+      "learning_rate": 4.514336917562724e-05,
+      "loss": 0.3186,
+      "step": 29050
+    },
+    {
+      "epoch": 187.48,
+      "learning_rate": 4.5139784946236566e-05,
+      "loss": 0.2893,
+      "step": 29060
+    },
+    {
+      "epoch": 187.55,
+      "learning_rate": 4.513620071684588e-05,
+      "loss": 0.2941,
+      "step": 29070
+    },
+    {
+      "epoch": 187.61,
+      "learning_rate": 4.5132616487455195e-05,
+      "loss": 0.3095,
+      "step": 29080
+    },
+    {
+      "epoch": 187.68,
+      "learning_rate": 4.512903225806452e-05,
+      "loss": 0.3508,
+      "step": 29090
+    },
+    {
+      "epoch": 187.74,
+      "learning_rate": 4.512544802867384e-05,
+      "loss": 0.3265,
+      "step": 29100
+    },
+    {
+      "epoch": 187.8,
+      "learning_rate": 4.512186379928316e-05,
+      "loss": 0.3547,
+      "step": 29110
+    },
+    {
+      "epoch": 187.87,
+      "learning_rate": 4.5118279569892475e-05,
+      "loss": 0.332,
+      "step": 29120
+    },
+    {
+      "epoch": 187.93,
+      "learning_rate": 4.511469534050179e-05,
+      "loss": 0.2878,
+      "step": 29130
+    },
+    {
+      "epoch": 188.0,
+      "learning_rate": 4.511111111111112e-05,
+      "loss": 0.2953,
+      "step": 29140
+    },
+    {
+      "epoch": 188.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0036320686340332,
+      "eval_runtime": 0.8615,
+      "eval_samples_per_second": 233.303,
+      "eval_steps_per_second": 8.125,
+      "step": 29140
+    },
+    {
+      "epoch": 188.06,
+      "learning_rate": 4.510752688172043e-05,
+      "loss": 0.3678,
+      "step": 29150
+    },
+    {
+      "epoch": 188.13,
+      "learning_rate": 4.510394265232975e-05,
+      "loss": 0.3447,
+      "step": 29160
+    },
+    {
+      "epoch": 188.19,
+      "learning_rate": 4.510035842293907e-05,
+      "loss": 0.3156,
+      "step": 29170
+    },
+    {
+      "epoch": 188.26,
+      "learning_rate": 4.509677419354839e-05,
+      "loss": 0.3348,
+      "step": 29180
+    },
+    {
+      "epoch": 188.32,
+      "learning_rate": 4.509318996415771e-05,
+      "loss": 0.3656,
+      "step": 29190
+    },
+    {
+      "epoch": 188.39,
+      "learning_rate": 4.508960573476703e-05,
+      "loss": 0.2948,
+      "step": 29200
+    },
+    {
+      "epoch": 188.45,
+      "learning_rate": 4.5086021505376346e-05,
+      "loss": 0.3474,
+      "step": 29210
+    },
+    {
+      "epoch": 188.51,
+      "learning_rate": 4.508243727598567e-05,
+      "loss": 0.3467,
+      "step": 29220
+    },
+    {
+      "epoch": 188.58,
+      "learning_rate": 4.507885304659498e-05,
+      "loss": 0.3124,
+      "step": 29230
+    },
+    {
+      "epoch": 188.64,
+      "learning_rate": 4.50752688172043e-05,
+      "loss": 0.3197,
+      "step": 29240
+    },
+    {
+      "epoch": 188.71,
+      "learning_rate": 4.5071684587813625e-05,
+      "loss": 0.3178,
+      "step": 29250
+    },
+    {
+      "epoch": 188.77,
+      "learning_rate": 4.506810035842294e-05,
+      "loss": 0.2879,
+      "step": 29260
+    },
+    {
+      "epoch": 188.84,
+      "learning_rate": 4.5064516129032255e-05,
+      "loss": 0.2894,
+      "step": 29270
+    },
+    {
+      "epoch": 188.9,
+      "learning_rate": 4.506093189964158e-05,
+      "loss": 0.2879,
+      "step": 29280
+    },
+    {
+      "epoch": 188.96,
+      "learning_rate": 4.50573476702509e-05,
+      "loss": 0.3601,
+      "step": 29290
+    },
+    {
+      "epoch": 189.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0178688764572144,
+      "eval_runtime": 0.8504,
+      "eval_samples_per_second": 236.36,
+      "eval_steps_per_second": 8.231,
+      "step": 29295
+    },
+    {
+      "epoch": 189.03,
+      "learning_rate": 4.5053763440860216e-05,
+      "loss": 0.3601,
+      "step": 29300
+    },
+    {
+      "epoch": 189.1,
+      "learning_rate": 4.5050179211469534e-05,
+      "loss": 0.2684,
+      "step": 29310
+    },
+    {
+      "epoch": 189.16,
+      "learning_rate": 4.504659498207885e-05,
+      "loss": 0.2908,
+      "step": 29320
+    },
+    {
+      "epoch": 189.23,
+      "learning_rate": 4.504301075268818e-05,
+      "loss": 0.3134,
+      "step": 29330
+    },
+    {
+      "epoch": 189.29,
+      "learning_rate": 4.5039426523297496e-05,
+      "loss": 0.3223,
+      "step": 29340
+    },
+    {
+      "epoch": 189.35,
+      "learning_rate": 4.503584229390681e-05,
+      "loss": 0.316,
+      "step": 29350
+    },
+    {
+      "epoch": 189.42,
+      "learning_rate": 4.503225806451613e-05,
+      "loss": 0.3463,
+      "step": 29360
+    },
+    {
+      "epoch": 189.48,
+      "learning_rate": 4.502867383512545e-05,
+      "loss": 0.3497,
+      "step": 29370
+    },
+    {
+      "epoch": 189.55,
+      "learning_rate": 4.502508960573477e-05,
+      "loss": 0.3671,
+      "step": 29380
+    },
+    {
+      "epoch": 189.61,
+      "learning_rate": 4.502150537634409e-05,
+      "loss": 0.3123,
+      "step": 29390
+    },
+    {
+      "epoch": 189.68,
+      "learning_rate": 4.5017921146953405e-05,
+      "loss": 0.3051,
+      "step": 29400
+    },
+    {
+      "epoch": 189.74,
+      "learning_rate": 4.501433691756273e-05,
+      "loss": 0.3532,
+      "step": 29410
+    },
+    {
+      "epoch": 189.8,
+      "learning_rate": 4.501075268817205e-05,
+      "loss": 0.3527,
+      "step": 29420
+    },
+    {
+      "epoch": 189.87,
+      "learning_rate": 4.500716845878136e-05,
+      "loss": 0.3481,
+      "step": 29430
+    },
+    {
+      "epoch": 189.93,
+      "learning_rate": 4.5003584229390685e-05,
+      "loss": 0.3638,
+      "step": 29440
+    },
+    {
+      "epoch": 190.0,
+      "learning_rate": 4.5e-05,
+      "loss": 0.345,
+      "step": 29450
+    },
+    {
+      "epoch": 190.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0146777629852295,
+      "eval_runtime": 0.8749,
+      "eval_samples_per_second": 229.728,
+      "eval_steps_per_second": 8.0,
+      "step": 29450
+    },
+    {
+      "epoch": 190.06,
+      "learning_rate": 4.499641577060932e-05,
+      "loss": 0.3556,
+      "step": 29460
+    },
+    {
+      "epoch": 190.13,
+      "learning_rate": 4.499283154121864e-05,
+      "loss": 0.3292,
+      "step": 29470
+    },
+    {
+      "epoch": 190.19,
+      "learning_rate": 4.498924731182796e-05,
+      "loss": 0.2627,
+      "step": 29480
+    },
+    {
+      "epoch": 190.26,
+      "learning_rate": 4.498566308243728e-05,
+      "loss": 0.3097,
+      "step": 29490
+    },
+    {
+      "epoch": 190.32,
+      "learning_rate": 4.49820788530466e-05,
+      "loss": 0.3315,
+      "step": 29500
+    },
+    {
+      "epoch": 190.39,
+      "learning_rate": 4.497849462365591e-05,
+      "loss": 0.2961,
+      "step": 29510
+    },
+    {
+      "epoch": 190.45,
+      "learning_rate": 4.497491039426524e-05,
+      "loss": 0.352,
+      "step": 29520
+    },
+    {
+      "epoch": 190.51,
+      "learning_rate": 4.4971326164874555e-05,
+      "loss": 0.2967,
+      "step": 29530
+    },
+    {
+      "epoch": 190.58,
+      "learning_rate": 4.4967741935483873e-05,
+      "loss": 0.3333,
+      "step": 29540
+    },
+    {
+      "epoch": 190.64,
+      "learning_rate": 4.496415770609319e-05,
+      "loss": 0.3572,
+      "step": 29550
+    },
+    {
+      "epoch": 190.71,
+      "learning_rate": 4.496057347670251e-05,
+      "loss": 0.3351,
+      "step": 29560
+    },
+    {
+      "epoch": 190.77,
+      "learning_rate": 4.4956989247311835e-05,
+      "loss": 0.3484,
+      "step": 29570
+    },
+    {
+      "epoch": 190.84,
+      "learning_rate": 4.4953405017921146e-05,
+      "loss": 0.3128,
+      "step": 29580
+    },
+    {
+      "epoch": 190.9,
+      "learning_rate": 4.4949820788530464e-05,
+      "loss": 0.3232,
+      "step": 29590
+    },
+    {
+      "epoch": 190.96,
+      "learning_rate": 4.494623655913979e-05,
+      "loss": 0.3569,
+      "step": 29600
+    },
+    {
+      "epoch": 191.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0248595476150513,
+      "eval_runtime": 0.8852,
+      "eval_samples_per_second": 227.059,
+      "eval_steps_per_second": 7.908,
+      "step": 29605
+    },
+    {
+      "epoch": 191.03,
+      "learning_rate": 4.494265232974911e-05,
+      "loss": 0.3473,
+      "step": 29610
+    },
+    {
+      "epoch": 191.1,
+      "learning_rate": 4.493906810035842e-05,
+      "loss": 0.2789,
+      "step": 29620
+    },
+    {
+      "epoch": 191.16,
+      "learning_rate": 4.4935483870967744e-05,
+      "loss": 0.3089,
+      "step": 29630
+    },
+    {
+      "epoch": 191.23,
+      "learning_rate": 4.493189964157706e-05,
+      "loss": 0.3191,
+      "step": 29640
+    },
+    {
+      "epoch": 191.29,
+      "learning_rate": 4.492831541218638e-05,
+      "loss": 0.3128,
+      "step": 29650
+    },
+    {
+      "epoch": 191.35,
+      "learning_rate": 4.49247311827957e-05,
+      "loss": 0.3039,
+      "step": 29660
+    },
+    {
+      "epoch": 191.42,
+      "learning_rate": 4.492114695340502e-05,
+      "loss": 0.3007,
+      "step": 29670
+    },
+    {
+      "epoch": 191.48,
+      "learning_rate": 4.491756272401434e-05,
+      "loss": 0.3157,
+      "step": 29680
+    },
+    {
+      "epoch": 191.55,
+      "learning_rate": 4.491397849462366e-05,
+      "loss": 0.2986,
+      "step": 29690
+    },
+    {
+      "epoch": 191.61,
+      "learning_rate": 4.491039426523297e-05,
+      "loss": 0.3529,
+      "step": 29700
+    },
+    {
+      "epoch": 191.68,
+      "learning_rate": 4.4906810035842296e-05,
+      "loss": 0.3575,
+      "step": 29710
+    },
+    {
+      "epoch": 191.74,
+      "learning_rate": 4.4903225806451615e-05,
+      "loss": 0.3355,
+      "step": 29720
+    },
+    {
+      "epoch": 191.8,
+      "learning_rate": 4.489964157706093e-05,
+      "loss": 0.253,
+      "step": 29730
+    },
+    {
+      "epoch": 191.87,
+      "learning_rate": 4.489605734767025e-05,
+      "loss": 0.3167,
+      "step": 29740
+    },
+    {
+      "epoch": 191.93,
+      "learning_rate": 4.489247311827957e-05,
+      "loss": 0.3438,
+      "step": 29750
+    },
+    {
+      "epoch": 192.0,
+      "learning_rate": 4.4888888888888894e-05,
+      "loss": 0.2725,
+      "step": 29760
+    },
+    {
+      "epoch": 192.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8727553486824036,
+      "eval_runtime": 0.8518,
+      "eval_samples_per_second": 235.984,
+      "eval_steps_per_second": 8.218,
+      "step": 29760
+    },
+    {
+      "epoch": 192.06,
+      "learning_rate": 4.488530465949821e-05,
+      "loss": 0.2996,
+      "step": 29770
+    },
+    {
+      "epoch": 192.13,
+      "learning_rate": 4.4881720430107524e-05,
+      "loss": 0.2735,
+      "step": 29780
+    },
+    {
+      "epoch": 192.19,
+      "learning_rate": 4.487813620071685e-05,
+      "loss": 0.3091,
+      "step": 29790
+    },
+    {
+      "epoch": 192.26,
+      "learning_rate": 4.487455197132617e-05,
+      "loss": 0.3159,
+      "step": 29800
+    },
+    {
+      "epoch": 192.32,
+      "learning_rate": 4.4870967741935485e-05,
+      "loss": 0.3147,
+      "step": 29810
+    },
+    {
+      "epoch": 192.39,
+      "learning_rate": 4.4867383512544804e-05,
+      "loss": 0.3111,
+      "step": 29820
+    },
+    {
+      "epoch": 192.45,
+      "learning_rate": 4.486379928315412e-05,
+      "loss": 0.2857,
+      "step": 29830
+    },
+    {
+      "epoch": 192.51,
+      "learning_rate": 4.486021505376345e-05,
+      "loss": 0.2965,
+      "step": 29840
+    },
+    {
+      "epoch": 192.58,
+      "learning_rate": 4.4856630824372765e-05,
+      "loss": 0.3531,
+      "step": 29850
+    },
+    {
+      "epoch": 192.64,
+      "learning_rate": 4.4853046594982076e-05,
+      "loss": 0.266,
+      "step": 29860
+    },
+    {
+      "epoch": 192.71,
+      "learning_rate": 4.48494623655914e-05,
+      "loss": 0.2779,
+      "step": 29870
+    },
+    {
+      "epoch": 192.77,
+      "learning_rate": 4.484587813620072e-05,
+      "loss": 0.3008,
+      "step": 29880
+    },
+    {
+      "epoch": 192.84,
+      "learning_rate": 4.484229390681004e-05,
+      "loss": 0.3079,
+      "step": 29890
+    },
+    {
+      "epoch": 192.9,
+      "learning_rate": 4.4838709677419356e-05,
+      "loss": 0.3801,
+      "step": 29900
+    },
+    {
+      "epoch": 192.96,
+      "learning_rate": 4.4835125448028674e-05,
+      "loss": 0.2786,
+      "step": 29910
+    },
+    {
+      "epoch": 193.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0356892347335815,
+      "eval_runtime": 0.8814,
+      "eval_samples_per_second": 228.045,
+      "eval_steps_per_second": 7.942,
+      "step": 29915
+    },
+    {
+      "epoch": 193.03,
+      "learning_rate": 4.483154121863799e-05,
+      "loss": 0.3064,
+      "step": 29920
+    },
+    {
+      "epoch": 193.1,
+      "learning_rate": 4.482795698924732e-05,
+      "loss": 0.353,
+      "step": 29930
+    },
+    {
+      "epoch": 193.16,
+      "learning_rate": 4.482437275985663e-05,
+      "loss": 0.3001,
+      "step": 29940
+    },
+    {
+      "epoch": 193.23,
+      "learning_rate": 4.4820788530465954e-05,
+      "loss": 0.2888,
+      "step": 29950
+    },
+    {
+      "epoch": 193.29,
+      "learning_rate": 4.481720430107527e-05,
+      "loss": 0.3453,
+      "step": 29960
+    },
+    {
+      "epoch": 193.35,
+      "learning_rate": 4.481362007168459e-05,
+      "loss": 0.2805,
+      "step": 29970
+    },
+    {
+      "epoch": 193.42,
+      "learning_rate": 4.481003584229391e-05,
+      "loss": 0.2611,
+      "step": 29980
+    },
+    {
+      "epoch": 193.48,
+      "learning_rate": 4.4806451612903227e-05,
+      "loss": 0.2695,
+      "step": 29990
+    },
+    {
+      "epoch": 193.55,
+      "learning_rate": 4.4802867383512545e-05,
+      "loss": 0.3247,
+      "step": 30000
+    },
+    {
+      "epoch": 193.61,
+      "learning_rate": 4.479928315412187e-05,
+      "loss": 0.3661,
+      "step": 30010
+    },
+    {
+      "epoch": 193.68,
+      "learning_rate": 4.479569892473118e-05,
+      "loss": 0.3064,
+      "step": 30020
+    },
+    {
+      "epoch": 193.74,
+      "learning_rate": 4.4792114695340506e-05,
+      "loss": 0.3174,
+      "step": 30030
+    },
+    {
+      "epoch": 193.8,
+      "learning_rate": 4.4788530465949824e-05,
+      "loss": 0.3038,
+      "step": 30040
+    },
+    {
+      "epoch": 193.87,
+      "learning_rate": 4.478494623655914e-05,
+      "loss": 0.2685,
+      "step": 30050
+    },
+    {
+      "epoch": 193.93,
+      "learning_rate": 4.478136200716846e-05,
+      "loss": 0.2705,
+      "step": 30060
+    },
+    {
+      "epoch": 194.0,
+      "learning_rate": 4.477777777777778e-05,
+      "loss": 0.3033,
+      "step": 30070
+    },
+    {
+      "epoch": 194.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.960686981678009,
+      "eval_runtime": 0.8743,
+      "eval_samples_per_second": 229.909,
+      "eval_steps_per_second": 8.007,
+      "step": 30070
+    },
+    {
+      "epoch": 194.06,
+      "learning_rate": 4.47741935483871e-05,
+      "loss": 0.3606,
+      "step": 30080
+    },
+    {
+      "epoch": 194.13,
+      "learning_rate": 4.477060931899642e-05,
+      "loss": 0.3137,
+      "step": 30090
+    },
+    {
+      "epoch": 194.19,
+      "learning_rate": 4.4767025089605734e-05,
+      "loss": 0.2577,
+      "step": 30100
+    },
+    {
+      "epoch": 194.26,
+      "learning_rate": 4.476344086021506e-05,
+      "loss": 0.3344,
+      "step": 30110
+    },
+    {
+      "epoch": 194.32,
+      "learning_rate": 4.475985663082438e-05,
+      "loss": 0.3454,
+      "step": 30120
+    },
+    {
+      "epoch": 194.39,
+      "learning_rate": 4.4756272401433695e-05,
+      "loss": 0.2683,
+      "step": 30130
+    },
+    {
+      "epoch": 194.45,
+      "learning_rate": 4.475268817204301e-05,
+      "loss": 0.2873,
+      "step": 30140
+    },
+    {
+      "epoch": 194.51,
+      "learning_rate": 4.474910394265233e-05,
+      "loss": 0.3043,
+      "step": 30150
+    },
+    {
+      "epoch": 194.58,
+      "learning_rate": 4.474551971326165e-05,
+      "loss": 0.2588,
+      "step": 30160
+    },
+    {
+      "epoch": 194.64,
+      "learning_rate": 4.4741935483870975e-05,
+      "loss": 0.3161,
+      "step": 30170
+    },
+    {
+      "epoch": 194.71,
+      "learning_rate": 4.4738351254480286e-05,
+      "loss": 0.3416,
+      "step": 30180
+    },
+    {
+      "epoch": 194.77,
+      "learning_rate": 4.473476702508961e-05,
+      "loss": 0.2719,
+      "step": 30190
+    },
+    {
+      "epoch": 194.84,
+      "learning_rate": 4.473118279569893e-05,
+      "loss": 0.3168,
+      "step": 30200
+    },
+    {
+      "epoch": 194.9,
+      "learning_rate": 4.472759856630825e-05,
+      "loss": 0.3168,
+      "step": 30210
+    },
+    {
+      "epoch": 194.96,
+      "learning_rate": 4.4724014336917566e-05,
+      "loss": 0.3403,
+      "step": 30220
+    },
+    {
+      "epoch": 195.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9680474996566772,
+      "eval_runtime": 0.8637,
+      "eval_samples_per_second": 232.718,
+      "eval_steps_per_second": 8.105,
+      "step": 30225
+    },
+    {
+      "epoch": 195.03,
+      "learning_rate": 4.4720430107526884e-05,
+      "loss": 0.3431,
+      "step": 30230
+    },
+    {
+      "epoch": 195.1,
+      "learning_rate": 4.47168458781362e-05,
+      "loss": 0.3665,
+      "step": 30240
+    },
+    {
+      "epoch": 195.16,
+      "learning_rate": 4.471326164874552e-05,
+      "loss": 0.3115,
+      "step": 30250
+    },
+    {
+      "epoch": 195.23,
+      "learning_rate": 4.470967741935484e-05,
+      "loss": 0.2835,
+      "step": 30260
+    },
+    {
+      "epoch": 195.29,
+      "learning_rate": 4.470609318996416e-05,
+      "loss": 0.3007,
+      "step": 30270
+    },
+    {
+      "epoch": 195.35,
+      "learning_rate": 4.470250896057348e-05,
+      "loss": 0.3146,
+      "step": 30280
+    },
+    {
+      "epoch": 195.42,
+      "learning_rate": 4.469892473118279e-05,
+      "loss": 0.2723,
+      "step": 30290
+    },
+    {
+      "epoch": 195.48,
+      "learning_rate": 4.469534050179212e-05,
+      "loss": 0.3121,
+      "step": 30300
+    },
+    {
+      "epoch": 195.55,
+      "learning_rate": 4.4691756272401436e-05,
+      "loss": 0.2641,
+      "step": 30310
+    },
+    {
+      "epoch": 195.61,
+      "learning_rate": 4.4688172043010754e-05,
+      "loss": 0.3317,
+      "step": 30320
+    },
+    {
+      "epoch": 195.68,
+      "learning_rate": 4.468458781362007e-05,
+      "loss": 0.2928,
+      "step": 30330
+    },
+    {
+      "epoch": 195.74,
+      "learning_rate": 4.468100358422939e-05,
+      "loss": 0.3402,
+      "step": 30340
+    },
+    {
+      "epoch": 195.8,
+      "learning_rate": 4.467741935483871e-05,
+      "loss": 0.3004,
+      "step": 30350
+    },
+    {
+      "epoch": 195.87,
+      "learning_rate": 4.4673835125448034e-05,
+      "loss": 0.3771,
+      "step": 30360
+    },
+    {
+      "epoch": 195.93,
+      "learning_rate": 4.4670250896057345e-05,
+      "loss": 0.3269,
+      "step": 30370
+    },
+    {
+      "epoch": 196.0,
+      "learning_rate": 4.466666666666667e-05,
+      "loss": 0.3013,
+      "step": 30380
+    },
+    {
+      "epoch": 196.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9250904321670532,
+      "eval_runtime": 0.8748,
+      "eval_samples_per_second": 229.764,
+      "eval_steps_per_second": 8.002,
+      "step": 30380
+    },
+    {
+      "epoch": 196.06,
+      "learning_rate": 4.466308243727599e-05,
+      "loss": 0.2811,
+      "step": 30390
+    },
+    {
+      "epoch": 196.13,
+      "learning_rate": 4.465949820788531e-05,
+      "loss": 0.3195,
+      "step": 30400
+    },
+    {
+      "epoch": 196.19,
+      "learning_rate": 4.4655913978494625e-05,
+      "loss": 0.2716,
+      "step": 30410
+    },
+    {
+      "epoch": 196.26,
+      "learning_rate": 4.465232974910394e-05,
+      "loss": 0.3561,
+      "step": 30420
+    },
+    {
+      "epoch": 196.32,
+      "learning_rate": 4.464874551971326e-05,
+      "loss": 0.3468,
+      "step": 30430
+    },
+    {
+      "epoch": 196.39,
+      "learning_rate": 4.4645161290322586e-05,
+      "loss": 0.3305,
+      "step": 30440
+    },
+    {
+      "epoch": 196.45,
+      "learning_rate": 4.46415770609319e-05,
+      "loss": 0.2813,
+      "step": 30450
+    },
+    {
+      "epoch": 196.51,
+      "learning_rate": 4.463799283154122e-05,
+      "loss": 0.3178,
+      "step": 30460
+    },
+    {
+      "epoch": 196.58,
+      "learning_rate": 4.463440860215054e-05,
+      "loss": 0.2408,
+      "step": 30470
+    },
+    {
+      "epoch": 196.64,
+      "learning_rate": 4.463082437275986e-05,
+      "loss": 0.2743,
+      "step": 30480
+    },
+    {
+      "epoch": 196.71,
+      "learning_rate": 4.462724014336918e-05,
+      "loss": 0.2636,
+      "step": 30490
+    },
+    {
+      "epoch": 196.77,
+      "learning_rate": 4.4623655913978496e-05,
+      "loss": 0.3521,
+      "step": 30500
+    },
+    {
+      "epoch": 196.84,
+      "learning_rate": 4.4620071684587814e-05,
+      "loss": 0.3493,
+      "step": 30510
+    },
+    {
+      "epoch": 196.9,
+      "learning_rate": 4.461648745519714e-05,
+      "loss": 0.323,
+      "step": 30520
+    },
+    {
+      "epoch": 196.96,
+      "learning_rate": 4.461290322580645e-05,
+      "loss": 0.2614,
+      "step": 30530
+    },
+    {
+      "epoch": 197.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8789924383163452,
+      "eval_runtime": 0.9127,
+      "eval_samples_per_second": 220.227,
+      "eval_steps_per_second": 7.67,
+      "step": 30535
+    },
+    {
+      "epoch": 197.03,
+      "learning_rate": 4.4609318996415775e-05,
+      "loss": 0.2922,
+      "step": 30540
+    },
+    {
+      "epoch": 197.1,
+      "learning_rate": 4.4605734767025093e-05,
+      "loss": 0.2653,
+      "step": 30550
+    },
+    {
+      "epoch": 197.16,
+      "learning_rate": 4.460215053763441e-05,
+      "loss": 0.3228,
+      "step": 30560
+    },
+    {
+      "epoch": 197.23,
+      "learning_rate": 4.459856630824373e-05,
+      "loss": 0.3393,
+      "step": 30570
+    },
+    {
+      "epoch": 197.29,
+      "learning_rate": 4.459498207885305e-05,
+      "loss": 0.2983,
+      "step": 30580
+    },
+    {
+      "epoch": 197.35,
+      "learning_rate": 4.4591397849462366e-05,
+      "loss": 0.289,
+      "step": 30590
+    },
+    {
+      "epoch": 197.42,
+      "learning_rate": 4.458781362007169e-05,
+      "loss": 0.3014,
+      "step": 30600
+    },
+    {
+      "epoch": 197.48,
+      "learning_rate": 4.4584229390681e-05,
+      "loss": 0.3003,
+      "step": 30610
+    },
+    {
+      "epoch": 197.55,
+      "learning_rate": 4.458064516129032e-05,
+      "loss": 0.347,
+      "step": 30620
+    },
+    {
+      "epoch": 197.61,
+      "learning_rate": 4.4577060931899646e-05,
+      "loss": 0.3345,
+      "step": 30630
+    },
+    {
+      "epoch": 197.68,
+      "learning_rate": 4.4573476702508964e-05,
+      "loss": 0.3391,
+      "step": 30640
+    },
+    {
+      "epoch": 197.74,
+      "learning_rate": 4.456989247311828e-05,
+      "loss": 0.319,
+      "step": 30650
+    },
+    {
+      "epoch": 197.8,
+      "learning_rate": 4.45663082437276e-05,
+      "loss": 0.2861,
+      "step": 30660
+    },
+    {
+      "epoch": 197.87,
+      "learning_rate": 4.456272401433692e-05,
+      "loss": 0.3161,
+      "step": 30670
+    },
+    {
+      "epoch": 197.93,
+      "learning_rate": 4.4559139784946244e-05,
+      "loss": 0.3259,
+      "step": 30680
+    },
+    {
+      "epoch": 198.0,
+      "learning_rate": 4.4555555555555555e-05,
+      "loss": 0.3184,
+      "step": 30690
+    },
+    {
+      "epoch": 198.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.8986311554908752,
+      "eval_runtime": 0.9409,
+      "eval_samples_per_second": 213.63,
+      "eval_steps_per_second": 7.44,
+      "step": 30690
+    },
+    {
+      "epoch": 198.06,
+      "learning_rate": 4.455197132616487e-05,
+      "loss": 0.3476,
+      "step": 30700
+    },
+    {
+      "epoch": 198.13,
+      "learning_rate": 4.45483870967742e-05,
+      "loss": 0.2953,
+      "step": 30710
+    },
+    {
+      "epoch": 198.19,
+      "learning_rate": 4.4544802867383517e-05,
+      "loss": 0.3289,
+      "step": 30720
+    },
+    {
+      "epoch": 198.26,
+      "learning_rate": 4.4541218637992835e-05,
+      "loss": 0.2991,
+      "step": 30730
+    },
+    {
+      "epoch": 198.32,
+      "learning_rate": 4.453763440860215e-05,
+      "loss": 0.3832,
+      "step": 30740
+    },
+    {
+      "epoch": 198.39,
+      "learning_rate": 4.453405017921147e-05,
+      "loss": 0.3108,
+      "step": 30750
+    },
+    {
+      "epoch": 198.45,
+      "learning_rate": 4.4530465949820796e-05,
+      "loss": 0.248,
+      "step": 30760
+    },
+    {
+      "epoch": 198.51,
+      "learning_rate": 4.452688172043011e-05,
+      "loss": 0.3042,
+      "step": 30770
+    },
+    {
+      "epoch": 198.58,
+      "learning_rate": 4.4523297491039426e-05,
+      "loss": 0.3083,
+      "step": 30780
+    },
+    {
+      "epoch": 198.64,
+      "learning_rate": 4.451971326164875e-05,
+      "loss": 0.2783,
+      "step": 30790
+    },
+    {
+      "epoch": 198.71,
+      "learning_rate": 4.451612903225807e-05,
+      "loss": 0.3054,
+      "step": 30800
+    },
+    {
+      "epoch": 198.77,
+      "learning_rate": 4.451254480286739e-05,
+      "loss": 0.3192,
+      "step": 30810
+    },
+    {
+      "epoch": 198.84,
+      "learning_rate": 4.4508960573476705e-05,
+      "loss": 0.3796,
+      "step": 30820
+    },
+    {
+      "epoch": 198.9,
+      "learning_rate": 4.4505376344086024e-05,
+      "loss": 0.3228,
+      "step": 30830
+    },
+    {
+      "epoch": 198.96,
+      "learning_rate": 4.450179211469535e-05,
+      "loss": 0.3177,
+      "step": 30840
+    },
+    {
+      "epoch": 199.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.8753862380981445,
+      "eval_runtime": 0.876,
+      "eval_samples_per_second": 229.464,
+      "eval_steps_per_second": 7.991,
+      "step": 30845
+    },
+    {
+      "epoch": 199.03,
+      "learning_rate": 4.449820788530466e-05,
+      "loss": 0.3471,
+      "step": 30850
+    },
+    {
+      "epoch": 199.1,
+      "learning_rate": 4.449462365591398e-05,
+      "loss": 0.2723,
+      "step": 30860
+    },
+    {
+      "epoch": 199.16,
+      "learning_rate": 4.44910394265233e-05,
+      "loss": 0.3172,
+      "step": 30870
+    },
+    {
+      "epoch": 199.23,
+      "learning_rate": 4.4487455197132615e-05,
+      "loss": 0.3365,
+      "step": 30880
+    },
+    {
+      "epoch": 199.29,
+      "learning_rate": 4.448387096774194e-05,
+      "loss": 0.2954,
+      "step": 30890
+    },
+    {
+      "epoch": 199.35,
+      "learning_rate": 4.448028673835126e-05,
+      "loss": 0.3297,
+      "step": 30900
+    },
+    {
+      "epoch": 199.42,
+      "learning_rate": 4.4476702508960576e-05,
+      "loss": 0.3105,
+      "step": 30910
+    },
+    {
+      "epoch": 199.48,
+      "learning_rate": 4.4473118279569894e-05,
+      "loss": 0.332,
+      "step": 30920
+    },
+    {
+      "epoch": 199.55,
+      "learning_rate": 4.446953405017921e-05,
+      "loss": 0.3935,
+      "step": 30930
+    },
+    {
+      "epoch": 199.61,
+      "learning_rate": 4.446594982078853e-05,
+      "loss": 0.3556,
+      "step": 30940
+    },
+    {
+      "epoch": 199.68,
+      "learning_rate": 4.4462365591397856e-05,
+      "loss": 0.2789,
+      "step": 30950
+    },
+    {
+      "epoch": 199.74,
+      "learning_rate": 4.445878136200717e-05,
+      "loss": 0.3225,
+      "step": 30960
+    },
+    {
+      "epoch": 199.8,
+      "learning_rate": 4.4455197132616485e-05,
+      "loss": 0.3277,
+      "step": 30970
+    },
+    {
+      "epoch": 199.87,
+      "learning_rate": 4.445161290322581e-05,
+      "loss": 0.3186,
+      "step": 30980
+    },
+    {
+      "epoch": 199.93,
+      "learning_rate": 4.444802867383513e-05,
+      "loss": 0.2884,
+      "step": 30990
+    },
+    {
+      "epoch": 200.0,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.3549,
+      "step": 31000
+    },
+    {
+      "epoch": 200.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9412870407104492,
+      "eval_runtime": 0.8739,
+      "eval_samples_per_second": 230.016,
+      "eval_steps_per_second": 8.011,
+      "step": 31000
+    },
+    {
+      "epoch": 200.06,
+      "learning_rate": 4.4440860215053765e-05,
+      "loss": 0.339,
+      "step": 31010
+    },
+    {
+      "epoch": 200.13,
+      "learning_rate": 4.443727598566308e-05,
+      "loss": 0.256,
+      "step": 31020
+    },
+    {
+      "epoch": 200.19,
+      "learning_rate": 4.443369175627241e-05,
+      "loss": 0.3426,
+      "step": 31030
+    },
+    {
+      "epoch": 200.26,
+      "learning_rate": 4.443010752688172e-05,
+      "loss": 0.3069,
+      "step": 31040
+    },
+    {
+      "epoch": 200.32,
+      "learning_rate": 4.442652329749104e-05,
+      "loss": 0.3114,
+      "step": 31050
+    },
+    {
+      "epoch": 200.39,
+      "learning_rate": 4.442293906810036e-05,
+      "loss": 0.3191,
+      "step": 31060
+    },
+    {
+      "epoch": 200.45,
+      "learning_rate": 4.441935483870968e-05,
+      "loss": 0.3355,
+      "step": 31070
+    },
+    {
+      "epoch": 200.51,
+      "learning_rate": 4.4415770609319e-05,
+      "loss": 0.3365,
+      "step": 31080
+    },
+    {
+      "epoch": 200.58,
+      "learning_rate": 4.441218637992832e-05,
+      "loss": 0.3121,
+      "step": 31090
+    },
+    {
+      "epoch": 200.64,
+      "learning_rate": 4.4408602150537635e-05,
+      "loss": 0.2947,
+      "step": 31100
+    },
+    {
+      "epoch": 200.71,
+      "learning_rate": 4.440501792114696e-05,
+      "loss": 0.2912,
+      "step": 31110
+    },
+    {
+      "epoch": 200.77,
+      "learning_rate": 4.440143369175627e-05,
+      "loss": 0.3064,
+      "step": 31120
+    },
+    {
+      "epoch": 200.84,
+      "learning_rate": 4.439784946236559e-05,
+      "loss": 0.3037,
+      "step": 31130
+    },
+    {
+      "epoch": 200.9,
+      "learning_rate": 4.4394265232974915e-05,
+      "loss": 0.3241,
+      "step": 31140
+    },
+    {
+      "epoch": 200.96,
+      "learning_rate": 4.439068100358423e-05,
+      "loss": 0.3772,
+      "step": 31150
+    },
+    {
+      "epoch": 201.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.889571487903595,
+      "eval_runtime": 0.9002,
+      "eval_samples_per_second": 223.275,
+      "eval_steps_per_second": 7.776,
+      "step": 31155
+    },
+    {
+      "epoch": 201.03,
+      "learning_rate": 4.438709677419355e-05,
+      "loss": 0.3563,
+      "step": 31160
+    },
+    {
+      "epoch": 201.1,
+      "learning_rate": 4.438351254480287e-05,
+      "loss": 0.3074,
+      "step": 31170
+    },
+    {
+      "epoch": 201.16,
+      "learning_rate": 4.437992831541219e-05,
+      "loss": 0.3038,
+      "step": 31180
+    },
+    {
+      "epoch": 201.23,
+      "learning_rate": 4.437634408602151e-05,
+      "loss": 0.2726,
+      "step": 31190
+    },
+    {
+      "epoch": 201.29,
+      "learning_rate": 4.4372759856630824e-05,
+      "loss": 0.349,
+      "step": 31200
+    },
+    {
+      "epoch": 201.35,
+      "learning_rate": 4.436917562724014e-05,
+      "loss": 0.345,
+      "step": 31210
+    },
+    {
+      "epoch": 201.42,
+      "learning_rate": 4.436559139784947e-05,
+      "loss": 0.2767,
+      "step": 31220
+    },
+    {
+      "epoch": 201.48,
+      "learning_rate": 4.4362007168458786e-05,
+      "loss": 0.3067,
+      "step": 31230
+    },
+    {
+      "epoch": 201.55,
+      "learning_rate": 4.4358422939068104e-05,
+      "loss": 0.2937,
+      "step": 31240
+    },
+    {
+      "epoch": 201.61,
+      "learning_rate": 4.435483870967742e-05,
+      "loss": 0.3302,
+      "step": 31250
+    },
+    {
+      "epoch": 201.68,
+      "learning_rate": 4.435125448028674e-05,
+      "loss": 0.3507,
+      "step": 31260
+    },
+    {
+      "epoch": 201.74,
+      "learning_rate": 4.434767025089606e-05,
+      "loss": 0.3004,
+      "step": 31270
+    },
+    {
+      "epoch": 201.8,
+      "learning_rate": 4.434408602150538e-05,
+      "loss": 0.302,
+      "step": 31280
+    },
+    {
+      "epoch": 201.87,
+      "learning_rate": 4.4340501792114695e-05,
+      "loss": 0.3161,
+      "step": 31290
+    },
+    {
+      "epoch": 201.93,
+      "learning_rate": 4.433691756272402e-05,
+      "loss": 0.3067,
+      "step": 31300
+    },
+    {
+      "epoch": 202.0,
+      "learning_rate": 4.433333333333334e-05,
+      "loss": 0.2922,
+      "step": 31310
+    },
+    {
+      "epoch": 202.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.8751164674758911,
+      "eval_runtime": 0.8863,
+      "eval_samples_per_second": 226.794,
+      "eval_steps_per_second": 7.898,
+      "step": 31310
+    },
+    {
+      "epoch": 202.06,
+      "learning_rate": 4.432974910394265e-05,
+      "loss": 0.3577,
+      "step": 31320
+    },
+    {
+      "epoch": 202.13,
+      "learning_rate": 4.4326164874551974e-05,
+      "loss": 0.3166,
+      "step": 31330
+    },
+    {
+      "epoch": 202.19,
+      "learning_rate": 4.432258064516129e-05,
+      "loss": 0.2926,
+      "step": 31340
+    },
+    {
+      "epoch": 202.26,
+      "learning_rate": 4.431899641577061e-05,
+      "loss": 0.3267,
+      "step": 31350
+    },
+    {
+      "epoch": 202.32,
+      "learning_rate": 4.431541218637993e-05,
+      "loss": 0.3029,
+      "step": 31360
+    },
+    {
+      "epoch": 202.39,
+      "learning_rate": 4.431182795698925e-05,
+      "loss": 0.2768,
+      "step": 31370
+    },
+    {
+      "epoch": 202.45,
+      "learning_rate": 4.430824372759857e-05,
+      "loss": 0.2615,
+      "step": 31380
+    },
+    {
+      "epoch": 202.51,
+      "learning_rate": 4.430465949820789e-05,
+      "loss": 0.3408,
+      "step": 31390
+    },
+    {
+      "epoch": 202.58,
+      "learning_rate": 4.43010752688172e-05,
+      "loss": 0.3178,
+      "step": 31400
+    },
+    {
+      "epoch": 202.64,
+      "learning_rate": 4.429749103942653e-05,
+      "loss": 0.2795,
+      "step": 31410
+    },
+    {
+      "epoch": 202.71,
+      "learning_rate": 4.4293906810035845e-05,
+      "loss": 0.2632,
+      "step": 31420
+    },
+    {
+      "epoch": 202.77,
+      "learning_rate": 4.429032258064516e-05,
+      "loss": 0.3271,
+      "step": 31430
+    },
+    {
+      "epoch": 202.84,
+      "learning_rate": 4.428673835125448e-05,
+      "loss": 0.284,
+      "step": 31440
+    },
+    {
+      "epoch": 202.9,
+      "learning_rate": 4.42831541218638e-05,
+      "loss": 0.392,
+      "step": 31450
+    },
+    {
+      "epoch": 202.96,
+      "learning_rate": 4.4279569892473125e-05,
+      "loss": 0.3202,
+      "step": 31460
+    },
+    {
+      "epoch": 203.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.924308717250824,
+      "eval_runtime": 0.869,
+      "eval_samples_per_second": 231.297,
+      "eval_steps_per_second": 8.055,
+      "step": 31465
+    },
+    {
+      "epoch": 203.03,
+      "learning_rate": 4.427598566308244e-05,
+      "loss": 0.3672,
+      "step": 31470
+    },
+    {
+      "epoch": 203.1,
+      "learning_rate": 4.4272401433691754e-05,
+      "loss": 0.392,
+      "step": 31480
+    },
+    {
+      "epoch": 203.16,
+      "learning_rate": 4.426881720430108e-05,
+      "loss": 0.3342,
+      "step": 31490
+    },
+    {
+      "epoch": 203.23,
+      "learning_rate": 4.42652329749104e-05,
+      "loss": 0.3168,
+      "step": 31500
+    },
+    {
+      "epoch": 203.29,
+      "learning_rate": 4.4261648745519716e-05,
+      "loss": 0.2768,
+      "step": 31510
+    },
+    {
+      "epoch": 203.35,
+      "learning_rate": 4.4258064516129034e-05,
+      "loss": 0.3335,
+      "step": 31520
+    },
+    {
+      "epoch": 203.42,
+      "learning_rate": 4.425448028673835e-05,
+      "loss": 0.3116,
+      "step": 31530
+    },
+    {
+      "epoch": 203.48,
+      "learning_rate": 4.425089605734768e-05,
+      "loss": 0.301,
+      "step": 31540
+    },
+    {
+      "epoch": 203.55,
+      "learning_rate": 4.424731182795699e-05,
+      "loss": 0.3101,
+      "step": 31550
+    },
+    {
+      "epoch": 203.61,
+      "learning_rate": 4.424372759856631e-05,
+      "loss": 0.3791,
+      "step": 31560
+    },
+    {
+      "epoch": 203.68,
+      "learning_rate": 4.424014336917563e-05,
+      "loss": 0.2582,
+      "step": 31570
+    },
+    {
+      "epoch": 203.74,
+      "learning_rate": 4.423655913978495e-05,
+      "loss": 0.2616,
+      "step": 31580
+    },
+    {
+      "epoch": 203.8,
+      "learning_rate": 4.423297491039427e-05,
+      "loss": 0.3092,
+      "step": 31590
+    },
+    {
+      "epoch": 203.87,
+      "learning_rate": 4.4229390681003586e-05,
+      "loss": 0.2936,
+      "step": 31600
+    },
+    {
+      "epoch": 203.93,
+      "learning_rate": 4.4225806451612905e-05,
+      "loss": 0.3353,
+      "step": 31610
+    },
+    {
+      "epoch": 204.0,
+      "learning_rate": 4.422222222222222e-05,
+      "loss": 0.338,
+      "step": 31620
+    },
+    {
+      "epoch": 204.0,
+      "eval_accuracy": 0.8009950248756219,
+      "eval_loss": 0.8652067184448242,
+      "eval_runtime": 0.8807,
+      "eval_samples_per_second": 228.237,
+      "eval_steps_per_second": 7.949,
+      "step": 31620
+    },
+    {
+      "epoch": 204.06,
+      "learning_rate": 4.421863799283154e-05,
+      "loss": 0.3456,
+      "step": 31630
+    },
+    {
+      "epoch": 204.13,
+      "learning_rate": 4.421505376344086e-05,
+      "loss": 0.3076,
+      "step": 31640
+    },
+    {
+      "epoch": 204.19,
+      "learning_rate": 4.4211469534050184e-05,
+      "loss": 0.3017,
+      "step": 31650
+    },
+    {
+      "epoch": 204.26,
+      "learning_rate": 4.42078853046595e-05,
+      "loss": 0.3141,
+      "step": 31660
+    },
+    {
+      "epoch": 204.32,
+      "learning_rate": 4.4204301075268814e-05,
+      "loss": 0.2915,
+      "step": 31670
+    },
+    {
+      "epoch": 204.39,
+      "learning_rate": 4.420071684587814e-05,
+      "loss": 0.2622,
+      "step": 31680
+    },
+    {
+      "epoch": 204.45,
+      "learning_rate": 4.419713261648746e-05,
+      "loss": 0.3675,
+      "step": 31690
+    },
+    {
+      "epoch": 204.51,
+      "learning_rate": 4.4193548387096775e-05,
+      "loss": 0.2938,
+      "step": 31700
+    },
+    {
+      "epoch": 204.58,
+      "learning_rate": 4.418996415770609e-05,
+      "loss": 0.269,
+      "step": 31710
+    },
+    {
+      "epoch": 204.64,
+      "learning_rate": 4.418637992831541e-05,
+      "loss": 0.3,
+      "step": 31720
+    },
+    {
+      "epoch": 204.71,
+      "learning_rate": 4.4182795698924737e-05,
+      "loss": 0.3635,
+      "step": 31730
+    },
+    {
+      "epoch": 204.77,
+      "learning_rate": 4.4179211469534055e-05,
+      "loss": 0.2925,
+      "step": 31740
+    },
+    {
+      "epoch": 204.84,
+      "learning_rate": 4.4175627240143366e-05,
+      "loss": 0.3023,
+      "step": 31750
+    },
+    {
+      "epoch": 204.9,
+      "learning_rate": 4.417204301075269e-05,
+      "loss": 0.2787,
+      "step": 31760
+    },
+    {
+      "epoch": 204.96,
+      "learning_rate": 4.416845878136201e-05,
+      "loss": 0.3249,
+      "step": 31770
+    },
+    {
+      "epoch": 205.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.8889145851135254,
+      "eval_runtime": 0.8912,
+      "eval_samples_per_second": 225.527,
+      "eval_steps_per_second": 7.854,
+      "step": 31775
+    },
+    {
+      "epoch": 205.03,
+      "learning_rate": 4.416487455197133e-05,
+      "loss": 0.3012,
+      "step": 31780
+    },
+    {
+      "epoch": 205.1,
+      "learning_rate": 4.4161290322580646e-05,
+      "loss": 0.3056,
+      "step": 31790
+    },
+    {
+      "epoch": 205.16,
+      "learning_rate": 4.4157706093189964e-05,
+      "loss": 0.3046,
+      "step": 31800
+    },
+    {
+      "epoch": 205.23,
+      "learning_rate": 4.415412186379929e-05,
+      "loss": 0.2675,
+      "step": 31810
+    },
+    {
+      "epoch": 205.29,
+      "learning_rate": 4.415053763440861e-05,
+      "loss": 0.3204,
+      "step": 31820
+    },
+    {
+      "epoch": 205.35,
+      "learning_rate": 4.414695340501792e-05,
+      "loss": 0.2854,
+      "step": 31830
+    },
+    {
+      "epoch": 205.42,
+      "learning_rate": 4.4143369175627244e-05,
+      "loss": 0.3359,
+      "step": 31840
+    },
+    {
+      "epoch": 205.48,
+      "learning_rate": 4.413978494623656e-05,
+      "loss": 0.291,
+      "step": 31850
+    },
+    {
+      "epoch": 205.55,
+      "learning_rate": 4.413620071684588e-05,
+      "loss": 0.3137,
+      "step": 31860
+    },
+    {
+      "epoch": 205.61,
+      "learning_rate": 4.41326164874552e-05,
+      "loss": 0.3287,
+      "step": 31870
+    },
+    {
+      "epoch": 205.68,
+      "learning_rate": 4.4129032258064516e-05,
+      "loss": 0.3148,
+      "step": 31880
+    },
+    {
+      "epoch": 205.74,
+      "learning_rate": 4.412544802867384e-05,
+      "loss": 0.3567,
+      "step": 31890
+    },
+    {
+      "epoch": 205.8,
+      "learning_rate": 4.412186379928316e-05,
+      "loss": 0.3123,
+      "step": 31900
+    },
+    {
+      "epoch": 205.87,
+      "learning_rate": 4.411827956989247e-05,
+      "loss": 0.3419,
+      "step": 31910
+    },
+    {
+      "epoch": 205.93,
+      "learning_rate": 4.4114695340501796e-05,
+      "loss": 0.3066,
+      "step": 31920
+    },
+    {
+      "epoch": 206.0,
+      "learning_rate": 4.4111111111111114e-05,
+      "loss": 0.3232,
+      "step": 31930
+    },
+    {
+      "epoch": 206.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8832225799560547,
+      "eval_runtime": 0.858,
+      "eval_samples_per_second": 234.268,
+      "eval_steps_per_second": 8.159,
+      "step": 31930
+    },
+    {
+      "epoch": 206.06,
+      "learning_rate": 4.410752688172043e-05,
+      "loss": 0.3321,
+      "step": 31940
+    },
+    {
+      "epoch": 206.13,
+      "learning_rate": 4.410394265232975e-05,
+      "loss": 0.3112,
+      "step": 31950
+    },
+    {
+      "epoch": 206.19,
+      "learning_rate": 4.410035842293907e-05,
+      "loss": 0.2546,
+      "step": 31960
+    },
+    {
+      "epoch": 206.26,
+      "learning_rate": 4.409677419354839e-05,
+      "loss": 0.2505,
+      "step": 31970
+    },
+    {
+      "epoch": 206.32,
+      "learning_rate": 4.409318996415771e-05,
+      "loss": 0.3048,
+      "step": 31980
+    },
+    {
+      "epoch": 206.39,
+      "learning_rate": 4.4089605734767023e-05,
+      "loss": 0.2721,
+      "step": 31990
+    },
+    {
+      "epoch": 206.45,
+      "learning_rate": 4.408602150537635e-05,
+      "loss": 0.2999,
+      "step": 32000
+    },
+    {
+      "epoch": 206.51,
+      "learning_rate": 4.4082437275985667e-05,
+      "loss": 0.3211,
+      "step": 32010
+    },
+    {
+      "epoch": 206.58,
+      "learning_rate": 4.4078853046594985e-05,
+      "loss": 0.2907,
+      "step": 32020
+    },
+    {
+      "epoch": 206.64,
+      "learning_rate": 4.40752688172043e-05,
+      "loss": 0.3561,
+      "step": 32030
+    },
+    {
+      "epoch": 206.71,
+      "learning_rate": 4.407168458781362e-05,
+      "loss": 0.3159,
+      "step": 32040
+    },
+    {
+      "epoch": 206.77,
+      "learning_rate": 4.406810035842294e-05,
+      "loss": 0.3052,
+      "step": 32050
+    },
+    {
+      "epoch": 206.84,
+      "learning_rate": 4.4064516129032264e-05,
+      "loss": 0.2859,
+      "step": 32060
+    },
+    {
+      "epoch": 206.9,
+      "learning_rate": 4.4060931899641576e-05,
+      "loss": 0.3236,
+      "step": 32070
+    },
+    {
+      "epoch": 206.96,
+      "learning_rate": 4.40573476702509e-05,
+      "loss": 0.2925,
+      "step": 32080
+    },
+    {
+      "epoch": 207.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9514356851577759,
+      "eval_runtime": 0.8614,
+      "eval_samples_per_second": 233.333,
+      "eval_steps_per_second": 8.126,
+      "step": 32085
+    },
+    {
+      "epoch": 207.03,
+      "learning_rate": 4.405376344086022e-05,
+      "loss": 0.3373,
+      "step": 32090
+    },
+    {
+      "epoch": 207.1,
+      "learning_rate": 4.405017921146954e-05,
+      "loss": 0.3288,
+      "step": 32100
+    },
+    {
+      "epoch": 207.16,
+      "learning_rate": 4.4046594982078855e-05,
+      "loss": 0.3429,
+      "step": 32110
+    },
+    {
+      "epoch": 207.23,
+      "learning_rate": 4.4043010752688174e-05,
+      "loss": 0.2771,
+      "step": 32120
+    },
+    {
+      "epoch": 207.29,
+      "learning_rate": 4.403942652329749e-05,
+      "loss": 0.2966,
+      "step": 32130
+    },
+    {
+      "epoch": 207.35,
+      "learning_rate": 4.403584229390681e-05,
+      "loss": 0.2729,
+      "step": 32140
+    },
+    {
+      "epoch": 207.42,
+      "learning_rate": 4.403225806451613e-05,
+      "loss": 0.3222,
+      "step": 32150
+    },
+    {
+      "epoch": 207.48,
+      "learning_rate": 4.402867383512545e-05,
+      "loss": 0.3268,
+      "step": 32160
+    },
+    {
+      "epoch": 207.55,
+      "learning_rate": 4.402508960573477e-05,
+      "loss": 0.3303,
+      "step": 32170
+    },
+    {
+      "epoch": 207.61,
+      "learning_rate": 4.402150537634408e-05,
+      "loss": 0.3028,
+      "step": 32180
+    },
+    {
+      "epoch": 207.68,
+      "learning_rate": 4.401792114695341e-05,
+      "loss": 0.2721,
+      "step": 32190
+    },
+    {
+      "epoch": 207.74,
+      "learning_rate": 4.4014336917562726e-05,
+      "loss": 0.3355,
+      "step": 32200
+    },
+    {
+      "epoch": 207.8,
+      "learning_rate": 4.4010752688172044e-05,
+      "loss": 0.2419,
+      "step": 32210
+    },
+    {
+      "epoch": 207.87,
+      "learning_rate": 4.400716845878136e-05,
+      "loss": 0.3458,
+      "step": 32220
+    },
+    {
+      "epoch": 207.93,
+      "learning_rate": 4.400358422939068e-05,
+      "loss": 0.3389,
+      "step": 32230
+    },
+    {
+      "epoch": 208.0,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.3318,
+      "step": 32240
+    },
+    {
+      "epoch": 208.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9700671434402466,
+      "eval_runtime": 0.8628,
+      "eval_samples_per_second": 232.95,
+      "eval_steps_per_second": 8.113,
+      "step": 32240
+    },
+    {
+      "epoch": 208.06,
+      "learning_rate": 4.3996415770609324e-05,
+      "loss": 0.2785,
+      "step": 32250
+    },
+    {
+      "epoch": 208.13,
+      "learning_rate": 4.3992831541218635e-05,
+      "loss": 0.2769,
+      "step": 32260
+    },
+    {
+      "epoch": 208.19,
+      "learning_rate": 4.398924731182796e-05,
+      "loss": 0.289,
+      "step": 32270
+    },
+    {
+      "epoch": 208.26,
+      "learning_rate": 4.398566308243728e-05,
+      "loss": 0.3268,
+      "step": 32280
+    },
+    {
+      "epoch": 208.32,
+      "learning_rate": 4.39820788530466e-05,
+      "loss": 0.3142,
+      "step": 32290
+    },
+    {
+      "epoch": 208.39,
+      "learning_rate": 4.3978494623655915e-05,
+      "loss": 0.3233,
+      "step": 32300
+    },
+    {
+      "epoch": 208.45,
+      "learning_rate": 4.397491039426523e-05,
+      "loss": 0.3635,
+      "step": 32310
+    },
+    {
+      "epoch": 208.51,
+      "learning_rate": 4.397132616487455e-05,
+      "loss": 0.3274,
+      "step": 32320
+    },
+    {
+      "epoch": 208.58,
+      "learning_rate": 4.3967741935483876e-05,
+      "loss": 0.2462,
+      "step": 32330
+    },
+    {
+      "epoch": 208.64,
+      "learning_rate": 4.396415770609319e-05,
+      "loss": 0.3243,
+      "step": 32340
+    },
+    {
+      "epoch": 208.71,
+      "learning_rate": 4.396057347670251e-05,
+      "loss": 0.3014,
+      "step": 32350
+    },
+    {
+      "epoch": 208.77,
+      "learning_rate": 4.395698924731183e-05,
+      "loss": 0.2669,
+      "step": 32360
+    },
+    {
+      "epoch": 208.84,
+      "learning_rate": 4.395340501792115e-05,
+      "loss": 0.2844,
+      "step": 32370
+    },
+    {
+      "epoch": 208.9,
+      "learning_rate": 4.394982078853047e-05,
+      "loss": 0.3244,
+      "step": 32380
+    },
+    {
+      "epoch": 208.96,
+      "learning_rate": 4.3946236559139785e-05,
+      "loss": 0.2688,
+      "step": 32390
+    },
+    {
+      "epoch": 209.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9754597544670105,
+      "eval_runtime": 0.8599,
+      "eval_samples_per_second": 233.752,
+      "eval_steps_per_second": 8.141,
+      "step": 32395
+    },
+    {
+      "epoch": 209.03,
+      "learning_rate": 4.3942652329749104e-05,
+      "loss": 0.3061,
+      "step": 32400
+    },
+    {
+      "epoch": 209.1,
+      "learning_rate": 4.393906810035843e-05,
+      "loss": 0.3053,
+      "step": 32410
+    },
+    {
+      "epoch": 209.16,
+      "learning_rate": 4.393548387096774e-05,
+      "loss": 0.2742,
+      "step": 32420
+    },
+    {
+      "epoch": 209.23,
+      "learning_rate": 4.3931899641577065e-05,
+      "loss": 0.369,
+      "step": 32430
+    },
+    {
+      "epoch": 209.29,
+      "learning_rate": 4.392831541218638e-05,
+      "loss": 0.277,
+      "step": 32440
+    },
+    {
+      "epoch": 209.35,
+      "learning_rate": 4.39247311827957e-05,
+      "loss": 0.336,
+      "step": 32450
+    },
+    {
+      "epoch": 209.42,
+      "learning_rate": 4.392114695340502e-05,
+      "loss": 0.3161,
+      "step": 32460
+    },
+    {
+      "epoch": 209.48,
+      "learning_rate": 4.391756272401434e-05,
+      "loss": 0.3704,
+      "step": 32470
+    },
+    {
+      "epoch": 209.55,
+      "learning_rate": 4.3913978494623656e-05,
+      "loss": 0.3198,
+      "step": 32480
+    },
+    {
+      "epoch": 209.61,
+      "learning_rate": 4.391039426523298e-05,
+      "loss": 0.3318,
+      "step": 32490
+    },
+    {
+      "epoch": 209.68,
+      "learning_rate": 4.390681003584229e-05,
+      "loss": 0.3027,
+      "step": 32500
+    },
+    {
+      "epoch": 209.74,
+      "learning_rate": 4.390322580645162e-05,
+      "loss": 0.2938,
+      "step": 32510
+    },
+    {
+      "epoch": 209.8,
+      "learning_rate": 4.3899641577060936e-05,
+      "loss": 0.2944,
+      "step": 32520
+    },
+    {
+      "epoch": 209.87,
+      "learning_rate": 4.3896057347670254e-05,
+      "loss": 0.2462,
+      "step": 32530
+    },
+    {
+      "epoch": 209.93,
+      "learning_rate": 4.389247311827957e-05,
+      "loss": 0.31,
+      "step": 32540
+    },
+    {
+      "epoch": 210.0,
+      "learning_rate": 4.388888888888889e-05,
+      "loss": 0.3354,
+      "step": 32550
+    },
+    {
+      "epoch": 210.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0299919843673706,
+      "eval_runtime": 0.8683,
+      "eval_samples_per_second": 231.495,
+      "eval_steps_per_second": 8.062,
+      "step": 32550
+    },
+    {
+      "epoch": 210.06,
+      "learning_rate": 4.388530465949821e-05,
+      "loss": 0.3295,
+      "step": 32560
+    },
+    {
+      "epoch": 210.13,
+      "learning_rate": 4.3881720430107534e-05,
+      "loss": 0.3135,
+      "step": 32570
+    },
+    {
+      "epoch": 210.19,
+      "learning_rate": 4.3878136200716845e-05,
+      "loss": 0.3037,
+      "step": 32580
+    },
+    {
+      "epoch": 210.26,
+      "learning_rate": 4.387455197132617e-05,
+      "loss": 0.2445,
+      "step": 32590
+    },
+    {
+      "epoch": 210.32,
+      "learning_rate": 4.387096774193549e-05,
+      "loss": 0.3604,
+      "step": 32600
+    },
+    {
+      "epoch": 210.39,
+      "learning_rate": 4.3867383512544806e-05,
+      "loss": 0.3284,
+      "step": 32610
+    },
+    {
+      "epoch": 210.45,
+      "learning_rate": 4.3863799283154125e-05,
+      "loss": 0.2886,
+      "step": 32620
+    },
+    {
+      "epoch": 210.51,
+      "learning_rate": 4.386021505376344e-05,
+      "loss": 0.261,
+      "step": 32630
+    },
+    {
+      "epoch": 210.58,
+      "learning_rate": 4.385663082437276e-05,
+      "loss": 0.2493,
+      "step": 32640
+    },
+    {
+      "epoch": 210.64,
+      "learning_rate": 4.3853046594982086e-05,
+      "loss": 0.3315,
+      "step": 32650
+    },
+    {
+      "epoch": 210.71,
+      "learning_rate": 4.38494623655914e-05,
+      "loss": 0.3426,
+      "step": 32660
+    },
+    {
+      "epoch": 210.77,
+      "learning_rate": 4.3845878136200716e-05,
+      "loss": 0.2871,
+      "step": 32670
+    },
+    {
+      "epoch": 210.84,
+      "learning_rate": 4.384229390681004e-05,
+      "loss": 0.3321,
+      "step": 32680
+    },
+    {
+      "epoch": 210.9,
+      "learning_rate": 4.383870967741936e-05,
+      "loss": 0.3256,
+      "step": 32690
+    },
+    {
+      "epoch": 210.96,
+      "learning_rate": 4.383512544802868e-05,
+      "loss": 0.3308,
+      "step": 32700
+    },
+    {
+      "epoch": 211.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0211849212646484,
+      "eval_runtime": 0.8398,
+      "eval_samples_per_second": 239.351,
+      "eval_steps_per_second": 8.336,
+      "step": 32705
+    },
+    {
+      "epoch": 211.03,
+      "learning_rate": 4.3831541218637995e-05,
+      "loss": 0.2823,
+      "step": 32710
+    },
+    {
+      "epoch": 211.1,
+      "learning_rate": 4.382795698924731e-05,
+      "loss": 0.2783,
+      "step": 32720
+    },
+    {
+      "epoch": 211.16,
+      "learning_rate": 4.382437275985664e-05,
+      "loss": 0.3311,
+      "step": 32730
+    },
+    {
+      "epoch": 211.23,
+      "learning_rate": 4.382078853046595e-05,
+      "loss": 0.2542,
+      "step": 32740
+    },
+    {
+      "epoch": 211.29,
+      "learning_rate": 4.381720430107527e-05,
+      "loss": 0.3377,
+      "step": 32750
+    },
+    {
+      "epoch": 211.35,
+      "learning_rate": 4.381362007168459e-05,
+      "loss": 0.3102,
+      "step": 32760
+    },
+    {
+      "epoch": 211.42,
+      "learning_rate": 4.381003584229391e-05,
+      "loss": 0.2747,
+      "step": 32770
+    },
+    {
+      "epoch": 211.48,
+      "learning_rate": 4.380645161290323e-05,
+      "loss": 0.321,
+      "step": 32780
+    },
+    {
+      "epoch": 211.55,
+      "learning_rate": 4.380286738351255e-05,
+      "loss": 0.286,
+      "step": 32790
+    },
+    {
+      "epoch": 211.61,
+      "learning_rate": 4.3799283154121866e-05,
+      "loss": 0.306,
+      "step": 32800
+    },
+    {
+      "epoch": 211.68,
+      "learning_rate": 4.3795698924731184e-05,
+      "loss": 0.2853,
+      "step": 32810
+    },
+    {
+      "epoch": 211.74,
+      "learning_rate": 4.37921146953405e-05,
+      "loss": 0.2704,
+      "step": 32820
+    },
+    {
+      "epoch": 211.8,
+      "learning_rate": 4.378853046594982e-05,
+      "loss": 0.3059,
+      "step": 32830
+    },
+    {
+      "epoch": 211.87,
+      "learning_rate": 4.3784946236559145e-05,
+      "loss": 0.3269,
+      "step": 32840
+    },
+    {
+      "epoch": 211.93,
+      "learning_rate": 4.378136200716846e-05,
+      "loss": 0.2978,
+      "step": 32850
+    },
+    {
+      "epoch": 212.0,
+      "learning_rate": 4.377777777777778e-05,
+      "loss": 0.2892,
+      "step": 32860
+    },
+    {
+      "epoch": 212.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0061239004135132,
+      "eval_runtime": 0.8533,
+      "eval_samples_per_second": 235.55,
+      "eval_steps_per_second": 8.203,
+      "step": 32860
+    },
+    {
+      "epoch": 212.06,
+      "learning_rate": 4.37741935483871e-05,
+      "loss": 0.2936,
+      "step": 32870
+    },
+    {
+      "epoch": 212.13,
+      "learning_rate": 4.377060931899642e-05,
+      "loss": 0.2792,
+      "step": 32880
+    },
+    {
+      "epoch": 212.19,
+      "learning_rate": 4.3767025089605736e-05,
+      "loss": 0.3028,
+      "step": 32890
+    },
+    {
+      "epoch": 212.26,
+      "learning_rate": 4.3763440860215055e-05,
+      "loss": 0.3112,
+      "step": 32900
+    },
+    {
+      "epoch": 212.32,
+      "learning_rate": 4.375985663082437e-05,
+      "loss": 0.2553,
+      "step": 32910
+    },
+    {
+      "epoch": 212.39,
+      "learning_rate": 4.37562724014337e-05,
+      "loss": 0.2962,
+      "step": 32920
+    },
+    {
+      "epoch": 212.45,
+      "learning_rate": 4.375268817204301e-05,
+      "loss": 0.3285,
+      "step": 32930
+    },
+    {
+      "epoch": 212.51,
+      "learning_rate": 4.3749103942652334e-05,
+      "loss": 0.343,
+      "step": 32940
+    },
+    {
+      "epoch": 212.58,
+      "learning_rate": 4.374551971326165e-05,
+      "loss": 0.2594,
+      "step": 32950
+    },
+    {
+      "epoch": 212.64,
+      "learning_rate": 4.374193548387097e-05,
+      "loss": 0.3001,
+      "step": 32960
+    },
+    {
+      "epoch": 212.71,
+      "learning_rate": 4.373835125448029e-05,
+      "loss": 0.2407,
+      "step": 32970
+    },
+    {
+      "epoch": 212.77,
+      "learning_rate": 4.373476702508961e-05,
+      "loss": 0.2847,
+      "step": 32980
+    },
+    {
+      "epoch": 212.84,
+      "learning_rate": 4.3731182795698925e-05,
+      "loss": 0.3341,
+      "step": 32990
+    },
+    {
+      "epoch": 212.9,
+      "learning_rate": 4.372759856630825e-05,
+      "loss": 0.3208,
+      "step": 33000
+    },
+    {
+      "epoch": 212.96,
+      "learning_rate": 4.372401433691756e-05,
+      "loss": 0.3221,
+      "step": 33010
+    },
+    {
+      "epoch": 213.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.052895426750183,
+      "eval_runtime": 0.8415,
+      "eval_samples_per_second": 238.867,
+      "eval_steps_per_second": 8.319,
+      "step": 33015
+    },
+    {
+      "epoch": 213.03,
+      "learning_rate": 4.372043010752688e-05,
+      "loss": 0.3465,
+      "step": 33020
+    },
+    {
+      "epoch": 213.1,
+      "learning_rate": 4.3716845878136205e-05,
+      "loss": 0.3071,
+      "step": 33030
+    },
+    {
+      "epoch": 213.16,
+      "learning_rate": 4.371326164874552e-05,
+      "loss": 0.3161,
+      "step": 33040
+    },
+    {
+      "epoch": 213.23,
+      "learning_rate": 4.370967741935484e-05,
+      "loss": 0.2248,
+      "step": 33050
+    },
+    {
+      "epoch": 213.29,
+      "learning_rate": 4.370609318996416e-05,
+      "loss": 0.2824,
+      "step": 33060
+    },
+    {
+      "epoch": 213.35,
+      "learning_rate": 4.370250896057348e-05,
+      "loss": 0.3121,
+      "step": 33070
+    },
+    {
+      "epoch": 213.42,
+      "learning_rate": 4.36989247311828e-05,
+      "loss": 0.2937,
+      "step": 33080
+    },
+    {
+      "epoch": 213.48,
+      "learning_rate": 4.3695340501792114e-05,
+      "loss": 0.2941,
+      "step": 33090
+    },
+    {
+      "epoch": 213.55,
+      "learning_rate": 4.369175627240143e-05,
+      "loss": 0.273,
+      "step": 33100
+    },
+    {
+      "epoch": 213.61,
+      "learning_rate": 4.368817204301076e-05,
+      "loss": 0.2715,
+      "step": 33110
+    },
+    {
+      "epoch": 213.68,
+      "learning_rate": 4.3684587813620075e-05,
+      "loss": 0.326,
+      "step": 33120
+    },
+    {
+      "epoch": 213.74,
+      "learning_rate": 4.3681003584229394e-05,
+      "loss": 0.2842,
+      "step": 33130
+    },
+    {
+      "epoch": 213.8,
+      "learning_rate": 4.367741935483871e-05,
+      "loss": 0.2953,
+      "step": 33140
+    },
+    {
+      "epoch": 213.87,
+      "learning_rate": 4.367383512544803e-05,
+      "loss": 0.3216,
+      "step": 33150
+    },
+    {
+      "epoch": 213.93,
+      "learning_rate": 4.3670250896057355e-05,
+      "loss": 0.2727,
+      "step": 33160
+    },
+    {
+      "epoch": 214.0,
+      "learning_rate": 4.3666666666666666e-05,
+      "loss": 0.3401,
+      "step": 33170
+    },
+    {
+      "epoch": 214.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 0.9682870507240295,
+      "eval_runtime": 0.8462,
+      "eval_samples_per_second": 237.543,
+      "eval_steps_per_second": 8.273,
+      "step": 33170
+    },
+    {
+      "epoch": 214.06,
+      "learning_rate": 4.3663082437275985e-05,
+      "loss": 0.2885,
+      "step": 33180
+    },
+    {
+      "epoch": 214.13,
+      "learning_rate": 4.365949820788531e-05,
+      "loss": 0.3271,
+      "step": 33190
+    },
+    {
+      "epoch": 214.19,
+      "learning_rate": 4.365591397849463e-05,
+      "loss": 0.3111,
+      "step": 33200
+    },
+    {
+      "epoch": 214.26,
+      "learning_rate": 4.3652329749103946e-05,
+      "loss": 0.3088,
+      "step": 33210
+    },
+    {
+      "epoch": 214.32,
+      "learning_rate": 4.3648745519713264e-05,
+      "loss": 0.3432,
+      "step": 33220
+    },
+    {
+      "epoch": 214.39,
+      "learning_rate": 4.364516129032258e-05,
+      "loss": 0.3129,
+      "step": 33230
+    },
+    {
+      "epoch": 214.45,
+      "learning_rate": 4.364157706093191e-05,
+      "loss": 0.2719,
+      "step": 33240
+    },
+    {
+      "epoch": 214.51,
+      "learning_rate": 4.363799283154122e-05,
+      "loss": 0.2813,
+      "step": 33250
+    },
+    {
+      "epoch": 214.58,
+      "learning_rate": 4.363440860215054e-05,
+      "loss": 0.3319,
+      "step": 33260
+    },
+    {
+      "epoch": 214.64,
+      "learning_rate": 4.363082437275986e-05,
+      "loss": 0.3313,
+      "step": 33270
+    },
+    {
+      "epoch": 214.71,
+      "learning_rate": 4.362724014336918e-05,
+      "loss": 0.2991,
+      "step": 33280
+    },
+    {
+      "epoch": 214.77,
+      "learning_rate": 4.362365591397849e-05,
+      "loss": 0.3248,
+      "step": 33290
+    },
+    {
+      "epoch": 214.84,
+      "learning_rate": 4.362007168458782e-05,
+      "loss": 0.2436,
+      "step": 33300
+    },
+    {
+      "epoch": 214.9,
+      "learning_rate": 4.3616487455197135e-05,
+      "loss": 0.3055,
+      "step": 33310
+    },
+    {
+      "epoch": 214.96,
+      "learning_rate": 4.361290322580645e-05,
+      "loss": 0.3467,
+      "step": 33320
+    },
+    {
+      "epoch": 215.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9288849234580994,
+      "eval_runtime": 0.8422,
+      "eval_samples_per_second": 238.65,
+      "eval_steps_per_second": 8.311,
+      "step": 33325
+    },
+    {
+      "epoch": 215.03,
+      "learning_rate": 4.360931899641577e-05,
+      "loss": 0.317,
+      "step": 33330
+    },
+    {
+      "epoch": 215.1,
+      "learning_rate": 4.360573476702509e-05,
+      "loss": 0.3109,
+      "step": 33340
+    },
+    {
+      "epoch": 215.16,
+      "learning_rate": 4.3602150537634414e-05,
+      "loss": 0.2717,
+      "step": 33350
+    },
+    {
+      "epoch": 215.23,
+      "learning_rate": 4.359856630824373e-05,
+      "loss": 0.3057,
+      "step": 33360
+    },
+    {
+      "epoch": 215.29,
+      "learning_rate": 4.3594982078853044e-05,
+      "loss": 0.2747,
+      "step": 33370
+    },
+    {
+      "epoch": 215.35,
+      "learning_rate": 4.359139784946237e-05,
+      "loss": 0.2718,
+      "step": 33380
+    },
+    {
+      "epoch": 215.42,
+      "learning_rate": 4.358781362007169e-05,
+      "loss": 0.3177,
+      "step": 33390
+    },
+    {
+      "epoch": 215.48,
+      "learning_rate": 4.3584229390681005e-05,
+      "loss": 0.3295,
+      "step": 33400
+    },
+    {
+      "epoch": 215.55,
+      "learning_rate": 4.3580645161290324e-05,
+      "loss": 0.3124,
+      "step": 33410
+    },
+    {
+      "epoch": 215.61,
+      "learning_rate": 4.357706093189964e-05,
+      "loss": 0.2904,
+      "step": 33420
+    },
+    {
+      "epoch": 215.68,
+      "learning_rate": 4.357347670250897e-05,
+      "loss": 0.3481,
+      "step": 33430
+    },
+    {
+      "epoch": 215.74,
+      "learning_rate": 4.356989247311828e-05,
+      "loss": 0.3326,
+      "step": 33440
+    },
+    {
+      "epoch": 215.8,
+      "learning_rate": 4.3566308243727597e-05,
+      "loss": 0.3345,
+      "step": 33450
+    },
+    {
+      "epoch": 215.87,
+      "learning_rate": 4.356272401433692e-05,
+      "loss": 0.3076,
+      "step": 33460
+    },
+    {
+      "epoch": 215.93,
+      "learning_rate": 4.355913978494624e-05,
+      "loss": 0.2905,
+      "step": 33470
+    },
+    {
+      "epoch": 216.0,
+      "learning_rate": 4.355555555555556e-05,
+      "loss": 0.2943,
+      "step": 33480
+    },
+    {
+      "epoch": 216.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.023049235343933,
+      "eval_runtime": 0.8436,
+      "eval_samples_per_second": 238.266,
+      "eval_steps_per_second": 8.298,
+      "step": 33480
+    },
+    {
+      "epoch": 216.06,
+      "learning_rate": 4.3551971326164876e-05,
+      "loss": 0.335,
+      "step": 33490
+    },
+    {
+      "epoch": 216.13,
+      "learning_rate": 4.3548387096774194e-05,
+      "loss": 0.3064,
+      "step": 33500
+    },
+    {
+      "epoch": 216.19,
+      "learning_rate": 4.354480286738352e-05,
+      "loss": 0.2732,
+      "step": 33510
+    },
+    {
+      "epoch": 216.26,
+      "learning_rate": 4.354121863799283e-05,
+      "loss": 0.2939,
+      "step": 33520
+    },
+    {
+      "epoch": 216.32,
+      "learning_rate": 4.353763440860215e-05,
+      "loss": 0.3356,
+      "step": 33530
+    },
+    {
+      "epoch": 216.39,
+      "learning_rate": 4.3534050179211474e-05,
+      "loss": 0.3077,
+      "step": 33540
+    },
+    {
+      "epoch": 216.45,
+      "learning_rate": 4.353046594982079e-05,
+      "loss": 0.324,
+      "step": 33550
+    },
+    {
+      "epoch": 216.51,
+      "learning_rate": 4.352688172043011e-05,
+      "loss": 0.3469,
+      "step": 33560
+    },
+    {
+      "epoch": 216.58,
+      "learning_rate": 4.352329749103943e-05,
+      "loss": 0.3664,
+      "step": 33570
+    },
+    {
+      "epoch": 216.64,
+      "learning_rate": 4.351971326164875e-05,
+      "loss": 0.3012,
+      "step": 33580
+    },
+    {
+      "epoch": 216.71,
+      "learning_rate": 4.351612903225807e-05,
+      "loss": 0.2246,
+      "step": 33590
+    },
+    {
+      "epoch": 216.77,
+      "learning_rate": 4.351254480286738e-05,
+      "loss": 0.3022,
+      "step": 33600
+    },
+    {
+      "epoch": 216.84,
+      "learning_rate": 4.35089605734767e-05,
+      "loss": 0.3292,
+      "step": 33610
+    },
+    {
+      "epoch": 216.9,
+      "learning_rate": 4.3505376344086026e-05,
+      "loss": 0.287,
+      "step": 33620
+    },
+    {
+      "epoch": 216.96,
+      "learning_rate": 4.3501792114695345e-05,
+      "loss": 0.326,
+      "step": 33630
+    },
+    {
+      "epoch": 217.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.957181990146637,
+      "eval_runtime": 0.8557,
+      "eval_samples_per_second": 234.907,
+      "eval_steps_per_second": 8.181,
+      "step": 33635
+    },
+    {
+      "epoch": 217.03,
+      "learning_rate": 4.3498207885304656e-05,
+      "loss": 0.254,
+      "step": 33640
+    },
+    {
+      "epoch": 217.1,
+      "learning_rate": 4.349462365591398e-05,
+      "loss": 0.3302,
+      "step": 33650
+    },
+    {
+      "epoch": 217.16,
+      "learning_rate": 4.34910394265233e-05,
+      "loss": 0.3204,
+      "step": 33660
+    },
+    {
+      "epoch": 217.23,
+      "learning_rate": 4.348745519713262e-05,
+      "loss": 0.3304,
+      "step": 33670
+    },
+    {
+      "epoch": 217.29,
+      "learning_rate": 4.3483870967741936e-05,
+      "loss": 0.2802,
+      "step": 33680
+    },
+    {
+      "epoch": 217.35,
+      "learning_rate": 4.3480286738351254e-05,
+      "loss": 0.3701,
+      "step": 33690
+    },
+    {
+      "epoch": 217.42,
+      "learning_rate": 4.347670250896058e-05,
+      "loss": 0.3284,
+      "step": 33700
+    },
+    {
+      "epoch": 217.48,
+      "learning_rate": 4.34731182795699e-05,
+      "loss": 0.278,
+      "step": 33710
+    },
+    {
+      "epoch": 217.55,
+      "learning_rate": 4.346953405017921e-05,
+      "loss": 0.2785,
+      "step": 33720
+    },
+    {
+      "epoch": 217.61,
+      "learning_rate": 4.346594982078853e-05,
+      "loss": 0.3328,
+      "step": 33730
+    },
+    {
+      "epoch": 217.68,
+      "learning_rate": 4.346236559139785e-05,
+      "loss": 0.3199,
+      "step": 33740
+    },
+    {
+      "epoch": 217.74,
+      "learning_rate": 4.345878136200717e-05,
+      "loss": 0.2606,
+      "step": 33750
+    },
+    {
+      "epoch": 217.8,
+      "learning_rate": 4.345519713261649e-05,
+      "loss": 0.3348,
+      "step": 33760
+    },
+    {
+      "epoch": 217.87,
+      "learning_rate": 4.3451612903225806e-05,
+      "loss": 0.2645,
+      "step": 33770
+    },
+    {
+      "epoch": 217.93,
+      "learning_rate": 4.344802867383513e-05,
+      "loss": 0.2941,
+      "step": 33780
+    },
+    {
+      "epoch": 218.0,
+      "learning_rate": 4.344444444444445e-05,
+      "loss": 0.3205,
+      "step": 33790
+    },
+    {
+      "epoch": 218.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.987675130367279,
+      "eval_runtime": 0.8782,
+      "eval_samples_per_second": 228.869,
+      "eval_steps_per_second": 7.971,
+      "step": 33790
+    },
+    {
+      "epoch": 218.06,
+      "learning_rate": 4.344086021505376e-05,
+      "loss": 0.3094,
+      "step": 33800
+    },
+    {
+      "epoch": 218.13,
+      "learning_rate": 4.3437275985663086e-05,
+      "loss": 0.2613,
+      "step": 33810
+    },
+    {
+      "epoch": 218.19,
+      "learning_rate": 4.3433691756272404e-05,
+      "loss": 0.3148,
+      "step": 33820
+    },
+    {
+      "epoch": 218.26,
+      "learning_rate": 4.343010752688172e-05,
+      "loss": 0.3703,
+      "step": 33830
+    },
+    {
+      "epoch": 218.32,
+      "learning_rate": 4.342652329749104e-05,
+      "loss": 0.2725,
+      "step": 33840
+    },
+    {
+      "epoch": 218.39,
+      "learning_rate": 4.342293906810036e-05,
+      "loss": 0.2644,
+      "step": 33850
+    },
+    {
+      "epoch": 218.45,
+      "learning_rate": 4.3419354838709684e-05,
+      "loss": 0.3144,
+      "step": 33860
+    },
+    {
+      "epoch": 218.51,
+      "learning_rate": 4.3415770609319e-05,
+      "loss": 0.2437,
+      "step": 33870
+    },
+    {
+      "epoch": 218.58,
+      "learning_rate": 4.341218637992831e-05,
+      "loss": 0.2916,
+      "step": 33880
+    },
+    {
+      "epoch": 218.64,
+      "learning_rate": 4.340860215053764e-05,
+      "loss": 0.3177,
+      "step": 33890
+    },
+    {
+      "epoch": 218.71,
+      "learning_rate": 4.3405017921146956e-05,
+      "loss": 0.3186,
+      "step": 33900
+    },
+    {
+      "epoch": 218.77,
+      "learning_rate": 4.3401433691756275e-05,
+      "loss": 0.3184,
+      "step": 33910
+    },
+    {
+      "epoch": 218.84,
+      "learning_rate": 4.339784946236559e-05,
+      "loss": 0.2931,
+      "step": 33920
+    },
+    {
+      "epoch": 218.9,
+      "learning_rate": 4.339426523297491e-05,
+      "loss": 0.328,
+      "step": 33930
+    },
+    {
+      "epoch": 218.96,
+      "learning_rate": 4.339068100358423e-05,
+      "loss": 0.287,
+      "step": 33940
+    },
+    {
+      "epoch": 219.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9707269668579102,
+      "eval_runtime": 0.8403,
+      "eval_samples_per_second": 239.2,
+      "eval_steps_per_second": 8.33,
+      "step": 33945
+    },
+    {
+      "epoch": 219.03,
+      "learning_rate": 4.3387096774193554e-05,
+      "loss": 0.3191,
+      "step": 33950
+    },
+    {
+      "epoch": 219.1,
+      "learning_rate": 4.3383512544802866e-05,
+      "loss": 0.3292,
+      "step": 33960
+    },
+    {
+      "epoch": 219.16,
+      "learning_rate": 4.337992831541219e-05,
+      "loss": 0.2447,
+      "step": 33970
+    },
+    {
+      "epoch": 219.23,
+      "learning_rate": 4.337634408602151e-05,
+      "loss": 0.2904,
+      "step": 33980
+    },
+    {
+      "epoch": 219.29,
+      "learning_rate": 4.337275985663083e-05,
+      "loss": 0.2823,
+      "step": 33990
+    },
+    {
+      "epoch": 219.35,
+      "learning_rate": 4.3369175627240145e-05,
+      "loss": 0.3207,
+      "step": 34000
+    },
+    {
+      "epoch": 219.42,
+      "learning_rate": 4.3365591397849463e-05,
+      "loss": 0.3596,
+      "step": 34010
+    },
+    {
+      "epoch": 219.48,
+      "learning_rate": 4.336200716845878e-05,
+      "loss": 0.3261,
+      "step": 34020
+    },
+    {
+      "epoch": 219.55,
+      "learning_rate": 4.335842293906811e-05,
+      "loss": 0.3151,
+      "step": 34030
+    },
+    {
+      "epoch": 219.61,
+      "learning_rate": 4.335483870967742e-05,
+      "loss": 0.3234,
+      "step": 34040
+    },
+    {
+      "epoch": 219.68,
+      "learning_rate": 4.335125448028674e-05,
+      "loss": 0.262,
+      "step": 34050
+    },
+    {
+      "epoch": 219.74,
+      "learning_rate": 4.334767025089606e-05,
+      "loss": 0.2931,
+      "step": 34060
+    },
+    {
+      "epoch": 219.8,
+      "learning_rate": 4.334408602150537e-05,
+      "loss": 0.3168,
+      "step": 34070
+    },
+    {
+      "epoch": 219.87,
+      "learning_rate": 4.33405017921147e-05,
+      "loss": 0.2641,
+      "step": 34080
+    },
+    {
+      "epoch": 219.93,
+      "learning_rate": 4.3336917562724016e-05,
+      "loss": 0.3132,
+      "step": 34090
+    },
+    {
+      "epoch": 220.0,
+      "learning_rate": 4.3333333333333334e-05,
+      "loss": 0.2821,
+      "step": 34100
+    },
+    {
+      "epoch": 220.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9981220960617065,
+      "eval_runtime": 0.8575,
+      "eval_samples_per_second": 234.392,
+      "eval_steps_per_second": 8.163,
+      "step": 34100
+    },
+    {
+      "epoch": 220.06,
+      "learning_rate": 4.332974910394265e-05,
+      "loss": 0.2853,
+      "step": 34110
+    },
+    {
+      "epoch": 220.13,
+      "learning_rate": 4.332616487455197e-05,
+      "loss": 0.2857,
+      "step": 34120
+    },
+    {
+      "epoch": 220.19,
+      "learning_rate": 4.3322580645161295e-05,
+      "loss": 0.3023,
+      "step": 34130
+    },
+    {
+      "epoch": 220.26,
+      "learning_rate": 4.3318996415770614e-05,
+      "loss": 0.3145,
+      "step": 34140
+    },
+    {
+      "epoch": 220.32,
+      "learning_rate": 4.3315412186379925e-05,
+      "loss": 0.2823,
+      "step": 34150
+    },
+    {
+      "epoch": 220.39,
+      "learning_rate": 4.331182795698925e-05,
+      "loss": 0.3379,
+      "step": 34160
+    },
+    {
+      "epoch": 220.45,
+      "learning_rate": 4.330824372759857e-05,
+      "loss": 0.2939,
+      "step": 34170
+    },
+    {
+      "epoch": 220.51,
+      "learning_rate": 4.3304659498207886e-05,
+      "loss": 0.3138,
+      "step": 34180
+    },
+    {
+      "epoch": 220.58,
+      "learning_rate": 4.3301075268817205e-05,
+      "loss": 0.2755,
+      "step": 34190
+    },
+    {
+      "epoch": 220.64,
+      "learning_rate": 4.329749103942652e-05,
+      "loss": 0.3515,
+      "step": 34200
+    },
+    {
+      "epoch": 220.71,
+      "learning_rate": 4.329390681003585e-05,
+      "loss": 0.3074,
+      "step": 34210
+    },
+    {
+      "epoch": 220.77,
+      "learning_rate": 4.3290322580645166e-05,
+      "loss": 0.2971,
+      "step": 34220
+    },
+    {
+      "epoch": 220.84,
+      "learning_rate": 4.328673835125448e-05,
+      "loss": 0.27,
+      "step": 34230
+    },
+    {
+      "epoch": 220.9,
+      "learning_rate": 4.32831541218638e-05,
+      "loss": 0.3153,
+      "step": 34240
+    },
+    {
+      "epoch": 220.96,
+      "learning_rate": 4.327956989247312e-05,
+      "loss": 0.3228,
+      "step": 34250
+    },
+    {
+      "epoch": 221.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9631350040435791,
+      "eval_runtime": 0.8496,
+      "eval_samples_per_second": 236.576,
+      "eval_steps_per_second": 8.239,
+      "step": 34255
+    },
+    {
+      "epoch": 221.03,
+      "learning_rate": 4.327598566308244e-05,
+      "loss": 0.3181,
+      "step": 34260
+    },
+    {
+      "epoch": 221.1,
+      "learning_rate": 4.327240143369176e-05,
+      "loss": 0.3163,
+      "step": 34270
+    },
+    {
+      "epoch": 221.16,
+      "learning_rate": 4.3268817204301075e-05,
+      "loss": 0.3043,
+      "step": 34280
+    },
+    {
+      "epoch": 221.23,
+      "learning_rate": 4.3265232974910394e-05,
+      "loss": 0.3261,
+      "step": 34290
+    },
+    {
+      "epoch": 221.29,
+      "learning_rate": 4.326164874551972e-05,
+      "loss": 0.3002,
+      "step": 34300
+    },
+    {
+      "epoch": 221.35,
+      "learning_rate": 4.325806451612903e-05,
+      "loss": 0.3052,
+      "step": 34310
+    },
+    {
+      "epoch": 221.42,
+      "learning_rate": 4.3254480286738355e-05,
+      "loss": 0.2426,
+      "step": 34320
+    },
+    {
+      "epoch": 221.48,
+      "learning_rate": 4.325089605734767e-05,
+      "loss": 0.3592,
+      "step": 34330
+    },
+    {
+      "epoch": 221.55,
+      "learning_rate": 4.324731182795699e-05,
+      "loss": 0.3092,
+      "step": 34340
+    },
+    {
+      "epoch": 221.61,
+      "learning_rate": 4.324372759856631e-05,
+      "loss": 0.2695,
+      "step": 34350
+    },
+    {
+      "epoch": 221.68,
+      "learning_rate": 4.324014336917563e-05,
+      "loss": 0.2746,
+      "step": 34360
+    },
+    {
+      "epoch": 221.74,
+      "learning_rate": 4.3236559139784946e-05,
+      "loss": 0.3442,
+      "step": 34370
+    },
+    {
+      "epoch": 221.8,
+      "learning_rate": 4.323297491039427e-05,
+      "loss": 0.3194,
+      "step": 34380
+    },
+    {
+      "epoch": 221.87,
+      "learning_rate": 4.322939068100358e-05,
+      "loss": 0.3137,
+      "step": 34390
+    },
+    {
+      "epoch": 221.93,
+      "learning_rate": 4.322580645161291e-05,
+      "loss": 0.3475,
+      "step": 34400
+    },
+    {
+      "epoch": 222.0,
+      "learning_rate": 4.3222222222222226e-05,
+      "loss": 0.3194,
+      "step": 34410
+    },
+    {
+      "epoch": 222.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9405515789985657,
+      "eval_runtime": 0.8545,
+      "eval_samples_per_second": 235.216,
+      "eval_steps_per_second": 8.192,
+      "step": 34410
+    },
+    {
+      "epoch": 222.06,
+      "learning_rate": 4.3218637992831544e-05,
+      "loss": 0.3163,
+      "step": 34420
+    },
+    {
+      "epoch": 222.13,
+      "learning_rate": 4.321505376344086e-05,
+      "loss": 0.2336,
+      "step": 34430
+    },
+    {
+      "epoch": 222.19,
+      "learning_rate": 4.321146953405018e-05,
+      "loss": 0.2734,
+      "step": 34440
+    },
+    {
+      "epoch": 222.26,
+      "learning_rate": 4.32078853046595e-05,
+      "loss": 0.3276,
+      "step": 34450
+    },
+    {
+      "epoch": 222.32,
+      "learning_rate": 4.320430107526882e-05,
+      "loss": 0.2582,
+      "step": 34460
+    },
+    {
+      "epoch": 222.39,
+      "learning_rate": 4.3200716845878135e-05,
+      "loss": 0.3673,
+      "step": 34470
+    },
+    {
+      "epoch": 222.45,
+      "learning_rate": 4.319713261648746e-05,
+      "loss": 0.2826,
+      "step": 34480
+    },
+    {
+      "epoch": 222.51,
+      "learning_rate": 4.319354838709678e-05,
+      "loss": 0.3201,
+      "step": 34490
+    },
+    {
+      "epoch": 222.58,
+      "learning_rate": 4.3189964157706096e-05,
+      "loss": 0.2886,
+      "step": 34500
+    },
+    {
+      "epoch": 222.64,
+      "learning_rate": 4.3186379928315414e-05,
+      "loss": 0.2576,
+      "step": 34510
+    },
+    {
+      "epoch": 222.71,
+      "learning_rate": 4.318279569892473e-05,
+      "loss": 0.2686,
+      "step": 34520
+    },
+    {
+      "epoch": 222.77,
+      "learning_rate": 4.317921146953405e-05,
+      "loss": 0.294,
+      "step": 34530
+    },
+    {
+      "epoch": 222.84,
+      "learning_rate": 4.3175627240143376e-05,
+      "loss": 0.2946,
+      "step": 34540
+    },
+    {
+      "epoch": 222.9,
+      "learning_rate": 4.317204301075269e-05,
+      "loss": 0.2769,
+      "step": 34550
+    },
+    {
+      "epoch": 222.96,
+      "learning_rate": 4.316845878136201e-05,
+      "loss": 0.2995,
+      "step": 34560
+    },
+    {
+      "epoch": 223.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9604700803756714,
+      "eval_runtime": 0.8617,
+      "eval_samples_per_second": 233.262,
+      "eval_steps_per_second": 8.124,
+      "step": 34565
+    },
+    {
+      "epoch": 223.03,
+      "learning_rate": 4.316487455197133e-05,
+      "loss": 0.3031,
+      "step": 34570
+    },
+    {
+      "epoch": 223.1,
+      "learning_rate": 4.316129032258065e-05,
+      "loss": 0.3081,
+      "step": 34580
+    },
+    {
+      "epoch": 223.16,
+      "learning_rate": 4.315770609318997e-05,
+      "loss": 0.2956,
+      "step": 34590
+    },
+    {
+      "epoch": 223.23,
+      "learning_rate": 4.3154121863799285e-05,
+      "loss": 0.2912,
+      "step": 34600
+    },
+    {
+      "epoch": 223.29,
+      "learning_rate": 4.31505376344086e-05,
+      "loss": 0.3158,
+      "step": 34610
+    },
+    {
+      "epoch": 223.35,
+      "learning_rate": 4.314695340501793e-05,
+      "loss": 0.303,
+      "step": 34620
+    },
+    {
+      "epoch": 223.42,
+      "learning_rate": 4.314336917562724e-05,
+      "loss": 0.326,
+      "step": 34630
+    },
+    {
+      "epoch": 223.48,
+      "learning_rate": 4.313978494623656e-05,
+      "loss": 0.2949,
+      "step": 34640
+    },
+    {
+      "epoch": 223.55,
+      "learning_rate": 4.313620071684588e-05,
+      "loss": 0.2825,
+      "step": 34650
+    },
+    {
+      "epoch": 223.61,
+      "learning_rate": 4.31326164874552e-05,
+      "loss": 0.3408,
+      "step": 34660
+    },
+    {
+      "epoch": 223.68,
+      "learning_rate": 4.312903225806452e-05,
+      "loss": 0.2761,
+      "step": 34670
+    },
+    {
+      "epoch": 223.74,
+      "learning_rate": 4.312544802867384e-05,
+      "loss": 0.3188,
+      "step": 34680
+    },
+    {
+      "epoch": 223.8,
+      "learning_rate": 4.3121863799283156e-05,
+      "loss": 0.329,
+      "step": 34690
+    },
+    {
+      "epoch": 223.87,
+      "learning_rate": 4.3118279569892474e-05,
+      "loss": 0.2647,
+      "step": 34700
+    },
+    {
+      "epoch": 223.93,
+      "learning_rate": 4.311469534050179e-05,
+      "loss": 0.3174,
+      "step": 34710
+    },
+    {
+      "epoch": 224.0,
+      "learning_rate": 4.311111111111111e-05,
+      "loss": 0.3162,
+      "step": 34720
+    },
+    {
+      "epoch": 224.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9422999620437622,
+      "eval_runtime": 0.8552,
+      "eval_samples_per_second": 235.04,
+      "eval_steps_per_second": 8.185,
+      "step": 34720
+    },
+    {
+      "epoch": 224.06,
+      "learning_rate": 4.3107526881720435e-05,
+      "loss": 0.3736,
+      "step": 34730
+    },
+    {
+      "epoch": 224.13,
+      "learning_rate": 4.3103942652329747e-05,
+      "loss": 0.3459,
+      "step": 34740
+    },
+    {
+      "epoch": 224.19,
+      "learning_rate": 4.310035842293907e-05,
+      "loss": 0.3058,
+      "step": 34750
+    },
+    {
+      "epoch": 224.26,
+      "learning_rate": 4.309677419354839e-05,
+      "loss": 0.293,
+      "step": 34760
+    },
+    {
+      "epoch": 224.32,
+      "learning_rate": 4.309318996415771e-05,
+      "loss": 0.3214,
+      "step": 34770
+    },
+    {
+      "epoch": 224.39,
+      "learning_rate": 4.3089605734767026e-05,
+      "loss": 0.3266,
+      "step": 34780
+    },
+    {
+      "epoch": 224.45,
+      "learning_rate": 4.3086021505376344e-05,
+      "loss": 0.2964,
+      "step": 34790
+    },
+    {
+      "epoch": 224.51,
+      "learning_rate": 4.308243727598566e-05,
+      "loss": 0.3023,
+      "step": 34800
+    },
+    {
+      "epoch": 224.58,
+      "learning_rate": 4.307885304659499e-05,
+      "loss": 0.2929,
+      "step": 34810
+    },
+    {
+      "epoch": 224.64,
+      "learning_rate": 4.30752688172043e-05,
+      "loss": 0.3474,
+      "step": 34820
+    },
+    {
+      "epoch": 224.71,
+      "learning_rate": 4.3071684587813624e-05,
+      "loss": 0.3393,
+      "step": 34830
+    },
+    {
+      "epoch": 224.77,
+      "learning_rate": 4.306810035842294e-05,
+      "loss": 0.2799,
+      "step": 34840
+    },
+    {
+      "epoch": 224.84,
+      "learning_rate": 4.306451612903226e-05,
+      "loss": 0.3203,
+      "step": 34850
+    },
+    {
+      "epoch": 224.9,
+      "learning_rate": 4.306093189964158e-05,
+      "loss": 0.2721,
+      "step": 34860
+    },
+    {
+      "epoch": 224.96,
+      "learning_rate": 4.30573476702509e-05,
+      "loss": 0.2861,
+      "step": 34870
+    },
+    {
+      "epoch": 225.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0068203210830688,
+      "eval_runtime": 0.8521,
+      "eval_samples_per_second": 235.893,
+      "eval_steps_per_second": 8.215,
+      "step": 34875
+    },
+    {
+      "epoch": 225.03,
+      "learning_rate": 4.3053763440860215e-05,
+      "loss": 0.3276,
+      "step": 34880
+    },
+    {
+      "epoch": 225.1,
+      "learning_rate": 4.305017921146954e-05,
+      "loss": 0.2677,
+      "step": 34890
+    },
+    {
+      "epoch": 225.16,
+      "learning_rate": 4.304659498207885e-05,
+      "loss": 0.2823,
+      "step": 34900
+    },
+    {
+      "epoch": 225.23,
+      "learning_rate": 4.3043010752688176e-05,
+      "loss": 0.3087,
+      "step": 34910
+    },
+    {
+      "epoch": 225.29,
+      "learning_rate": 4.3039426523297495e-05,
+      "loss": 0.3649,
+      "step": 34920
+    },
+    {
+      "epoch": 225.35,
+      "learning_rate": 4.303584229390681e-05,
+      "loss": 0.3031,
+      "step": 34930
+    },
+    {
+      "epoch": 225.42,
+      "learning_rate": 4.303225806451613e-05,
+      "loss": 0.309,
+      "step": 34940
+    },
+    {
+      "epoch": 225.48,
+      "learning_rate": 4.302867383512545e-05,
+      "loss": 0.3383,
+      "step": 34950
+    },
+    {
+      "epoch": 225.55,
+      "learning_rate": 4.302508960573477e-05,
+      "loss": 0.2932,
+      "step": 34960
+    },
+    {
+      "epoch": 225.61,
+      "learning_rate": 4.302150537634409e-05,
+      "loss": 0.313,
+      "step": 34970
+    },
+    {
+      "epoch": 225.68,
+      "learning_rate": 4.3017921146953404e-05,
+      "loss": 0.2896,
+      "step": 34980
+    },
+    {
+      "epoch": 225.74,
+      "learning_rate": 4.301433691756272e-05,
+      "loss": 0.3,
+      "step": 34990
+    },
+    {
+      "epoch": 225.8,
+      "learning_rate": 4.301075268817205e-05,
+      "loss": 0.2594,
+      "step": 35000
+    },
+    {
+      "epoch": 225.87,
+      "learning_rate": 4.3007168458781365e-05,
+      "loss": 0.2933,
+      "step": 35010
+    },
+    {
+      "epoch": 225.93,
+      "learning_rate": 4.3003584229390683e-05,
+      "loss": 0.2359,
+      "step": 35020
+    },
+    {
+      "epoch": 226.0,
+      "learning_rate": 4.3e-05,
+      "loss": 0.3032,
+      "step": 35030
+    },
+    {
+      "epoch": 226.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9880895614624023,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 233.112,
+      "eval_steps_per_second": 8.118,
+      "step": 35030
+    },
+    {
+      "epoch": 226.06,
+      "learning_rate": 4.299641577060932e-05,
+      "loss": 0.2995,
+      "step": 35040
+    },
+    {
+      "epoch": 226.13,
+      "learning_rate": 4.2992831541218645e-05,
+      "loss": 0.2727,
+      "step": 35050
+    },
+    {
+      "epoch": 226.19,
+      "learning_rate": 4.2989247311827956e-05,
+      "loss": 0.2784,
+      "step": 35060
+    },
+    {
+      "epoch": 226.26,
+      "learning_rate": 4.2985663082437274e-05,
+      "loss": 0.3076,
+      "step": 35070
+    },
+    {
+      "epoch": 226.32,
+      "learning_rate": 4.29820788530466e-05,
+      "loss": 0.2852,
+      "step": 35080
+    },
+    {
+      "epoch": 226.39,
+      "learning_rate": 4.297849462365592e-05,
+      "loss": 0.2785,
+      "step": 35090
+    },
+    {
+      "epoch": 226.45,
+      "learning_rate": 4.2974910394265236e-05,
+      "loss": 0.3204,
+      "step": 35100
+    },
+    {
+      "epoch": 226.51,
+      "learning_rate": 4.2971326164874554e-05,
+      "loss": 0.2682,
+      "step": 35110
+    },
+    {
+      "epoch": 226.58,
+      "learning_rate": 4.296774193548387e-05,
+      "loss": 0.2782,
+      "step": 35120
+    },
+    {
+      "epoch": 226.64,
+      "learning_rate": 4.29641577060932e-05,
+      "loss": 0.2667,
+      "step": 35130
+    },
+    {
+      "epoch": 226.71,
+      "learning_rate": 4.296057347670251e-05,
+      "loss": 0.2867,
+      "step": 35140
+    },
+    {
+      "epoch": 226.77,
+      "learning_rate": 4.295698924731183e-05,
+      "loss": 0.3045,
+      "step": 35150
+    },
+    {
+      "epoch": 226.84,
+      "learning_rate": 4.295340501792115e-05,
+      "loss": 0.3066,
+      "step": 35160
+    },
+    {
+      "epoch": 226.9,
+      "learning_rate": 4.294982078853047e-05,
+      "loss": 0.3109,
+      "step": 35170
+    },
+    {
+      "epoch": 226.96,
+      "learning_rate": 4.294623655913979e-05,
+      "loss": 0.2871,
+      "step": 35180
+    },
+    {
+      "epoch": 227.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0278968811035156,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.386,
+      "eval_steps_per_second": 8.058,
+      "step": 35185
+    },
+    {
+      "epoch": 227.03,
+      "learning_rate": 4.2942652329749106e-05,
+      "loss": 0.3219,
+      "step": 35190
+    },
+    {
+      "epoch": 227.1,
+      "learning_rate": 4.2939068100358425e-05,
+      "loss": 0.2592,
+      "step": 35200
+    },
+    {
+      "epoch": 227.16,
+      "learning_rate": 4.293548387096775e-05,
+      "loss": 0.2737,
+      "step": 35210
+    },
+    {
+      "epoch": 227.23,
+      "learning_rate": 4.293189964157706e-05,
+      "loss": 0.2745,
+      "step": 35220
+    },
+    {
+      "epoch": 227.29,
+      "learning_rate": 4.292831541218638e-05,
+      "loss": 0.3153,
+      "step": 35230
+    },
+    {
+      "epoch": 227.35,
+      "learning_rate": 4.2924731182795704e-05,
+      "loss": 0.3208,
+      "step": 35240
+    },
+    {
+      "epoch": 227.42,
+      "learning_rate": 4.292114695340502e-05,
+      "loss": 0.282,
+      "step": 35250
+    },
+    {
+      "epoch": 227.48,
+      "learning_rate": 4.291756272401434e-05,
+      "loss": 0.2882,
+      "step": 35260
+    },
+    {
+      "epoch": 227.55,
+      "learning_rate": 4.291397849462366e-05,
+      "loss": 0.3102,
+      "step": 35270
+    },
+    {
+      "epoch": 227.61,
+      "learning_rate": 4.291039426523298e-05,
+      "loss": 0.288,
+      "step": 35280
+    },
+    {
+      "epoch": 227.68,
+      "learning_rate": 4.2906810035842295e-05,
+      "loss": 0.3094,
+      "step": 35290
+    },
+    {
+      "epoch": 227.74,
+      "learning_rate": 4.2903225806451614e-05,
+      "loss": 0.3214,
+      "step": 35300
+    },
+    {
+      "epoch": 227.8,
+      "learning_rate": 4.289964157706093e-05,
+      "loss": 0.3005,
+      "step": 35310
+    },
+    {
+      "epoch": 227.87,
+      "learning_rate": 4.289605734767026e-05,
+      "loss": 0.3513,
+      "step": 35320
+    },
+    {
+      "epoch": 227.93,
+      "learning_rate": 4.2892473118279575e-05,
+      "loss": 0.2744,
+      "step": 35330
+    },
+    {
+      "epoch": 228.0,
+      "learning_rate": 4.2888888888888886e-05,
+      "loss": 0.2779,
+      "step": 35340
+    },
+    {
+      "epoch": 228.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9253153800964355,
+      "eval_runtime": 0.8608,
+      "eval_samples_per_second": 233.502,
+      "eval_steps_per_second": 8.132,
+      "step": 35340
+    },
+    {
+      "epoch": 228.06,
+      "learning_rate": 4.288530465949821e-05,
+      "loss": 0.3304,
+      "step": 35350
+    },
+    {
+      "epoch": 228.13,
+      "learning_rate": 4.288172043010753e-05,
+      "loss": 0.2769,
+      "step": 35360
+    },
+    {
+      "epoch": 228.19,
+      "learning_rate": 4.287813620071685e-05,
+      "loss": 0.2725,
+      "step": 35370
+    },
+    {
+      "epoch": 228.26,
+      "learning_rate": 4.2874551971326166e-05,
+      "loss": 0.3175,
+      "step": 35380
+    },
+    {
+      "epoch": 228.32,
+      "learning_rate": 4.2870967741935484e-05,
+      "loss": 0.2883,
+      "step": 35390
+    },
+    {
+      "epoch": 228.39,
+      "learning_rate": 4.286738351254481e-05,
+      "loss": 0.3259,
+      "step": 35400
+    },
+    {
+      "epoch": 228.45,
+      "learning_rate": 4.286379928315412e-05,
+      "loss": 0.2636,
+      "step": 35410
+    },
+    {
+      "epoch": 228.51,
+      "learning_rate": 4.286021505376344e-05,
+      "loss": 0.3216,
+      "step": 35420
+    },
+    {
+      "epoch": 228.58,
+      "learning_rate": 4.2856630824372764e-05,
+      "loss": 0.3039,
+      "step": 35430
+    },
+    {
+      "epoch": 228.64,
+      "learning_rate": 4.285304659498208e-05,
+      "loss": 0.2871,
+      "step": 35440
+    },
+    {
+      "epoch": 228.71,
+      "learning_rate": 4.28494623655914e-05,
+      "loss": 0.2808,
+      "step": 35450
+    },
+    {
+      "epoch": 228.77,
+      "learning_rate": 4.284587813620072e-05,
+      "loss": 0.2813,
+      "step": 35460
+    },
+    {
+      "epoch": 228.84,
+      "learning_rate": 4.2842293906810037e-05,
+      "loss": 0.2705,
+      "step": 35470
+    },
+    {
+      "epoch": 228.9,
+      "learning_rate": 4.283870967741936e-05,
+      "loss": 0.2631,
+      "step": 35480
+    },
+    {
+      "epoch": 228.96,
+      "learning_rate": 4.283512544802867e-05,
+      "loss": 0.3245,
+      "step": 35490
+    },
+    {
+      "epoch": 229.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9292686581611633,
+      "eval_runtime": 0.8682,
+      "eval_samples_per_second": 231.51,
+      "eval_steps_per_second": 8.063,
+      "step": 35495
+    },
+    {
+      "epoch": 229.03,
+      "learning_rate": 4.283154121863799e-05,
+      "loss": 0.3127,
+      "step": 35500
+    },
+    {
+      "epoch": 229.1,
+      "learning_rate": 4.2827956989247316e-05,
+      "loss": 0.3083,
+      "step": 35510
+    },
+    {
+      "epoch": 229.16,
+      "learning_rate": 4.2824372759856634e-05,
+      "loss": 0.3643,
+      "step": 35520
+    },
+    {
+      "epoch": 229.23,
+      "learning_rate": 4.282078853046595e-05,
+      "loss": 0.2764,
+      "step": 35530
+    },
+    {
+      "epoch": 229.29,
+      "learning_rate": 4.281720430107527e-05,
+      "loss": 0.2894,
+      "step": 35540
+    },
+    {
+      "epoch": 229.35,
+      "learning_rate": 4.281362007168459e-05,
+      "loss": 0.3094,
+      "step": 35550
+    },
+    {
+      "epoch": 229.42,
+      "learning_rate": 4.2810035842293914e-05,
+      "loss": 0.3311,
+      "step": 35560
+    },
+    {
+      "epoch": 229.48,
+      "learning_rate": 4.2806451612903225e-05,
+      "loss": 0.2774,
+      "step": 35570
+    },
+    {
+      "epoch": 229.55,
+      "learning_rate": 4.2802867383512544e-05,
+      "loss": 0.309,
+      "step": 35580
+    },
+    {
+      "epoch": 229.61,
+      "learning_rate": 4.279928315412187e-05,
+      "loss": 0.2699,
+      "step": 35590
+    },
+    {
+      "epoch": 229.68,
+      "learning_rate": 4.279569892473119e-05,
+      "loss": 0.2919,
+      "step": 35600
+    },
+    {
+      "epoch": 229.74,
+      "learning_rate": 4.2792114695340505e-05,
+      "loss": 0.3005,
+      "step": 35610
+    },
+    {
+      "epoch": 229.8,
+      "learning_rate": 4.278853046594982e-05,
+      "loss": 0.2764,
+      "step": 35620
+    },
+    {
+      "epoch": 229.87,
+      "learning_rate": 4.278494623655914e-05,
+      "loss": 0.3017,
+      "step": 35630
+    },
+    {
+      "epoch": 229.93,
+      "learning_rate": 4.278136200716846e-05,
+      "loss": 0.2559,
+      "step": 35640
+    },
+    {
+      "epoch": 230.0,
+      "learning_rate": 4.277777777777778e-05,
+      "loss": 0.3067,
+      "step": 35650
+    },
+    {
+      "epoch": 230.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9739165902137756,
+      "eval_runtime": 0.9338,
+      "eval_samples_per_second": 215.255,
+      "eval_steps_per_second": 7.496,
+      "step": 35650
+    },
+    {
+      "epoch": 230.06,
+      "learning_rate": 4.2774193548387096e-05,
+      "loss": 0.3612,
+      "step": 35660
+    },
+    {
+      "epoch": 230.13,
+      "learning_rate": 4.277060931899642e-05,
+      "loss": 0.2671,
+      "step": 35670
+    },
+    {
+      "epoch": 230.19,
+      "learning_rate": 4.276702508960574e-05,
+      "loss": 0.2599,
+      "step": 35680
+    },
+    {
+      "epoch": 230.26,
+      "learning_rate": 4.276344086021505e-05,
+      "loss": 0.282,
+      "step": 35690
+    },
+    {
+      "epoch": 230.32,
+      "learning_rate": 4.2759856630824376e-05,
+      "loss": 0.2983,
+      "step": 35700
+    },
+    {
+      "epoch": 230.39,
+      "learning_rate": 4.2756272401433694e-05,
+      "loss": 0.2529,
+      "step": 35710
+    },
+    {
+      "epoch": 230.45,
+      "learning_rate": 4.275268817204301e-05,
+      "loss": 0.2801,
+      "step": 35720
+    },
+    {
+      "epoch": 230.51,
+      "learning_rate": 4.274910394265233e-05,
+      "loss": 0.2489,
+      "step": 35730
+    },
+    {
+      "epoch": 230.58,
+      "learning_rate": 4.274551971326165e-05,
+      "loss": 0.3862,
+      "step": 35740
+    },
+    {
+      "epoch": 230.64,
+      "learning_rate": 4.2741935483870973e-05,
+      "loss": 0.2686,
+      "step": 35750
+    },
+    {
+      "epoch": 230.71,
+      "learning_rate": 4.273835125448029e-05,
+      "loss": 0.3109,
+      "step": 35760
+    },
+    {
+      "epoch": 230.77,
+      "learning_rate": 4.27347670250896e-05,
+      "loss": 0.2864,
+      "step": 35770
+    },
+    {
+      "epoch": 230.84,
+      "learning_rate": 4.273118279569893e-05,
+      "loss": 0.2656,
+      "step": 35780
+    },
+    {
+      "epoch": 230.9,
+      "learning_rate": 4.2727598566308246e-05,
+      "loss": 0.2835,
+      "step": 35790
+    },
+    {
+      "epoch": 230.96,
+      "learning_rate": 4.2724014336917564e-05,
+      "loss": 0.2536,
+      "step": 35800
+    },
+    {
+      "epoch": 231.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9622406363487244,
+      "eval_runtime": 0.8644,
+      "eval_samples_per_second": 232.535,
+      "eval_steps_per_second": 8.098,
+      "step": 35805
+    },
+    {
+      "epoch": 231.03,
+      "learning_rate": 4.272043010752688e-05,
+      "loss": 0.2733,
+      "step": 35810
+    },
+    {
+      "epoch": 231.1,
+      "learning_rate": 4.27168458781362e-05,
+      "loss": 0.3163,
+      "step": 35820
+    },
+    {
+      "epoch": 231.16,
+      "learning_rate": 4.2713261648745526e-05,
+      "loss": 0.2984,
+      "step": 35830
+    },
+    {
+      "epoch": 231.23,
+      "learning_rate": 4.2709677419354844e-05,
+      "loss": 0.3186,
+      "step": 35840
+    },
+    {
+      "epoch": 231.29,
+      "learning_rate": 4.2706093189964155e-05,
+      "loss": 0.2963,
+      "step": 35850
+    },
+    {
+      "epoch": 231.35,
+      "learning_rate": 4.270250896057348e-05,
+      "loss": 0.3072,
+      "step": 35860
+    },
+    {
+      "epoch": 231.42,
+      "learning_rate": 4.26989247311828e-05,
+      "loss": 0.29,
+      "step": 35870
+    },
+    {
+      "epoch": 231.48,
+      "learning_rate": 4.269534050179212e-05,
+      "loss": 0.3414,
+      "step": 35880
+    },
+    {
+      "epoch": 231.55,
+      "learning_rate": 4.2691756272401435e-05,
+      "loss": 0.2471,
+      "step": 35890
+    },
+    {
+      "epoch": 231.61,
+      "learning_rate": 4.268817204301075e-05,
+      "loss": 0.3216,
+      "step": 35900
+    },
+    {
+      "epoch": 231.68,
+      "learning_rate": 4.268458781362008e-05,
+      "loss": 0.3226,
+      "step": 35910
+    },
+    {
+      "epoch": 231.74,
+      "learning_rate": 4.2681003584229396e-05,
+      "loss": 0.2813,
+      "step": 35920
+    },
+    {
+      "epoch": 231.8,
+      "learning_rate": 4.267741935483871e-05,
+      "loss": 0.2739,
+      "step": 35930
+    },
+    {
+      "epoch": 231.87,
+      "learning_rate": 4.267383512544803e-05,
+      "loss": 0.297,
+      "step": 35940
+    },
+    {
+      "epoch": 231.93,
+      "learning_rate": 4.267025089605735e-05,
+      "loss": 0.3017,
+      "step": 35950
+    },
+    {
+      "epoch": 232.0,
+      "learning_rate": 4.266666666666667e-05,
+      "loss": 0.3165,
+      "step": 35960
+    },
+    {
+      "epoch": 232.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9978538155555725,
+      "eval_runtime": 0.8691,
+      "eval_samples_per_second": 231.285,
+      "eval_steps_per_second": 8.055,
+      "step": 35960
+    },
+    {
+      "epoch": 232.06,
+      "learning_rate": 4.266308243727599e-05,
+      "loss": 0.287,
+      "step": 35970
+    },
+    {
+      "epoch": 232.13,
+      "learning_rate": 4.2659498207885306e-05,
+      "loss": 0.2853,
+      "step": 35980
+    },
+    {
+      "epoch": 232.19,
+      "learning_rate": 4.2655913978494624e-05,
+      "loss": 0.3022,
+      "step": 35990
+    },
+    {
+      "epoch": 232.26,
+      "learning_rate": 4.265232974910394e-05,
+      "loss": 0.2618,
+      "step": 36000
+    },
+    {
+      "epoch": 232.32,
+      "learning_rate": 4.264874551971326e-05,
+      "loss": 0.3052,
+      "step": 36010
+    },
+    {
+      "epoch": 232.39,
+      "learning_rate": 4.2645161290322585e-05,
+      "loss": 0.3628,
+      "step": 36020
+    },
+    {
+      "epoch": 232.45,
+      "learning_rate": 4.2641577060931903e-05,
+      "loss": 0.3089,
+      "step": 36030
+    },
+    {
+      "epoch": 232.51,
+      "learning_rate": 4.2637992831541215e-05,
+      "loss": 0.3056,
+      "step": 36040
+    },
+    {
+      "epoch": 232.58,
+      "learning_rate": 4.263440860215054e-05,
+      "loss": 0.3026,
+      "step": 36050
+    },
+    {
+      "epoch": 232.64,
+      "learning_rate": 4.263082437275986e-05,
+      "loss": 0.3292,
+      "step": 36060
+    },
+    {
+      "epoch": 232.71,
+      "learning_rate": 4.2627240143369176e-05,
+      "loss": 0.3405,
+      "step": 36070
+    },
+    {
+      "epoch": 232.77,
+      "learning_rate": 4.2623655913978494e-05,
+      "loss": 0.2631,
+      "step": 36080
+    },
+    {
+      "epoch": 232.84,
+      "learning_rate": 4.262007168458781e-05,
+      "loss": 0.2641,
+      "step": 36090
+    },
+    {
+      "epoch": 232.9,
+      "learning_rate": 4.261648745519714e-05,
+      "loss": 0.2891,
+      "step": 36100
+    },
+    {
+      "epoch": 232.96,
+      "learning_rate": 4.2612903225806456e-05,
+      "loss": 0.298,
+      "step": 36110
+    },
+    {
+      "epoch": 233.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9121502041816711,
+      "eval_runtime": 0.8683,
+      "eval_samples_per_second": 231.488,
+      "eval_steps_per_second": 8.062,
+      "step": 36115
+    },
+    {
+      "epoch": 233.03,
+      "learning_rate": 4.260931899641577e-05,
+      "loss": 0.2832,
+      "step": 36120
+    },
+    {
+      "epoch": 233.1,
+      "learning_rate": 4.260573476702509e-05,
+      "loss": 0.3015,
+      "step": 36130
+    },
+    {
+      "epoch": 233.16,
+      "learning_rate": 4.260215053763441e-05,
+      "loss": 0.2822,
+      "step": 36140
+    },
+    {
+      "epoch": 233.23,
+      "learning_rate": 4.259856630824373e-05,
+      "loss": 0.2593,
+      "step": 36150
+    },
+    {
+      "epoch": 233.29,
+      "learning_rate": 4.259498207885305e-05,
+      "loss": 0.3439,
+      "step": 36160
+    },
+    {
+      "epoch": 233.35,
+      "learning_rate": 4.2591397849462365e-05,
+      "loss": 0.3162,
+      "step": 36170
+    },
+    {
+      "epoch": 233.42,
+      "learning_rate": 4.258781362007169e-05,
+      "loss": 0.263,
+      "step": 36180
+    },
+    {
+      "epoch": 233.48,
+      "learning_rate": 4.258422939068101e-05,
+      "loss": 0.2763,
+      "step": 36190
+    },
+    {
+      "epoch": 233.55,
+      "learning_rate": 4.258064516129032e-05,
+      "loss": 0.2853,
+      "step": 36200
+    },
+    {
+      "epoch": 233.61,
+      "learning_rate": 4.2577060931899645e-05,
+      "loss": 0.2632,
+      "step": 36210
+    },
+    {
+      "epoch": 233.68,
+      "learning_rate": 4.257347670250896e-05,
+      "loss": 0.2897,
+      "step": 36220
+    },
+    {
+      "epoch": 233.74,
+      "learning_rate": 4.256989247311828e-05,
+      "loss": 0.2563,
+      "step": 36230
+    },
+    {
+      "epoch": 233.8,
+      "learning_rate": 4.25663082437276e-05,
+      "loss": 0.2507,
+      "step": 36240
+    },
+    {
+      "epoch": 233.87,
+      "learning_rate": 4.256272401433692e-05,
+      "loss": 0.2885,
+      "step": 36250
+    },
+    {
+      "epoch": 233.93,
+      "learning_rate": 4.255913978494624e-05,
+      "loss": 0.3045,
+      "step": 36260
+    },
+    {
+      "epoch": 234.0,
+      "learning_rate": 4.255555555555556e-05,
+      "loss": 0.2884,
+      "step": 36270
+    },
+    {
+      "epoch": 234.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9410242438316345,
+      "eval_runtime": 0.8588,
+      "eval_samples_per_second": 234.048,
+      "eval_steps_per_second": 8.151,
+      "step": 36270
+    },
+    {
+      "epoch": 234.06,
+      "learning_rate": 4.255197132616487e-05,
+      "loss": 0.3038,
+      "step": 36280
+    },
+    {
+      "epoch": 234.13,
+      "learning_rate": 4.25483870967742e-05,
+      "loss": 0.2658,
+      "step": 36290
+    },
+    {
+      "epoch": 234.19,
+      "learning_rate": 4.2544802867383515e-05,
+      "loss": 0.2933,
+      "step": 36300
+    },
+    {
+      "epoch": 234.26,
+      "learning_rate": 4.2541218637992834e-05,
+      "loss": 0.2738,
+      "step": 36310
+    },
+    {
+      "epoch": 234.32,
+      "learning_rate": 4.253763440860215e-05,
+      "loss": 0.3108,
+      "step": 36320
+    },
+    {
+      "epoch": 234.39,
+      "learning_rate": 4.253405017921147e-05,
+      "loss": 0.3335,
+      "step": 36330
+    },
+    {
+      "epoch": 234.45,
+      "learning_rate": 4.253046594982079e-05,
+      "loss": 0.3168,
+      "step": 36340
+    },
+    {
+      "epoch": 234.51,
+      "learning_rate": 4.252688172043011e-05,
+      "loss": 0.2617,
+      "step": 36350
+    },
+    {
+      "epoch": 234.58,
+      "learning_rate": 4.2523297491039425e-05,
+      "loss": 0.2971,
+      "step": 36360
+    },
+    {
+      "epoch": 234.64,
+      "learning_rate": 4.251971326164875e-05,
+      "loss": 0.3505,
+      "step": 36370
+    },
+    {
+      "epoch": 234.71,
+      "learning_rate": 4.251612903225807e-05,
+      "loss": 0.3728,
+      "step": 36380
+    },
+    {
+      "epoch": 234.77,
+      "learning_rate": 4.2512544802867386e-05,
+      "loss": 0.2793,
+      "step": 36390
+    },
+    {
+      "epoch": 234.84,
+      "learning_rate": 4.2508960573476704e-05,
+      "loss": 0.3066,
+      "step": 36400
+    },
+    {
+      "epoch": 234.9,
+      "learning_rate": 4.250537634408602e-05,
+      "loss": 0.3139,
+      "step": 36410
+    },
+    {
+      "epoch": 234.96,
+      "learning_rate": 4.250179211469534e-05,
+      "loss": 0.3259,
+      "step": 36420
+    },
+    {
+      "epoch": 235.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9545983672142029,
+      "eval_runtime": 0.8669,
+      "eval_samples_per_second": 231.854,
+      "eval_steps_per_second": 8.075,
+      "step": 36425
+    },
+    {
+      "epoch": 235.03,
+      "learning_rate": 4.2498207885304666e-05,
+      "loss": 0.2966,
+      "step": 36430
+    },
+    {
+      "epoch": 235.1,
+      "learning_rate": 4.249462365591398e-05,
+      "loss": 0.3132,
+      "step": 36440
+    },
+    {
+      "epoch": 235.16,
+      "learning_rate": 4.24910394265233e-05,
+      "loss": 0.2775,
+      "step": 36450
+    },
+    {
+      "epoch": 235.23,
+      "learning_rate": 4.248745519713262e-05,
+      "loss": 0.2879,
+      "step": 36460
+    },
+    {
+      "epoch": 235.29,
+      "learning_rate": 4.248387096774194e-05,
+      "loss": 0.3558,
+      "step": 36470
+    },
+    {
+      "epoch": 235.35,
+      "learning_rate": 4.2480286738351257e-05,
+      "loss": 0.2462,
+      "step": 36480
+    },
+    {
+      "epoch": 235.42,
+      "learning_rate": 4.2476702508960575e-05,
+      "loss": 0.2937,
+      "step": 36490
+    },
+    {
+      "epoch": 235.48,
+      "learning_rate": 4.247311827956989e-05,
+      "loss": 0.3112,
+      "step": 36500
+    },
+    {
+      "epoch": 235.55,
+      "learning_rate": 4.246953405017922e-05,
+      "loss": 0.308,
+      "step": 36510
+    },
+    {
+      "epoch": 235.61,
+      "learning_rate": 4.246594982078853e-05,
+      "loss": 0.2875,
+      "step": 36520
+    },
+    {
+      "epoch": 235.68,
+      "learning_rate": 4.2462365591397854e-05,
+      "loss": 0.2613,
+      "step": 36530
+    },
+    {
+      "epoch": 235.74,
+      "learning_rate": 4.245878136200717e-05,
+      "loss": 0.3164,
+      "step": 36540
+    },
+    {
+      "epoch": 235.8,
+      "learning_rate": 4.245519713261649e-05,
+      "loss": 0.2567,
+      "step": 36550
+    },
+    {
+      "epoch": 235.87,
+      "learning_rate": 4.245161290322581e-05,
+      "loss": 0.2632,
+      "step": 36560
+    },
+    {
+      "epoch": 235.93,
+      "learning_rate": 4.244802867383513e-05,
+      "loss": 0.2981,
+      "step": 36570
+    },
+    {
+      "epoch": 236.0,
+      "learning_rate": 4.2444444444444445e-05,
+      "loss": 0.2656,
+      "step": 36580
+    },
+    {
+      "epoch": 236.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9543556571006775,
+      "eval_runtime": 0.8707,
+      "eval_samples_per_second": 230.837,
+      "eval_steps_per_second": 8.039,
+      "step": 36580
+    },
+    {
+      "epoch": 236.06,
+      "learning_rate": 4.244086021505377e-05,
+      "loss": 0.2972,
+      "step": 36590
+    },
+    {
+      "epoch": 236.13,
+      "learning_rate": 4.243727598566308e-05,
+      "loss": 0.3179,
+      "step": 36600
+    },
+    {
+      "epoch": 236.19,
+      "learning_rate": 4.243369175627241e-05,
+      "loss": 0.2832,
+      "step": 36610
+    },
+    {
+      "epoch": 236.26,
+      "learning_rate": 4.2430107526881725e-05,
+      "loss": 0.3179,
+      "step": 36620
+    },
+    {
+      "epoch": 236.32,
+      "learning_rate": 4.2426523297491036e-05,
+      "loss": 0.2903,
+      "step": 36630
+    },
+    {
+      "epoch": 236.39,
+      "learning_rate": 4.242293906810036e-05,
+      "loss": 0.2773,
+      "step": 36640
+    },
+    {
+      "epoch": 236.45,
+      "learning_rate": 4.241935483870968e-05,
+      "loss": 0.3066,
+      "step": 36650
+    },
+    {
+      "epoch": 236.51,
+      "learning_rate": 4.2415770609319e-05,
+      "loss": 0.3011,
+      "step": 36660
+    },
+    {
+      "epoch": 236.58,
+      "learning_rate": 4.2412186379928316e-05,
+      "loss": 0.2596,
+      "step": 36670
+    },
+    {
+      "epoch": 236.64,
+      "learning_rate": 4.2408602150537634e-05,
+      "loss": 0.276,
+      "step": 36680
+    },
+    {
+      "epoch": 236.71,
+      "learning_rate": 4.240501792114695e-05,
+      "loss": 0.3299,
+      "step": 36690
+    },
+    {
+      "epoch": 236.77,
+      "learning_rate": 4.240143369175628e-05,
+      "loss": 0.3027,
+      "step": 36700
+    },
+    {
+      "epoch": 236.84,
+      "learning_rate": 4.239784946236559e-05,
+      "loss": 0.2982,
+      "step": 36710
+    },
+    {
+      "epoch": 236.9,
+      "learning_rate": 4.2394265232974914e-05,
+      "loss": 0.3223,
+      "step": 36720
+    },
+    {
+      "epoch": 236.96,
+      "learning_rate": 4.239068100358423e-05,
+      "loss": 0.2654,
+      "step": 36730
+    },
+    {
+      "epoch": 237.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9625423550605774,
+      "eval_runtime": 0.863,
+      "eval_samples_per_second": 232.902,
+      "eval_steps_per_second": 8.111,
+      "step": 36735
+    },
+    {
+      "epoch": 237.03,
+      "learning_rate": 4.238709677419355e-05,
+      "loss": 0.321,
+      "step": 36740
+    },
+    {
+      "epoch": 237.1,
+      "learning_rate": 4.238351254480287e-05,
+      "loss": 0.3166,
+      "step": 36750
+    },
+    {
+      "epoch": 237.16,
+      "learning_rate": 4.237992831541219e-05,
+      "loss": 0.3106,
+      "step": 36760
+    },
+    {
+      "epoch": 237.23,
+      "learning_rate": 4.2376344086021505e-05,
+      "loss": 0.2913,
+      "step": 36770
+    },
+    {
+      "epoch": 237.29,
+      "learning_rate": 4.237275985663083e-05,
+      "loss": 0.2621,
+      "step": 36780
+    },
+    {
+      "epoch": 237.35,
+      "learning_rate": 4.236917562724014e-05,
+      "loss": 0.2717,
+      "step": 36790
+    },
+    {
+      "epoch": 237.42,
+      "learning_rate": 4.2365591397849466e-05,
+      "loss": 0.2658,
+      "step": 36800
+    },
+    {
+      "epoch": 237.48,
+      "learning_rate": 4.2362007168458784e-05,
+      "loss": 0.2954,
+      "step": 36810
+    },
+    {
+      "epoch": 237.55,
+      "learning_rate": 4.23584229390681e-05,
+      "loss": 0.3154,
+      "step": 36820
+    },
+    {
+      "epoch": 237.61,
+      "learning_rate": 4.235483870967742e-05,
+      "loss": 0.2996,
+      "step": 36830
+    },
+    {
+      "epoch": 237.68,
+      "learning_rate": 4.235125448028674e-05,
+      "loss": 0.2761,
+      "step": 36840
+    },
+    {
+      "epoch": 237.74,
+      "learning_rate": 4.234767025089606e-05,
+      "loss": 0.2852,
+      "step": 36850
+    },
+    {
+      "epoch": 237.8,
+      "learning_rate": 4.234408602150538e-05,
+      "loss": 0.2795,
+      "step": 36860
+    },
+    {
+      "epoch": 237.87,
+      "learning_rate": 4.2340501792114694e-05,
+      "loss": 0.2857,
+      "step": 36870
+    },
+    {
+      "epoch": 237.93,
+      "learning_rate": 4.233691756272402e-05,
+      "loss": 0.2592,
+      "step": 36880
+    },
+    {
+      "epoch": 238.0,
+      "learning_rate": 4.233333333333334e-05,
+      "loss": 0.2905,
+      "step": 36890
+    },
+    {
+      "epoch": 238.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9876997470855713,
+      "eval_runtime": 0.8554,
+      "eval_samples_per_second": 234.988,
+      "eval_steps_per_second": 8.184,
+      "step": 36890
+    },
+    {
+      "epoch": 238.06,
+      "learning_rate": 4.2329749103942655e-05,
+      "loss": 0.2573,
+      "step": 36900
+    },
+    {
+      "epoch": 238.13,
+      "learning_rate": 4.232616487455197e-05,
+      "loss": 0.3084,
+      "step": 36910
+    },
+    {
+      "epoch": 238.19,
+      "learning_rate": 4.232258064516129e-05,
+      "loss": 0.269,
+      "step": 36920
+    },
+    {
+      "epoch": 238.26,
+      "learning_rate": 4.231899641577061e-05,
+      "loss": 0.294,
+      "step": 36930
+    },
+    {
+      "epoch": 238.32,
+      "learning_rate": 4.2315412186379935e-05,
+      "loss": 0.3121,
+      "step": 36940
+    },
+    {
+      "epoch": 238.39,
+      "learning_rate": 4.2311827956989246e-05,
+      "loss": 0.3013,
+      "step": 36950
+    },
+    {
+      "epoch": 238.45,
+      "learning_rate": 4.230824372759857e-05,
+      "loss": 0.2749,
+      "step": 36960
+    },
+    {
+      "epoch": 238.51,
+      "learning_rate": 4.230465949820789e-05,
+      "loss": 0.2866,
+      "step": 36970
+    },
+    {
+      "epoch": 238.58,
+      "learning_rate": 4.230107526881721e-05,
+      "loss": 0.2621,
+      "step": 36980
+    },
+    {
+      "epoch": 238.64,
+      "learning_rate": 4.2297491039426526e-05,
+      "loss": 0.2883,
+      "step": 36990
+    },
+    {
+      "epoch": 238.71,
+      "learning_rate": 4.2293906810035844e-05,
+      "loss": 0.2478,
+      "step": 37000
+    },
+    {
+      "epoch": 238.77,
+      "learning_rate": 4.229032258064516e-05,
+      "loss": 0.2742,
+      "step": 37010
+    },
+    {
+      "epoch": 238.84,
+      "learning_rate": 4.228673835125449e-05,
+      "loss": 0.3115,
+      "step": 37020
+    },
+    {
+      "epoch": 238.9,
+      "learning_rate": 4.22831541218638e-05,
+      "loss": 0.256,
+      "step": 37030
+    },
+    {
+      "epoch": 238.96,
+      "learning_rate": 4.227956989247312e-05,
+      "loss": 0.28,
+      "step": 37040
+    },
+    {
+      "epoch": 239.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9474425911903381,
+      "eval_runtime": 0.8485,
+      "eval_samples_per_second": 236.889,
+      "eval_steps_per_second": 8.25,
+      "step": 37045
+    },
+    {
+      "epoch": 239.03,
+      "learning_rate": 4.227598566308244e-05,
+      "loss": 0.2357,
+      "step": 37050
+    },
+    {
+      "epoch": 239.1,
+      "learning_rate": 4.227240143369176e-05,
+      "loss": 0.2951,
+      "step": 37060
+    },
+    {
+      "epoch": 239.16,
+      "learning_rate": 4.226881720430108e-05,
+      "loss": 0.2959,
+      "step": 37070
+    },
+    {
+      "epoch": 239.23,
+      "learning_rate": 4.2265232974910396e-05,
+      "loss": 0.2803,
+      "step": 37080
+    },
+    {
+      "epoch": 239.29,
+      "learning_rate": 4.2261648745519715e-05,
+      "loss": 0.2423,
+      "step": 37090
+    },
+    {
+      "epoch": 239.35,
+      "learning_rate": 4.225806451612904e-05,
+      "loss": 0.3404,
+      "step": 37100
+    },
+    {
+      "epoch": 239.42,
+      "learning_rate": 4.225448028673835e-05,
+      "loss": 0.2716,
+      "step": 37110
+    },
+    {
+      "epoch": 239.48,
+      "learning_rate": 4.225089605734767e-05,
+      "loss": 0.2839,
+      "step": 37120
+    },
+    {
+      "epoch": 239.55,
+      "learning_rate": 4.2247311827956994e-05,
+      "loss": 0.3101,
+      "step": 37130
+    },
+    {
+      "epoch": 239.61,
+      "learning_rate": 4.224372759856631e-05,
+      "loss": 0.2717,
+      "step": 37140
+    },
+    {
+      "epoch": 239.68,
+      "learning_rate": 4.224014336917563e-05,
+      "loss": 0.3078,
+      "step": 37150
+    },
+    {
+      "epoch": 239.74,
+      "learning_rate": 4.223655913978495e-05,
+      "loss": 0.2763,
+      "step": 37160
+    },
+    {
+      "epoch": 239.8,
+      "learning_rate": 4.223297491039427e-05,
+      "loss": 0.3436,
+      "step": 37170
+    },
+    {
+      "epoch": 239.87,
+      "learning_rate": 4.222939068100359e-05,
+      "loss": 0.2586,
+      "step": 37180
+    },
+    {
+      "epoch": 239.93,
+      "learning_rate": 4.22258064516129e-05,
+      "loss": 0.3111,
+      "step": 37190
+    },
+    {
+      "epoch": 240.0,
+      "learning_rate": 4.222222222222222e-05,
+      "loss": 0.3256,
+      "step": 37200
+    },
+    {
+      "epoch": 240.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.944776713848114,
+      "eval_runtime": 0.8846,
+      "eval_samples_per_second": 227.221,
+      "eval_steps_per_second": 7.913,
+      "step": 37200
+    },
+    {
+      "epoch": 240.06,
+      "learning_rate": 4.2218637992831547e-05,
+      "loss": 0.2634,
+      "step": 37210
+    },
+    {
+      "epoch": 240.13,
+      "learning_rate": 4.2215053763440865e-05,
+      "loss": 0.2802,
+      "step": 37220
+    },
+    {
+      "epoch": 240.19,
+      "learning_rate": 4.221146953405018e-05,
+      "loss": 0.3203,
+      "step": 37230
+    },
+    {
+      "epoch": 240.26,
+      "learning_rate": 4.22078853046595e-05,
+      "loss": 0.292,
+      "step": 37240
+    },
+    {
+      "epoch": 240.32,
+      "learning_rate": 4.220430107526882e-05,
+      "loss": 0.2417,
+      "step": 37250
+    },
+    {
+      "epoch": 240.39,
+      "learning_rate": 4.220071684587814e-05,
+      "loss": 0.3187,
+      "step": 37260
+    },
+    {
+      "epoch": 240.45,
+      "learning_rate": 4.2197132616487456e-05,
+      "loss": 0.298,
+      "step": 37270
+    },
+    {
+      "epoch": 240.51,
+      "learning_rate": 4.2193548387096774e-05,
+      "loss": 0.298,
+      "step": 37280
+    },
+    {
+      "epoch": 240.58,
+      "learning_rate": 4.21899641577061e-05,
+      "loss": 0.2939,
+      "step": 37290
+    },
+    {
+      "epoch": 240.64,
+      "learning_rate": 4.218637992831541e-05,
+      "loss": 0.282,
+      "step": 37300
+    },
+    {
+      "epoch": 240.71,
+      "learning_rate": 4.218279569892473e-05,
+      "loss": 0.2838,
+      "step": 37310
+    },
+    {
+      "epoch": 240.77,
+      "learning_rate": 4.2179211469534054e-05,
+      "loss": 0.309,
+      "step": 37320
+    },
+    {
+      "epoch": 240.84,
+      "learning_rate": 4.217562724014337e-05,
+      "loss": 0.2462,
+      "step": 37330
+    },
+    {
+      "epoch": 240.9,
+      "learning_rate": 4.217204301075269e-05,
+      "loss": 0.3086,
+      "step": 37340
+    },
+    {
+      "epoch": 240.96,
+      "learning_rate": 4.216845878136201e-05,
+      "loss": 0.2996,
+      "step": 37350
+    },
+    {
+      "epoch": 241.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9167143702507019,
+      "eval_runtime": 0.8603,
+      "eval_samples_per_second": 233.645,
+      "eval_steps_per_second": 8.137,
+      "step": 37355
+    },
+    {
+      "epoch": 241.03,
+      "learning_rate": 4.2164874551971326e-05,
+      "loss": 0.2841,
+      "step": 37360
+    },
+    {
+      "epoch": 241.1,
+      "learning_rate": 4.216129032258065e-05,
+      "loss": 0.3095,
+      "step": 37370
+    },
+    {
+      "epoch": 241.16,
+      "learning_rate": 4.215770609318996e-05,
+      "loss": 0.3121,
+      "step": 37380
+    },
+    {
+      "epoch": 241.23,
+      "learning_rate": 4.215412186379928e-05,
+      "loss": 0.3018,
+      "step": 37390
+    },
+    {
+      "epoch": 241.29,
+      "learning_rate": 4.2150537634408606e-05,
+      "loss": 0.2756,
+      "step": 37400
+    },
+    {
+      "epoch": 241.35,
+      "learning_rate": 4.2146953405017924e-05,
+      "loss": 0.2629,
+      "step": 37410
+    },
+    {
+      "epoch": 241.42,
+      "learning_rate": 4.214336917562724e-05,
+      "loss": 0.324,
+      "step": 37420
+    },
+    {
+      "epoch": 241.48,
+      "learning_rate": 4.213978494623656e-05,
+      "loss": 0.2636,
+      "step": 37430
+    },
+    {
+      "epoch": 241.55,
+      "learning_rate": 4.213620071684588e-05,
+      "loss": 0.2897,
+      "step": 37440
+    },
+    {
+      "epoch": 241.61,
+      "learning_rate": 4.2132616487455204e-05,
+      "loss": 0.3139,
+      "step": 37450
+    },
+    {
+      "epoch": 241.68,
+      "learning_rate": 4.2129032258064515e-05,
+      "loss": 0.2857,
+      "step": 37460
+    },
+    {
+      "epoch": 241.74,
+      "learning_rate": 4.2125448028673833e-05,
+      "loss": 0.409,
+      "step": 37470
+    },
+    {
+      "epoch": 241.8,
+      "learning_rate": 4.212186379928316e-05,
+      "loss": 0.2766,
+      "step": 37480
+    },
+    {
+      "epoch": 241.87,
+      "learning_rate": 4.2118279569892477e-05,
+      "loss": 0.289,
+      "step": 37490
+    },
+    {
+      "epoch": 241.93,
+      "learning_rate": 4.2114695340501795e-05,
+      "loss": 0.3181,
+      "step": 37500
+    },
+    {
+      "epoch": 242.0,
+      "learning_rate": 4.211111111111111e-05,
+      "loss": 0.3087,
+      "step": 37510
+    },
+    {
+      "epoch": 242.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9232776165008545,
+      "eval_runtime": 0.8806,
+      "eval_samples_per_second": 228.251,
+      "eval_steps_per_second": 7.949,
+      "step": 37510
+    },
+    {
+      "epoch": 242.06,
+      "learning_rate": 4.210752688172043e-05,
+      "loss": 0.318,
+      "step": 37520
+    },
+    {
+      "epoch": 242.13,
+      "learning_rate": 4.2103942652329756e-05,
+      "loss": 0.317,
+      "step": 37530
+    },
+    {
+      "epoch": 242.19,
+      "learning_rate": 4.210035842293907e-05,
+      "loss": 0.2864,
+      "step": 37540
+    },
+    {
+      "epoch": 242.26,
+      "learning_rate": 4.2096774193548386e-05,
+      "loss": 0.2756,
+      "step": 37550
+    },
+    {
+      "epoch": 242.32,
+      "learning_rate": 4.209318996415771e-05,
+      "loss": 0.283,
+      "step": 37560
+    },
+    {
+      "epoch": 242.39,
+      "learning_rate": 4.208960573476703e-05,
+      "loss": 0.2976,
+      "step": 37570
+    },
+    {
+      "epoch": 242.45,
+      "learning_rate": 4.208602150537635e-05,
+      "loss": 0.3004,
+      "step": 37580
+    },
+    {
+      "epoch": 242.51,
+      "learning_rate": 4.2082437275985665e-05,
+      "loss": 0.27,
+      "step": 37590
+    },
+    {
+      "epoch": 242.58,
+      "learning_rate": 4.2078853046594984e-05,
+      "loss": 0.2771,
+      "step": 37600
+    },
+    {
+      "epoch": 242.64,
+      "learning_rate": 4.207526881720431e-05,
+      "loss": 0.2988,
+      "step": 37610
+    },
+    {
+      "epoch": 242.71,
+      "learning_rate": 4.207168458781362e-05,
+      "loss": 0.2975,
+      "step": 37620
+    },
+    {
+      "epoch": 242.77,
+      "learning_rate": 4.206810035842294e-05,
+      "loss": 0.2816,
+      "step": 37630
+    },
+    {
+      "epoch": 242.84,
+      "learning_rate": 4.206451612903226e-05,
+      "loss": 0.2853,
+      "step": 37640
+    },
+    {
+      "epoch": 242.9,
+      "learning_rate": 4.206093189964158e-05,
+      "loss": 0.2996,
+      "step": 37650
+    },
+    {
+      "epoch": 242.96,
+      "learning_rate": 4.205734767025089e-05,
+      "loss": 0.2908,
+      "step": 37660
+    },
+    {
+      "epoch": 243.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.977866530418396,
+      "eval_runtime": 0.886,
+      "eval_samples_per_second": 226.853,
+      "eval_steps_per_second": 7.9,
+      "step": 37665
+    },
+    {
+      "epoch": 243.03,
+      "learning_rate": 4.205376344086022e-05,
+      "loss": 0.3242,
+      "step": 37670
+    },
+    {
+      "epoch": 243.1,
+      "learning_rate": 4.2050179211469536e-05,
+      "loss": 0.3031,
+      "step": 37680
+    },
+    {
+      "epoch": 243.16,
+      "learning_rate": 4.2046594982078854e-05,
+      "loss": 0.2849,
+      "step": 37690
+    },
+    {
+      "epoch": 243.23,
+      "learning_rate": 4.204301075268817e-05,
+      "loss": 0.3192,
+      "step": 37700
+    },
+    {
+      "epoch": 243.29,
+      "learning_rate": 4.203942652329749e-05,
+      "loss": 0.2244,
+      "step": 37710
+    },
+    {
+      "epoch": 243.35,
+      "learning_rate": 4.2035842293906816e-05,
+      "loss": 0.2512,
+      "step": 37720
+    },
+    {
+      "epoch": 243.42,
+      "learning_rate": 4.2032258064516134e-05,
+      "loss": 0.2966,
+      "step": 37730
+    },
+    {
+      "epoch": 243.48,
+      "learning_rate": 4.2028673835125445e-05,
+      "loss": 0.238,
+      "step": 37740
+    },
+    {
+      "epoch": 243.55,
+      "learning_rate": 4.202508960573477e-05,
+      "loss": 0.3029,
+      "step": 37750
+    },
+    {
+      "epoch": 243.61,
+      "learning_rate": 4.202150537634409e-05,
+      "loss": 0.3428,
+      "step": 37760
+    },
+    {
+      "epoch": 243.68,
+      "learning_rate": 4.201792114695341e-05,
+      "loss": 0.2759,
+      "step": 37770
+    },
+    {
+      "epoch": 243.74,
+      "learning_rate": 4.2014336917562725e-05,
+      "loss": 0.3196,
+      "step": 37780
+    },
+    {
+      "epoch": 243.8,
+      "learning_rate": 4.201075268817204e-05,
+      "loss": 0.2663,
+      "step": 37790
+    },
+    {
+      "epoch": 243.87,
+      "learning_rate": 4.200716845878137e-05,
+      "loss": 0.2913,
+      "step": 37800
+    },
+    {
+      "epoch": 243.93,
+      "learning_rate": 4.2003584229390686e-05,
+      "loss": 0.2502,
+      "step": 37810
+    },
+    {
+      "epoch": 244.0,
+      "learning_rate": 4.2e-05,
+      "loss": 0.3209,
+      "step": 37820
+    },
+    {
+      "epoch": 244.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0140814781188965,
+      "eval_runtime": 0.8562,
+      "eval_samples_per_second": 234.764,
+      "eval_steps_per_second": 8.176,
+      "step": 37820
+    },
+    {
+      "epoch": 244.06,
+      "learning_rate": 4.199641577060932e-05,
+      "loss": 0.3133,
+      "step": 37830
+    },
+    {
+      "epoch": 244.13,
+      "learning_rate": 4.199283154121864e-05,
+      "loss": 0.2916,
+      "step": 37840
+    },
+    {
+      "epoch": 244.19,
+      "learning_rate": 4.198924731182796e-05,
+      "loss": 0.2854,
+      "step": 37850
+    },
+    {
+      "epoch": 244.26,
+      "learning_rate": 4.198566308243728e-05,
+      "loss": 0.3112,
+      "step": 37860
+    },
+    {
+      "epoch": 244.32,
+      "learning_rate": 4.1982078853046595e-05,
+      "loss": 0.3122,
+      "step": 37870
+    },
+    {
+      "epoch": 244.39,
+      "learning_rate": 4.197849462365592e-05,
+      "loss": 0.2543,
+      "step": 37880
+    },
+    {
+      "epoch": 244.45,
+      "learning_rate": 4.197491039426524e-05,
+      "loss": 0.2646,
+      "step": 37890
+    },
+    {
+      "epoch": 244.51,
+      "learning_rate": 4.197132616487455e-05,
+      "loss": 0.2537,
+      "step": 37900
+    },
+    {
+      "epoch": 244.58,
+      "learning_rate": 4.1967741935483875e-05,
+      "loss": 0.2895,
+      "step": 37910
+    },
+    {
+      "epoch": 244.64,
+      "learning_rate": 4.196415770609319e-05,
+      "loss": 0.3076,
+      "step": 37920
+    },
+    {
+      "epoch": 244.71,
+      "learning_rate": 4.196057347670251e-05,
+      "loss": 0.2697,
+      "step": 37930
+    },
+    {
+      "epoch": 244.77,
+      "learning_rate": 4.195698924731183e-05,
+      "loss": 0.3198,
+      "step": 37940
+    },
+    {
+      "epoch": 244.84,
+      "learning_rate": 4.195340501792115e-05,
+      "loss": 0.2903,
+      "step": 37950
+    },
+    {
+      "epoch": 244.9,
+      "learning_rate": 4.1949820788530466e-05,
+      "loss": 0.2924,
+      "step": 37960
+    },
+    {
+      "epoch": 244.96,
+      "learning_rate": 4.1946236559139784e-05,
+      "loss": 0.2401,
+      "step": 37970
+    },
+    {
+      "epoch": 245.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9130392670631409,
+      "eval_runtime": 0.8888,
+      "eval_samples_per_second": 226.15,
+      "eval_steps_per_second": 7.876,
+      "step": 37975
+    },
+    {
+      "epoch": 245.03,
+      "learning_rate": 4.19426523297491e-05,
+      "loss": 0.2983,
+      "step": 37980
+    },
+    {
+      "epoch": 245.1,
+      "learning_rate": 4.193906810035843e-05,
+      "loss": 0.2655,
+      "step": 37990
+    },
+    {
+      "epoch": 245.16,
+      "learning_rate": 4.1935483870967746e-05,
+      "loss": 0.2921,
+      "step": 38000
+    },
+    {
+      "epoch": 245.23,
+      "learning_rate": 4.193189964157706e-05,
+      "loss": 0.3413,
+      "step": 38010
+    },
+    {
+      "epoch": 245.29,
+      "learning_rate": 4.192831541218638e-05,
+      "loss": 0.3536,
+      "step": 38020
+    },
+    {
+      "epoch": 245.35,
+      "learning_rate": 4.19247311827957e-05,
+      "loss": 0.2658,
+      "step": 38030
+    },
+    {
+      "epoch": 245.42,
+      "learning_rate": 4.192114695340502e-05,
+      "loss": 0.2473,
+      "step": 38040
+    },
+    {
+      "epoch": 245.48,
+      "learning_rate": 4.191756272401434e-05,
+      "loss": 0.3365,
+      "step": 38050
+    },
+    {
+      "epoch": 245.55,
+      "learning_rate": 4.1913978494623655e-05,
+      "loss": 0.3014,
+      "step": 38060
+    },
+    {
+      "epoch": 245.61,
+      "learning_rate": 4.191039426523298e-05,
+      "loss": 0.2773,
+      "step": 38070
+    },
+    {
+      "epoch": 245.68,
+      "learning_rate": 4.19068100358423e-05,
+      "loss": 0.2564,
+      "step": 38080
+    },
+    {
+      "epoch": 245.74,
+      "learning_rate": 4.190322580645161e-05,
+      "loss": 0.2492,
+      "step": 38090
+    },
+    {
+      "epoch": 245.8,
+      "learning_rate": 4.1899641577060935e-05,
+      "loss": 0.264,
+      "step": 38100
+    },
+    {
+      "epoch": 245.87,
+      "learning_rate": 4.189605734767025e-05,
+      "loss": 0.271,
+      "step": 38110
+    },
+    {
+      "epoch": 245.93,
+      "learning_rate": 4.189247311827957e-05,
+      "loss": 0.2691,
+      "step": 38120
+    },
+    {
+      "epoch": 246.0,
+      "learning_rate": 4.188888888888889e-05,
+      "loss": 0.2774,
+      "step": 38130
+    },
+    {
+      "epoch": 246.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.910443902015686,
+      "eval_runtime": 0.8473,
+      "eval_samples_per_second": 237.236,
+      "eval_steps_per_second": 8.262,
+      "step": 38130
+    },
+    {
+      "epoch": 246.06,
+      "learning_rate": 4.188530465949821e-05,
+      "loss": 0.3011,
+      "step": 38140
+    },
+    {
+      "epoch": 246.13,
+      "learning_rate": 4.188172043010753e-05,
+      "loss": 0.3098,
+      "step": 38150
+    },
+    {
+      "epoch": 246.19,
+      "learning_rate": 4.187813620071685e-05,
+      "loss": 0.3484,
+      "step": 38160
+    },
+    {
+      "epoch": 246.26,
+      "learning_rate": 4.187455197132616e-05,
+      "loss": 0.2991,
+      "step": 38170
+    },
+    {
+      "epoch": 246.32,
+      "learning_rate": 4.187096774193549e-05,
+      "loss": 0.3186,
+      "step": 38180
+    },
+    {
+      "epoch": 246.39,
+      "learning_rate": 4.1867383512544805e-05,
+      "loss": 0.302,
+      "step": 38190
+    },
+    {
+      "epoch": 246.45,
+      "learning_rate": 4.186379928315412e-05,
+      "loss": 0.2695,
+      "step": 38200
+    },
+    {
+      "epoch": 246.51,
+      "learning_rate": 4.186021505376344e-05,
+      "loss": 0.281,
+      "step": 38210
+    },
+    {
+      "epoch": 246.58,
+      "learning_rate": 4.185663082437276e-05,
+      "loss": 0.3039,
+      "step": 38220
+    },
+    {
+      "epoch": 246.64,
+      "learning_rate": 4.1853046594982085e-05,
+      "loss": 0.3492,
+      "step": 38230
+    },
+    {
+      "epoch": 246.71,
+      "learning_rate": 4.18494623655914e-05,
+      "loss": 0.295,
+      "step": 38240
+    },
+    {
+      "epoch": 246.77,
+      "learning_rate": 4.1845878136200714e-05,
+      "loss": 0.3129,
+      "step": 38250
+    },
+    {
+      "epoch": 246.84,
+      "learning_rate": 4.184229390681004e-05,
+      "loss": 0.2767,
+      "step": 38260
+    },
+    {
+      "epoch": 246.9,
+      "learning_rate": 4.183870967741936e-05,
+      "loss": 0.3057,
+      "step": 38270
+    },
+    {
+      "epoch": 246.96,
+      "learning_rate": 4.1835125448028676e-05,
+      "loss": 0.2789,
+      "step": 38280
+    },
+    {
+      "epoch": 247.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9962852001190186,
+      "eval_runtime": 0.8574,
+      "eval_samples_per_second": 234.429,
+      "eval_steps_per_second": 8.164,
+      "step": 38285
+    },
+    {
+      "epoch": 247.03,
+      "learning_rate": 4.1831541218637994e-05,
+      "loss": 0.3031,
+      "step": 38290
+    },
+    {
+      "epoch": 247.1,
+      "learning_rate": 4.182795698924731e-05,
+      "loss": 0.3049,
+      "step": 38300
+    },
+    {
+      "epoch": 247.16,
+      "learning_rate": 4.182437275985663e-05,
+      "loss": 0.332,
+      "step": 38310
+    },
+    {
+      "epoch": 247.23,
+      "learning_rate": 4.1820788530465955e-05,
+      "loss": 0.3043,
+      "step": 38320
+    },
+    {
+      "epoch": 247.29,
+      "learning_rate": 4.181720430107527e-05,
+      "loss": 0.2544,
+      "step": 38330
+    },
+    {
+      "epoch": 247.35,
+      "learning_rate": 4.181362007168459e-05,
+      "loss": 0.2861,
+      "step": 38340
+    },
+    {
+      "epoch": 247.42,
+      "learning_rate": 4.181003584229391e-05,
+      "loss": 0.2964,
+      "step": 38350
+    },
+    {
+      "epoch": 247.48,
+      "learning_rate": 4.180645161290323e-05,
+      "loss": 0.3022,
+      "step": 38360
+    },
+    {
+      "epoch": 247.55,
+      "learning_rate": 4.1802867383512546e-05,
+      "loss": 0.281,
+      "step": 38370
+    },
+    {
+      "epoch": 247.61,
+      "learning_rate": 4.1799283154121865e-05,
+      "loss": 0.3364,
+      "step": 38380
+    },
+    {
+      "epoch": 247.68,
+      "learning_rate": 4.179569892473118e-05,
+      "loss": 0.2659,
+      "step": 38390
+    },
+    {
+      "epoch": 247.74,
+      "learning_rate": 4.179211469534051e-05,
+      "loss": 0.3622,
+      "step": 38400
+    },
+    {
+      "epoch": 247.8,
+      "learning_rate": 4.178853046594982e-05,
+      "loss": 0.2729,
+      "step": 38410
+    },
+    {
+      "epoch": 247.87,
+      "learning_rate": 4.1784946236559144e-05,
+      "loss": 0.2943,
+      "step": 38420
+    },
+    {
+      "epoch": 247.93,
+      "learning_rate": 4.178136200716846e-05,
+      "loss": 0.2468,
+      "step": 38430
+    },
+    {
+      "epoch": 248.0,
+      "learning_rate": 4.177777777777778e-05,
+      "loss": 0.2989,
+      "step": 38440
+    },
+    {
+      "epoch": 248.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9978200197219849,
+      "eval_runtime": 0.8592,
+      "eval_samples_per_second": 233.939,
+      "eval_steps_per_second": 8.147,
+      "step": 38440
+    },
+    {
+      "epoch": 248.06,
+      "learning_rate": 4.17741935483871e-05,
+      "loss": 0.3274,
+      "step": 38450
+    },
+    {
+      "epoch": 248.13,
+      "learning_rate": 4.177060931899642e-05,
+      "loss": 0.3091,
+      "step": 38460
+    },
+    {
+      "epoch": 248.19,
+      "learning_rate": 4.1767025089605735e-05,
+      "loss": 0.3011,
+      "step": 38470
+    },
+    {
+      "epoch": 248.26,
+      "learning_rate": 4.176344086021506e-05,
+      "loss": 0.2416,
+      "step": 38480
+    },
+    {
+      "epoch": 248.32,
+      "learning_rate": 4.175985663082437e-05,
+      "loss": 0.2728,
+      "step": 38490
+    },
+    {
+      "epoch": 248.39,
+      "learning_rate": 4.1756272401433697e-05,
+      "loss": 0.2849,
+      "step": 38500
+    },
+    {
+      "epoch": 248.45,
+      "learning_rate": 4.1752688172043015e-05,
+      "loss": 0.2919,
+      "step": 38510
+    },
+    {
+      "epoch": 248.51,
+      "learning_rate": 4.174910394265233e-05,
+      "loss": 0.2999,
+      "step": 38520
+    },
+    {
+      "epoch": 248.58,
+      "learning_rate": 4.174551971326165e-05,
+      "loss": 0.3196,
+      "step": 38530
+    },
+    {
+      "epoch": 248.64,
+      "learning_rate": 4.174193548387097e-05,
+      "loss": 0.301,
+      "step": 38540
+    },
+    {
+      "epoch": 248.71,
+      "learning_rate": 4.173835125448029e-05,
+      "loss": 0.2796,
+      "step": 38550
+    },
+    {
+      "epoch": 248.77,
+      "learning_rate": 4.1734767025089606e-05,
+      "loss": 0.2997,
+      "step": 38560
+    },
+    {
+      "epoch": 248.84,
+      "learning_rate": 4.1731182795698924e-05,
+      "loss": 0.2577,
+      "step": 38570
+    },
+    {
+      "epoch": 248.9,
+      "learning_rate": 4.172759856630825e-05,
+      "loss": 0.2978,
+      "step": 38580
+    },
+    {
+      "epoch": 248.96,
+      "learning_rate": 4.172401433691757e-05,
+      "loss": 0.2172,
+      "step": 38590
+    },
+    {
+      "epoch": 249.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9269899725914001,
+      "eval_runtime": 0.8883,
+      "eval_samples_per_second": 226.281,
+      "eval_steps_per_second": 7.88,
+      "step": 38595
+    },
+    {
+      "epoch": 249.03,
+      "learning_rate": 4.172043010752688e-05,
+      "loss": 0.2873,
+      "step": 38600
+    },
+    {
+      "epoch": 249.1,
+      "learning_rate": 4.1716845878136204e-05,
+      "loss": 0.2605,
+      "step": 38610
+    },
+    {
+      "epoch": 249.16,
+      "learning_rate": 4.171326164874552e-05,
+      "loss": 0.3069,
+      "step": 38620
+    },
+    {
+      "epoch": 249.23,
+      "learning_rate": 4.170967741935484e-05,
+      "loss": 0.3102,
+      "step": 38630
+    },
+    {
+      "epoch": 249.29,
+      "learning_rate": 4.170609318996416e-05,
+      "loss": 0.2885,
+      "step": 38640
+    },
+    {
+      "epoch": 249.35,
+      "learning_rate": 4.1702508960573476e-05,
+      "loss": 0.3395,
+      "step": 38650
+    },
+    {
+      "epoch": 249.42,
+      "learning_rate": 4.1698924731182795e-05,
+      "loss": 0.3403,
+      "step": 38660
+    },
+    {
+      "epoch": 249.48,
+      "learning_rate": 4.169534050179212e-05,
+      "loss": 0.2385,
+      "step": 38670
+    },
+    {
+      "epoch": 249.55,
+      "learning_rate": 4.169175627240143e-05,
+      "loss": 0.2909,
+      "step": 38680
+    },
+    {
+      "epoch": 249.61,
+      "learning_rate": 4.1688172043010756e-05,
+      "loss": 0.3125,
+      "step": 38690
+    },
+    {
+      "epoch": 249.68,
+      "learning_rate": 4.1684587813620074e-05,
+      "loss": 0.2669,
+      "step": 38700
+    },
+    {
+      "epoch": 249.74,
+      "learning_rate": 4.168100358422939e-05,
+      "loss": 0.2576,
+      "step": 38710
+    },
+    {
+      "epoch": 249.8,
+      "learning_rate": 4.167741935483871e-05,
+      "loss": 0.2903,
+      "step": 38720
+    },
+    {
+      "epoch": 249.87,
+      "learning_rate": 4.167383512544803e-05,
+      "loss": 0.2774,
+      "step": 38730
+    },
+    {
+      "epoch": 249.93,
+      "learning_rate": 4.167025089605735e-05,
+      "loss": 0.2821,
+      "step": 38740
+    },
+    {
+      "epoch": 250.0,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 0.2572,
+      "step": 38750
+    },
+    {
+      "epoch": 250.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9663498997688293,
+      "eval_runtime": 0.8725,
+      "eval_samples_per_second": 230.362,
+      "eval_steps_per_second": 8.023,
+      "step": 38750
+    },
+    {
+      "epoch": 250.06,
+      "learning_rate": 4.1663082437275983e-05,
+      "loss": 0.3564,
+      "step": 38760
+    },
+    {
+      "epoch": 250.13,
+      "learning_rate": 4.165949820788531e-05,
+      "loss": 0.233,
+      "step": 38770
+    },
+    {
+      "epoch": 250.19,
+      "learning_rate": 4.165591397849463e-05,
+      "loss": 0.2724,
+      "step": 38780
+    },
+    {
+      "epoch": 250.26,
+      "learning_rate": 4.1652329749103945e-05,
+      "loss": 0.3072,
+      "step": 38790
+    },
+    {
+      "epoch": 250.32,
+      "learning_rate": 4.164874551971326e-05,
+      "loss": 0.345,
+      "step": 38800
+    },
+    {
+      "epoch": 250.39,
+      "learning_rate": 4.164516129032258e-05,
+      "loss": 0.3015,
+      "step": 38810
+    },
+    {
+      "epoch": 250.45,
+      "learning_rate": 4.16415770609319e-05,
+      "loss": 0.3223,
+      "step": 38820
+    },
+    {
+      "epoch": 250.51,
+      "learning_rate": 4.1637992831541224e-05,
+      "loss": 0.3065,
+      "step": 38830
+    },
+    {
+      "epoch": 250.58,
+      "learning_rate": 4.1634408602150536e-05,
+      "loss": 0.2894,
+      "step": 38840
+    },
+    {
+      "epoch": 250.64,
+      "learning_rate": 4.163082437275986e-05,
+      "loss": 0.2361,
+      "step": 38850
+    },
+    {
+      "epoch": 250.71,
+      "learning_rate": 4.162724014336918e-05,
+      "loss": 0.2678,
+      "step": 38860
+    },
+    {
+      "epoch": 250.77,
+      "learning_rate": 4.16236559139785e-05,
+      "loss": 0.3192,
+      "step": 38870
+    },
+    {
+      "epoch": 250.84,
+      "learning_rate": 4.1620071684587815e-05,
+      "loss": 0.3232,
+      "step": 38880
+    },
+    {
+      "epoch": 250.9,
+      "learning_rate": 4.1616487455197134e-05,
+      "loss": 0.2781,
+      "step": 38890
+    },
+    {
+      "epoch": 250.96,
+      "learning_rate": 4.161290322580645e-05,
+      "loss": 0.2758,
+      "step": 38900
+    },
+    {
+      "epoch": 251.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9681599736213684,
+      "eval_runtime": 0.8693,
+      "eval_samples_per_second": 231.227,
+      "eval_steps_per_second": 8.053,
+      "step": 38905
+    },
+    {
+      "epoch": 251.03,
+      "learning_rate": 4.160931899641578e-05,
+      "loss": 0.3348,
+      "step": 38910
+    },
+    {
+      "epoch": 251.1,
+      "learning_rate": 4.160573476702509e-05,
+      "loss": 0.2629,
+      "step": 38920
+    },
+    {
+      "epoch": 251.16,
+      "learning_rate": 4.160215053763441e-05,
+      "loss": 0.245,
+      "step": 38930
+    },
+    {
+      "epoch": 251.23,
+      "learning_rate": 4.159856630824373e-05,
+      "loss": 0.3172,
+      "step": 38940
+    },
+    {
+      "epoch": 251.29,
+      "learning_rate": 4.159498207885305e-05,
+      "loss": 0.2752,
+      "step": 38950
+    },
+    {
+      "epoch": 251.35,
+      "learning_rate": 4.159139784946237e-05,
+      "loss": 0.3065,
+      "step": 38960
+    },
+    {
+      "epoch": 251.42,
+      "learning_rate": 4.1587813620071686e-05,
+      "loss": 0.2632,
+      "step": 38970
+    },
+    {
+      "epoch": 251.48,
+      "learning_rate": 4.1584229390681004e-05,
+      "loss": 0.2595,
+      "step": 38980
+    },
+    {
+      "epoch": 251.55,
+      "learning_rate": 4.158064516129033e-05,
+      "loss": 0.2787,
+      "step": 38990
+    },
+    {
+      "epoch": 251.61,
+      "learning_rate": 4.157706093189964e-05,
+      "loss": 0.2499,
+      "step": 39000
+    },
+    {
+      "epoch": 251.68,
+      "learning_rate": 4.157347670250896e-05,
+      "loss": 0.2682,
+      "step": 39010
+    },
+    {
+      "epoch": 251.74,
+      "learning_rate": 4.1569892473118284e-05,
+      "loss": 0.3328,
+      "step": 39020
+    },
+    {
+      "epoch": 251.8,
+      "learning_rate": 4.15663082437276e-05,
+      "loss": 0.2671,
+      "step": 39030
+    },
+    {
+      "epoch": 251.87,
+      "learning_rate": 4.156272401433692e-05,
+      "loss": 0.3038,
+      "step": 39040
+    },
+    {
+      "epoch": 251.93,
+      "learning_rate": 4.155913978494624e-05,
+      "loss": 0.3393,
+      "step": 39050
+    },
+    {
+      "epoch": 252.0,
+      "learning_rate": 4.155555555555556e-05,
+      "loss": 0.2704,
+      "step": 39060
+    },
+    {
+      "epoch": 252.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0670104026794434,
+      "eval_runtime": 0.862,
+      "eval_samples_per_second": 233.167,
+      "eval_steps_per_second": 8.12,
+      "step": 39060
+    },
+    {
+      "epoch": 252.06,
+      "learning_rate": 4.155197132616488e-05,
+      "loss": 0.3246,
+      "step": 39070
+    },
+    {
+      "epoch": 252.13,
+      "learning_rate": 4.154838709677419e-05,
+      "loss": 0.3658,
+      "step": 39080
+    },
+    {
+      "epoch": 252.19,
+      "learning_rate": 4.154480286738351e-05,
+      "loss": 0.2611,
+      "step": 39090
+    },
+    {
+      "epoch": 252.26,
+      "learning_rate": 4.1541218637992836e-05,
+      "loss": 0.2854,
+      "step": 39100
+    },
+    {
+      "epoch": 252.32,
+      "learning_rate": 4.1537634408602155e-05,
+      "loss": 0.2714,
+      "step": 39110
+    },
+    {
+      "epoch": 252.39,
+      "learning_rate": 4.153405017921147e-05,
+      "loss": 0.2613,
+      "step": 39120
+    },
+    {
+      "epoch": 252.45,
+      "learning_rate": 4.153046594982079e-05,
+      "loss": 0.3263,
+      "step": 39130
+    },
+    {
+      "epoch": 252.51,
+      "learning_rate": 4.152688172043011e-05,
+      "loss": 0.2657,
+      "step": 39140
+    },
+    {
+      "epoch": 252.58,
+      "learning_rate": 4.1523297491039434e-05,
+      "loss": 0.2933,
+      "step": 39150
+    },
+    {
+      "epoch": 252.64,
+      "learning_rate": 4.1519713261648746e-05,
+      "loss": 0.2477,
+      "step": 39160
+    },
+    {
+      "epoch": 252.71,
+      "learning_rate": 4.1516129032258064e-05,
+      "loss": 0.2665,
+      "step": 39170
+    },
+    {
+      "epoch": 252.77,
+      "learning_rate": 4.151254480286739e-05,
+      "loss": 0.267,
+      "step": 39180
+    },
+    {
+      "epoch": 252.84,
+      "learning_rate": 4.15089605734767e-05,
+      "loss": 0.3222,
+      "step": 39190
+    },
+    {
+      "epoch": 252.9,
+      "learning_rate": 4.1505376344086025e-05,
+      "loss": 0.2978,
+      "step": 39200
+    },
+    {
+      "epoch": 252.96,
+      "learning_rate": 4.150179211469534e-05,
+      "loss": 0.3037,
+      "step": 39210
+    },
+    {
+      "epoch": 253.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0696662664413452,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 231.242,
+      "eval_steps_per_second": 8.053,
+      "step": 39215
+    },
+    {
+      "epoch": 253.03,
+      "learning_rate": 4.149820788530466e-05,
+      "loss": 0.3095,
+      "step": 39220
+    },
+    {
+      "epoch": 253.1,
+      "learning_rate": 4.149462365591398e-05,
+      "loss": 0.2833,
+      "step": 39230
+    },
+    {
+      "epoch": 253.16,
+      "learning_rate": 4.14910394265233e-05,
+      "loss": 0.2397,
+      "step": 39240
+    },
+    {
+      "epoch": 253.23,
+      "learning_rate": 4.1487455197132616e-05,
+      "loss": 0.2901,
+      "step": 39250
+    },
+    {
+      "epoch": 253.29,
+      "learning_rate": 4.148387096774194e-05,
+      "loss": 0.3403,
+      "step": 39260
+    },
+    {
+      "epoch": 253.35,
+      "learning_rate": 4.148028673835125e-05,
+      "loss": 0.2919,
+      "step": 39270
+    },
+    {
+      "epoch": 253.42,
+      "learning_rate": 4.147670250896058e-05,
+      "loss": 0.2683,
+      "step": 39280
+    },
+    {
+      "epoch": 253.48,
+      "learning_rate": 4.1473118279569896e-05,
+      "loss": 0.2793,
+      "step": 39290
+    },
+    {
+      "epoch": 253.55,
+      "learning_rate": 4.1469534050179214e-05,
+      "loss": 0.2569,
+      "step": 39300
+    },
+    {
+      "epoch": 253.61,
+      "learning_rate": 4.146594982078853e-05,
+      "loss": 0.3079,
+      "step": 39310
+    },
+    {
+      "epoch": 253.68,
+      "learning_rate": 4.146236559139785e-05,
+      "loss": 0.3014,
+      "step": 39320
+    },
+    {
+      "epoch": 253.74,
+      "learning_rate": 4.145878136200717e-05,
+      "loss": 0.3,
+      "step": 39330
+    },
+    {
+      "epoch": 253.8,
+      "learning_rate": 4.1455197132616494e-05,
+      "loss": 0.2848,
+      "step": 39340
+    },
+    {
+      "epoch": 253.87,
+      "learning_rate": 4.1451612903225805e-05,
+      "loss": 0.2689,
+      "step": 39350
+    },
+    {
+      "epoch": 253.93,
+      "learning_rate": 4.144802867383512e-05,
+      "loss": 0.2637,
+      "step": 39360
+    },
+    {
+      "epoch": 254.0,
+      "learning_rate": 4.144444444444445e-05,
+      "loss": 0.2826,
+      "step": 39370
+    },
+    {
+      "epoch": 254.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0464991331100464,
+      "eval_runtime": 0.849,
+      "eval_samples_per_second": 236.739,
+      "eval_steps_per_second": 8.245,
+      "step": 39370
+    },
+    {
+      "epoch": 254.06,
+      "learning_rate": 4.1440860215053766e-05,
+      "loss": 0.2837,
+      "step": 39380
+    },
+    {
+      "epoch": 254.13,
+      "learning_rate": 4.1437275985663085e-05,
+      "loss": 0.2504,
+      "step": 39390
+    },
+    {
+      "epoch": 254.19,
+      "learning_rate": 4.14336917562724e-05,
+      "loss": 0.3533,
+      "step": 39400
+    },
+    {
+      "epoch": 254.26,
+      "learning_rate": 4.143010752688172e-05,
+      "loss": 0.253,
+      "step": 39410
+    },
+    {
+      "epoch": 254.32,
+      "learning_rate": 4.1426523297491046e-05,
+      "loss": 0.2811,
+      "step": 39420
+    },
+    {
+      "epoch": 254.39,
+      "learning_rate": 4.142293906810036e-05,
+      "loss": 0.278,
+      "step": 39430
+    },
+    {
+      "epoch": 254.45,
+      "learning_rate": 4.1419354838709676e-05,
+      "loss": 0.3352,
+      "step": 39440
+    },
+    {
+      "epoch": 254.51,
+      "learning_rate": 4.1415770609319e-05,
+      "loss": 0.3424,
+      "step": 39450
+    },
+    {
+      "epoch": 254.58,
+      "learning_rate": 4.141218637992832e-05,
+      "loss": 0.3144,
+      "step": 39460
+    },
+    {
+      "epoch": 254.64,
+      "learning_rate": 4.140860215053764e-05,
+      "loss": 0.2818,
+      "step": 39470
+    },
+    {
+      "epoch": 254.71,
+      "learning_rate": 4.1405017921146955e-05,
+      "loss": 0.2753,
+      "step": 39480
+    },
+    {
+      "epoch": 254.77,
+      "learning_rate": 4.1401433691756273e-05,
+      "loss": 0.2838,
+      "step": 39490
+    },
+    {
+      "epoch": 254.84,
+      "learning_rate": 4.13978494623656e-05,
+      "loss": 0.2863,
+      "step": 39500
+    },
+    {
+      "epoch": 254.9,
+      "learning_rate": 4.139426523297491e-05,
+      "loss": 0.2863,
+      "step": 39510
+    },
+    {
+      "epoch": 254.96,
+      "learning_rate": 4.139068100358423e-05,
+      "loss": 0.2992,
+      "step": 39520
+    },
+    {
+      "epoch": 255.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0168235301971436,
+      "eval_runtime": 0.862,
+      "eval_samples_per_second": 233.182,
+      "eval_steps_per_second": 8.121,
+      "step": 39525
+    },
+    {
+      "epoch": 255.03,
+      "learning_rate": 4.138709677419355e-05,
+      "loss": 0.3041,
+      "step": 39530
+    },
+    {
+      "epoch": 255.1,
+      "learning_rate": 4.138351254480287e-05,
+      "loss": 0.2878,
+      "step": 39540
+    },
+    {
+      "epoch": 255.16,
+      "learning_rate": 4.137992831541219e-05,
+      "loss": 0.2828,
+      "step": 39550
+    },
+    {
+      "epoch": 255.23,
+      "learning_rate": 4.137634408602151e-05,
+      "loss": 0.2724,
+      "step": 39560
+    },
+    {
+      "epoch": 255.29,
+      "learning_rate": 4.1372759856630826e-05,
+      "loss": 0.2553,
+      "step": 39570
+    },
+    {
+      "epoch": 255.35,
+      "learning_rate": 4.136917562724015e-05,
+      "loss": 0.2614,
+      "step": 39580
+    },
+    {
+      "epoch": 255.42,
+      "learning_rate": 4.136559139784946e-05,
+      "loss": 0.2537,
+      "step": 39590
+    },
+    {
+      "epoch": 255.48,
+      "learning_rate": 4.136200716845878e-05,
+      "loss": 0.2663,
+      "step": 39600
+    },
+    {
+      "epoch": 255.55,
+      "learning_rate": 4.1358422939068105e-05,
+      "loss": 0.3107,
+      "step": 39610
+    },
+    {
+      "epoch": 255.61,
+      "learning_rate": 4.1354838709677424e-05,
+      "loss": 0.2747,
+      "step": 39620
+    },
+    {
+      "epoch": 255.68,
+      "learning_rate": 4.135125448028674e-05,
+      "loss": 0.2648,
+      "step": 39630
+    },
+    {
+      "epoch": 255.74,
+      "learning_rate": 4.134767025089606e-05,
+      "loss": 0.2692,
+      "step": 39640
+    },
+    {
+      "epoch": 255.8,
+      "learning_rate": 4.134408602150538e-05,
+      "loss": 0.2911,
+      "step": 39650
+    },
+    {
+      "epoch": 255.87,
+      "learning_rate": 4.1340501792114696e-05,
+      "loss": 0.3434,
+      "step": 39660
+    },
+    {
+      "epoch": 255.93,
+      "learning_rate": 4.1336917562724015e-05,
+      "loss": 0.2632,
+      "step": 39670
+    },
+    {
+      "epoch": 256.0,
+      "learning_rate": 4.133333333333333e-05,
+      "loss": 0.2967,
+      "step": 39680
+    },
+    {
+      "epoch": 256.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9484884738922119,
+      "eval_runtime": 0.8569,
+      "eval_samples_per_second": 234.555,
+      "eval_steps_per_second": 8.169,
+      "step": 39680
+    },
+    {
+      "epoch": 256.06,
+      "learning_rate": 4.132974910394266e-05,
+      "loss": 0.2874,
+      "step": 39690
+    },
+    {
+      "epoch": 256.13,
+      "learning_rate": 4.1326164874551976e-05,
+      "loss": 0.3373,
+      "step": 39700
+    },
+    {
+      "epoch": 256.19,
+      "learning_rate": 4.132258064516129e-05,
+      "loss": 0.2652,
+      "step": 39710
+    },
+    {
+      "epoch": 256.26,
+      "learning_rate": 4.131899641577061e-05,
+      "loss": 0.2908,
+      "step": 39720
+    },
+    {
+      "epoch": 256.32,
+      "learning_rate": 4.131541218637993e-05,
+      "loss": 0.24,
+      "step": 39730
+    },
+    {
+      "epoch": 256.39,
+      "learning_rate": 4.131182795698925e-05,
+      "loss": 0.272,
+      "step": 39740
+    },
+    {
+      "epoch": 256.45,
+      "learning_rate": 4.130824372759857e-05,
+      "loss": 0.2505,
+      "step": 39750
+    },
+    {
+      "epoch": 256.51,
+      "learning_rate": 4.1304659498207885e-05,
+      "loss": 0.3031,
+      "step": 39760
+    },
+    {
+      "epoch": 256.58,
+      "learning_rate": 4.130107526881721e-05,
+      "loss": 0.2515,
+      "step": 39770
+    },
+    {
+      "epoch": 256.64,
+      "learning_rate": 4.129749103942653e-05,
+      "loss": 0.3433,
+      "step": 39780
+    },
+    {
+      "epoch": 256.71,
+      "learning_rate": 4.129390681003584e-05,
+      "loss": 0.2765,
+      "step": 39790
+    },
+    {
+      "epoch": 256.77,
+      "learning_rate": 4.1290322580645165e-05,
+      "loss": 0.2979,
+      "step": 39800
+    },
+    {
+      "epoch": 256.84,
+      "learning_rate": 4.128673835125448e-05,
+      "loss": 0.2742,
+      "step": 39810
+    },
+    {
+      "epoch": 256.9,
+      "learning_rate": 4.12831541218638e-05,
+      "loss": 0.2684,
+      "step": 39820
+    },
+    {
+      "epoch": 256.96,
+      "learning_rate": 4.127956989247312e-05,
+      "loss": 0.2657,
+      "step": 39830
+    },
+    {
+      "epoch": 257.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9528272747993469,
+      "eval_runtime": 0.862,
+      "eval_samples_per_second": 233.18,
+      "eval_steps_per_second": 8.121,
+      "step": 39835
+    },
+    {
+      "epoch": 257.03,
+      "learning_rate": 4.127598566308244e-05,
+      "loss": 0.2619,
+      "step": 39840
+    },
+    {
+      "epoch": 257.1,
+      "learning_rate": 4.127240143369176e-05,
+      "loss": 0.2978,
+      "step": 39850
+    },
+    {
+      "epoch": 257.16,
+      "learning_rate": 4.1268817204301074e-05,
+      "loss": 0.2769,
+      "step": 39860
+    },
+    {
+      "epoch": 257.23,
+      "learning_rate": 4.126523297491039e-05,
+      "loss": 0.2816,
+      "step": 39870
+    },
+    {
+      "epoch": 257.29,
+      "learning_rate": 4.126164874551972e-05,
+      "loss": 0.2851,
+      "step": 39880
+    },
+    {
+      "epoch": 257.35,
+      "learning_rate": 4.1258064516129036e-05,
+      "loss": 0.3206,
+      "step": 39890
+    },
+    {
+      "epoch": 257.42,
+      "learning_rate": 4.1254480286738354e-05,
+      "loss": 0.3046,
+      "step": 39900
+    },
+    {
+      "epoch": 257.48,
+      "learning_rate": 4.125089605734767e-05,
+      "loss": 0.3091,
+      "step": 39910
+    },
+    {
+      "epoch": 257.55,
+      "learning_rate": 4.124731182795699e-05,
+      "loss": 0.339,
+      "step": 39920
+    },
+    {
+      "epoch": 257.61,
+      "learning_rate": 4.1243727598566315e-05,
+      "loss": 0.273,
+      "step": 39930
+    },
+    {
+      "epoch": 257.68,
+      "learning_rate": 4.1240143369175627e-05,
+      "loss": 0.2621,
+      "step": 39940
+    },
+    {
+      "epoch": 257.74,
+      "learning_rate": 4.1236559139784945e-05,
+      "loss": 0.2698,
+      "step": 39950
+    },
+    {
+      "epoch": 257.8,
+      "learning_rate": 4.123297491039427e-05,
+      "loss": 0.2463,
+      "step": 39960
+    },
+    {
+      "epoch": 257.87,
+      "learning_rate": 4.122939068100359e-05,
+      "loss": 0.298,
+      "step": 39970
+    },
+    {
+      "epoch": 257.93,
+      "learning_rate": 4.1225806451612906e-05,
+      "loss": 0.3168,
+      "step": 39980
+    },
+    {
+      "epoch": 258.0,
+      "learning_rate": 4.1222222222222224e-05,
+      "loss": 0.2472,
+      "step": 39990
+    },
+    {
+      "epoch": 258.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.936992347240448,
+      "eval_runtime": 0.8743,
+      "eval_samples_per_second": 229.888,
+      "eval_steps_per_second": 8.006,
+      "step": 39990
+    },
+    {
+      "epoch": 258.06,
+      "learning_rate": 4.121863799283154e-05,
+      "loss": 0.272,
+      "step": 40000
+    },
+    {
+      "epoch": 258.13,
+      "learning_rate": 4.121505376344086e-05,
+      "loss": 0.255,
+      "step": 40010
+    },
+    {
+      "epoch": 258.19,
+      "learning_rate": 4.121146953405018e-05,
+      "loss": 0.327,
+      "step": 40020
+    },
+    {
+      "epoch": 258.26,
+      "learning_rate": 4.12078853046595e-05,
+      "loss": 0.29,
+      "step": 40030
+    },
+    {
+      "epoch": 258.32,
+      "learning_rate": 4.120430107526882e-05,
+      "loss": 0.2887,
+      "step": 40040
+    },
+    {
+      "epoch": 258.39,
+      "learning_rate": 4.120071684587814e-05,
+      "loss": 0.2745,
+      "step": 40050
+    },
+    {
+      "epoch": 258.45,
+      "learning_rate": 4.119713261648745e-05,
+      "loss": 0.255,
+      "step": 40060
+    },
+    {
+      "epoch": 258.51,
+      "learning_rate": 4.119354838709678e-05,
+      "loss": 0.2383,
+      "step": 40070
+    },
+    {
+      "epoch": 258.58,
+      "learning_rate": 4.1189964157706095e-05,
+      "loss": 0.281,
+      "step": 40080
+    },
+    {
+      "epoch": 258.64,
+      "learning_rate": 4.118637992831541e-05,
+      "loss": 0.2604,
+      "step": 40090
+    },
+    {
+      "epoch": 258.71,
+      "learning_rate": 4.118279569892473e-05,
+      "loss": 0.2859,
+      "step": 40100
+    },
+    {
+      "epoch": 258.77,
+      "learning_rate": 4.117921146953405e-05,
+      "loss": 0.3131,
+      "step": 40110
+    },
+    {
+      "epoch": 258.84,
+      "learning_rate": 4.1175627240143375e-05,
+      "loss": 0.2846,
+      "step": 40120
+    },
+    {
+      "epoch": 258.9,
+      "learning_rate": 4.117204301075269e-05,
+      "loss": 0.2753,
+      "step": 40130
+    },
+    {
+      "epoch": 258.96,
+      "learning_rate": 4.1168458781362004e-05,
+      "loss": 0.2443,
+      "step": 40140
+    },
+    {
+      "epoch": 259.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.011531949043274,
+      "eval_runtime": 0.8563,
+      "eval_samples_per_second": 234.719,
+      "eval_steps_per_second": 8.174,
+      "step": 40145
+    },
+    {
+      "epoch": 259.03,
+      "learning_rate": 4.116487455197133e-05,
+      "loss": 0.3181,
+      "step": 40150
+    },
+    {
+      "epoch": 259.1,
+      "learning_rate": 4.116129032258065e-05,
+      "loss": 0.2696,
+      "step": 40160
+    },
+    {
+      "epoch": 259.16,
+      "learning_rate": 4.1157706093189966e-05,
+      "loss": 0.3151,
+      "step": 40170
+    },
+    {
+      "epoch": 259.23,
+      "learning_rate": 4.1154121863799284e-05,
+      "loss": 0.2767,
+      "step": 40180
+    },
+    {
+      "epoch": 259.29,
+      "learning_rate": 4.11505376344086e-05,
+      "loss": 0.2434,
+      "step": 40190
+    },
+    {
+      "epoch": 259.35,
+      "learning_rate": 4.114695340501793e-05,
+      "loss": 0.2505,
+      "step": 40200
+    },
+    {
+      "epoch": 259.42,
+      "learning_rate": 4.1143369175627245e-05,
+      "loss": 0.3174,
+      "step": 40210
+    },
+    {
+      "epoch": 259.48,
+      "learning_rate": 4.1139784946236557e-05,
+      "loss": 0.3152,
+      "step": 40220
+    },
+    {
+      "epoch": 259.55,
+      "learning_rate": 4.113620071684588e-05,
+      "loss": 0.2892,
+      "step": 40230
+    },
+    {
+      "epoch": 259.61,
+      "learning_rate": 4.11326164874552e-05,
+      "loss": 0.2852,
+      "step": 40240
+    },
+    {
+      "epoch": 259.68,
+      "learning_rate": 4.112903225806452e-05,
+      "loss": 0.307,
+      "step": 40250
+    },
+    {
+      "epoch": 259.74,
+      "learning_rate": 4.1125448028673836e-05,
+      "loss": 0.3207,
+      "step": 40260
+    },
+    {
+      "epoch": 259.8,
+      "learning_rate": 4.1121863799283154e-05,
+      "loss": 0.3281,
+      "step": 40270
+    },
+    {
+      "epoch": 259.87,
+      "learning_rate": 4.111827956989248e-05,
+      "loss": 0.271,
+      "step": 40280
+    },
+    {
+      "epoch": 259.93,
+      "learning_rate": 4.11146953405018e-05,
+      "loss": 0.3128,
+      "step": 40290
+    },
+    {
+      "epoch": 260.0,
+      "learning_rate": 4.111111111111111e-05,
+      "loss": 0.318,
+      "step": 40300
+    },
+    {
+      "epoch": 260.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9992966055870056,
+      "eval_runtime": 0.8647,
+      "eval_samples_per_second": 232.455,
+      "eval_steps_per_second": 8.095,
+      "step": 40300
+    },
+    {
+      "epoch": 260.06,
+      "learning_rate": 4.1107526881720434e-05,
+      "loss": 0.3167,
+      "step": 40310
+    },
+    {
+      "epoch": 260.13,
+      "learning_rate": 4.110394265232975e-05,
+      "loss": 0.3087,
+      "step": 40320
+    },
+    {
+      "epoch": 260.19,
+      "learning_rate": 4.110035842293907e-05,
+      "loss": 0.2219,
+      "step": 40330
+    },
+    {
+      "epoch": 260.26,
+      "learning_rate": 4.109677419354839e-05,
+      "loss": 0.3445,
+      "step": 40340
+    },
+    {
+      "epoch": 260.32,
+      "learning_rate": 4.109318996415771e-05,
+      "loss": 0.2904,
+      "step": 40350
+    },
+    {
+      "epoch": 260.39,
+      "learning_rate": 4.1089605734767025e-05,
+      "loss": 0.2919,
+      "step": 40360
+    },
+    {
+      "epoch": 260.45,
+      "learning_rate": 4.108602150537635e-05,
+      "loss": 0.2697,
+      "step": 40370
+    },
+    {
+      "epoch": 260.51,
+      "learning_rate": 4.108243727598566e-05,
+      "loss": 0.3371,
+      "step": 40380
+    },
+    {
+      "epoch": 260.58,
+      "learning_rate": 4.1078853046594986e-05,
+      "loss": 0.2287,
+      "step": 40390
+    },
+    {
+      "epoch": 260.64,
+      "learning_rate": 4.1075268817204305e-05,
+      "loss": 0.3088,
+      "step": 40400
+    },
+    {
+      "epoch": 260.71,
+      "learning_rate": 4.107168458781362e-05,
+      "loss": 0.2825,
+      "step": 40410
+    },
+    {
+      "epoch": 260.77,
+      "learning_rate": 4.106810035842294e-05,
+      "loss": 0.2818,
+      "step": 40420
+    },
+    {
+      "epoch": 260.84,
+      "learning_rate": 4.106451612903226e-05,
+      "loss": 0.3076,
+      "step": 40430
+    },
+    {
+      "epoch": 260.9,
+      "learning_rate": 4.106093189964158e-05,
+      "loss": 0.2633,
+      "step": 40440
+    },
+    {
+      "epoch": 260.96,
+      "learning_rate": 4.10573476702509e-05,
+      "loss": 0.2786,
+      "step": 40450
+    },
+    {
+      "epoch": 261.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0467208623886108,
+      "eval_runtime": 0.8968,
+      "eval_samples_per_second": 224.128,
+      "eval_steps_per_second": 7.805,
+      "step": 40455
+    },
+    {
+      "epoch": 261.03,
+      "learning_rate": 4.1053763440860214e-05,
+      "loss": 0.2964,
+      "step": 40460
+    },
+    {
+      "epoch": 261.1,
+      "learning_rate": 4.105017921146954e-05,
+      "loss": 0.2232,
+      "step": 40470
+    },
+    {
+      "epoch": 261.16,
+      "learning_rate": 4.104659498207886e-05,
+      "loss": 0.2708,
+      "step": 40480
+    },
+    {
+      "epoch": 261.23,
+      "learning_rate": 4.104301075268817e-05,
+      "loss": 0.2611,
+      "step": 40490
+    },
+    {
+      "epoch": 261.29,
+      "learning_rate": 4.1039426523297493e-05,
+      "loss": 0.2755,
+      "step": 40500
+    },
+    {
+      "epoch": 261.35,
+      "learning_rate": 4.103584229390681e-05,
+      "loss": 0.2516,
+      "step": 40510
+    },
+    {
+      "epoch": 261.42,
+      "learning_rate": 4.103225806451613e-05,
+      "loss": 0.247,
+      "step": 40520
+    },
+    {
+      "epoch": 261.48,
+      "learning_rate": 4.102867383512545e-05,
+      "loss": 0.2823,
+      "step": 40530
+    },
+    {
+      "epoch": 261.55,
+      "learning_rate": 4.1025089605734766e-05,
+      "loss": 0.3101,
+      "step": 40540
+    },
+    {
+      "epoch": 261.61,
+      "learning_rate": 4.102150537634409e-05,
+      "loss": 0.2992,
+      "step": 40550
+    },
+    {
+      "epoch": 261.68,
+      "learning_rate": 4.101792114695341e-05,
+      "loss": 0.3076,
+      "step": 40560
+    },
+    {
+      "epoch": 261.74,
+      "learning_rate": 4.101433691756272e-05,
+      "loss": 0.267,
+      "step": 40570
+    },
+    {
+      "epoch": 261.8,
+      "learning_rate": 4.1010752688172046e-05,
+      "loss": 0.2501,
+      "step": 40580
+    },
+    {
+      "epoch": 261.87,
+      "learning_rate": 4.1007168458781364e-05,
+      "loss": 0.2948,
+      "step": 40590
+    },
+    {
+      "epoch": 261.93,
+      "learning_rate": 4.100358422939068e-05,
+      "loss": 0.2771,
+      "step": 40600
+    },
+    {
+      "epoch": 262.0,
+      "learning_rate": 4.1e-05,
+      "loss": 0.2812,
+      "step": 40610
+    },
+    {
+      "epoch": 262.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0275951623916626,
+      "eval_runtime": 0.8487,
+      "eval_samples_per_second": 236.842,
+      "eval_steps_per_second": 8.248,
+      "step": 40610
+    },
+    {
+      "epoch": 262.06,
+      "learning_rate": 4.099641577060932e-05,
+      "loss": 0.2724,
+      "step": 40620
+    },
+    {
+      "epoch": 262.13,
+      "learning_rate": 4.0992831541218644e-05,
+      "loss": 0.2683,
+      "step": 40630
+    },
+    {
+      "epoch": 262.19,
+      "learning_rate": 4.098924731182796e-05,
+      "loss": 0.2541,
+      "step": 40640
+    },
+    {
+      "epoch": 262.26,
+      "learning_rate": 4.098566308243727e-05,
+      "loss": 0.2731,
+      "step": 40650
+    },
+    {
+      "epoch": 262.32,
+      "learning_rate": 4.09820788530466e-05,
+      "loss": 0.2544,
+      "step": 40660
+    },
+    {
+      "epoch": 262.39,
+      "learning_rate": 4.0978494623655916e-05,
+      "loss": 0.2562,
+      "step": 40670
+    },
+    {
+      "epoch": 262.45,
+      "learning_rate": 4.0974910394265235e-05,
+      "loss": 0.3103,
+      "step": 40680
+    },
+    {
+      "epoch": 262.51,
+      "learning_rate": 4.097132616487455e-05,
+      "loss": 0.3117,
+      "step": 40690
+    },
+    {
+      "epoch": 262.58,
+      "learning_rate": 4.096774193548387e-05,
+      "loss": 0.2645,
+      "step": 40700
+    },
+    {
+      "epoch": 262.64,
+      "learning_rate": 4.096415770609319e-05,
+      "loss": 0.2539,
+      "step": 40710
+    },
+    {
+      "epoch": 262.71,
+      "learning_rate": 4.0960573476702514e-05,
+      "loss": 0.2913,
+      "step": 40720
+    },
+    {
+      "epoch": 262.77,
+      "learning_rate": 4.0956989247311826e-05,
+      "loss": 0.2888,
+      "step": 40730
+    },
+    {
+      "epoch": 262.84,
+      "learning_rate": 4.095340501792115e-05,
+      "loss": 0.2456,
+      "step": 40740
+    },
+    {
+      "epoch": 262.9,
+      "learning_rate": 4.094982078853047e-05,
+      "loss": 0.2915,
+      "step": 40750
+    },
+    {
+      "epoch": 262.96,
+      "learning_rate": 4.094623655913979e-05,
+      "loss": 0.3537,
+      "step": 40760
+    },
+    {
+      "epoch": 263.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.1189571619033813,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.116,
+      "eval_steps_per_second": 8.153,
+      "step": 40765
+    },
+    {
+      "epoch": 263.03,
+      "learning_rate": 4.0942652329749105e-05,
+      "loss": 0.3251,
+      "step": 40770
+    },
+    {
+      "epoch": 263.1,
+      "learning_rate": 4.0939068100358424e-05,
+      "loss": 0.3125,
+      "step": 40780
+    },
+    {
+      "epoch": 263.16,
+      "learning_rate": 4.093548387096774e-05,
+      "loss": 0.2875,
+      "step": 40790
+    },
+    {
+      "epoch": 263.23,
+      "learning_rate": 4.093189964157707e-05,
+      "loss": 0.2864,
+      "step": 40800
+    },
+    {
+      "epoch": 263.29,
+      "learning_rate": 4.092831541218638e-05,
+      "loss": 0.2704,
+      "step": 40810
+    },
+    {
+      "epoch": 263.35,
+      "learning_rate": 4.09247311827957e-05,
+      "loss": 0.311,
+      "step": 40820
+    },
+    {
+      "epoch": 263.42,
+      "learning_rate": 4.092114695340502e-05,
+      "loss": 0.2589,
+      "step": 40830
+    },
+    {
+      "epoch": 263.48,
+      "learning_rate": 4.091756272401434e-05,
+      "loss": 0.313,
+      "step": 40840
+    },
+    {
+      "epoch": 263.55,
+      "learning_rate": 4.091397849462366e-05,
+      "loss": 0.3114,
+      "step": 40850
+    },
+    {
+      "epoch": 263.61,
+      "learning_rate": 4.0910394265232976e-05,
+      "loss": 0.319,
+      "step": 40860
+    },
+    {
+      "epoch": 263.68,
+      "learning_rate": 4.0906810035842294e-05,
+      "loss": 0.2721,
+      "step": 40870
+    },
+    {
+      "epoch": 263.74,
+      "learning_rate": 4.090322580645162e-05,
+      "loss": 0.3145,
+      "step": 40880
+    },
+    {
+      "epoch": 263.8,
+      "learning_rate": 4.089964157706093e-05,
+      "loss": 0.3502,
+      "step": 40890
+    },
+    {
+      "epoch": 263.87,
+      "learning_rate": 4.0896057347670256e-05,
+      "loss": 0.292,
+      "step": 40900
+    },
+    {
+      "epoch": 263.93,
+      "learning_rate": 4.0892473118279574e-05,
+      "loss": 0.2262,
+      "step": 40910
+    },
+    {
+      "epoch": 264.0,
+      "learning_rate": 4.088888888888889e-05,
+      "loss": 0.3076,
+      "step": 40920
+    },
+    {
+      "epoch": 264.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 1.0411542654037476,
+      "eval_runtime": 0.8653,
+      "eval_samples_per_second": 232.298,
+      "eval_steps_per_second": 8.09,
+      "step": 40920
+    },
+    {
+      "epoch": 264.06,
+      "learning_rate": 4.088530465949821e-05,
+      "loss": 0.3258,
+      "step": 40930
+    },
+    {
+      "epoch": 264.13,
+      "learning_rate": 4.088172043010753e-05,
+      "loss": 0.2724,
+      "step": 40940
+    },
+    {
+      "epoch": 264.19,
+      "learning_rate": 4.0878136200716847e-05,
+      "loss": 0.2799,
+      "step": 40950
+    },
+    {
+      "epoch": 264.26,
+      "learning_rate": 4.087455197132617e-05,
+      "loss": 0.2864,
+      "step": 40960
+    },
+    {
+      "epoch": 264.32,
+      "learning_rate": 4.087096774193548e-05,
+      "loss": 0.2689,
+      "step": 40970
+    },
+    {
+      "epoch": 264.39,
+      "learning_rate": 4.086738351254481e-05,
+      "loss": 0.2897,
+      "step": 40980
+    },
+    {
+      "epoch": 264.45,
+      "learning_rate": 4.0863799283154126e-05,
+      "loss": 0.3314,
+      "step": 40990
+    },
+    {
+      "epoch": 264.51,
+      "learning_rate": 4.0860215053763444e-05,
+      "loss": 0.3569,
+      "step": 41000
+    },
+    {
+      "epoch": 264.58,
+      "learning_rate": 4.085663082437276e-05,
+      "loss": 0.2944,
+      "step": 41010
+    },
+    {
+      "epoch": 264.64,
+      "learning_rate": 4.085304659498208e-05,
+      "loss": 0.2952,
+      "step": 41020
+    },
+    {
+      "epoch": 264.71,
+      "learning_rate": 4.08494623655914e-05,
+      "loss": 0.259,
+      "step": 41030
+    },
+    {
+      "epoch": 264.77,
+      "learning_rate": 4.0845878136200724e-05,
+      "loss": 0.2544,
+      "step": 41040
+    },
+    {
+      "epoch": 264.84,
+      "learning_rate": 4.0842293906810035e-05,
+      "loss": 0.2699,
+      "step": 41050
+    },
+    {
+      "epoch": 264.9,
+      "learning_rate": 4.0838709677419354e-05,
+      "loss": 0.3212,
+      "step": 41060
+    },
+    {
+      "epoch": 264.96,
+      "learning_rate": 4.083512544802868e-05,
+      "loss": 0.2766,
+      "step": 41070
+    },
+    {
+      "epoch": 265.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.1231948137283325,
+      "eval_runtime": 0.878,
+      "eval_samples_per_second": 228.922,
+      "eval_steps_per_second": 7.972,
+      "step": 41075
+    },
+    {
+      "epoch": 265.03,
+      "learning_rate": 4.0831541218638e-05,
+      "loss": 0.2831,
+      "step": 41080
+    },
+    {
+      "epoch": 265.1,
+      "learning_rate": 4.0827956989247315e-05,
+      "loss": 0.2437,
+      "step": 41090
+    },
+    {
+      "epoch": 265.16,
+      "learning_rate": 4.082437275985663e-05,
+      "loss": 0.2469,
+      "step": 41100
+    },
+    {
+      "epoch": 265.23,
+      "learning_rate": 4.082078853046595e-05,
+      "loss": 0.3254,
+      "step": 41110
+    },
+    {
+      "epoch": 265.29,
+      "learning_rate": 4.081720430107527e-05,
+      "loss": 0.274,
+      "step": 41120
+    },
+    {
+      "epoch": 265.35,
+      "learning_rate": 4.081362007168459e-05,
+      "loss": 0.2615,
+      "step": 41130
+    },
+    {
+      "epoch": 265.42,
+      "learning_rate": 4.0810035842293906e-05,
+      "loss": 0.2957,
+      "step": 41140
+    },
+    {
+      "epoch": 265.48,
+      "learning_rate": 4.080645161290323e-05,
+      "loss": 0.2798,
+      "step": 41150
+    },
+    {
+      "epoch": 265.55,
+      "learning_rate": 4.080286738351254e-05,
+      "loss": 0.3234,
+      "step": 41160
+    },
+    {
+      "epoch": 265.61,
+      "learning_rate": 4.079928315412187e-05,
+      "loss": 0.3086,
+      "step": 41170
+    },
+    {
+      "epoch": 265.68,
+      "learning_rate": 4.0795698924731186e-05,
+      "loss": 0.2808,
+      "step": 41180
+    },
+    {
+      "epoch": 265.74,
+      "learning_rate": 4.0792114695340504e-05,
+      "loss": 0.2431,
+      "step": 41190
+    },
+    {
+      "epoch": 265.8,
+      "learning_rate": 4.078853046594982e-05,
+      "loss": 0.286,
+      "step": 41200
+    },
+    {
+      "epoch": 265.87,
+      "learning_rate": 4.078494623655914e-05,
+      "loss": 0.3455,
+      "step": 41210
+    },
+    {
+      "epoch": 265.93,
+      "learning_rate": 4.078136200716846e-05,
+      "loss": 0.2908,
+      "step": 41220
+    },
+    {
+      "epoch": 266.0,
+      "learning_rate": 4.0777777777777783e-05,
+      "loss": 0.2552,
+      "step": 41230
+    },
+    {
+      "epoch": 266.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0698556900024414,
+      "eval_runtime": 0.8541,
+      "eval_samples_per_second": 235.331,
+      "eval_steps_per_second": 8.196,
+      "step": 41230
+    },
+    {
+      "epoch": 266.06,
+      "learning_rate": 4.0774193548387095e-05,
+      "loss": 0.2725,
+      "step": 41240
+    },
+    {
+      "epoch": 266.13,
+      "learning_rate": 4.077060931899642e-05,
+      "loss": 0.1796,
+      "step": 41250
+    },
+    {
+      "epoch": 266.19,
+      "learning_rate": 4.076702508960574e-05,
+      "loss": 0.2647,
+      "step": 41260
+    },
+    {
+      "epoch": 266.26,
+      "learning_rate": 4.0763440860215056e-05,
+      "loss": 0.2815,
+      "step": 41270
+    },
+    {
+      "epoch": 266.32,
+      "learning_rate": 4.0759856630824374e-05,
+      "loss": 0.3412,
+      "step": 41280
+    },
+    {
+      "epoch": 266.39,
+      "learning_rate": 4.075627240143369e-05,
+      "loss": 0.3366,
+      "step": 41290
+    },
+    {
+      "epoch": 266.45,
+      "learning_rate": 4.075268817204301e-05,
+      "loss": 0.2879,
+      "step": 41300
+    },
+    {
+      "epoch": 266.51,
+      "learning_rate": 4.0749103942652336e-05,
+      "loss": 0.3075,
+      "step": 41310
+    },
+    {
+      "epoch": 266.58,
+      "learning_rate": 4.074551971326165e-05,
+      "loss": 0.2584,
+      "step": 41320
+    },
+    {
+      "epoch": 266.64,
+      "learning_rate": 4.0741935483870965e-05,
+      "loss": 0.2817,
+      "step": 41330
+    },
+    {
+      "epoch": 266.71,
+      "learning_rate": 4.073835125448029e-05,
+      "loss": 0.2835,
+      "step": 41340
+    },
+    {
+      "epoch": 266.77,
+      "learning_rate": 4.073476702508961e-05,
+      "loss": 0.2724,
+      "step": 41350
+    },
+    {
+      "epoch": 266.84,
+      "learning_rate": 4.073118279569893e-05,
+      "loss": 0.2607,
+      "step": 41360
+    },
+    {
+      "epoch": 266.9,
+      "learning_rate": 4.0727598566308245e-05,
+      "loss": 0.2982,
+      "step": 41370
+    },
+    {
+      "epoch": 266.96,
+      "learning_rate": 4.072401433691756e-05,
+      "loss": 0.2666,
+      "step": 41380
+    },
+    {
+      "epoch": 267.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.060293436050415,
+      "eval_runtime": 0.8802,
+      "eval_samples_per_second": 228.358,
+      "eval_steps_per_second": 7.953,
+      "step": 41385
+    },
+    {
+      "epoch": 267.03,
+      "learning_rate": 4.072043010752689e-05,
+      "loss": 0.3092,
+      "step": 41390
+    },
+    {
+      "epoch": 267.1,
+      "learning_rate": 4.07168458781362e-05,
+      "loss": 0.278,
+      "step": 41400
+    },
+    {
+      "epoch": 267.16,
+      "learning_rate": 4.071326164874552e-05,
+      "loss": 0.2862,
+      "step": 41410
+    },
+    {
+      "epoch": 267.23,
+      "learning_rate": 4.070967741935484e-05,
+      "loss": 0.3027,
+      "step": 41420
+    },
+    {
+      "epoch": 267.29,
+      "learning_rate": 4.070609318996416e-05,
+      "loss": 0.2913,
+      "step": 41430
+    },
+    {
+      "epoch": 267.35,
+      "learning_rate": 4.070250896057348e-05,
+      "loss": 0.2946,
+      "step": 41440
+    },
+    {
+      "epoch": 267.42,
+      "learning_rate": 4.06989247311828e-05,
+      "loss": 0.3428,
+      "step": 41450
+    },
+    {
+      "epoch": 267.48,
+      "learning_rate": 4.0695340501792116e-05,
+      "loss": 0.3235,
+      "step": 41460
+    },
+    {
+      "epoch": 267.55,
+      "learning_rate": 4.069175627240144e-05,
+      "loss": 0.2607,
+      "step": 41470
+    },
+    {
+      "epoch": 267.61,
+      "learning_rate": 4.068817204301075e-05,
+      "loss": 0.2413,
+      "step": 41480
+    },
+    {
+      "epoch": 267.68,
+      "learning_rate": 4.068458781362007e-05,
+      "loss": 0.309,
+      "step": 41490
+    },
+    {
+      "epoch": 267.74,
+      "learning_rate": 4.0681003584229395e-05,
+      "loss": 0.2696,
+      "step": 41500
+    },
+    {
+      "epoch": 267.8,
+      "learning_rate": 4.0677419354838713e-05,
+      "loss": 0.2941,
+      "step": 41510
+    },
+    {
+      "epoch": 267.87,
+      "learning_rate": 4.067383512544803e-05,
+      "loss": 0.2319,
+      "step": 41520
+    },
+    {
+      "epoch": 267.93,
+      "learning_rate": 4.067025089605735e-05,
+      "loss": 0.2413,
+      "step": 41530
+    },
+    {
+      "epoch": 268.0,
+      "learning_rate": 4.066666666666667e-05,
+      "loss": 0.3115,
+      "step": 41540
+    },
+    {
+      "epoch": 268.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0449004173278809,
+      "eval_runtime": 0.8714,
+      "eval_samples_per_second": 230.667,
+      "eval_steps_per_second": 8.033,
+      "step": 41540
+    },
+    {
+      "epoch": 268.06,
+      "learning_rate": 4.066308243727599e-05,
+      "loss": 0.2984,
+      "step": 41550
+    },
+    {
+      "epoch": 268.13,
+      "learning_rate": 4.0659498207885304e-05,
+      "loss": 0.2534,
+      "step": 41560
+    },
+    {
+      "epoch": 268.19,
+      "learning_rate": 4.065591397849462e-05,
+      "loss": 0.2641,
+      "step": 41570
+    },
+    {
+      "epoch": 268.26,
+      "learning_rate": 4.065232974910395e-05,
+      "loss": 0.2973,
+      "step": 41580
+    },
+    {
+      "epoch": 268.32,
+      "learning_rate": 4.0648745519713266e-05,
+      "loss": 0.2856,
+      "step": 41590
+    },
+    {
+      "epoch": 268.39,
+      "learning_rate": 4.0645161290322584e-05,
+      "loss": 0.2631,
+      "step": 41600
+    },
+    {
+      "epoch": 268.45,
+      "learning_rate": 4.06415770609319e-05,
+      "loss": 0.2547,
+      "step": 41610
+    },
+    {
+      "epoch": 268.51,
+      "learning_rate": 4.063799283154122e-05,
+      "loss": 0.2741,
+      "step": 41620
+    },
+    {
+      "epoch": 268.58,
+      "learning_rate": 4.0634408602150545e-05,
+      "loss": 0.2934,
+      "step": 41630
+    },
+    {
+      "epoch": 268.64,
+      "learning_rate": 4.063082437275986e-05,
+      "loss": 0.2833,
+      "step": 41640
+    },
+    {
+      "epoch": 268.71,
+      "learning_rate": 4.0627240143369175e-05,
+      "loss": 0.2782,
+      "step": 41650
+    },
+    {
+      "epoch": 268.77,
+      "learning_rate": 4.06236559139785e-05,
+      "loss": 0.3051,
+      "step": 41660
+    },
+    {
+      "epoch": 268.84,
+      "learning_rate": 4.062007168458782e-05,
+      "loss": 0.2588,
+      "step": 41670
+    },
+    {
+      "epoch": 268.9,
+      "learning_rate": 4.061648745519713e-05,
+      "loss": 0.2735,
+      "step": 41680
+    },
+    {
+      "epoch": 268.96,
+      "learning_rate": 4.0612903225806455e-05,
+      "loss": 0.3252,
+      "step": 41690
+    },
+    {
+      "epoch": 269.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9591686725616455,
+      "eval_runtime": 0.8766,
+      "eval_samples_per_second": 229.296,
+      "eval_steps_per_second": 7.985,
+      "step": 41695
+    },
+    {
+      "epoch": 269.03,
+      "learning_rate": 4.060931899641577e-05,
+      "loss": 0.3401,
+      "step": 41700
+    },
+    {
+      "epoch": 269.1,
+      "learning_rate": 4.060573476702509e-05,
+      "loss": 0.2862,
+      "step": 41710
+    },
+    {
+      "epoch": 269.16,
+      "learning_rate": 4.060215053763441e-05,
+      "loss": 0.2579,
+      "step": 41720
+    },
+    {
+      "epoch": 269.23,
+      "learning_rate": 4.059856630824373e-05,
+      "loss": 0.2742,
+      "step": 41730
+    },
+    {
+      "epoch": 269.29,
+      "learning_rate": 4.059498207885305e-05,
+      "loss": 0.3058,
+      "step": 41740
+    },
+    {
+      "epoch": 269.35,
+      "learning_rate": 4.0591397849462364e-05,
+      "loss": 0.3193,
+      "step": 41750
+    },
+    {
+      "epoch": 269.42,
+      "learning_rate": 4.058781362007168e-05,
+      "loss": 0.31,
+      "step": 41760
+    },
+    {
+      "epoch": 269.48,
+      "learning_rate": 4.058422939068101e-05,
+      "loss": 0.306,
+      "step": 41770
+    },
+    {
+      "epoch": 269.55,
+      "learning_rate": 4.0580645161290325e-05,
+      "loss": 0.286,
+      "step": 41780
+    },
+    {
+      "epoch": 269.61,
+      "learning_rate": 4.0577060931899644e-05,
+      "loss": 0.2465,
+      "step": 41790
+    },
+    {
+      "epoch": 269.68,
+      "learning_rate": 4.057347670250896e-05,
+      "loss": 0.2868,
+      "step": 41800
+    },
+    {
+      "epoch": 269.74,
+      "learning_rate": 4.056989247311828e-05,
+      "loss": 0.2719,
+      "step": 41810
+    },
+    {
+      "epoch": 269.8,
+      "learning_rate": 4.0566308243727605e-05,
+      "loss": 0.272,
+      "step": 41820
+    },
+    {
+      "epoch": 269.87,
+      "learning_rate": 4.0562724014336916e-05,
+      "loss": 0.2691,
+      "step": 41830
+    },
+    {
+      "epoch": 269.93,
+      "learning_rate": 4.0559139784946235e-05,
+      "loss": 0.336,
+      "step": 41840
+    },
+    {
+      "epoch": 270.0,
+      "learning_rate": 4.055555555555556e-05,
+      "loss": 0.2777,
+      "step": 41850
+    },
+    {
+      "epoch": 270.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0255988836288452,
+      "eval_runtime": 0.872,
+      "eval_samples_per_second": 230.497,
+      "eval_steps_per_second": 8.027,
+      "step": 41850
+    },
+    {
+      "epoch": 270.06,
+      "learning_rate": 4.055197132616488e-05,
+      "loss": 0.3303,
+      "step": 41860
+    },
+    {
+      "epoch": 270.13,
+      "learning_rate": 4.0548387096774196e-05,
+      "loss": 0.2832,
+      "step": 41870
+    },
+    {
+      "epoch": 270.19,
+      "learning_rate": 4.0544802867383514e-05,
+      "loss": 0.2707,
+      "step": 41880
+    },
+    {
+      "epoch": 270.26,
+      "learning_rate": 4.054121863799283e-05,
+      "loss": 0.2448,
+      "step": 41890
+    },
+    {
+      "epoch": 270.32,
+      "learning_rate": 4.053763440860216e-05,
+      "loss": 0.2818,
+      "step": 41900
+    },
+    {
+      "epoch": 270.39,
+      "learning_rate": 4.053405017921147e-05,
+      "loss": 0.26,
+      "step": 41910
+    },
+    {
+      "epoch": 270.45,
+      "learning_rate": 4.053046594982079e-05,
+      "loss": 0.28,
+      "step": 41920
+    },
+    {
+      "epoch": 270.51,
+      "learning_rate": 4.052688172043011e-05,
+      "loss": 0.2834,
+      "step": 41930
+    },
+    {
+      "epoch": 270.58,
+      "learning_rate": 4.052329749103943e-05,
+      "loss": 0.2635,
+      "step": 41940
+    },
+    {
+      "epoch": 270.64,
+      "learning_rate": 4.051971326164875e-05,
+      "loss": 0.2519,
+      "step": 41950
+    },
+    {
+      "epoch": 270.71,
+      "learning_rate": 4.0516129032258067e-05,
+      "loss": 0.2882,
+      "step": 41960
+    },
+    {
+      "epoch": 270.77,
+      "learning_rate": 4.0512544802867385e-05,
+      "loss": 0.287,
+      "step": 41970
+    },
+    {
+      "epoch": 270.84,
+      "learning_rate": 4.050896057347671e-05,
+      "loss": 0.3262,
+      "step": 41980
+    },
+    {
+      "epoch": 270.9,
+      "learning_rate": 4.050537634408602e-05,
+      "loss": 0.3234,
+      "step": 41990
+    },
+    {
+      "epoch": 270.96,
+      "learning_rate": 4.050179211469534e-05,
+      "loss": 0.3495,
+      "step": 42000
+    },
+    {
+      "epoch": 271.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.006808876991272,
+      "eval_runtime": 0.871,
+      "eval_samples_per_second": 230.769,
+      "eval_steps_per_second": 8.037,
+      "step": 42005
+    },
+    {
+      "epoch": 271.03,
+      "learning_rate": 4.0498207885304664e-05,
+      "loss": 0.3319,
+      "step": 42010
+    },
+    {
+      "epoch": 271.1,
+      "learning_rate": 4.049462365591398e-05,
+      "loss": 0.3283,
+      "step": 42020
+    },
+    {
+      "epoch": 271.16,
+      "learning_rate": 4.0491039426523294e-05,
+      "loss": 0.2702,
+      "step": 42030
+    },
+    {
+      "epoch": 271.23,
+      "learning_rate": 4.048745519713262e-05,
+      "loss": 0.2653,
+      "step": 42040
+    },
+    {
+      "epoch": 271.29,
+      "learning_rate": 4.048387096774194e-05,
+      "loss": 0.2903,
+      "step": 42050
+    },
+    {
+      "epoch": 271.35,
+      "learning_rate": 4.0480286738351255e-05,
+      "loss": 0.2904,
+      "step": 42060
+    },
+    {
+      "epoch": 271.42,
+      "learning_rate": 4.0476702508960574e-05,
+      "loss": 0.2568,
+      "step": 42070
+    },
+    {
+      "epoch": 271.48,
+      "learning_rate": 4.047311827956989e-05,
+      "loss": 0.2704,
+      "step": 42080
+    },
+    {
+      "epoch": 271.55,
+      "learning_rate": 4.046953405017922e-05,
+      "loss": 0.2455,
+      "step": 42090
+    },
+    {
+      "epoch": 271.61,
+      "learning_rate": 4.0465949820788535e-05,
+      "loss": 0.2752,
+      "step": 42100
+    },
+    {
+      "epoch": 271.68,
+      "learning_rate": 4.0462365591397846e-05,
+      "loss": 0.2766,
+      "step": 42110
+    },
+    {
+      "epoch": 271.74,
+      "learning_rate": 4.045878136200717e-05,
+      "loss": 0.26,
+      "step": 42120
+    },
+    {
+      "epoch": 271.8,
+      "learning_rate": 4.045519713261649e-05,
+      "loss": 0.2753,
+      "step": 42130
+    },
+    {
+      "epoch": 271.87,
+      "learning_rate": 4.045161290322581e-05,
+      "loss": 0.2823,
+      "step": 42140
+    },
+    {
+      "epoch": 271.93,
+      "learning_rate": 4.0448028673835126e-05,
+      "loss": 0.2865,
+      "step": 42150
+    },
+    {
+      "epoch": 272.0,
+      "learning_rate": 4.0444444444444444e-05,
+      "loss": 0.3245,
+      "step": 42160
+    },
+    {
+      "epoch": 272.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0269964933395386,
+      "eval_runtime": 0.8763,
+      "eval_samples_per_second": 229.37,
+      "eval_steps_per_second": 7.988,
+      "step": 42160
+    },
+    {
+      "epoch": 272.06,
+      "learning_rate": 4.044086021505377e-05,
+      "loss": 0.3393,
+      "step": 42170
+    },
+    {
+      "epoch": 272.13,
+      "learning_rate": 4.043727598566309e-05,
+      "loss": 0.269,
+      "step": 42180
+    },
+    {
+      "epoch": 272.19,
+      "learning_rate": 4.04336917562724e-05,
+      "loss": 0.2638,
+      "step": 42190
+    },
+    {
+      "epoch": 272.26,
+      "learning_rate": 4.0430107526881724e-05,
+      "loss": 0.2679,
+      "step": 42200
+    },
+    {
+      "epoch": 272.32,
+      "learning_rate": 4.042652329749104e-05,
+      "loss": 0.2877,
+      "step": 42210
+    },
+    {
+      "epoch": 272.39,
+      "learning_rate": 4.042293906810036e-05,
+      "loss": 0.269,
+      "step": 42220
+    },
+    {
+      "epoch": 272.45,
+      "learning_rate": 4.041935483870968e-05,
+      "loss": 0.3029,
+      "step": 42230
+    },
+    {
+      "epoch": 272.51,
+      "learning_rate": 4.0415770609319e-05,
+      "loss": 0.2859,
+      "step": 42240
+    },
+    {
+      "epoch": 272.58,
+      "learning_rate": 4.041218637992832e-05,
+      "loss": 0.2848,
+      "step": 42250
+    },
+    {
+      "epoch": 272.64,
+      "learning_rate": 4.040860215053764e-05,
+      "loss": 0.292,
+      "step": 42260
+    },
+    {
+      "epoch": 272.71,
+      "learning_rate": 4.040501792114695e-05,
+      "loss": 0.2405,
+      "step": 42270
+    },
+    {
+      "epoch": 272.77,
+      "learning_rate": 4.0401433691756276e-05,
+      "loss": 0.3395,
+      "step": 42280
+    },
+    {
+      "epoch": 272.84,
+      "learning_rate": 4.0397849462365594e-05,
+      "loss": 0.3035,
+      "step": 42290
+    },
+    {
+      "epoch": 272.9,
+      "learning_rate": 4.039426523297491e-05,
+      "loss": 0.2578,
+      "step": 42300
+    },
+    {
+      "epoch": 272.96,
+      "learning_rate": 4.039068100358423e-05,
+      "loss": 0.2963,
+      "step": 42310
+    },
+    {
+      "epoch": 273.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.1183547973632812,
+      "eval_runtime": 0.8963,
+      "eval_samples_per_second": 224.263,
+      "eval_steps_per_second": 7.81,
+      "step": 42315
+    },
+    {
+      "epoch": 273.03,
+      "learning_rate": 4.038709677419355e-05,
+      "loss": 0.3796,
+      "step": 42320
+    },
+    {
+      "epoch": 273.1,
+      "learning_rate": 4.038351254480287e-05,
+      "loss": 0.2418,
+      "step": 42330
+    },
+    {
+      "epoch": 273.16,
+      "learning_rate": 4.037992831541219e-05,
+      "loss": 0.2648,
+      "step": 42340
+    },
+    {
+      "epoch": 273.23,
+      "learning_rate": 4.0376344086021504e-05,
+      "loss": 0.2337,
+      "step": 42350
+    },
+    {
+      "epoch": 273.29,
+      "learning_rate": 4.037275985663083e-05,
+      "loss": 0.2649,
+      "step": 42360
+    },
+    {
+      "epoch": 273.35,
+      "learning_rate": 4.036917562724015e-05,
+      "loss": 0.3136,
+      "step": 42370
+    },
+    {
+      "epoch": 273.42,
+      "learning_rate": 4.036559139784946e-05,
+      "loss": 0.2824,
+      "step": 42380
+    },
+    {
+      "epoch": 273.48,
+      "learning_rate": 4.036200716845878e-05,
+      "loss": 0.2713,
+      "step": 42390
+    },
+    {
+      "epoch": 273.55,
+      "learning_rate": 4.03584229390681e-05,
+      "loss": 0.2983,
+      "step": 42400
+    },
+    {
+      "epoch": 273.61,
+      "learning_rate": 4.035483870967742e-05,
+      "loss": 0.2923,
+      "step": 42410
+    },
+    {
+      "epoch": 273.68,
+      "learning_rate": 4.035125448028674e-05,
+      "loss": 0.3354,
+      "step": 42420
+    },
+    {
+      "epoch": 273.74,
+      "learning_rate": 4.0347670250896056e-05,
+      "loss": 0.2533,
+      "step": 42430
+    },
+    {
+      "epoch": 273.8,
+      "learning_rate": 4.034408602150538e-05,
+      "loss": 0.274,
+      "step": 42440
+    },
+    {
+      "epoch": 273.87,
+      "learning_rate": 4.03405017921147e-05,
+      "loss": 0.313,
+      "step": 42450
+    },
+    {
+      "epoch": 273.93,
+      "learning_rate": 4.033691756272401e-05,
+      "loss": 0.2723,
+      "step": 42460
+    },
+    {
+      "epoch": 274.0,
+      "learning_rate": 4.0333333333333336e-05,
+      "loss": 0.2836,
+      "step": 42470
+    },
+    {
+      "epoch": 274.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0326701402664185,
+      "eval_runtime": 0.8852,
+      "eval_samples_per_second": 227.057,
+      "eval_steps_per_second": 7.907,
+      "step": 42470
+    },
+    {
+      "epoch": 274.06,
+      "learning_rate": 4.0329749103942654e-05,
+      "loss": 0.2941,
+      "step": 42480
+    },
+    {
+      "epoch": 274.13,
+      "learning_rate": 4.032616487455197e-05,
+      "loss": 0.3171,
+      "step": 42490
+    },
+    {
+      "epoch": 274.19,
+      "learning_rate": 4.032258064516129e-05,
+      "loss": 0.2871,
+      "step": 42500
+    },
+    {
+      "epoch": 274.26,
+      "learning_rate": 4.031899641577061e-05,
+      "loss": 0.2906,
+      "step": 42510
+    },
+    {
+      "epoch": 274.32,
+      "learning_rate": 4.0315412186379933e-05,
+      "loss": 0.2943,
+      "step": 42520
+    },
+    {
+      "epoch": 274.39,
+      "learning_rate": 4.031182795698925e-05,
+      "loss": 0.299,
+      "step": 42530
+    },
+    {
+      "epoch": 274.45,
+      "learning_rate": 4.030824372759856e-05,
+      "loss": 0.2853,
+      "step": 42540
+    },
+    {
+      "epoch": 274.51,
+      "learning_rate": 4.030465949820789e-05,
+      "loss": 0.3674,
+      "step": 42550
+    },
+    {
+      "epoch": 274.58,
+      "learning_rate": 4.0301075268817206e-05,
+      "loss": 0.2417,
+      "step": 42560
+    },
+    {
+      "epoch": 274.64,
+      "learning_rate": 4.0297491039426524e-05,
+      "loss": 0.2429,
+      "step": 42570
+    },
+    {
+      "epoch": 274.71,
+      "learning_rate": 4.029390681003584e-05,
+      "loss": 0.2869,
+      "step": 42580
+    },
+    {
+      "epoch": 274.77,
+      "learning_rate": 4.029032258064516e-05,
+      "loss": 0.3143,
+      "step": 42590
+    },
+    {
+      "epoch": 274.84,
+      "learning_rate": 4.0286738351254486e-05,
+      "loss": 0.2598,
+      "step": 42600
+    },
+    {
+      "epoch": 274.9,
+      "learning_rate": 4.0283154121863804e-05,
+      "loss": 0.3021,
+      "step": 42610
+    },
+    {
+      "epoch": 274.96,
+      "learning_rate": 4.0279569892473116e-05,
+      "loss": 0.2963,
+      "step": 42620
+    },
+    {
+      "epoch": 275.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0137828588485718,
+      "eval_runtime": 0.8784,
+      "eval_samples_per_second": 228.827,
+      "eval_steps_per_second": 7.969,
+      "step": 42625
+    },
+    {
+      "epoch": 275.03,
+      "learning_rate": 4.027598566308244e-05,
+      "loss": 0.2956,
+      "step": 42630
+    },
+    {
+      "epoch": 275.1,
+      "learning_rate": 4.027240143369176e-05,
+      "loss": 0.2659,
+      "step": 42640
+    },
+    {
+      "epoch": 275.16,
+      "learning_rate": 4.026881720430108e-05,
+      "loss": 0.2995,
+      "step": 42650
+    },
+    {
+      "epoch": 275.23,
+      "learning_rate": 4.0265232974910395e-05,
+      "loss": 0.3259,
+      "step": 42660
+    },
+    {
+      "epoch": 275.29,
+      "learning_rate": 4.026164874551971e-05,
+      "loss": 0.281,
+      "step": 42670
+    },
+    {
+      "epoch": 275.35,
+      "learning_rate": 4.025806451612903e-05,
+      "loss": 0.2539,
+      "step": 42680
+    },
+    {
+      "epoch": 275.42,
+      "learning_rate": 4.0254480286738357e-05,
+      "loss": 0.3375,
+      "step": 42690
+    },
+    {
+      "epoch": 275.48,
+      "learning_rate": 4.025089605734767e-05,
+      "loss": 0.2776,
+      "step": 42700
+    },
+    {
+      "epoch": 275.55,
+      "learning_rate": 4.024731182795699e-05,
+      "loss": 0.2708,
+      "step": 42710
+    },
+    {
+      "epoch": 275.61,
+      "learning_rate": 4.024372759856631e-05,
+      "loss": 0.2928,
+      "step": 42720
+    },
+    {
+      "epoch": 275.68,
+      "learning_rate": 4.024014336917563e-05,
+      "loss": 0.3094,
+      "step": 42730
+    },
+    {
+      "epoch": 275.74,
+      "learning_rate": 4.023655913978495e-05,
+      "loss": 0.2595,
+      "step": 42740
+    },
+    {
+      "epoch": 275.8,
+      "learning_rate": 4.0232974910394266e-05,
+      "loss": 0.2743,
+      "step": 42750
+    },
+    {
+      "epoch": 275.87,
+      "learning_rate": 4.0229390681003584e-05,
+      "loss": 0.2907,
+      "step": 42760
+    },
+    {
+      "epoch": 275.93,
+      "learning_rate": 4.022580645161291e-05,
+      "loss": 0.2875,
+      "step": 42770
+    },
+    {
+      "epoch": 276.0,
+      "learning_rate": 4.022222222222222e-05,
+      "loss": 0.2797,
+      "step": 42780
+    },
+    {
+      "epoch": 276.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9684038162231445,
+      "eval_runtime": 0.8632,
+      "eval_samples_per_second": 232.846,
+      "eval_steps_per_second": 8.109,
+      "step": 42780
+    },
+    {
+      "epoch": 276.06,
+      "learning_rate": 4.0218637992831545e-05,
+      "loss": 0.2841,
+      "step": 42790
+    },
+    {
+      "epoch": 276.13,
+      "learning_rate": 4.0215053763440864e-05,
+      "loss": 0.2611,
+      "step": 42800
+    },
+    {
+      "epoch": 276.19,
+      "learning_rate": 4.021146953405018e-05,
+      "loss": 0.2692,
+      "step": 42810
+    },
+    {
+      "epoch": 276.26,
+      "learning_rate": 4.02078853046595e-05,
+      "loss": 0.2749,
+      "step": 42820
+    },
+    {
+      "epoch": 276.32,
+      "learning_rate": 4.020430107526882e-05,
+      "loss": 0.2556,
+      "step": 42830
+    },
+    {
+      "epoch": 276.39,
+      "learning_rate": 4.0200716845878136e-05,
+      "loss": 0.2831,
+      "step": 42840
+    },
+    {
+      "epoch": 276.45,
+      "learning_rate": 4.019713261648746e-05,
+      "loss": 0.2856,
+      "step": 42850
+    },
+    {
+      "epoch": 276.51,
+      "learning_rate": 4.019354838709677e-05,
+      "loss": 0.2499,
+      "step": 42860
+    },
+    {
+      "epoch": 276.58,
+      "learning_rate": 4.01899641577061e-05,
+      "loss": 0.3067,
+      "step": 42870
+    },
+    {
+      "epoch": 276.64,
+      "learning_rate": 4.0186379928315416e-05,
+      "loss": 0.2734,
+      "step": 42880
+    },
+    {
+      "epoch": 276.71,
+      "learning_rate": 4.0182795698924734e-05,
+      "loss": 0.3341,
+      "step": 42890
+    },
+    {
+      "epoch": 276.77,
+      "learning_rate": 4.017921146953405e-05,
+      "loss": 0.3186,
+      "step": 42900
+    },
+    {
+      "epoch": 276.84,
+      "learning_rate": 4.017562724014337e-05,
+      "loss": 0.2727,
+      "step": 42910
+    },
+    {
+      "epoch": 276.9,
+      "learning_rate": 4.017204301075269e-05,
+      "loss": 0.2537,
+      "step": 42920
+    },
+    {
+      "epoch": 276.96,
+      "learning_rate": 4.0168458781362014e-05,
+      "loss": 0.262,
+      "step": 42930
+    },
+    {
+      "epoch": 277.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0011177062988281,
+      "eval_runtime": 0.8546,
+      "eval_samples_per_second": 235.206,
+      "eval_steps_per_second": 8.191,
+      "step": 42935
+    },
+    {
+      "epoch": 277.03,
+      "learning_rate": 4.0164874551971325e-05,
+      "loss": 0.2631,
+      "step": 42940
+    },
+    {
+      "epoch": 277.1,
+      "learning_rate": 4.016129032258065e-05,
+      "loss": 0.2577,
+      "step": 42950
+    },
+    {
+      "epoch": 277.16,
+      "learning_rate": 4.015770609318997e-05,
+      "loss": 0.2968,
+      "step": 42960
+    },
+    {
+      "epoch": 277.23,
+      "learning_rate": 4.0154121863799287e-05,
+      "loss": 0.2515,
+      "step": 42970
+    },
+    {
+      "epoch": 277.29,
+      "learning_rate": 4.0150537634408605e-05,
+      "loss": 0.311,
+      "step": 42980
+    },
+    {
+      "epoch": 277.35,
+      "learning_rate": 4.014695340501792e-05,
+      "loss": 0.313,
+      "step": 42990
+    },
+    {
+      "epoch": 277.42,
+      "learning_rate": 4.014336917562724e-05,
+      "loss": 0.296,
+      "step": 43000
+    },
+    {
+      "epoch": 277.48,
+      "learning_rate": 4.0139784946236566e-05,
+      "loss": 0.268,
+      "step": 43010
+    },
+    {
+      "epoch": 277.55,
+      "learning_rate": 4.013620071684588e-05,
+      "loss": 0.3261,
+      "step": 43020
+    },
+    {
+      "epoch": 277.61,
+      "learning_rate": 4.0132616487455196e-05,
+      "loss": 0.2355,
+      "step": 43030
+    },
+    {
+      "epoch": 277.68,
+      "learning_rate": 4.012903225806452e-05,
+      "loss": 0.2941,
+      "step": 43040
+    },
+    {
+      "epoch": 277.74,
+      "learning_rate": 4.012544802867383e-05,
+      "loss": 0.3157,
+      "step": 43050
+    },
+    {
+      "epoch": 277.8,
+      "learning_rate": 4.012186379928316e-05,
+      "loss": 0.3169,
+      "step": 43060
+    },
+    {
+      "epoch": 277.87,
+      "learning_rate": 4.0118279569892475e-05,
+      "loss": 0.2741,
+      "step": 43070
+    },
+    {
+      "epoch": 277.93,
+      "learning_rate": 4.0114695340501794e-05,
+      "loss": 0.2347,
+      "step": 43080
+    },
+    {
+      "epoch": 278.0,
+      "learning_rate": 4.011111111111111e-05,
+      "loss": 0.2528,
+      "step": 43090
+    },
+    {
+      "epoch": 278.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.970280110836029,
+      "eval_runtime": 0.8599,
+      "eval_samples_per_second": 233.745,
+      "eval_steps_per_second": 8.14,
+      "step": 43090
+    },
+    {
+      "epoch": 278.06,
+      "learning_rate": 4.010752688172043e-05,
+      "loss": 0.296,
+      "step": 43100
+    },
+    {
+      "epoch": 278.13,
+      "learning_rate": 4.010394265232975e-05,
+      "loss": 0.2615,
+      "step": 43110
+    },
+    {
+      "epoch": 278.19,
+      "learning_rate": 4.010035842293907e-05,
+      "loss": 0.2529,
+      "step": 43120
+    },
+    {
+      "epoch": 278.26,
+      "learning_rate": 4.0096774193548385e-05,
+      "loss": 0.2751,
+      "step": 43130
+    },
+    {
+      "epoch": 278.32,
+      "learning_rate": 4.009318996415771e-05,
+      "loss": 0.2555,
+      "step": 43140
+    },
+    {
+      "epoch": 278.39,
+      "learning_rate": 4.008960573476703e-05,
+      "loss": 0.24,
+      "step": 43150
+    },
+    {
+      "epoch": 278.45,
+      "learning_rate": 4.0086021505376346e-05,
+      "loss": 0.2897,
+      "step": 43160
+    },
+    {
+      "epoch": 278.51,
+      "learning_rate": 4.0082437275985664e-05,
+      "loss": 0.2558,
+      "step": 43170
+    },
+    {
+      "epoch": 278.58,
+      "learning_rate": 4.007885304659498e-05,
+      "loss": 0.3018,
+      "step": 43180
+    },
+    {
+      "epoch": 278.64,
+      "learning_rate": 4.00752688172043e-05,
+      "loss": 0.2931,
+      "step": 43190
+    },
+    {
+      "epoch": 278.71,
+      "learning_rate": 4.0071684587813626e-05,
+      "loss": 0.238,
+      "step": 43200
+    },
+    {
+      "epoch": 278.77,
+      "learning_rate": 4.006810035842294e-05,
+      "loss": 0.2855,
+      "step": 43210
+    },
+    {
+      "epoch": 278.84,
+      "learning_rate": 4.006451612903226e-05,
+      "loss": 0.2992,
+      "step": 43220
+    },
+    {
+      "epoch": 278.9,
+      "learning_rate": 4.006093189964158e-05,
+      "loss": 0.2507,
+      "step": 43230
+    },
+    {
+      "epoch": 278.96,
+      "learning_rate": 4.00573476702509e-05,
+      "loss": 0.2699,
+      "step": 43240
+    },
+    {
+      "epoch": 279.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.011295199394226,
+      "eval_runtime": 0.8694,
+      "eval_samples_per_second": 231.187,
+      "eval_steps_per_second": 8.051,
+      "step": 43245
+    },
+    {
+      "epoch": 279.03,
+      "learning_rate": 4.005376344086022e-05,
+      "loss": 0.3057,
+      "step": 43250
+    },
+    {
+      "epoch": 279.1,
+      "learning_rate": 4.0050179211469535e-05,
+      "loss": 0.2786,
+      "step": 43260
+    },
+    {
+      "epoch": 279.16,
+      "learning_rate": 4.004659498207885e-05,
+      "loss": 0.3445,
+      "step": 43270
+    },
+    {
+      "epoch": 279.23,
+      "learning_rate": 4.004301075268818e-05,
+      "loss": 0.2745,
+      "step": 43280
+    },
+    {
+      "epoch": 279.29,
+      "learning_rate": 4.003942652329749e-05,
+      "loss": 0.2326,
+      "step": 43290
+    },
+    {
+      "epoch": 279.35,
+      "learning_rate": 4.0035842293906814e-05,
+      "loss": 0.2871,
+      "step": 43300
+    },
+    {
+      "epoch": 279.42,
+      "learning_rate": 4.003225806451613e-05,
+      "loss": 0.2958,
+      "step": 43310
+    },
+    {
+      "epoch": 279.48,
+      "learning_rate": 4.002867383512545e-05,
+      "loss": 0.305,
+      "step": 43320
+    },
+    {
+      "epoch": 279.55,
+      "learning_rate": 4.002508960573477e-05,
+      "loss": 0.3012,
+      "step": 43330
+    },
+    {
+      "epoch": 279.61,
+      "learning_rate": 4.002150537634409e-05,
+      "loss": 0.2622,
+      "step": 43340
+    },
+    {
+      "epoch": 279.68,
+      "learning_rate": 4.0017921146953405e-05,
+      "loss": 0.2776,
+      "step": 43350
+    },
+    {
+      "epoch": 279.74,
+      "learning_rate": 4.001433691756273e-05,
+      "loss": 0.2231,
+      "step": 43360
+    },
+    {
+      "epoch": 279.8,
+      "learning_rate": 4.001075268817204e-05,
+      "loss": 0.3419,
+      "step": 43370
+    },
+    {
+      "epoch": 279.87,
+      "learning_rate": 4.000716845878136e-05,
+      "loss": 0.2501,
+      "step": 43380
+    },
+    {
+      "epoch": 279.93,
+      "learning_rate": 4.0003584229390685e-05,
+      "loss": 0.2982,
+      "step": 43390
+    },
+    {
+      "epoch": 280.0,
+      "learning_rate": 4e-05,
+      "loss": 0.2963,
+      "step": 43400
+    },
+    {
+      "epoch": 280.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9901048541069031,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.836,
+      "eval_steps_per_second": 8.074,
+      "step": 43400
+    },
+    {
+      "epoch": 280.06,
+      "learning_rate": 3.999641577060932e-05,
+      "loss": 0.4195,
+      "step": 43410
+    },
+    {
+      "epoch": 280.13,
+      "learning_rate": 3.999283154121864e-05,
+      "loss": 0.3262,
+      "step": 43420
+    },
+    {
+      "epoch": 280.19,
+      "learning_rate": 3.998924731182796e-05,
+      "loss": 0.2856,
+      "step": 43430
+    },
+    {
+      "epoch": 280.26,
+      "learning_rate": 3.998566308243728e-05,
+      "loss": 0.3023,
+      "step": 43440
+    },
+    {
+      "epoch": 280.32,
+      "learning_rate": 3.9982078853046594e-05,
+      "loss": 0.2116,
+      "step": 43450
+    },
+    {
+      "epoch": 280.39,
+      "learning_rate": 3.997849462365591e-05,
+      "loss": 0.2584,
+      "step": 43460
+    },
+    {
+      "epoch": 280.45,
+      "learning_rate": 3.997491039426524e-05,
+      "loss": 0.3723,
+      "step": 43470
+    },
+    {
+      "epoch": 280.51,
+      "learning_rate": 3.9971326164874556e-05,
+      "loss": 0.2068,
+      "step": 43480
+    },
+    {
+      "epoch": 280.58,
+      "learning_rate": 3.9967741935483874e-05,
+      "loss": 0.2433,
+      "step": 43490
+    },
+    {
+      "epoch": 280.64,
+      "learning_rate": 3.996415770609319e-05,
+      "loss": 0.2434,
+      "step": 43500
+    },
+    {
+      "epoch": 280.71,
+      "learning_rate": 3.996057347670251e-05,
+      "loss": 0.3186,
+      "step": 43510
+    },
+    {
+      "epoch": 280.77,
+      "learning_rate": 3.9956989247311835e-05,
+      "loss": 0.2605,
+      "step": 43520
+    },
+    {
+      "epoch": 280.84,
+      "learning_rate": 3.995340501792115e-05,
+      "loss": 0.28,
+      "step": 43530
+    },
+    {
+      "epoch": 280.9,
+      "learning_rate": 3.9949820788530465e-05,
+      "loss": 0.2984,
+      "step": 43540
+    },
+    {
+      "epoch": 280.96,
+      "learning_rate": 3.994623655913979e-05,
+      "loss": 0.2399,
+      "step": 43550
+    },
+    {
+      "epoch": 281.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0606499910354614,
+      "eval_runtime": 0.8607,
+      "eval_samples_per_second": 233.522,
+      "eval_steps_per_second": 8.133,
+      "step": 43555
+    },
+    {
+      "epoch": 281.03,
+      "learning_rate": 3.994265232974911e-05,
+      "loss": 0.2678,
+      "step": 43560
+    },
+    {
+      "epoch": 281.1,
+      "learning_rate": 3.9939068100358426e-05,
+      "loss": 0.2483,
+      "step": 43570
+    },
+    {
+      "epoch": 281.16,
+      "learning_rate": 3.9935483870967745e-05,
+      "loss": 0.2635,
+      "step": 43580
+    },
+    {
+      "epoch": 281.23,
+      "learning_rate": 3.993189964157706e-05,
+      "loss": 0.2691,
+      "step": 43590
+    },
+    {
+      "epoch": 281.29,
+      "learning_rate": 3.992831541218639e-05,
+      "loss": 0.2551,
+      "step": 43600
+    },
+    {
+      "epoch": 281.35,
+      "learning_rate": 3.99247311827957e-05,
+      "loss": 0.2525,
+      "step": 43610
+    },
+    {
+      "epoch": 281.42,
+      "learning_rate": 3.992114695340502e-05,
+      "loss": 0.2666,
+      "step": 43620
+    },
+    {
+      "epoch": 281.48,
+      "learning_rate": 3.991756272401434e-05,
+      "loss": 0.2143,
+      "step": 43630
+    },
+    {
+      "epoch": 281.55,
+      "learning_rate": 3.991397849462366e-05,
+      "loss": 0.2599,
+      "step": 43640
+    },
+    {
+      "epoch": 281.61,
+      "learning_rate": 3.991039426523298e-05,
+      "loss": 0.2954,
+      "step": 43650
+    },
+    {
+      "epoch": 281.68,
+      "learning_rate": 3.99068100358423e-05,
+      "loss": 0.3143,
+      "step": 43660
+    },
+    {
+      "epoch": 281.74,
+      "learning_rate": 3.9903225806451615e-05,
+      "loss": 0.2489,
+      "step": 43670
+    },
+    {
+      "epoch": 281.8,
+      "learning_rate": 3.989964157706093e-05,
+      "loss": 0.2648,
+      "step": 43680
+    },
+    {
+      "epoch": 281.87,
+      "learning_rate": 3.989605734767025e-05,
+      "loss": 0.2582,
+      "step": 43690
+    },
+    {
+      "epoch": 281.93,
+      "learning_rate": 3.989247311827957e-05,
+      "loss": 0.2855,
+      "step": 43700
+    },
+    {
+      "epoch": 282.0,
+      "learning_rate": 3.9888888888888895e-05,
+      "loss": 0.2367,
+      "step": 43710
+    },
+    {
+      "epoch": 282.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0254015922546387,
+      "eval_runtime": 0.8502,
+      "eval_samples_per_second": 236.418,
+      "eval_steps_per_second": 8.233,
+      "step": 43710
+    },
+    {
+      "epoch": 282.06,
+      "learning_rate": 3.9885304659498206e-05,
+      "loss": 0.3094,
+      "step": 43720
+    },
+    {
+      "epoch": 282.13,
+      "learning_rate": 3.9881720430107524e-05,
+      "loss": 0.2689,
+      "step": 43730
+    },
+    {
+      "epoch": 282.19,
+      "learning_rate": 3.987813620071685e-05,
+      "loss": 0.2827,
+      "step": 43740
+    },
+    {
+      "epoch": 282.26,
+      "learning_rate": 3.987455197132617e-05,
+      "loss": 0.2984,
+      "step": 43750
+    },
+    {
+      "epoch": 282.32,
+      "learning_rate": 3.9870967741935486e-05,
+      "loss": 0.3005,
+      "step": 43760
+    },
+    {
+      "epoch": 282.39,
+      "learning_rate": 3.9867383512544804e-05,
+      "loss": 0.2732,
+      "step": 43770
+    },
+    {
+      "epoch": 282.45,
+      "learning_rate": 3.986379928315412e-05,
+      "loss": 0.2101,
+      "step": 43780
+    },
+    {
+      "epoch": 282.51,
+      "learning_rate": 3.986021505376345e-05,
+      "loss": 0.2838,
+      "step": 43790
+    },
+    {
+      "epoch": 282.58,
+      "learning_rate": 3.985663082437276e-05,
+      "loss": 0.2398,
+      "step": 43800
+    },
+    {
+      "epoch": 282.64,
+      "learning_rate": 3.985304659498208e-05,
+      "loss": 0.31,
+      "step": 43810
+    },
+    {
+      "epoch": 282.71,
+      "learning_rate": 3.98494623655914e-05,
+      "loss": 0.282,
+      "step": 43820
+    },
+    {
+      "epoch": 282.77,
+      "learning_rate": 3.984587813620072e-05,
+      "loss": 0.2554,
+      "step": 43830
+    },
+    {
+      "epoch": 282.84,
+      "learning_rate": 3.984229390681004e-05,
+      "loss": 0.2553,
+      "step": 43840
+    },
+    {
+      "epoch": 282.9,
+      "learning_rate": 3.9838709677419356e-05,
+      "loss": 0.265,
+      "step": 43850
+    },
+    {
+      "epoch": 282.96,
+      "learning_rate": 3.9835125448028675e-05,
+      "loss": 0.2358,
+      "step": 43860
+    },
+    {
+      "epoch": 283.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0002617835998535,
+      "eval_runtime": 0.8638,
+      "eval_samples_per_second": 232.689,
+      "eval_steps_per_second": 8.104,
+      "step": 43865
+    },
+    {
+      "epoch": 283.03,
+      "learning_rate": 3.9831541218638e-05,
+      "loss": 0.282,
+      "step": 43870
+    },
+    {
+      "epoch": 283.1,
+      "learning_rate": 3.982795698924731e-05,
+      "loss": 0.2763,
+      "step": 43880
+    },
+    {
+      "epoch": 283.16,
+      "learning_rate": 3.982437275985663e-05,
+      "loss": 0.238,
+      "step": 43890
+    },
+    {
+      "epoch": 283.23,
+      "learning_rate": 3.9820788530465954e-05,
+      "loss": 0.2743,
+      "step": 43900
+    },
+    {
+      "epoch": 283.29,
+      "learning_rate": 3.981720430107527e-05,
+      "loss": 0.2812,
+      "step": 43910
+    },
+    {
+      "epoch": 283.35,
+      "learning_rate": 3.981362007168459e-05,
+      "loss": 0.2843,
+      "step": 43920
+    },
+    {
+      "epoch": 283.42,
+      "learning_rate": 3.981003584229391e-05,
+      "loss": 0.2732,
+      "step": 43930
+    },
+    {
+      "epoch": 283.48,
+      "learning_rate": 3.980645161290323e-05,
+      "loss": 0.2871,
+      "step": 43940
+    },
+    {
+      "epoch": 283.55,
+      "learning_rate": 3.980286738351255e-05,
+      "loss": 0.2742,
+      "step": 43950
+    },
+    {
+      "epoch": 283.61,
+      "learning_rate": 3.9799283154121863e-05,
+      "loss": 0.3164,
+      "step": 43960
+    },
+    {
+      "epoch": 283.68,
+      "learning_rate": 3.979569892473118e-05,
+      "loss": 0.2789,
+      "step": 43970
+    },
+    {
+      "epoch": 283.74,
+      "learning_rate": 3.9792114695340507e-05,
+      "loss": 0.2828,
+      "step": 43980
+    },
+    {
+      "epoch": 283.8,
+      "learning_rate": 3.9788530465949825e-05,
+      "loss": 0.2313,
+      "step": 43990
+    },
+    {
+      "epoch": 283.87,
+      "learning_rate": 3.978494623655914e-05,
+      "loss": 0.2668,
+      "step": 44000
+    },
+    {
+      "epoch": 283.93,
+      "learning_rate": 3.978136200716846e-05,
+      "loss": 0.2722,
+      "step": 44010
+    },
+    {
+      "epoch": 284.0,
+      "learning_rate": 3.977777777777778e-05,
+      "loss": 0.3575,
+      "step": 44020
+    },
+    {
+      "epoch": 284.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0392097234725952,
+      "eval_runtime": 0.8737,
+      "eval_samples_per_second": 230.055,
+      "eval_steps_per_second": 8.012,
+      "step": 44020
+    },
+    {
+      "epoch": 284.06,
+      "learning_rate": 3.97741935483871e-05,
+      "loss": 0.323,
+      "step": 44030
+    },
+    {
+      "epoch": 284.13,
+      "learning_rate": 3.9770609318996416e-05,
+      "loss": 0.2836,
+      "step": 44040
+    },
+    {
+      "epoch": 284.19,
+      "learning_rate": 3.9767025089605734e-05,
+      "loss": 0.2863,
+      "step": 44050
+    },
+    {
+      "epoch": 284.26,
+      "learning_rate": 3.976344086021506e-05,
+      "loss": 0.2829,
+      "step": 44060
+    },
+    {
+      "epoch": 284.32,
+      "learning_rate": 3.975985663082438e-05,
+      "loss": 0.2371,
+      "step": 44070
+    },
+    {
+      "epoch": 284.39,
+      "learning_rate": 3.975627240143369e-05,
+      "loss": 0.2416,
+      "step": 44080
+    },
+    {
+      "epoch": 284.45,
+      "learning_rate": 3.9752688172043014e-05,
+      "loss": 0.2554,
+      "step": 44090
+    },
+    {
+      "epoch": 284.51,
+      "learning_rate": 3.974910394265233e-05,
+      "loss": 0.2679,
+      "step": 44100
+    },
+    {
+      "epoch": 284.58,
+      "learning_rate": 3.974551971326165e-05,
+      "loss": 0.2805,
+      "step": 44110
+    },
+    {
+      "epoch": 284.64,
+      "learning_rate": 3.974193548387097e-05,
+      "loss": 0.2688,
+      "step": 44120
+    },
+    {
+      "epoch": 284.71,
+      "learning_rate": 3.9738351254480286e-05,
+      "loss": 0.3123,
+      "step": 44130
+    },
+    {
+      "epoch": 284.77,
+      "learning_rate": 3.973476702508961e-05,
+      "loss": 0.3072,
+      "step": 44140
+    },
+    {
+      "epoch": 284.84,
+      "learning_rate": 3.973118279569893e-05,
+      "loss": 0.2624,
+      "step": 44150
+    },
+    {
+      "epoch": 284.9,
+      "learning_rate": 3.972759856630824e-05,
+      "loss": 0.3092,
+      "step": 44160
+    },
+    {
+      "epoch": 284.96,
+      "learning_rate": 3.9724014336917566e-05,
+      "loss": 0.3035,
+      "step": 44170
+    },
+    {
+      "epoch": 285.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.009652853012085,
+      "eval_runtime": 0.8486,
+      "eval_samples_per_second": 236.853,
+      "eval_steps_per_second": 8.249,
+      "step": 44175
+    },
+    {
+      "epoch": 285.03,
+      "learning_rate": 3.9720430107526884e-05,
+      "loss": 0.3111,
+      "step": 44180
+    },
+    {
+      "epoch": 285.1,
+      "learning_rate": 3.97168458781362e-05,
+      "loss": 0.2856,
+      "step": 44190
+    },
+    {
+      "epoch": 285.16,
+      "learning_rate": 3.971326164874552e-05,
+      "loss": 0.2888,
+      "step": 44200
+    },
+    {
+      "epoch": 285.23,
+      "learning_rate": 3.970967741935484e-05,
+      "loss": 0.2334,
+      "step": 44210
+    },
+    {
+      "epoch": 285.29,
+      "learning_rate": 3.9706093189964164e-05,
+      "loss": 0.2469,
+      "step": 44220
+    },
+    {
+      "epoch": 285.35,
+      "learning_rate": 3.970250896057348e-05,
+      "loss": 0.2389,
+      "step": 44230
+    },
+    {
+      "epoch": 285.42,
+      "learning_rate": 3.9698924731182793e-05,
+      "loss": 0.2759,
+      "step": 44240
+    },
+    {
+      "epoch": 285.48,
+      "learning_rate": 3.969534050179212e-05,
+      "loss": 0.2759,
+      "step": 44250
+    },
+    {
+      "epoch": 285.55,
+      "learning_rate": 3.969175627240144e-05,
+      "loss": 0.2762,
+      "step": 44260
+    },
+    {
+      "epoch": 285.61,
+      "learning_rate": 3.9688172043010755e-05,
+      "loss": 0.2907,
+      "step": 44270
+    },
+    {
+      "epoch": 285.68,
+      "learning_rate": 3.968458781362007e-05,
+      "loss": 0.256,
+      "step": 44280
+    },
+    {
+      "epoch": 285.74,
+      "learning_rate": 3.968100358422939e-05,
+      "loss": 0.2671,
+      "step": 44290
+    },
+    {
+      "epoch": 285.8,
+      "learning_rate": 3.9677419354838716e-05,
+      "loss": 0.2817,
+      "step": 44300
+    },
+    {
+      "epoch": 285.87,
+      "learning_rate": 3.967383512544803e-05,
+      "loss": 0.3169,
+      "step": 44310
+    },
+    {
+      "epoch": 285.93,
+      "learning_rate": 3.9670250896057346e-05,
+      "loss": 0.2753,
+      "step": 44320
+    },
+    {
+      "epoch": 286.0,
+      "learning_rate": 3.966666666666667e-05,
+      "loss": 0.2316,
+      "step": 44330
+    },
+    {
+      "epoch": 286.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.1243842840194702,
+      "eval_runtime": 0.8524,
+      "eval_samples_per_second": 235.797,
+      "eval_steps_per_second": 8.212,
+      "step": 44330
+    },
+    {
+      "epoch": 286.06,
+      "learning_rate": 3.966308243727599e-05,
+      "loss": 0.2443,
+      "step": 44340
+    },
+    {
+      "epoch": 286.13,
+      "learning_rate": 3.965949820788531e-05,
+      "loss": 0.2705,
+      "step": 44350
+    },
+    {
+      "epoch": 286.19,
+      "learning_rate": 3.9655913978494625e-05,
+      "loss": 0.2636,
+      "step": 44360
+    },
+    {
+      "epoch": 286.26,
+      "learning_rate": 3.9652329749103944e-05,
+      "loss": 0.2619,
+      "step": 44370
+    },
+    {
+      "epoch": 286.32,
+      "learning_rate": 3.964874551971326e-05,
+      "loss": 0.2755,
+      "step": 44380
+    },
+    {
+      "epoch": 286.39,
+      "learning_rate": 3.964516129032258e-05,
+      "loss": 0.3138,
+      "step": 44390
+    },
+    {
+      "epoch": 286.45,
+      "learning_rate": 3.96415770609319e-05,
+      "loss": 0.2841,
+      "step": 44400
+    },
+    {
+      "epoch": 286.51,
+      "learning_rate": 3.963799283154122e-05,
+      "loss": 0.3092,
+      "step": 44410
+    },
+    {
+      "epoch": 286.58,
+      "learning_rate": 3.963440860215054e-05,
+      "loss": 0.2788,
+      "step": 44420
+    },
+    {
+      "epoch": 286.64,
+      "learning_rate": 3.963082437275985e-05,
+      "loss": 0.2986,
+      "step": 44430
+    },
+    {
+      "epoch": 286.71,
+      "learning_rate": 3.962724014336918e-05,
+      "loss": 0.2872,
+      "step": 44440
+    },
+    {
+      "epoch": 286.77,
+      "learning_rate": 3.9623655913978496e-05,
+      "loss": 0.3094,
+      "step": 44450
+    },
+    {
+      "epoch": 286.84,
+      "learning_rate": 3.9620071684587814e-05,
+      "loss": 0.2998,
+      "step": 44460
+    },
+    {
+      "epoch": 286.9,
+      "learning_rate": 3.961648745519713e-05,
+      "loss": 0.3034,
+      "step": 44470
+    },
+    {
+      "epoch": 286.96,
+      "learning_rate": 3.961290322580645e-05,
+      "loss": 0.2549,
+      "step": 44480
+    },
+    {
+      "epoch": 287.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0686671733856201,
+      "eval_runtime": 0.8526,
+      "eval_samples_per_second": 235.751,
+      "eval_steps_per_second": 8.21,
+      "step": 44485
+    },
+    {
+      "epoch": 287.03,
+      "learning_rate": 3.9609318996415776e-05,
+      "loss": 0.3435,
+      "step": 44490
+    },
+    {
+      "epoch": 287.1,
+      "learning_rate": 3.9605734767025094e-05,
+      "loss": 0.3014,
+      "step": 44500
+    },
+    {
+      "epoch": 287.16,
+      "learning_rate": 3.9602150537634405e-05,
+      "loss": 0.2815,
+      "step": 44510
+    },
+    {
+      "epoch": 287.23,
+      "learning_rate": 3.959856630824373e-05,
+      "loss": 0.2818,
+      "step": 44520
+    },
+    {
+      "epoch": 287.29,
+      "learning_rate": 3.959498207885305e-05,
+      "loss": 0.2024,
+      "step": 44530
+    },
+    {
+      "epoch": 287.35,
+      "learning_rate": 3.959139784946237e-05,
+      "loss": 0.3045,
+      "step": 44540
+    },
+    {
+      "epoch": 287.42,
+      "learning_rate": 3.9587813620071685e-05,
+      "loss": 0.2514,
+      "step": 44550
+    },
+    {
+      "epoch": 287.48,
+      "learning_rate": 3.9584229390681e-05,
+      "loss": 0.3522,
+      "step": 44560
+    },
+    {
+      "epoch": 287.55,
+      "learning_rate": 3.958064516129033e-05,
+      "loss": 0.2564,
+      "step": 44570
+    },
+    {
+      "epoch": 287.61,
+      "learning_rate": 3.9577060931899646e-05,
+      "loss": 0.2928,
+      "step": 44580
+    },
+    {
+      "epoch": 287.68,
+      "learning_rate": 3.957347670250896e-05,
+      "loss": 0.2577,
+      "step": 44590
+    },
+    {
+      "epoch": 287.74,
+      "learning_rate": 3.956989247311828e-05,
+      "loss": 0.3016,
+      "step": 44600
+    },
+    {
+      "epoch": 287.8,
+      "learning_rate": 3.95663082437276e-05,
+      "loss": 0.2543,
+      "step": 44610
+    },
+    {
+      "epoch": 287.87,
+      "learning_rate": 3.956272401433692e-05,
+      "loss": 0.2901,
+      "step": 44620
+    },
+    {
+      "epoch": 287.93,
+      "learning_rate": 3.955913978494624e-05,
+      "loss": 0.2657,
+      "step": 44630
+    },
+    {
+      "epoch": 288.0,
+      "learning_rate": 3.9555555555555556e-05,
+      "loss": 0.2365,
+      "step": 44640
+    },
+    {
+      "epoch": 288.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.1068649291992188,
+      "eval_runtime": 0.9113,
+      "eval_samples_per_second": 220.552,
+      "eval_steps_per_second": 7.681,
+      "step": 44640
+    },
+    {
+      "epoch": 288.06,
+      "learning_rate": 3.955197132616488e-05,
+      "loss": 0.2947,
+      "step": 44650
+    },
+    {
+      "epoch": 288.13,
+      "learning_rate": 3.95483870967742e-05,
+      "loss": 0.2303,
+      "step": 44660
+    },
+    {
+      "epoch": 288.19,
+      "learning_rate": 3.954480286738351e-05,
+      "loss": 0.2765,
+      "step": 44670
+    },
+    {
+      "epoch": 288.26,
+      "learning_rate": 3.9541218637992835e-05,
+      "loss": 0.2476,
+      "step": 44680
+    },
+    {
+      "epoch": 288.32,
+      "learning_rate": 3.953763440860215e-05,
+      "loss": 0.224,
+      "step": 44690
+    },
+    {
+      "epoch": 288.39,
+      "learning_rate": 3.953405017921147e-05,
+      "loss": 0.2161,
+      "step": 44700
+    },
+    {
+      "epoch": 288.45,
+      "learning_rate": 3.953046594982079e-05,
+      "loss": 0.2864,
+      "step": 44710
+    },
+    {
+      "epoch": 288.51,
+      "learning_rate": 3.952688172043011e-05,
+      "loss": 0.3165,
+      "step": 44720
+    },
+    {
+      "epoch": 288.58,
+      "learning_rate": 3.9523297491039426e-05,
+      "loss": 0.3001,
+      "step": 44730
+    },
+    {
+      "epoch": 288.64,
+      "learning_rate": 3.951971326164875e-05,
+      "loss": 0.2711,
+      "step": 44740
+    },
+    {
+      "epoch": 288.71,
+      "learning_rate": 3.951612903225806e-05,
+      "loss": 0.3103,
+      "step": 44750
+    },
+    {
+      "epoch": 288.77,
+      "learning_rate": 3.951254480286739e-05,
+      "loss": 0.2267,
+      "step": 44760
+    },
+    {
+      "epoch": 288.84,
+      "learning_rate": 3.9508960573476706e-05,
+      "loss": 0.2921,
+      "step": 44770
+    },
+    {
+      "epoch": 288.9,
+      "learning_rate": 3.9505376344086024e-05,
+      "loss": 0.345,
+      "step": 44780
+    },
+    {
+      "epoch": 288.96,
+      "learning_rate": 3.950179211469534e-05,
+      "loss": 0.3168,
+      "step": 44790
+    },
+    {
+      "epoch": 289.0,
+      "eval_accuracy": 0.7313432835820896,
+      "eval_loss": 1.1021668910980225,
+      "eval_runtime": 0.8518,
+      "eval_samples_per_second": 235.973,
+      "eval_steps_per_second": 8.218,
+      "step": 44795
+    },
+    {
+      "epoch": 289.03,
+      "learning_rate": 3.949820788530466e-05,
+      "loss": 0.2941,
+      "step": 44800
+    },
+    {
+      "epoch": 289.1,
+      "learning_rate": 3.949462365591398e-05,
+      "loss": 0.2925,
+      "step": 44810
+    },
+    {
+      "epoch": 289.16,
+      "learning_rate": 3.9491039426523304e-05,
+      "loss": 0.2745,
+      "step": 44820
+    },
+    {
+      "epoch": 289.23,
+      "learning_rate": 3.9487455197132615e-05,
+      "loss": 0.2827,
+      "step": 44830
+    },
+    {
+      "epoch": 289.29,
+      "learning_rate": 3.948387096774194e-05,
+      "loss": 0.312,
+      "step": 44840
+    },
+    {
+      "epoch": 289.35,
+      "learning_rate": 3.948028673835126e-05,
+      "loss": 0.2765,
+      "step": 44850
+    },
+    {
+      "epoch": 289.42,
+      "learning_rate": 3.9476702508960576e-05,
+      "loss": 0.2732,
+      "step": 44860
+    },
+    {
+      "epoch": 289.48,
+      "learning_rate": 3.9473118279569895e-05,
+      "loss": 0.2838,
+      "step": 44870
+    },
+    {
+      "epoch": 289.55,
+      "learning_rate": 3.946953405017921e-05,
+      "loss": 0.2504,
+      "step": 44880
+    },
+    {
+      "epoch": 289.61,
+      "learning_rate": 3.946594982078853e-05,
+      "loss": 0.2935,
+      "step": 44890
+    },
+    {
+      "epoch": 289.68,
+      "learning_rate": 3.9462365591397856e-05,
+      "loss": 0.2636,
+      "step": 44900
+    },
+    {
+      "epoch": 289.74,
+      "learning_rate": 3.945878136200717e-05,
+      "loss": 0.2307,
+      "step": 44910
+    },
+    {
+      "epoch": 289.8,
+      "learning_rate": 3.945519713261649e-05,
+      "loss": 0.3078,
+      "step": 44920
+    },
+    {
+      "epoch": 289.87,
+      "learning_rate": 3.945161290322581e-05,
+      "loss": 0.2433,
+      "step": 44930
+    },
+    {
+      "epoch": 289.93,
+      "learning_rate": 3.944802867383512e-05,
+      "loss": 0.2382,
+      "step": 44940
+    },
+    {
+      "epoch": 290.0,
+      "learning_rate": 3.944444444444445e-05,
+      "loss": 0.3088,
+      "step": 44950
+    },
+    {
+      "epoch": 290.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9550564885139465,
+      "eval_runtime": 0.8733,
+      "eval_samples_per_second": 230.161,
+      "eval_steps_per_second": 8.016,
+      "step": 44950
+    },
+    {
+      "epoch": 290.06,
+      "learning_rate": 3.9440860215053765e-05,
+      "loss": 0.277,
+      "step": 44960
+    },
+    {
+      "epoch": 290.13,
+      "learning_rate": 3.9437275985663083e-05,
+      "loss": 0.2226,
+      "step": 44970
+    },
+    {
+      "epoch": 290.19,
+      "learning_rate": 3.94336917562724e-05,
+      "loss": 0.2443,
+      "step": 44980
+    },
+    {
+      "epoch": 290.26,
+      "learning_rate": 3.943010752688172e-05,
+      "loss": 0.2455,
+      "step": 44990
+    },
+    {
+      "epoch": 290.32,
+      "learning_rate": 3.9426523297491045e-05,
+      "loss": 0.2089,
+      "step": 45000
+    },
+    {
+      "epoch": 290.39,
+      "learning_rate": 3.942293906810036e-05,
+      "loss": 0.2523,
+      "step": 45010
+    },
+    {
+      "epoch": 290.45,
+      "learning_rate": 3.9419354838709674e-05,
+      "loss": 0.279,
+      "step": 45020
+    },
+    {
+      "epoch": 290.51,
+      "learning_rate": 3.9415770609319e-05,
+      "loss": 0.2702,
+      "step": 45030
+    },
+    {
+      "epoch": 290.58,
+      "learning_rate": 3.941218637992832e-05,
+      "loss": 0.2625,
+      "step": 45040
+    },
+    {
+      "epoch": 290.64,
+      "learning_rate": 3.9408602150537636e-05,
+      "loss": 0.2777,
+      "step": 45050
+    },
+    {
+      "epoch": 290.71,
+      "learning_rate": 3.9405017921146954e-05,
+      "loss": 0.2733,
+      "step": 45060
+    },
+    {
+      "epoch": 290.77,
+      "learning_rate": 3.940143369175627e-05,
+      "loss": 0.2981,
+      "step": 45070
+    },
+    {
+      "epoch": 290.84,
+      "learning_rate": 3.939784946236559e-05,
+      "loss": 0.2817,
+      "step": 45080
+    },
+    {
+      "epoch": 290.9,
+      "learning_rate": 3.9394265232974915e-05,
+      "loss": 0.2871,
+      "step": 45090
+    },
+    {
+      "epoch": 290.96,
+      "learning_rate": 3.939068100358423e-05,
+      "loss": 0.3264,
+      "step": 45100
+    },
+    {
+      "epoch": 291.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0183805227279663,
+      "eval_runtime": 0.8586,
+      "eval_samples_per_second": 234.109,
+      "eval_steps_per_second": 8.153,
+      "step": 45105
+    },
+    {
+      "epoch": 291.03,
+      "learning_rate": 3.938709677419355e-05,
+      "loss": 0.2964,
+      "step": 45110
+    },
+    {
+      "epoch": 291.1,
+      "learning_rate": 3.938351254480287e-05,
+      "loss": 0.2375,
+      "step": 45120
+    },
+    {
+      "epoch": 291.16,
+      "learning_rate": 3.937992831541219e-05,
+      "loss": 0.2601,
+      "step": 45130
+    },
+    {
+      "epoch": 291.23,
+      "learning_rate": 3.9376344086021506e-05,
+      "loss": 0.277,
+      "step": 45140
+    },
+    {
+      "epoch": 291.29,
+      "learning_rate": 3.9372759856630825e-05,
+      "loss": 0.2278,
+      "step": 45150
+    },
+    {
+      "epoch": 291.35,
+      "learning_rate": 3.936917562724014e-05,
+      "loss": 0.3303,
+      "step": 45160
+    },
+    {
+      "epoch": 291.42,
+      "learning_rate": 3.936559139784947e-05,
+      "loss": 0.2265,
+      "step": 45170
+    },
+    {
+      "epoch": 291.48,
+      "learning_rate": 3.936200716845878e-05,
+      "loss": 0.3053,
+      "step": 45180
+    },
+    {
+      "epoch": 291.55,
+      "learning_rate": 3.9358422939068104e-05,
+      "loss": 0.2953,
+      "step": 45190
+    },
+    {
+      "epoch": 291.61,
+      "learning_rate": 3.935483870967742e-05,
+      "loss": 0.3073,
+      "step": 45200
+    },
+    {
+      "epoch": 291.68,
+      "learning_rate": 3.935125448028674e-05,
+      "loss": 0.2796,
+      "step": 45210
+    },
+    {
+      "epoch": 291.74,
+      "learning_rate": 3.934767025089606e-05,
+      "loss": 0.3094,
+      "step": 45220
+    },
+    {
+      "epoch": 291.8,
+      "learning_rate": 3.934408602150538e-05,
+      "loss": 0.2468,
+      "step": 45230
+    },
+    {
+      "epoch": 291.87,
+      "learning_rate": 3.9340501792114695e-05,
+      "loss": 0.3052,
+      "step": 45240
+    },
+    {
+      "epoch": 291.93,
+      "learning_rate": 3.933691756272402e-05,
+      "loss": 0.3223,
+      "step": 45250
+    },
+    {
+      "epoch": 292.0,
+      "learning_rate": 3.933333333333333e-05,
+      "loss": 0.2933,
+      "step": 45260
+    },
+    {
+      "epoch": 292.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.014046549797058,
+      "eval_runtime": 0.8569,
+      "eval_samples_per_second": 234.573,
+      "eval_steps_per_second": 8.169,
+      "step": 45260
+    },
+    {
+      "epoch": 292.06,
+      "learning_rate": 3.932974910394266e-05,
+      "loss": 0.2316,
+      "step": 45270
+    },
+    {
+      "epoch": 292.13,
+      "learning_rate": 3.9326164874551975e-05,
+      "loss": 0.2791,
+      "step": 45280
+    },
+    {
+      "epoch": 292.19,
+      "learning_rate": 3.932258064516129e-05,
+      "loss": 0.279,
+      "step": 45290
+    },
+    {
+      "epoch": 292.26,
+      "learning_rate": 3.931899641577061e-05,
+      "loss": 0.2507,
+      "step": 45300
+    },
+    {
+      "epoch": 292.32,
+      "learning_rate": 3.931541218637993e-05,
+      "loss": 0.2909,
+      "step": 45310
+    },
+    {
+      "epoch": 292.39,
+      "learning_rate": 3.931182795698925e-05,
+      "loss": 0.2444,
+      "step": 45320
+    },
+    {
+      "epoch": 292.45,
+      "learning_rate": 3.930824372759857e-05,
+      "loss": 0.2927,
+      "step": 45330
+    },
+    {
+      "epoch": 292.51,
+      "learning_rate": 3.9304659498207884e-05,
+      "loss": 0.2947,
+      "step": 45340
+    },
+    {
+      "epoch": 292.58,
+      "learning_rate": 3.93010752688172e-05,
+      "loss": 0.2414,
+      "step": 45350
+    },
+    {
+      "epoch": 292.64,
+      "learning_rate": 3.929749103942653e-05,
+      "loss": 0.3126,
+      "step": 45360
+    },
+    {
+      "epoch": 292.71,
+      "learning_rate": 3.9293906810035846e-05,
+      "loss": 0.3004,
+      "step": 45370
+    },
+    {
+      "epoch": 292.77,
+      "learning_rate": 3.9290322580645164e-05,
+      "loss": 0.2639,
+      "step": 45380
+    },
+    {
+      "epoch": 292.84,
+      "learning_rate": 3.928673835125448e-05,
+      "loss": 0.2996,
+      "step": 45390
+    },
+    {
+      "epoch": 292.9,
+      "learning_rate": 3.92831541218638e-05,
+      "loss": 0.2804,
+      "step": 45400
+    },
+    {
+      "epoch": 292.96,
+      "learning_rate": 3.9279569892473125e-05,
+      "loss": 0.2558,
+      "step": 45410
+    },
+    {
+      "epoch": 293.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.91397625207901,
+      "eval_runtime": 0.8732,
+      "eval_samples_per_second": 230.194,
+      "eval_steps_per_second": 8.017,
+      "step": 45415
+    },
+    {
+      "epoch": 293.03,
+      "learning_rate": 3.9275985663082437e-05,
+      "loss": 0.2812,
+      "step": 45420
+    },
+    {
+      "epoch": 293.1,
+      "learning_rate": 3.9272401433691755e-05,
+      "loss": 0.263,
+      "step": 45430
+    },
+    {
+      "epoch": 293.16,
+      "learning_rate": 3.926881720430108e-05,
+      "loss": 0.3145,
+      "step": 45440
+    },
+    {
+      "epoch": 293.23,
+      "learning_rate": 3.92652329749104e-05,
+      "loss": 0.2778,
+      "step": 45450
+    },
+    {
+      "epoch": 293.29,
+      "learning_rate": 3.9261648745519716e-05,
+      "loss": 0.2537,
+      "step": 45460
+    },
+    {
+      "epoch": 293.35,
+      "learning_rate": 3.9258064516129034e-05,
+      "loss": 0.2134,
+      "step": 45470
+    },
+    {
+      "epoch": 293.42,
+      "learning_rate": 3.925448028673835e-05,
+      "loss": 0.2236,
+      "step": 45480
+    },
+    {
+      "epoch": 293.48,
+      "learning_rate": 3.925089605734768e-05,
+      "loss": 0.237,
+      "step": 45490
+    },
+    {
+      "epoch": 293.55,
+      "learning_rate": 3.924731182795699e-05,
+      "loss": 0.2683,
+      "step": 45500
+    },
+    {
+      "epoch": 293.61,
+      "learning_rate": 3.924372759856631e-05,
+      "loss": 0.3204,
+      "step": 45510
+    },
+    {
+      "epoch": 293.68,
+      "learning_rate": 3.924014336917563e-05,
+      "loss": 0.3084,
+      "step": 45520
+    },
+    {
+      "epoch": 293.74,
+      "learning_rate": 3.923655913978495e-05,
+      "loss": 0.3019,
+      "step": 45530
+    },
+    {
+      "epoch": 293.8,
+      "learning_rate": 3.923297491039427e-05,
+      "loss": 0.2746,
+      "step": 45540
+    },
+    {
+      "epoch": 293.87,
+      "learning_rate": 3.922939068100359e-05,
+      "loss": 0.2591,
+      "step": 45550
+    },
+    {
+      "epoch": 293.93,
+      "learning_rate": 3.9225806451612905e-05,
+      "loss": 0.2409,
+      "step": 45560
+    },
+    {
+      "epoch": 294.0,
+      "learning_rate": 3.922222222222223e-05,
+      "loss": 0.2519,
+      "step": 45570
+    },
+    {
+      "epoch": 294.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9851414561271667,
+      "eval_runtime": 0.8527,
+      "eval_samples_per_second": 235.71,
+      "eval_steps_per_second": 8.209,
+      "step": 45570
+    },
+    {
+      "epoch": 294.06,
+      "learning_rate": 3.921863799283154e-05,
+      "loss": 0.2617,
+      "step": 45580
+    },
+    {
+      "epoch": 294.13,
+      "learning_rate": 3.921505376344086e-05,
+      "loss": 0.2703,
+      "step": 45590
+    },
+    {
+      "epoch": 294.19,
+      "learning_rate": 3.9211469534050185e-05,
+      "loss": 0.2663,
+      "step": 45600
+    },
+    {
+      "epoch": 294.26,
+      "learning_rate": 3.9207885304659496e-05,
+      "loss": 0.2463,
+      "step": 45610
+    },
+    {
+      "epoch": 294.32,
+      "learning_rate": 3.920430107526882e-05,
+      "loss": 0.2809,
+      "step": 45620
+    },
+    {
+      "epoch": 294.39,
+      "learning_rate": 3.920071684587814e-05,
+      "loss": 0.2579,
+      "step": 45630
+    },
+    {
+      "epoch": 294.45,
+      "learning_rate": 3.919713261648746e-05,
+      "loss": 0.2889,
+      "step": 45640
+    },
+    {
+      "epoch": 294.51,
+      "learning_rate": 3.9193548387096776e-05,
+      "loss": 0.3449,
+      "step": 45650
+    },
+    {
+      "epoch": 294.58,
+      "learning_rate": 3.9189964157706094e-05,
+      "loss": 0.3053,
+      "step": 45660
+    },
+    {
+      "epoch": 294.64,
+      "learning_rate": 3.918637992831541e-05,
+      "loss": 0.2976,
+      "step": 45670
+    },
+    {
+      "epoch": 294.71,
+      "learning_rate": 3.918279569892474e-05,
+      "loss": 0.2491,
+      "step": 45680
+    },
+    {
+      "epoch": 294.77,
+      "learning_rate": 3.917921146953405e-05,
+      "loss": 0.2365,
+      "step": 45690
+    },
+    {
+      "epoch": 294.84,
+      "learning_rate": 3.9175627240143367e-05,
+      "loss": 0.2569,
+      "step": 45700
+    },
+    {
+      "epoch": 294.9,
+      "learning_rate": 3.917204301075269e-05,
+      "loss": 0.2492,
+      "step": 45710
+    },
+    {
+      "epoch": 294.96,
+      "learning_rate": 3.916845878136201e-05,
+      "loss": 0.2245,
+      "step": 45720
+    },
+    {
+      "epoch": 295.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9946480393409729,
+      "eval_runtime": 0.8727,
+      "eval_samples_per_second": 230.31,
+      "eval_steps_per_second": 8.021,
+      "step": 45725
+    },
+    {
+      "epoch": 295.03,
+      "learning_rate": 3.916487455197133e-05,
+      "loss": 0.325,
+      "step": 45730
+    },
+    {
+      "epoch": 295.1,
+      "learning_rate": 3.9161290322580646e-05,
+      "loss": 0.2828,
+      "step": 45740
+    },
+    {
+      "epoch": 295.16,
+      "learning_rate": 3.9157706093189964e-05,
+      "loss": 0.2978,
+      "step": 45750
+    },
+    {
+      "epoch": 295.23,
+      "learning_rate": 3.915412186379929e-05,
+      "loss": 0.2473,
+      "step": 45760
+    },
+    {
+      "epoch": 295.29,
+      "learning_rate": 3.91505376344086e-05,
+      "loss": 0.2568,
+      "step": 45770
+    },
+    {
+      "epoch": 295.35,
+      "learning_rate": 3.914695340501792e-05,
+      "loss": 0.2637,
+      "step": 45780
+    },
+    {
+      "epoch": 295.42,
+      "learning_rate": 3.9143369175627244e-05,
+      "loss": 0.3121,
+      "step": 45790
+    },
+    {
+      "epoch": 295.48,
+      "learning_rate": 3.913978494623656e-05,
+      "loss": 0.2904,
+      "step": 45800
+    },
+    {
+      "epoch": 295.55,
+      "learning_rate": 3.913620071684588e-05,
+      "loss": 0.2529,
+      "step": 45810
+    },
+    {
+      "epoch": 295.61,
+      "learning_rate": 3.91326164874552e-05,
+      "loss": 0.2887,
+      "step": 45820
+    },
+    {
+      "epoch": 295.68,
+      "learning_rate": 3.912903225806452e-05,
+      "loss": 0.2621,
+      "step": 45830
+    },
+    {
+      "epoch": 295.74,
+      "learning_rate": 3.912544802867384e-05,
+      "loss": 0.2905,
+      "step": 45840
+    },
+    {
+      "epoch": 295.8,
+      "learning_rate": 3.912186379928315e-05,
+      "loss": 0.2256,
+      "step": 45850
+    },
+    {
+      "epoch": 295.87,
+      "learning_rate": 3.911827956989247e-05,
+      "loss": 0.2536,
+      "step": 45860
+    },
+    {
+      "epoch": 295.93,
+      "learning_rate": 3.9114695340501796e-05,
+      "loss": 0.2552,
+      "step": 45870
+    },
+    {
+      "epoch": 296.0,
+      "learning_rate": 3.9111111111111115e-05,
+      "loss": 0.3261,
+      "step": 45880
+    },
+    {
+      "epoch": 296.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9653016924858093,
+      "eval_runtime": 0.8802,
+      "eval_samples_per_second": 228.368,
+      "eval_steps_per_second": 7.953,
+      "step": 45880
+    },
+    {
+      "epoch": 296.06,
+      "learning_rate": 3.910752688172043e-05,
+      "loss": 0.2762,
+      "step": 45890
+    },
+    {
+      "epoch": 296.13,
+      "learning_rate": 3.910394265232975e-05,
+      "loss": 0.2597,
+      "step": 45900
+    },
+    {
+      "epoch": 296.19,
+      "learning_rate": 3.910035842293907e-05,
+      "loss": 0.2546,
+      "step": 45910
+    },
+    {
+      "epoch": 296.26,
+      "learning_rate": 3.9096774193548394e-05,
+      "loss": 0.2813,
+      "step": 45920
+    },
+    {
+      "epoch": 296.32,
+      "learning_rate": 3.9093189964157706e-05,
+      "loss": 0.2748,
+      "step": 45930
+    },
+    {
+      "epoch": 296.39,
+      "learning_rate": 3.9089605734767024e-05,
+      "loss": 0.2573,
+      "step": 45940
+    },
+    {
+      "epoch": 296.45,
+      "learning_rate": 3.908602150537635e-05,
+      "loss": 0.2316,
+      "step": 45950
+    },
+    {
+      "epoch": 296.51,
+      "learning_rate": 3.908243727598567e-05,
+      "loss": 0.2964,
+      "step": 45960
+    },
+    {
+      "epoch": 296.58,
+      "learning_rate": 3.9078853046594985e-05,
+      "loss": 0.2969,
+      "step": 45970
+    },
+    {
+      "epoch": 296.64,
+      "learning_rate": 3.9075268817204303e-05,
+      "loss": 0.2512,
+      "step": 45980
+    },
+    {
+      "epoch": 296.71,
+      "learning_rate": 3.907168458781362e-05,
+      "loss": 0.2823,
+      "step": 45990
+    },
+    {
+      "epoch": 296.77,
+      "learning_rate": 3.906810035842295e-05,
+      "loss": 0.2588,
+      "step": 46000
+    },
+    {
+      "epoch": 296.84,
+      "learning_rate": 3.906451612903226e-05,
+      "loss": 0.2904,
+      "step": 46010
+    },
+    {
+      "epoch": 296.9,
+      "learning_rate": 3.9060931899641576e-05,
+      "loss": 0.2772,
+      "step": 46020
+    },
+    {
+      "epoch": 296.96,
+      "learning_rate": 3.90573476702509e-05,
+      "loss": 0.2379,
+      "step": 46030
+    },
+    {
+      "epoch": 297.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9851051568984985,
+      "eval_runtime": 0.8555,
+      "eval_samples_per_second": 234.939,
+      "eval_steps_per_second": 8.182,
+      "step": 46035
+    },
+    {
+      "epoch": 297.03,
+      "learning_rate": 3.905376344086022e-05,
+      "loss": 0.2445,
+      "step": 46040
+    },
+    {
+      "epoch": 297.1,
+      "learning_rate": 3.905017921146953e-05,
+      "loss": 0.2253,
+      "step": 46050
+    },
+    {
+      "epoch": 297.16,
+      "learning_rate": 3.9046594982078856e-05,
+      "loss": 0.2852,
+      "step": 46060
+    },
+    {
+      "epoch": 297.23,
+      "learning_rate": 3.9043010752688174e-05,
+      "loss": 0.2687,
+      "step": 46070
+    },
+    {
+      "epoch": 297.29,
+      "learning_rate": 3.903942652329749e-05,
+      "loss": 0.2765,
+      "step": 46080
+    },
+    {
+      "epoch": 297.35,
+      "learning_rate": 3.903584229390681e-05,
+      "loss": 0.3216,
+      "step": 46090
+    },
+    {
+      "epoch": 297.42,
+      "learning_rate": 3.903225806451613e-05,
+      "loss": 0.2912,
+      "step": 46100
+    },
+    {
+      "epoch": 297.48,
+      "learning_rate": 3.9028673835125454e-05,
+      "loss": 0.3165,
+      "step": 46110
+    },
+    {
+      "epoch": 297.55,
+      "learning_rate": 3.902508960573477e-05,
+      "loss": 0.2652,
+      "step": 46120
+    },
+    {
+      "epoch": 297.61,
+      "learning_rate": 3.902150537634408e-05,
+      "loss": 0.234,
+      "step": 46130
+    },
+    {
+      "epoch": 297.68,
+      "learning_rate": 3.901792114695341e-05,
+      "loss": 0.2992,
+      "step": 46140
+    },
+    {
+      "epoch": 297.74,
+      "learning_rate": 3.9014336917562726e-05,
+      "loss": 0.2995,
+      "step": 46150
+    },
+    {
+      "epoch": 297.8,
+      "learning_rate": 3.9010752688172045e-05,
+      "loss": 0.2646,
+      "step": 46160
+    },
+    {
+      "epoch": 297.87,
+      "learning_rate": 3.900716845878136e-05,
+      "loss": 0.2632,
+      "step": 46170
+    },
+    {
+      "epoch": 297.93,
+      "learning_rate": 3.900358422939068e-05,
+      "loss": 0.2728,
+      "step": 46180
+    },
+    {
+      "epoch": 298.0,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.2546,
+      "step": 46190
+    },
+    {
+      "epoch": 298.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.032558560371399,
+      "eval_runtime": 0.8571,
+      "eval_samples_per_second": 234.508,
+      "eval_steps_per_second": 8.167,
+      "step": 46190
+    },
+    {
+      "epoch": 298.06,
+      "learning_rate": 3.8996415770609324e-05,
+      "loss": 0.2689,
+      "step": 46200
+    },
+    {
+      "epoch": 298.13,
+      "learning_rate": 3.8992831541218636e-05,
+      "loss": 0.3241,
+      "step": 46210
+    },
+    {
+      "epoch": 298.19,
+      "learning_rate": 3.898924731182796e-05,
+      "loss": 0.2386,
+      "step": 46220
+    },
+    {
+      "epoch": 298.26,
+      "learning_rate": 3.898566308243728e-05,
+      "loss": 0.2759,
+      "step": 46230
+    },
+    {
+      "epoch": 298.32,
+      "learning_rate": 3.89820788530466e-05,
+      "loss": 0.2567,
+      "step": 46240
+    },
+    {
+      "epoch": 298.39,
+      "learning_rate": 3.8978494623655915e-05,
+      "loss": 0.2323,
+      "step": 46250
+    },
+    {
+      "epoch": 298.45,
+      "learning_rate": 3.8974910394265234e-05,
+      "loss": 0.272,
+      "step": 46260
+    },
+    {
+      "epoch": 298.51,
+      "learning_rate": 3.897132616487456e-05,
+      "loss": 0.2872,
+      "step": 46270
+    },
+    {
+      "epoch": 298.58,
+      "learning_rate": 3.896774193548387e-05,
+      "loss": 0.2312,
+      "step": 46280
+    },
+    {
+      "epoch": 298.64,
+      "learning_rate": 3.896415770609319e-05,
+      "loss": 0.2669,
+      "step": 46290
+    },
+    {
+      "epoch": 298.71,
+      "learning_rate": 3.896057347670251e-05,
+      "loss": 0.3066,
+      "step": 46300
+    },
+    {
+      "epoch": 298.77,
+      "learning_rate": 3.895698924731183e-05,
+      "loss": 0.2761,
+      "step": 46310
+    },
+    {
+      "epoch": 298.84,
+      "learning_rate": 3.895340501792115e-05,
+      "loss": 0.2765,
+      "step": 46320
+    },
+    {
+      "epoch": 298.9,
+      "learning_rate": 3.894982078853047e-05,
+      "loss": 0.2803,
+      "step": 46330
+    },
+    {
+      "epoch": 298.96,
+      "learning_rate": 3.8946236559139786e-05,
+      "loss": 0.2357,
+      "step": 46340
+    },
+    {
+      "epoch": 299.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0820916891098022,
+      "eval_runtime": 0.8644,
+      "eval_samples_per_second": 232.523,
+      "eval_steps_per_second": 8.098,
+      "step": 46345
+    },
+    {
+      "epoch": 299.03,
+      "learning_rate": 3.8942652329749104e-05,
+      "loss": 0.2986,
+      "step": 46350
+    },
+    {
+      "epoch": 299.1,
+      "learning_rate": 3.893906810035842e-05,
+      "loss": 0.3149,
+      "step": 46360
+    },
+    {
+      "epoch": 299.16,
+      "learning_rate": 3.893548387096774e-05,
+      "loss": 0.2296,
+      "step": 46370
+    },
+    {
+      "epoch": 299.23,
+      "learning_rate": 3.8931899641577066e-05,
+      "loss": 0.313,
+      "step": 46380
+    },
+    {
+      "epoch": 299.29,
+      "learning_rate": 3.8928315412186384e-05,
+      "loss": 0.2417,
+      "step": 46390
+    },
+    {
+      "epoch": 299.35,
+      "learning_rate": 3.8924731182795695e-05,
+      "loss": 0.2071,
+      "step": 46400
+    },
+    {
+      "epoch": 299.42,
+      "learning_rate": 3.892114695340502e-05,
+      "loss": 0.2682,
+      "step": 46410
+    },
+    {
+      "epoch": 299.48,
+      "learning_rate": 3.891756272401434e-05,
+      "loss": 0.2866,
+      "step": 46420
+    },
+    {
+      "epoch": 299.55,
+      "learning_rate": 3.8913978494623657e-05,
+      "loss": 0.2206,
+      "step": 46430
+    },
+    {
+      "epoch": 299.61,
+      "learning_rate": 3.8910394265232975e-05,
+      "loss": 0.2583,
+      "step": 46440
+    },
+    {
+      "epoch": 299.68,
+      "learning_rate": 3.890681003584229e-05,
+      "loss": 0.2037,
+      "step": 46450
+    },
+    {
+      "epoch": 299.74,
+      "learning_rate": 3.890322580645162e-05,
+      "loss": 0.2583,
+      "step": 46460
+    },
+    {
+      "epoch": 299.8,
+      "learning_rate": 3.8899641577060936e-05,
+      "loss": 0.2617,
+      "step": 46470
+    },
+    {
+      "epoch": 299.87,
+      "learning_rate": 3.889605734767025e-05,
+      "loss": 0.2307,
+      "step": 46480
+    },
+    {
+      "epoch": 299.93,
+      "learning_rate": 3.889247311827957e-05,
+      "loss": 0.2677,
+      "step": 46490
+    },
+    {
+      "epoch": 300.0,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 0.2678,
+      "step": 46500
+    },
+    {
+      "epoch": 300.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0351133346557617,
+      "eval_runtime": 0.8581,
+      "eval_samples_per_second": 234.247,
+      "eval_steps_per_second": 8.158,
+      "step": 46500
+    },
+    {
+      "epoch": 300.06,
+      "learning_rate": 3.888530465949821e-05,
+      "loss": 0.2335,
+      "step": 46510
+    },
+    {
+      "epoch": 300.13,
+      "learning_rate": 3.888172043010753e-05,
+      "loss": 0.2709,
+      "step": 46520
+    },
+    {
+      "epoch": 300.19,
+      "learning_rate": 3.8878136200716845e-05,
+      "loss": 0.2845,
+      "step": 46530
+    },
+    {
+      "epoch": 300.26,
+      "learning_rate": 3.887455197132617e-05,
+      "loss": 0.3238,
+      "step": 46540
+    },
+    {
+      "epoch": 300.32,
+      "learning_rate": 3.887096774193549e-05,
+      "loss": 0.2661,
+      "step": 46550
+    },
+    {
+      "epoch": 300.39,
+      "learning_rate": 3.88673835125448e-05,
+      "loss": 0.2657,
+      "step": 46560
+    },
+    {
+      "epoch": 300.45,
+      "learning_rate": 3.8863799283154125e-05,
+      "loss": 0.2436,
+      "step": 46570
+    },
+    {
+      "epoch": 300.51,
+      "learning_rate": 3.886021505376344e-05,
+      "loss": 0.2208,
+      "step": 46580
+    },
+    {
+      "epoch": 300.58,
+      "learning_rate": 3.885663082437276e-05,
+      "loss": 0.2735,
+      "step": 46590
+    },
+    {
+      "epoch": 300.64,
+      "learning_rate": 3.885304659498208e-05,
+      "loss": 0.2671,
+      "step": 46600
+    },
+    {
+      "epoch": 300.71,
+      "learning_rate": 3.88494623655914e-05,
+      "loss": 0.2356,
+      "step": 46610
+    },
+    {
+      "epoch": 300.77,
+      "learning_rate": 3.884587813620072e-05,
+      "loss": 0.2567,
+      "step": 46620
+    },
+    {
+      "epoch": 300.84,
+      "learning_rate": 3.884229390681004e-05,
+      "loss": 0.3162,
+      "step": 46630
+    },
+    {
+      "epoch": 300.9,
+      "learning_rate": 3.883870967741935e-05,
+      "loss": 0.2921,
+      "step": 46640
+    },
+    {
+      "epoch": 300.96,
+      "learning_rate": 3.883512544802868e-05,
+      "loss": 0.248,
+      "step": 46650
+    },
+    {
+      "epoch": 301.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0845893621444702,
+      "eval_runtime": 0.8613,
+      "eval_samples_per_second": 233.365,
+      "eval_steps_per_second": 8.127,
+      "step": 46655
+    },
+    {
+      "epoch": 301.03,
+      "learning_rate": 3.8831541218637996e-05,
+      "loss": 0.2979,
+      "step": 46660
+    },
+    {
+      "epoch": 301.1,
+      "learning_rate": 3.8827956989247314e-05,
+      "loss": 0.2771,
+      "step": 46670
+    },
+    {
+      "epoch": 301.16,
+      "learning_rate": 3.882437275985663e-05,
+      "loss": 0.2218,
+      "step": 46680
+    },
+    {
+      "epoch": 301.23,
+      "learning_rate": 3.882078853046595e-05,
+      "loss": 0.3379,
+      "step": 46690
+    },
+    {
+      "epoch": 301.29,
+      "learning_rate": 3.881720430107527e-05,
+      "loss": 0.2709,
+      "step": 46700
+    },
+    {
+      "epoch": 301.35,
+      "learning_rate": 3.881362007168459e-05,
+      "loss": 0.2897,
+      "step": 46710
+    },
+    {
+      "epoch": 301.42,
+      "learning_rate": 3.8810035842293905e-05,
+      "loss": 0.2367,
+      "step": 46720
+    },
+    {
+      "epoch": 301.48,
+      "learning_rate": 3.880645161290323e-05,
+      "loss": 0.2394,
+      "step": 46730
+    },
+    {
+      "epoch": 301.55,
+      "learning_rate": 3.880286738351255e-05,
+      "loss": 0.2083,
+      "step": 46740
+    },
+    {
+      "epoch": 301.61,
+      "learning_rate": 3.8799283154121866e-05,
+      "loss": 0.2288,
+      "step": 46750
+    },
+    {
+      "epoch": 301.68,
+      "learning_rate": 3.8795698924731184e-05,
+      "loss": 0.2277,
+      "step": 46760
+    },
+    {
+      "epoch": 301.74,
+      "learning_rate": 3.87921146953405e-05,
+      "loss": 0.2596,
+      "step": 46770
+    },
+    {
+      "epoch": 301.8,
+      "learning_rate": 3.878853046594982e-05,
+      "loss": 0.2977,
+      "step": 46780
+    },
+    {
+      "epoch": 301.87,
+      "learning_rate": 3.8784946236559146e-05,
+      "loss": 0.2479,
+      "step": 46790
+    },
+    {
+      "epoch": 301.93,
+      "learning_rate": 3.878136200716846e-05,
+      "loss": 0.3291,
+      "step": 46800
+    },
+    {
+      "epoch": 302.0,
+      "learning_rate": 3.877777777777778e-05,
+      "loss": 0.3174,
+      "step": 46810
+    },
+    {
+      "epoch": 302.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0461586713790894,
+      "eval_runtime": 0.8513,
+      "eval_samples_per_second": 236.115,
+      "eval_steps_per_second": 8.223,
+      "step": 46810
+    },
+    {
+      "epoch": 302.06,
+      "learning_rate": 3.87741935483871e-05,
+      "loss": 0.3033,
+      "step": 46820
+    },
+    {
+      "epoch": 302.13,
+      "learning_rate": 3.877060931899642e-05,
+      "loss": 0.2816,
+      "step": 46830
+    },
+    {
+      "epoch": 302.19,
+      "learning_rate": 3.876702508960574e-05,
+      "loss": 0.2761,
+      "step": 46840
+    },
+    {
+      "epoch": 302.26,
+      "learning_rate": 3.8763440860215055e-05,
+      "loss": 0.3024,
+      "step": 46850
+    },
+    {
+      "epoch": 302.32,
+      "learning_rate": 3.875985663082437e-05,
+      "loss": 0.2953,
+      "step": 46860
+    },
+    {
+      "epoch": 302.39,
+      "learning_rate": 3.875627240143369e-05,
+      "loss": 0.204,
+      "step": 46870
+    },
+    {
+      "epoch": 302.45,
+      "learning_rate": 3.875268817204301e-05,
+      "loss": 0.2551,
+      "step": 46880
+    },
+    {
+      "epoch": 302.51,
+      "learning_rate": 3.8749103942652335e-05,
+      "loss": 0.3063,
+      "step": 46890
+    },
+    {
+      "epoch": 302.58,
+      "learning_rate": 3.874551971326165e-05,
+      "loss": 0.2829,
+      "step": 46900
+    },
+    {
+      "epoch": 302.64,
+      "learning_rate": 3.8741935483870964e-05,
+      "loss": 0.2812,
+      "step": 46910
+    },
+    {
+      "epoch": 302.71,
+      "learning_rate": 3.873835125448029e-05,
+      "loss": 0.2763,
+      "step": 46920
+    },
+    {
+      "epoch": 302.77,
+      "learning_rate": 3.873476702508961e-05,
+      "loss": 0.3135,
+      "step": 46930
+    },
+    {
+      "epoch": 302.84,
+      "learning_rate": 3.8731182795698926e-05,
+      "loss": 0.2303,
+      "step": 46940
+    },
+    {
+      "epoch": 302.9,
+      "learning_rate": 3.8727598566308244e-05,
+      "loss": 0.2494,
+      "step": 46950
+    },
+    {
+      "epoch": 302.96,
+      "learning_rate": 3.872401433691756e-05,
+      "loss": 0.2232,
+      "step": 46960
+    },
+    {
+      "epoch": 303.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.120409369468689,
+      "eval_runtime": 0.8539,
+      "eval_samples_per_second": 235.391,
+      "eval_steps_per_second": 8.198,
+      "step": 46965
+    },
+    {
+      "epoch": 303.03,
+      "learning_rate": 3.872043010752689e-05,
+      "loss": 0.2998,
+      "step": 46970
+    },
+    {
+      "epoch": 303.1,
+      "learning_rate": 3.8716845878136205e-05,
+      "loss": 0.2894,
+      "step": 46980
+    },
+    {
+      "epoch": 303.16,
+      "learning_rate": 3.871326164874552e-05,
+      "loss": 0.2841,
+      "step": 46990
+    },
+    {
+      "epoch": 303.23,
+      "learning_rate": 3.870967741935484e-05,
+      "loss": 0.277,
+      "step": 47000
+    },
+    {
+      "epoch": 303.29,
+      "learning_rate": 3.870609318996416e-05,
+      "loss": 0.2951,
+      "step": 47010
+    },
+    {
+      "epoch": 303.35,
+      "learning_rate": 3.870250896057348e-05,
+      "loss": 0.2538,
+      "step": 47020
+    },
+    {
+      "epoch": 303.42,
+      "learning_rate": 3.8698924731182796e-05,
+      "loss": 0.2541,
+      "step": 47030
+    },
+    {
+      "epoch": 303.48,
+      "learning_rate": 3.8695340501792114e-05,
+      "loss": 0.2476,
+      "step": 47040
+    },
+    {
+      "epoch": 303.55,
+      "learning_rate": 3.869175627240143e-05,
+      "loss": 0.2843,
+      "step": 47050
+    },
+    {
+      "epoch": 303.61,
+      "learning_rate": 3.868817204301076e-05,
+      "loss": 0.2879,
+      "step": 47060
+    },
+    {
+      "epoch": 303.68,
+      "learning_rate": 3.868458781362007e-05,
+      "loss": 0.2517,
+      "step": 47070
+    },
+    {
+      "epoch": 303.74,
+      "learning_rate": 3.8681003584229394e-05,
+      "loss": 0.29,
+      "step": 47080
+    },
+    {
+      "epoch": 303.8,
+      "learning_rate": 3.867741935483871e-05,
+      "loss": 0.2404,
+      "step": 47090
+    },
+    {
+      "epoch": 303.87,
+      "learning_rate": 3.867383512544803e-05,
+      "loss": 0.2633,
+      "step": 47100
+    },
+    {
+      "epoch": 303.93,
+      "learning_rate": 3.867025089605735e-05,
+      "loss": 0.2809,
+      "step": 47110
+    },
+    {
+      "epoch": 304.0,
+      "learning_rate": 3.866666666666667e-05,
+      "loss": 0.3091,
+      "step": 47120
+    },
+    {
+      "epoch": 304.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.1094586849212646,
+      "eval_runtime": 0.8779,
+      "eval_samples_per_second": 228.945,
+      "eval_steps_per_second": 7.973,
+      "step": 47120
+    },
+    {
+      "epoch": 304.06,
+      "learning_rate": 3.8663082437275985e-05,
+      "loss": 0.2606,
+      "step": 47130
+    },
+    {
+      "epoch": 304.13,
+      "learning_rate": 3.865949820788531e-05,
+      "loss": 0.3113,
+      "step": 47140
+    },
+    {
+      "epoch": 304.19,
+      "learning_rate": 3.865591397849462e-05,
+      "loss": 0.3285,
+      "step": 47150
+    },
+    {
+      "epoch": 304.26,
+      "learning_rate": 3.8652329749103946e-05,
+      "loss": 0.3021,
+      "step": 47160
+    },
+    {
+      "epoch": 304.32,
+      "learning_rate": 3.8648745519713265e-05,
+      "loss": 0.3452,
+      "step": 47170
+    },
+    {
+      "epoch": 304.39,
+      "learning_rate": 3.864516129032258e-05,
+      "loss": 0.2758,
+      "step": 47180
+    },
+    {
+      "epoch": 304.45,
+      "learning_rate": 3.86415770609319e-05,
+      "loss": 0.2817,
+      "step": 47190
+    },
+    {
+      "epoch": 304.51,
+      "learning_rate": 3.863799283154122e-05,
+      "loss": 0.2811,
+      "step": 47200
+    },
+    {
+      "epoch": 304.58,
+      "learning_rate": 3.863440860215054e-05,
+      "loss": 0.2853,
+      "step": 47210
+    },
+    {
+      "epoch": 304.64,
+      "learning_rate": 3.863082437275986e-05,
+      "loss": 0.3258,
+      "step": 47220
+    },
+    {
+      "epoch": 304.71,
+      "learning_rate": 3.8627240143369174e-05,
+      "loss": 0.2647,
+      "step": 47230
+    },
+    {
+      "epoch": 304.77,
+      "learning_rate": 3.86236559139785e-05,
+      "loss": 0.2657,
+      "step": 47240
+    },
+    {
+      "epoch": 304.84,
+      "learning_rate": 3.862007168458782e-05,
+      "loss": 0.2419,
+      "step": 47250
+    },
+    {
+      "epoch": 304.9,
+      "learning_rate": 3.8616487455197135e-05,
+      "loss": 0.2703,
+      "step": 47260
+    },
+    {
+      "epoch": 304.96,
+      "learning_rate": 3.8612903225806454e-05,
+      "loss": 0.2605,
+      "step": 47270
+    },
+    {
+      "epoch": 305.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0886881351470947,
+      "eval_runtime": 0.8393,
+      "eval_samples_per_second": 239.497,
+      "eval_steps_per_second": 8.341,
+      "step": 47275
+    },
+    {
+      "epoch": 305.03,
+      "learning_rate": 3.860931899641577e-05,
+      "loss": 0.3109,
+      "step": 47280
+    },
+    {
+      "epoch": 305.1,
+      "learning_rate": 3.860573476702509e-05,
+      "loss": 0.2692,
+      "step": 47290
+    },
+    {
+      "epoch": 305.16,
+      "learning_rate": 3.8602150537634415e-05,
+      "loss": 0.3095,
+      "step": 47300
+    },
+    {
+      "epoch": 305.23,
+      "learning_rate": 3.8598566308243726e-05,
+      "loss": 0.2743,
+      "step": 47310
+    },
+    {
+      "epoch": 305.29,
+      "learning_rate": 3.859498207885305e-05,
+      "loss": 0.2904,
+      "step": 47320
+    },
+    {
+      "epoch": 305.35,
+      "learning_rate": 3.859139784946237e-05,
+      "loss": 0.2184,
+      "step": 47330
+    },
+    {
+      "epoch": 305.42,
+      "learning_rate": 3.858781362007169e-05,
+      "loss": 0.2727,
+      "step": 47340
+    },
+    {
+      "epoch": 305.48,
+      "learning_rate": 3.8584229390681006e-05,
+      "loss": 0.2884,
+      "step": 47350
+    },
+    {
+      "epoch": 305.55,
+      "learning_rate": 3.8580645161290324e-05,
+      "loss": 0.2507,
+      "step": 47360
+    },
+    {
+      "epoch": 305.61,
+      "learning_rate": 3.857706093189964e-05,
+      "loss": 0.2524,
+      "step": 47370
+    },
+    {
+      "epoch": 305.68,
+      "learning_rate": 3.857347670250897e-05,
+      "loss": 0.2683,
+      "step": 47380
+    },
+    {
+      "epoch": 305.74,
+      "learning_rate": 3.856989247311828e-05,
+      "loss": 0.2588,
+      "step": 47390
+    },
+    {
+      "epoch": 305.8,
+      "learning_rate": 3.85663082437276e-05,
+      "loss": 0.2481,
+      "step": 47400
+    },
+    {
+      "epoch": 305.87,
+      "learning_rate": 3.856272401433692e-05,
+      "loss": 0.2508,
+      "step": 47410
+    },
+    {
+      "epoch": 305.93,
+      "learning_rate": 3.855913978494624e-05,
+      "loss": 0.2704,
+      "step": 47420
+    },
+    {
+      "epoch": 306.0,
+      "learning_rate": 3.855555555555556e-05,
+      "loss": 0.2816,
+      "step": 47430
+    },
+    {
+      "epoch": 306.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.001700758934021,
+      "eval_runtime": 0.8611,
+      "eval_samples_per_second": 233.419,
+      "eval_steps_per_second": 8.129,
+      "step": 47430
+    },
+    {
+      "epoch": 306.06,
+      "learning_rate": 3.8551971326164877e-05,
+      "loss": 0.2764,
+      "step": 47440
+    },
+    {
+      "epoch": 306.13,
+      "learning_rate": 3.8548387096774195e-05,
+      "loss": 0.2795,
+      "step": 47450
+    },
+    {
+      "epoch": 306.19,
+      "learning_rate": 3.854480286738352e-05,
+      "loss": 0.247,
+      "step": 47460
+    },
+    {
+      "epoch": 306.26,
+      "learning_rate": 3.854121863799283e-05,
+      "loss": 0.2684,
+      "step": 47470
+    },
+    {
+      "epoch": 306.32,
+      "learning_rate": 3.853763440860215e-05,
+      "loss": 0.2734,
+      "step": 47480
+    },
+    {
+      "epoch": 306.39,
+      "learning_rate": 3.8534050179211474e-05,
+      "loss": 0.2876,
+      "step": 47490
+    },
+    {
+      "epoch": 306.45,
+      "learning_rate": 3.8530465949820786e-05,
+      "loss": 0.3119,
+      "step": 47500
+    },
+    {
+      "epoch": 306.51,
+      "learning_rate": 3.852688172043011e-05,
+      "loss": 0.3025,
+      "step": 47510
+    },
+    {
+      "epoch": 306.58,
+      "learning_rate": 3.852329749103943e-05,
+      "loss": 0.324,
+      "step": 47520
+    },
+    {
+      "epoch": 306.64,
+      "learning_rate": 3.851971326164875e-05,
+      "loss": 0.2502,
+      "step": 47530
+    },
+    {
+      "epoch": 306.71,
+      "learning_rate": 3.8516129032258065e-05,
+      "loss": 0.2804,
+      "step": 47540
+    },
+    {
+      "epoch": 306.77,
+      "learning_rate": 3.8512544802867384e-05,
+      "loss": 0.2792,
+      "step": 47550
+    },
+    {
+      "epoch": 306.84,
+      "learning_rate": 3.85089605734767e-05,
+      "loss": 0.3049,
+      "step": 47560
+    },
+    {
+      "epoch": 306.9,
+      "learning_rate": 3.850537634408603e-05,
+      "loss": 0.2396,
+      "step": 47570
+    },
+    {
+      "epoch": 306.96,
+      "learning_rate": 3.850179211469534e-05,
+      "loss": 0.272,
+      "step": 47580
+    },
+    {
+      "epoch": 307.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.064512848854065,
+      "eval_runtime": 0.8615,
+      "eval_samples_per_second": 233.321,
+      "eval_steps_per_second": 8.126,
+      "step": 47585
+    },
+    {
+      "epoch": 307.03,
+      "learning_rate": 3.849820788530466e-05,
+      "loss": 0.2785,
+      "step": 47590
+    },
+    {
+      "epoch": 307.1,
+      "learning_rate": 3.849462365591398e-05,
+      "loss": 0.2596,
+      "step": 47600
+    },
+    {
+      "epoch": 307.16,
+      "learning_rate": 3.84910394265233e-05,
+      "loss": 0.2504,
+      "step": 47610
+    },
+    {
+      "epoch": 307.23,
+      "learning_rate": 3.848745519713262e-05,
+      "loss": 0.3005,
+      "step": 47620
+    },
+    {
+      "epoch": 307.29,
+      "learning_rate": 3.8483870967741936e-05,
+      "loss": 0.2484,
+      "step": 47630
+    },
+    {
+      "epoch": 307.35,
+      "learning_rate": 3.8480286738351254e-05,
+      "loss": 0.3235,
+      "step": 47640
+    },
+    {
+      "epoch": 307.42,
+      "learning_rate": 3.847670250896058e-05,
+      "loss": 0.263,
+      "step": 47650
+    },
+    {
+      "epoch": 307.48,
+      "learning_rate": 3.847311827956989e-05,
+      "loss": 0.2798,
+      "step": 47660
+    },
+    {
+      "epoch": 307.55,
+      "learning_rate": 3.8469534050179216e-05,
+      "loss": 0.2366,
+      "step": 47670
+    },
+    {
+      "epoch": 307.61,
+      "learning_rate": 3.8465949820788534e-05,
+      "loss": 0.3287,
+      "step": 47680
+    },
+    {
+      "epoch": 307.68,
+      "learning_rate": 3.846236559139785e-05,
+      "loss": 0.2652,
+      "step": 47690
+    },
+    {
+      "epoch": 307.74,
+      "learning_rate": 3.845878136200717e-05,
+      "loss": 0.2575,
+      "step": 47700
+    },
+    {
+      "epoch": 307.8,
+      "learning_rate": 3.845519713261649e-05,
+      "loss": 0.2567,
+      "step": 47710
+    },
+    {
+      "epoch": 307.87,
+      "learning_rate": 3.845161290322581e-05,
+      "loss": 0.236,
+      "step": 47720
+    },
+    {
+      "epoch": 307.93,
+      "learning_rate": 3.844802867383513e-05,
+      "loss": 0.2777,
+      "step": 47730
+    },
+    {
+      "epoch": 308.0,
+      "learning_rate": 3.844444444444444e-05,
+      "loss": 0.3205,
+      "step": 47740
+    },
+    {
+      "epoch": 308.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0545997619628906,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 230.727,
+      "eval_steps_per_second": 8.035,
+      "step": 47740
+    },
+    {
+      "epoch": 308.06,
+      "learning_rate": 3.844086021505376e-05,
+      "loss": 0.2709,
+      "step": 47750
+    },
+    {
+      "epoch": 308.13,
+      "learning_rate": 3.8437275985663086e-05,
+      "loss": 0.2902,
+      "step": 47760
+    },
+    {
+      "epoch": 308.19,
+      "learning_rate": 3.8433691756272404e-05,
+      "loss": 0.2478,
+      "step": 47770
+    },
+    {
+      "epoch": 308.26,
+      "learning_rate": 3.843010752688172e-05,
+      "loss": 0.2979,
+      "step": 47780
+    },
+    {
+      "epoch": 308.32,
+      "learning_rate": 3.842652329749104e-05,
+      "loss": 0.3162,
+      "step": 47790
+    },
+    {
+      "epoch": 308.39,
+      "learning_rate": 3.842293906810036e-05,
+      "loss": 0.2414,
+      "step": 47800
+    },
+    {
+      "epoch": 308.45,
+      "learning_rate": 3.8419354838709684e-05,
+      "loss": 0.3075,
+      "step": 47810
+    },
+    {
+      "epoch": 308.51,
+      "learning_rate": 3.8415770609318995e-05,
+      "loss": 0.2685,
+      "step": 47820
+    },
+    {
+      "epoch": 308.58,
+      "learning_rate": 3.8412186379928314e-05,
+      "loss": 0.2703,
+      "step": 47830
+    },
+    {
+      "epoch": 308.64,
+      "learning_rate": 3.840860215053764e-05,
+      "loss": 0.2802,
+      "step": 47840
+    },
+    {
+      "epoch": 308.71,
+      "learning_rate": 3.840501792114696e-05,
+      "loss": 0.3119,
+      "step": 47850
+    },
+    {
+      "epoch": 308.77,
+      "learning_rate": 3.8401433691756275e-05,
+      "loss": 0.3135,
+      "step": 47860
+    },
+    {
+      "epoch": 308.84,
+      "learning_rate": 3.839784946236559e-05,
+      "loss": 0.2771,
+      "step": 47870
+    },
+    {
+      "epoch": 308.9,
+      "learning_rate": 3.839426523297491e-05,
+      "loss": 0.2766,
+      "step": 47880
+    },
+    {
+      "epoch": 308.96,
+      "learning_rate": 3.8390681003584236e-05,
+      "loss": 0.2589,
+      "step": 47890
+    },
+    {
+      "epoch": 309.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0237172842025757,
+      "eval_runtime": 0.8554,
+      "eval_samples_per_second": 234.974,
+      "eval_steps_per_second": 8.183,
+      "step": 47895
+    },
+    {
+      "epoch": 309.03,
+      "learning_rate": 3.838709677419355e-05,
+      "loss": 0.2503,
+      "step": 47900
+    },
+    {
+      "epoch": 309.1,
+      "learning_rate": 3.8383512544802866e-05,
+      "loss": 0.2434,
+      "step": 47910
+    },
+    {
+      "epoch": 309.16,
+      "learning_rate": 3.837992831541219e-05,
+      "loss": 0.2929,
+      "step": 47920
+    },
+    {
+      "epoch": 309.23,
+      "learning_rate": 3.837634408602151e-05,
+      "loss": 0.2612,
+      "step": 47930
+    },
+    {
+      "epoch": 309.29,
+      "learning_rate": 3.837275985663083e-05,
+      "loss": 0.2596,
+      "step": 47940
+    },
+    {
+      "epoch": 309.35,
+      "learning_rate": 3.8369175627240146e-05,
+      "loss": 0.2811,
+      "step": 47950
+    },
+    {
+      "epoch": 309.42,
+      "learning_rate": 3.8365591397849464e-05,
+      "loss": 0.2582,
+      "step": 47960
+    },
+    {
+      "epoch": 309.48,
+      "learning_rate": 3.836200716845879e-05,
+      "loss": 0.2553,
+      "step": 47970
+    },
+    {
+      "epoch": 309.55,
+      "learning_rate": 3.83584229390681e-05,
+      "loss": 0.2535,
+      "step": 47980
+    },
+    {
+      "epoch": 309.61,
+      "learning_rate": 3.835483870967742e-05,
+      "loss": 0.2558,
+      "step": 47990
+    },
+    {
+      "epoch": 309.68,
+      "learning_rate": 3.8351254480286743e-05,
+      "loss": 0.2932,
+      "step": 48000
+    },
+    {
+      "epoch": 309.74,
+      "learning_rate": 3.834767025089606e-05,
+      "loss": 0.254,
+      "step": 48010
+    },
+    {
+      "epoch": 309.8,
+      "learning_rate": 3.834408602150538e-05,
+      "loss": 0.2545,
+      "step": 48020
+    },
+    {
+      "epoch": 309.87,
+      "learning_rate": 3.83405017921147e-05,
+      "loss": 0.2719,
+      "step": 48030
+    },
+    {
+      "epoch": 309.93,
+      "learning_rate": 3.8336917562724016e-05,
+      "loss": 0.2663,
+      "step": 48040
+    },
+    {
+      "epoch": 310.0,
+      "learning_rate": 3.8333333333333334e-05,
+      "loss": 0.2841,
+      "step": 48050
+    },
+    {
+      "epoch": 310.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.048534631729126,
+      "eval_runtime": 0.8638,
+      "eval_samples_per_second": 232.682,
+      "eval_steps_per_second": 8.103,
+      "step": 48050
+    },
+    {
+      "epoch": 310.06,
+      "learning_rate": 3.832974910394265e-05,
+      "loss": 0.2773,
+      "step": 48060
+    },
+    {
+      "epoch": 310.13,
+      "learning_rate": 3.832616487455197e-05,
+      "loss": 0.2949,
+      "step": 48070
+    },
+    {
+      "epoch": 310.19,
+      "learning_rate": 3.8322580645161296e-05,
+      "loss": 0.2485,
+      "step": 48080
+    },
+    {
+      "epoch": 310.26,
+      "learning_rate": 3.8318996415770614e-05,
+      "loss": 0.2079,
+      "step": 48090
+    },
+    {
+      "epoch": 310.32,
+      "learning_rate": 3.8315412186379926e-05,
+      "loss": 0.2356,
+      "step": 48100
+    },
+    {
+      "epoch": 310.39,
+      "learning_rate": 3.831182795698925e-05,
+      "loss": 0.2843,
+      "step": 48110
+    },
+    {
+      "epoch": 310.45,
+      "learning_rate": 3.830824372759857e-05,
+      "loss": 0.2599,
+      "step": 48120
+    },
+    {
+      "epoch": 310.51,
+      "learning_rate": 3.830465949820789e-05,
+      "loss": 0.2646,
+      "step": 48130
+    },
+    {
+      "epoch": 310.58,
+      "learning_rate": 3.8301075268817205e-05,
+      "loss": 0.2787,
+      "step": 48140
+    },
+    {
+      "epoch": 310.64,
+      "learning_rate": 3.829749103942652e-05,
+      "loss": 0.2615,
+      "step": 48150
+    },
+    {
+      "epoch": 310.71,
+      "learning_rate": 3.829390681003585e-05,
+      "loss": 0.2674,
+      "step": 48160
+    },
+    {
+      "epoch": 310.77,
+      "learning_rate": 3.829032258064516e-05,
+      "loss": 0.2646,
+      "step": 48170
+    },
+    {
+      "epoch": 310.84,
+      "learning_rate": 3.828673835125448e-05,
+      "loss": 0.2546,
+      "step": 48180
+    },
+    {
+      "epoch": 310.9,
+      "learning_rate": 3.82831541218638e-05,
+      "loss": 0.2583,
+      "step": 48190
+    },
+    {
+      "epoch": 310.96,
+      "learning_rate": 3.827956989247312e-05,
+      "loss": 0.3025,
+      "step": 48200
+    },
+    {
+      "epoch": 311.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0197347402572632,
+      "eval_runtime": 0.8496,
+      "eval_samples_per_second": 236.591,
+      "eval_steps_per_second": 8.239,
+      "step": 48205
+    },
+    {
+      "epoch": 311.03,
+      "learning_rate": 3.827598566308244e-05,
+      "loss": 0.2946,
+      "step": 48210
+    },
+    {
+      "epoch": 311.1,
+      "learning_rate": 3.827240143369176e-05,
+      "loss": 0.2452,
+      "step": 48220
+    },
+    {
+      "epoch": 311.16,
+      "learning_rate": 3.8268817204301076e-05,
+      "loss": 0.2974,
+      "step": 48230
+    },
+    {
+      "epoch": 311.23,
+      "learning_rate": 3.82652329749104e-05,
+      "loss": 0.275,
+      "step": 48240
+    },
+    {
+      "epoch": 311.29,
+      "learning_rate": 3.826164874551971e-05,
+      "loss": 0.2905,
+      "step": 48250
+    },
+    {
+      "epoch": 311.35,
+      "learning_rate": 3.825806451612903e-05,
+      "loss": 0.3023,
+      "step": 48260
+    },
+    {
+      "epoch": 311.42,
+      "learning_rate": 3.8254480286738355e-05,
+      "loss": 0.2516,
+      "step": 48270
+    },
+    {
+      "epoch": 311.48,
+      "learning_rate": 3.8250896057347674e-05,
+      "loss": 0.2744,
+      "step": 48280
+    },
+    {
+      "epoch": 311.55,
+      "learning_rate": 3.824731182795699e-05,
+      "loss": 0.3041,
+      "step": 48290
+    },
+    {
+      "epoch": 311.61,
+      "learning_rate": 3.824372759856631e-05,
+      "loss": 0.2951,
+      "step": 48300
+    },
+    {
+      "epoch": 311.68,
+      "learning_rate": 3.824014336917563e-05,
+      "loss": 0.2822,
+      "step": 48310
+    },
+    {
+      "epoch": 311.74,
+      "learning_rate": 3.823655913978495e-05,
+      "loss": 0.2099,
+      "step": 48320
+    },
+    {
+      "epoch": 311.8,
+      "learning_rate": 3.8232974910394265e-05,
+      "loss": 0.2637,
+      "step": 48330
+    },
+    {
+      "epoch": 311.87,
+      "learning_rate": 3.822939068100358e-05,
+      "loss": 0.2786,
+      "step": 48340
+    },
+    {
+      "epoch": 311.93,
+      "learning_rate": 3.822580645161291e-05,
+      "loss": 0.3205,
+      "step": 48350
+    },
+    {
+      "epoch": 312.0,
+      "learning_rate": 3.8222222222222226e-05,
+      "loss": 0.2926,
+      "step": 48360
+    },
+    {
+      "epoch": 312.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0425233840942383,
+      "eval_runtime": 0.8579,
+      "eval_samples_per_second": 234.3,
+      "eval_steps_per_second": 8.16,
+      "step": 48360
+    },
+    {
+      "epoch": 312.06,
+      "learning_rate": 3.8218637992831544e-05,
+      "loss": 0.2685,
+      "step": 48370
+    },
+    {
+      "epoch": 312.13,
+      "learning_rate": 3.821505376344086e-05,
+      "loss": 0.2359,
+      "step": 48380
+    },
+    {
+      "epoch": 312.19,
+      "learning_rate": 3.821146953405018e-05,
+      "loss": 0.2561,
+      "step": 48390
+    },
+    {
+      "epoch": 312.26,
+      "learning_rate": 3.82078853046595e-05,
+      "loss": 0.3064,
+      "step": 48400
+    },
+    {
+      "epoch": 312.32,
+      "learning_rate": 3.820430107526882e-05,
+      "loss": 0.2175,
+      "step": 48410
+    },
+    {
+      "epoch": 312.39,
+      "learning_rate": 3.8200716845878135e-05,
+      "loss": 0.2423,
+      "step": 48420
+    },
+    {
+      "epoch": 312.45,
+      "learning_rate": 3.819713261648746e-05,
+      "loss": 0.2846,
+      "step": 48430
+    },
+    {
+      "epoch": 312.51,
+      "learning_rate": 3.819354838709678e-05,
+      "loss": 0.2834,
+      "step": 48440
+    },
+    {
+      "epoch": 312.58,
+      "learning_rate": 3.818996415770609e-05,
+      "loss": 0.2321,
+      "step": 48450
+    },
+    {
+      "epoch": 312.64,
+      "learning_rate": 3.8186379928315415e-05,
+      "loss": 0.3262,
+      "step": 48460
+    },
+    {
+      "epoch": 312.71,
+      "learning_rate": 3.818279569892473e-05,
+      "loss": 0.2673,
+      "step": 48470
+    },
+    {
+      "epoch": 312.77,
+      "learning_rate": 3.817921146953405e-05,
+      "loss": 0.3061,
+      "step": 48480
+    },
+    {
+      "epoch": 312.84,
+      "learning_rate": 3.817562724014337e-05,
+      "loss": 0.2488,
+      "step": 48490
+    },
+    {
+      "epoch": 312.9,
+      "learning_rate": 3.817204301075269e-05,
+      "loss": 0.2488,
+      "step": 48500
+    },
+    {
+      "epoch": 312.96,
+      "learning_rate": 3.816845878136201e-05,
+      "loss": 0.2339,
+      "step": 48510
+    },
+    {
+      "epoch": 313.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8882163763046265,
+      "eval_runtime": 0.9006,
+      "eval_samples_per_second": 223.173,
+      "eval_steps_per_second": 7.772,
+      "step": 48515
+    },
+    {
+      "epoch": 313.03,
+      "learning_rate": 3.816487455197133e-05,
+      "loss": 0.2778,
+      "step": 48520
+    },
+    {
+      "epoch": 313.1,
+      "learning_rate": 3.816129032258064e-05,
+      "loss": 0.2517,
+      "step": 48530
+    },
+    {
+      "epoch": 313.16,
+      "learning_rate": 3.815770609318997e-05,
+      "loss": 0.2898,
+      "step": 48540
+    },
+    {
+      "epoch": 313.23,
+      "learning_rate": 3.8154121863799285e-05,
+      "loss": 0.2641,
+      "step": 48550
+    },
+    {
+      "epoch": 313.29,
+      "learning_rate": 3.8150537634408604e-05,
+      "loss": 0.2219,
+      "step": 48560
+    },
+    {
+      "epoch": 313.35,
+      "learning_rate": 3.814695340501792e-05,
+      "loss": 0.2673,
+      "step": 48570
+    },
+    {
+      "epoch": 313.42,
+      "learning_rate": 3.814336917562724e-05,
+      "loss": 0.2725,
+      "step": 48580
+    },
+    {
+      "epoch": 313.48,
+      "learning_rate": 3.8139784946236565e-05,
+      "loss": 0.2535,
+      "step": 48590
+    },
+    {
+      "epoch": 313.55,
+      "learning_rate": 3.813620071684588e-05,
+      "loss": 0.3119,
+      "step": 48600
+    },
+    {
+      "epoch": 313.61,
+      "learning_rate": 3.8132616487455195e-05,
+      "loss": 0.2526,
+      "step": 48610
+    },
+    {
+      "epoch": 313.68,
+      "learning_rate": 3.812903225806452e-05,
+      "loss": 0.3199,
+      "step": 48620
+    },
+    {
+      "epoch": 313.74,
+      "learning_rate": 3.812544802867384e-05,
+      "loss": 0.2934,
+      "step": 48630
+    },
+    {
+      "epoch": 313.8,
+      "learning_rate": 3.8121863799283156e-05,
+      "loss": 0.2688,
+      "step": 48640
+    },
+    {
+      "epoch": 313.87,
+      "learning_rate": 3.8118279569892474e-05,
+      "loss": 0.3169,
+      "step": 48650
+    },
+    {
+      "epoch": 313.93,
+      "learning_rate": 3.811469534050179e-05,
+      "loss": 0.2421,
+      "step": 48660
+    },
+    {
+      "epoch": 314.0,
+      "learning_rate": 3.811111111111112e-05,
+      "loss": 0.2661,
+      "step": 48670
+    },
+    {
+      "epoch": 314.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.986229419708252,
+      "eval_runtime": 0.8775,
+      "eval_samples_per_second": 229.057,
+      "eval_steps_per_second": 7.977,
+      "step": 48670
+    },
+    {
+      "epoch": 314.06,
+      "learning_rate": 3.8107526881720436e-05,
+      "loss": 0.2855,
+      "step": 48680
+    },
+    {
+      "epoch": 314.13,
+      "learning_rate": 3.810394265232975e-05,
+      "loss": 0.3087,
+      "step": 48690
+    },
+    {
+      "epoch": 314.19,
+      "learning_rate": 3.810035842293907e-05,
+      "loss": 0.2411,
+      "step": 48700
+    },
+    {
+      "epoch": 314.26,
+      "learning_rate": 3.809677419354839e-05,
+      "loss": 0.2222,
+      "step": 48710
+    },
+    {
+      "epoch": 314.32,
+      "learning_rate": 3.809318996415771e-05,
+      "loss": 0.2509,
+      "step": 48720
+    },
+    {
+      "epoch": 314.39,
+      "learning_rate": 3.808960573476703e-05,
+      "loss": 0.2546,
+      "step": 48730
+    },
+    {
+      "epoch": 314.45,
+      "learning_rate": 3.8086021505376345e-05,
+      "loss": 0.2325,
+      "step": 48740
+    },
+    {
+      "epoch": 314.51,
+      "learning_rate": 3.808243727598566e-05,
+      "loss": 0.2764,
+      "step": 48750
+    },
+    {
+      "epoch": 314.58,
+      "learning_rate": 3.807885304659499e-05,
+      "loss": 0.2473,
+      "step": 48760
+    },
+    {
+      "epoch": 314.64,
+      "learning_rate": 3.80752688172043e-05,
+      "loss": 0.2542,
+      "step": 48770
+    },
+    {
+      "epoch": 314.71,
+      "learning_rate": 3.8071684587813624e-05,
+      "loss": 0.2618,
+      "step": 48780
+    },
+    {
+      "epoch": 314.77,
+      "learning_rate": 3.806810035842294e-05,
+      "loss": 0.2965,
+      "step": 48790
+    },
+    {
+      "epoch": 314.84,
+      "learning_rate": 3.8064516129032254e-05,
+      "loss": 0.2497,
+      "step": 48800
+    },
+    {
+      "epoch": 314.9,
+      "learning_rate": 3.806093189964158e-05,
+      "loss": 0.3191,
+      "step": 48810
+    },
+    {
+      "epoch": 314.96,
+      "learning_rate": 3.80573476702509e-05,
+      "loss": 0.2669,
+      "step": 48820
+    },
+    {
+      "epoch": 315.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.035635232925415,
+      "eval_runtime": 0.8618,
+      "eval_samples_per_second": 233.225,
+      "eval_steps_per_second": 8.122,
+      "step": 48825
+    },
+    {
+      "epoch": 315.03,
+      "learning_rate": 3.8053763440860215e-05,
+      "loss": 0.2402,
+      "step": 48830
+    },
+    {
+      "epoch": 315.1,
+      "learning_rate": 3.8050179211469534e-05,
+      "loss": 0.2865,
+      "step": 48840
+    },
+    {
+      "epoch": 315.16,
+      "learning_rate": 3.804659498207885e-05,
+      "loss": 0.3079,
+      "step": 48850
+    },
+    {
+      "epoch": 315.23,
+      "learning_rate": 3.804301075268818e-05,
+      "loss": 0.278,
+      "step": 48860
+    },
+    {
+      "epoch": 315.29,
+      "learning_rate": 3.8039426523297495e-05,
+      "loss": 0.2523,
+      "step": 48870
+    },
+    {
+      "epoch": 315.35,
+      "learning_rate": 3.8035842293906806e-05,
+      "loss": 0.2391,
+      "step": 48880
+    },
+    {
+      "epoch": 315.42,
+      "learning_rate": 3.803225806451613e-05,
+      "loss": 0.2401,
+      "step": 48890
+    },
+    {
+      "epoch": 315.48,
+      "learning_rate": 3.802867383512545e-05,
+      "loss": 0.2746,
+      "step": 48900
+    },
+    {
+      "epoch": 315.55,
+      "learning_rate": 3.802508960573477e-05,
+      "loss": 0.2716,
+      "step": 48910
+    },
+    {
+      "epoch": 315.61,
+      "learning_rate": 3.8021505376344086e-05,
+      "loss": 0.2914,
+      "step": 48920
+    },
+    {
+      "epoch": 315.68,
+      "learning_rate": 3.8017921146953404e-05,
+      "loss": 0.2607,
+      "step": 48930
+    },
+    {
+      "epoch": 315.74,
+      "learning_rate": 3.801433691756273e-05,
+      "loss": 0.2562,
+      "step": 48940
+    },
+    {
+      "epoch": 315.8,
+      "learning_rate": 3.801075268817205e-05,
+      "loss": 0.2799,
+      "step": 48950
+    },
+    {
+      "epoch": 315.87,
+      "learning_rate": 3.800716845878136e-05,
+      "loss": 0.262,
+      "step": 48960
+    },
+    {
+      "epoch": 315.93,
+      "learning_rate": 3.8003584229390684e-05,
+      "loss": 0.3004,
+      "step": 48970
+    },
+    {
+      "epoch": 316.0,
+      "learning_rate": 3.8e-05,
+      "loss": 0.3174,
+      "step": 48980
+    },
+    {
+      "epoch": 316.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0152045488357544,
+      "eval_runtime": 0.8774,
+      "eval_samples_per_second": 229.084,
+      "eval_steps_per_second": 7.978,
+      "step": 48980
+    },
+    {
+      "epoch": 316.06,
+      "learning_rate": 3.799641577060932e-05,
+      "loss": 0.3492,
+      "step": 48990
+    },
+    {
+      "epoch": 316.13,
+      "learning_rate": 3.799283154121864e-05,
+      "loss": 0.3577,
+      "step": 49000
+    },
+    {
+      "epoch": 316.19,
+      "learning_rate": 3.798924731182796e-05,
+      "loss": 0.3278,
+      "step": 49010
+    },
+    {
+      "epoch": 316.26,
+      "learning_rate": 3.798566308243728e-05,
+      "loss": 0.3596,
+      "step": 49020
+    },
+    {
+      "epoch": 316.32,
+      "learning_rate": 3.79820788530466e-05,
+      "loss": 0.2972,
+      "step": 49030
+    },
+    {
+      "epoch": 316.39,
+      "learning_rate": 3.797849462365591e-05,
+      "loss": 0.364,
+      "step": 49040
+    },
+    {
+      "epoch": 316.45,
+      "learning_rate": 3.7974910394265236e-05,
+      "loss": 0.3243,
+      "step": 49050
+    },
+    {
+      "epoch": 316.51,
+      "learning_rate": 3.7971326164874555e-05,
+      "loss": 0.3212,
+      "step": 49060
+    },
+    {
+      "epoch": 316.58,
+      "learning_rate": 3.796774193548387e-05,
+      "loss": 0.2794,
+      "step": 49070
+    },
+    {
+      "epoch": 316.64,
+      "learning_rate": 3.796415770609319e-05,
+      "loss": 0.3528,
+      "step": 49080
+    },
+    {
+      "epoch": 316.71,
+      "learning_rate": 3.796057347670251e-05,
+      "loss": 0.276,
+      "step": 49090
+    },
+    {
+      "epoch": 316.77,
+      "learning_rate": 3.795698924731183e-05,
+      "loss": 0.391,
+      "step": 49100
+    },
+    {
+      "epoch": 316.84,
+      "learning_rate": 3.795340501792115e-05,
+      "loss": 0.2744,
+      "step": 49110
+    },
+    {
+      "epoch": 316.9,
+      "learning_rate": 3.7949820788530464e-05,
+      "loss": 0.2907,
+      "step": 49120
+    },
+    {
+      "epoch": 316.96,
+      "learning_rate": 3.794623655913979e-05,
+      "loss": 0.3458,
+      "step": 49130
+    },
+    {
+      "epoch": 317.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0714484453201294,
+      "eval_runtime": 0.8483,
+      "eval_samples_per_second": 236.943,
+      "eval_steps_per_second": 8.252,
+      "step": 49135
+    },
+    {
+      "epoch": 317.03,
+      "learning_rate": 3.794265232974911e-05,
+      "loss": 0.3522,
+      "step": 49140
+    },
+    {
+      "epoch": 317.1,
+      "learning_rate": 3.7939068100358425e-05,
+      "loss": 0.3076,
+      "step": 49150
+    },
+    {
+      "epoch": 317.16,
+      "learning_rate": 3.793548387096774e-05,
+      "loss": 0.3441,
+      "step": 49160
+    },
+    {
+      "epoch": 317.23,
+      "learning_rate": 3.793189964157706e-05,
+      "loss": 0.3586,
+      "step": 49170
+    },
+    {
+      "epoch": 317.29,
+      "learning_rate": 3.792831541218638e-05,
+      "loss": 0.3202,
+      "step": 49180
+    },
+    {
+      "epoch": 317.35,
+      "learning_rate": 3.7924731182795705e-05,
+      "loss": 0.3162,
+      "step": 49190
+    },
+    {
+      "epoch": 317.42,
+      "learning_rate": 3.7921146953405016e-05,
+      "loss": 0.2976,
+      "step": 49200
+    },
+    {
+      "epoch": 317.48,
+      "learning_rate": 3.791756272401434e-05,
+      "loss": 0.2908,
+      "step": 49210
+    },
+    {
+      "epoch": 317.55,
+      "learning_rate": 3.791397849462366e-05,
+      "loss": 0.2634,
+      "step": 49220
+    },
+    {
+      "epoch": 317.61,
+      "learning_rate": 3.791039426523298e-05,
+      "loss": 0.2929,
+      "step": 49230
+    },
+    {
+      "epoch": 317.68,
+      "learning_rate": 3.7906810035842296e-05,
+      "loss": 0.3026,
+      "step": 49240
+    },
+    {
+      "epoch": 317.74,
+      "learning_rate": 3.7903225806451614e-05,
+      "loss": 0.2842,
+      "step": 49250
+    },
+    {
+      "epoch": 317.8,
+      "learning_rate": 3.789964157706093e-05,
+      "loss": 0.2475,
+      "step": 49260
+    },
+    {
+      "epoch": 317.87,
+      "learning_rate": 3.789605734767026e-05,
+      "loss": 0.3119,
+      "step": 49270
+    },
+    {
+      "epoch": 317.93,
+      "learning_rate": 3.789247311827957e-05,
+      "loss": 0.2491,
+      "step": 49280
+    },
+    {
+      "epoch": 318.0,
+      "learning_rate": 3.7888888888888894e-05,
+      "loss": 0.3176,
+      "step": 49290
+    },
+    {
+      "epoch": 318.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0448318719863892,
+      "eval_runtime": 0.8968,
+      "eval_samples_per_second": 224.137,
+      "eval_steps_per_second": 7.806,
+      "step": 49290
+    },
+    {
+      "epoch": 318.06,
+      "learning_rate": 3.788530465949821e-05,
+      "loss": 0.3261,
+      "step": 49300
+    },
+    {
+      "epoch": 318.13,
+      "learning_rate": 3.788172043010753e-05,
+      "loss": 0.29,
+      "step": 49310
+    },
+    {
+      "epoch": 318.19,
+      "learning_rate": 3.787813620071685e-05,
+      "loss": 0.2923,
+      "step": 49320
+    },
+    {
+      "epoch": 318.26,
+      "learning_rate": 3.7874551971326166e-05,
+      "loss": 0.2831,
+      "step": 49330
+    },
+    {
+      "epoch": 318.32,
+      "learning_rate": 3.7870967741935485e-05,
+      "loss": 0.3823,
+      "step": 49340
+    },
+    {
+      "epoch": 318.39,
+      "learning_rate": 3.786738351254481e-05,
+      "loss": 0.2675,
+      "step": 49350
+    },
+    {
+      "epoch": 318.45,
+      "learning_rate": 3.786379928315412e-05,
+      "loss": 0.2534,
+      "step": 49360
+    },
+    {
+      "epoch": 318.51,
+      "learning_rate": 3.786021505376344e-05,
+      "loss": 0.2828,
+      "step": 49370
+    },
+    {
+      "epoch": 318.58,
+      "learning_rate": 3.7856630824372764e-05,
+      "loss": 0.2917,
+      "step": 49380
+    },
+    {
+      "epoch": 318.64,
+      "learning_rate": 3.785304659498208e-05,
+      "loss": 0.2886,
+      "step": 49390
+    },
+    {
+      "epoch": 318.71,
+      "learning_rate": 3.78494623655914e-05,
+      "loss": 0.3371,
+      "step": 49400
+    },
+    {
+      "epoch": 318.77,
+      "learning_rate": 3.784587813620072e-05,
+      "loss": 0.2745,
+      "step": 49410
+    },
+    {
+      "epoch": 318.84,
+      "learning_rate": 3.784229390681004e-05,
+      "loss": 0.3233,
+      "step": 49420
+    },
+    {
+      "epoch": 318.9,
+      "learning_rate": 3.7838709677419355e-05,
+      "loss": 0.3006,
+      "step": 49430
+    },
+    {
+      "epoch": 318.96,
+      "learning_rate": 3.7835125448028673e-05,
+      "loss": 0.2607,
+      "step": 49440
+    },
+    {
+      "epoch": 319.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0315192937850952,
+      "eval_runtime": 0.864,
+      "eval_samples_per_second": 232.644,
+      "eval_steps_per_second": 8.102,
+      "step": 49445
+    },
+    {
+      "epoch": 319.03,
+      "learning_rate": 3.783154121863799e-05,
+      "loss": 0.3568,
+      "step": 49450
+    },
+    {
+      "epoch": 319.1,
+      "learning_rate": 3.7827956989247317e-05,
+      "loss": 0.2881,
+      "step": 49460
+    },
+    {
+      "epoch": 319.16,
+      "learning_rate": 3.782437275985663e-05,
+      "loss": 0.339,
+      "step": 49470
+    },
+    {
+      "epoch": 319.23,
+      "learning_rate": 3.782078853046595e-05,
+      "loss": 0.2811,
+      "step": 49480
+    },
+    {
+      "epoch": 319.29,
+      "learning_rate": 3.781720430107527e-05,
+      "loss": 0.2608,
+      "step": 49490
+    },
+    {
+      "epoch": 319.35,
+      "learning_rate": 3.781362007168459e-05,
+      "loss": 0.3261,
+      "step": 49500
+    },
+    {
+      "epoch": 319.42,
+      "learning_rate": 3.781003584229391e-05,
+      "loss": 0.3112,
+      "step": 49510
+    },
+    {
+      "epoch": 319.48,
+      "learning_rate": 3.7806451612903226e-05,
+      "loss": 0.3591,
+      "step": 49520
+    },
+    {
+      "epoch": 319.55,
+      "learning_rate": 3.7802867383512544e-05,
+      "loss": 0.3148,
+      "step": 49530
+    },
+    {
+      "epoch": 319.61,
+      "learning_rate": 3.779928315412187e-05,
+      "loss": 0.3607,
+      "step": 49540
+    },
+    {
+      "epoch": 319.68,
+      "learning_rate": 3.779569892473118e-05,
+      "loss": 0.2794,
+      "step": 49550
+    },
+    {
+      "epoch": 319.74,
+      "learning_rate": 3.7792114695340505e-05,
+      "loss": 0.3009,
+      "step": 49560
+    },
+    {
+      "epoch": 319.8,
+      "learning_rate": 3.7788530465949824e-05,
+      "loss": 0.3021,
+      "step": 49570
+    },
+    {
+      "epoch": 319.87,
+      "learning_rate": 3.778494623655914e-05,
+      "loss": 0.2975,
+      "step": 49580
+    },
+    {
+      "epoch": 319.93,
+      "learning_rate": 3.778136200716846e-05,
+      "loss": 0.2931,
+      "step": 49590
+    },
+    {
+      "epoch": 320.0,
+      "learning_rate": 3.777777777777778e-05,
+      "loss": 0.3629,
+      "step": 49600
+    },
+    {
+      "epoch": 320.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.023464560508728,
+      "eval_runtime": 0.8537,
+      "eval_samples_per_second": 235.457,
+      "eval_steps_per_second": 8.2,
+      "step": 49600
+    },
+    {
+      "epoch": 320.06,
+      "learning_rate": 3.7774193548387096e-05,
+      "loss": 0.2747,
+      "step": 49610
+    },
+    {
+      "epoch": 320.13,
+      "learning_rate": 3.777060931899642e-05,
+      "loss": 0.3227,
+      "step": 49620
+    },
+    {
+      "epoch": 320.19,
+      "learning_rate": 3.776702508960573e-05,
+      "loss": 0.2837,
+      "step": 49630
+    },
+    {
+      "epoch": 320.26,
+      "learning_rate": 3.776344086021506e-05,
+      "loss": 0.2977,
+      "step": 49640
+    },
+    {
+      "epoch": 320.32,
+      "learning_rate": 3.7759856630824376e-05,
+      "loss": 0.3208,
+      "step": 49650
+    },
+    {
+      "epoch": 320.39,
+      "learning_rate": 3.7756272401433694e-05,
+      "loss": 0.3551,
+      "step": 49660
+    },
+    {
+      "epoch": 320.45,
+      "learning_rate": 3.775268817204301e-05,
+      "loss": 0.2966,
+      "step": 49670
+    },
+    {
+      "epoch": 320.51,
+      "learning_rate": 3.774910394265233e-05,
+      "loss": 0.2706,
+      "step": 49680
+    },
+    {
+      "epoch": 320.58,
+      "learning_rate": 3.774551971326165e-05,
+      "loss": 0.29,
+      "step": 49690
+    },
+    {
+      "epoch": 320.64,
+      "learning_rate": 3.7741935483870974e-05,
+      "loss": 0.263,
+      "step": 49700
+    },
+    {
+      "epoch": 320.71,
+      "learning_rate": 3.7738351254480285e-05,
+      "loss": 0.3132,
+      "step": 49710
+    },
+    {
+      "epoch": 320.77,
+      "learning_rate": 3.7734767025089603e-05,
+      "loss": 0.3077,
+      "step": 49720
+    },
+    {
+      "epoch": 320.84,
+      "learning_rate": 3.773118279569893e-05,
+      "loss": 0.3212,
+      "step": 49730
+    },
+    {
+      "epoch": 320.9,
+      "learning_rate": 3.772759856630825e-05,
+      "loss": 0.2846,
+      "step": 49740
+    },
+    {
+      "epoch": 320.96,
+      "learning_rate": 3.7724014336917565e-05,
+      "loss": 0.355,
+      "step": 49750
+    },
+    {
+      "epoch": 321.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.160177230834961,
+      "eval_runtime": 0.8807,
+      "eval_samples_per_second": 228.232,
+      "eval_steps_per_second": 7.948,
+      "step": 49755
+    },
+    {
+      "epoch": 321.03,
+      "learning_rate": 3.772043010752688e-05,
+      "loss": 0.3247,
+      "step": 49760
+    },
+    {
+      "epoch": 321.1,
+      "learning_rate": 3.77168458781362e-05,
+      "loss": 0.2882,
+      "step": 49770
+    },
+    {
+      "epoch": 321.16,
+      "learning_rate": 3.7713261648745526e-05,
+      "loss": 0.3024,
+      "step": 49780
+    },
+    {
+      "epoch": 321.23,
+      "learning_rate": 3.770967741935484e-05,
+      "loss": 0.2835,
+      "step": 49790
+    },
+    {
+      "epoch": 321.29,
+      "learning_rate": 3.7706093189964156e-05,
+      "loss": 0.3265,
+      "step": 49800
+    },
+    {
+      "epoch": 321.35,
+      "learning_rate": 3.770250896057348e-05,
+      "loss": 0.3224,
+      "step": 49810
+    },
+    {
+      "epoch": 321.42,
+      "learning_rate": 3.76989247311828e-05,
+      "loss": 0.3,
+      "step": 49820
+    },
+    {
+      "epoch": 321.48,
+      "learning_rate": 3.769534050179212e-05,
+      "loss": 0.3139,
+      "step": 49830
+    },
+    {
+      "epoch": 321.55,
+      "learning_rate": 3.7691756272401435e-05,
+      "loss": 0.2414,
+      "step": 49840
+    },
+    {
+      "epoch": 321.61,
+      "learning_rate": 3.7688172043010754e-05,
+      "loss": 0.3552,
+      "step": 49850
+    },
+    {
+      "epoch": 321.68,
+      "learning_rate": 3.768458781362008e-05,
+      "loss": 0.2979,
+      "step": 49860
+    },
+    {
+      "epoch": 321.74,
+      "learning_rate": 3.768100358422939e-05,
+      "loss": 0.3732,
+      "step": 49870
+    },
+    {
+      "epoch": 321.8,
+      "learning_rate": 3.767741935483871e-05,
+      "loss": 0.3333,
+      "step": 49880
+    },
+    {
+      "epoch": 321.87,
+      "learning_rate": 3.767383512544803e-05,
+      "loss": 0.334,
+      "step": 49890
+    },
+    {
+      "epoch": 321.93,
+      "learning_rate": 3.767025089605735e-05,
+      "loss": 0.3007,
+      "step": 49900
+    },
+    {
+      "epoch": 322.0,
+      "learning_rate": 3.766666666666667e-05,
+      "loss": 0.295,
+      "step": 49910
+    },
+    {
+      "epoch": 322.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0522838830947876,
+      "eval_runtime": 0.872,
+      "eval_samples_per_second": 230.51,
+      "eval_steps_per_second": 8.028,
+      "step": 49910
+    },
+    {
+      "epoch": 322.06,
+      "learning_rate": 3.766308243727599e-05,
+      "loss": 0.3155,
+      "step": 49920
+    },
+    {
+      "epoch": 322.13,
+      "learning_rate": 3.7659498207885306e-05,
+      "loss": 0.361,
+      "step": 49930
+    },
+    {
+      "epoch": 322.19,
+      "learning_rate": 3.765591397849463e-05,
+      "loss": 0.3138,
+      "step": 49940
+    },
+    {
+      "epoch": 322.26,
+      "learning_rate": 3.765232974910394e-05,
+      "loss": 0.2886,
+      "step": 49950
+    },
+    {
+      "epoch": 322.32,
+      "learning_rate": 3.764874551971326e-05,
+      "loss": 0.3509,
+      "step": 49960
+    },
+    {
+      "epoch": 322.39,
+      "learning_rate": 3.7645161290322586e-05,
+      "loss": 0.3289,
+      "step": 49970
+    },
+    {
+      "epoch": 322.45,
+      "learning_rate": 3.7641577060931904e-05,
+      "loss": 0.3309,
+      "step": 49980
+    },
+    {
+      "epoch": 322.51,
+      "learning_rate": 3.763799283154122e-05,
+      "loss": 0.3036,
+      "step": 49990
+    },
+    {
+      "epoch": 322.58,
+      "learning_rate": 3.763440860215054e-05,
+      "loss": 0.3273,
+      "step": 50000
+    },
+    {
+      "epoch": 322.64,
+      "learning_rate": 3.763082437275986e-05,
+      "loss": 0.2799,
+      "step": 50010
+    },
+    {
+      "epoch": 322.71,
+      "learning_rate": 3.7627240143369183e-05,
+      "loss": 0.2316,
+      "step": 50020
+    },
+    {
+      "epoch": 322.77,
+      "learning_rate": 3.7623655913978495e-05,
+      "loss": 0.337,
+      "step": 50030
+    },
+    {
+      "epoch": 322.84,
+      "learning_rate": 3.762007168458781e-05,
+      "loss": 0.333,
+      "step": 50040
+    },
+    {
+      "epoch": 322.9,
+      "learning_rate": 3.761648745519714e-05,
+      "loss": 0.2957,
+      "step": 50050
+    },
+    {
+      "epoch": 322.96,
+      "learning_rate": 3.761290322580645e-05,
+      "loss": 0.349,
+      "step": 50060
+    },
+    {
+      "epoch": 323.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.911620557308197,
+      "eval_runtime": 0.8408,
+      "eval_samples_per_second": 239.059,
+      "eval_steps_per_second": 8.325,
+      "step": 50065
+    },
+    {
+      "epoch": 323.03,
+      "learning_rate": 3.760931899641577e-05,
+      "loss": 0.2988,
+      "step": 50070
+    },
+    {
+      "epoch": 323.1,
+      "learning_rate": 3.760573476702509e-05,
+      "loss": 0.3401,
+      "step": 50080
+    },
+    {
+      "epoch": 323.16,
+      "learning_rate": 3.760215053763441e-05,
+      "loss": 0.3466,
+      "step": 50090
+    },
+    {
+      "epoch": 323.23,
+      "learning_rate": 3.759856630824373e-05,
+      "loss": 0.2882,
+      "step": 50100
+    },
+    {
+      "epoch": 323.29,
+      "learning_rate": 3.759498207885305e-05,
+      "loss": 0.2896,
+      "step": 50110
+    },
+    {
+      "epoch": 323.35,
+      "learning_rate": 3.7591397849462366e-05,
+      "loss": 0.2703,
+      "step": 50120
+    },
+    {
+      "epoch": 323.42,
+      "learning_rate": 3.758781362007169e-05,
+      "loss": 0.2778,
+      "step": 50130
+    },
+    {
+      "epoch": 323.48,
+      "learning_rate": 3.7584229390681e-05,
+      "loss": 0.3098,
+      "step": 50140
+    },
+    {
+      "epoch": 323.55,
+      "learning_rate": 3.758064516129032e-05,
+      "loss": 0.3393,
+      "step": 50150
+    },
+    {
+      "epoch": 323.61,
+      "learning_rate": 3.7577060931899645e-05,
+      "loss": 0.3049,
+      "step": 50160
+    },
+    {
+      "epoch": 323.68,
+      "learning_rate": 3.757347670250896e-05,
+      "loss": 0.3088,
+      "step": 50170
+    },
+    {
+      "epoch": 323.74,
+      "learning_rate": 3.756989247311828e-05,
+      "loss": 0.3242,
+      "step": 50180
+    },
+    {
+      "epoch": 323.8,
+      "learning_rate": 3.75663082437276e-05,
+      "loss": 0.3096,
+      "step": 50190
+    },
+    {
+      "epoch": 323.87,
+      "learning_rate": 3.756272401433692e-05,
+      "loss": 0.2961,
+      "step": 50200
+    },
+    {
+      "epoch": 323.93,
+      "learning_rate": 3.755913978494624e-05,
+      "loss": 0.2992,
+      "step": 50210
+    },
+    {
+      "epoch": 324.0,
+      "learning_rate": 3.7555555555555554e-05,
+      "loss": 0.2869,
+      "step": 50220
+    },
+    {
+      "epoch": 324.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.060489535331726,
+      "eval_runtime": 0.8754,
+      "eval_samples_per_second": 229.597,
+      "eval_steps_per_second": 7.996,
+      "step": 50220
+    },
+    {
+      "epoch": 324.06,
+      "learning_rate": 3.755197132616487e-05,
+      "loss": 0.3127,
+      "step": 50230
+    },
+    {
+      "epoch": 324.13,
+      "learning_rate": 3.75483870967742e-05,
+      "loss": 0.2815,
+      "step": 50240
+    },
+    {
+      "epoch": 324.19,
+      "learning_rate": 3.7544802867383516e-05,
+      "loss": 0.2951,
+      "step": 50250
+    },
+    {
+      "epoch": 324.26,
+      "learning_rate": 3.7541218637992834e-05,
+      "loss": 0.2921,
+      "step": 50260
+    },
+    {
+      "epoch": 324.32,
+      "learning_rate": 3.753763440860215e-05,
+      "loss": 0.2521,
+      "step": 50270
+    },
+    {
+      "epoch": 324.39,
+      "learning_rate": 3.753405017921147e-05,
+      "loss": 0.2807,
+      "step": 50280
+    },
+    {
+      "epoch": 324.45,
+      "learning_rate": 3.7530465949820795e-05,
+      "loss": 0.3258,
+      "step": 50290
+    },
+    {
+      "epoch": 324.51,
+      "learning_rate": 3.752688172043011e-05,
+      "loss": 0.301,
+      "step": 50300
+    },
+    {
+      "epoch": 324.58,
+      "learning_rate": 3.7523297491039425e-05,
+      "loss": 0.2905,
+      "step": 50310
+    },
+    {
+      "epoch": 324.64,
+      "learning_rate": 3.751971326164875e-05,
+      "loss": 0.3325,
+      "step": 50320
+    },
+    {
+      "epoch": 324.71,
+      "learning_rate": 3.751612903225807e-05,
+      "loss": 0.3063,
+      "step": 50330
+    },
+    {
+      "epoch": 324.77,
+      "learning_rate": 3.7512544802867386e-05,
+      "loss": 0.2879,
+      "step": 50340
+    },
+    {
+      "epoch": 324.84,
+      "learning_rate": 3.7508960573476705e-05,
+      "loss": 0.3306,
+      "step": 50350
+    },
+    {
+      "epoch": 324.9,
+      "learning_rate": 3.750537634408602e-05,
+      "loss": 0.2994,
+      "step": 50360
+    },
+    {
+      "epoch": 324.96,
+      "learning_rate": 3.750179211469534e-05,
+      "loss": 0.2771,
+      "step": 50370
+    },
+    {
+      "epoch": 325.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.06228506565094,
+      "eval_runtime": 0.8645,
+      "eval_samples_per_second": 232.514,
+      "eval_steps_per_second": 8.097,
+      "step": 50375
+    },
+    {
+      "epoch": 325.03,
+      "learning_rate": 3.749820788530466e-05,
+      "loss": 0.3391,
+      "step": 50380
+    },
+    {
+      "epoch": 325.1,
+      "learning_rate": 3.749462365591398e-05,
+      "loss": 0.313,
+      "step": 50390
+    },
+    {
+      "epoch": 325.16,
+      "learning_rate": 3.74910394265233e-05,
+      "loss": 0.3108,
+      "step": 50400
+    },
+    {
+      "epoch": 325.23,
+      "learning_rate": 3.748745519713262e-05,
+      "loss": 0.2833,
+      "step": 50410
+    },
+    {
+      "epoch": 325.29,
+      "learning_rate": 3.748387096774193e-05,
+      "loss": 0.3248,
+      "step": 50420
+    },
+    {
+      "epoch": 325.35,
+      "learning_rate": 3.748028673835126e-05,
+      "loss": 0.3122,
+      "step": 50430
+    },
+    {
+      "epoch": 325.42,
+      "learning_rate": 3.7476702508960575e-05,
+      "loss": 0.3309,
+      "step": 50440
+    },
+    {
+      "epoch": 325.48,
+      "learning_rate": 3.7473118279569893e-05,
+      "loss": 0.3214,
+      "step": 50450
+    },
+    {
+      "epoch": 325.55,
+      "learning_rate": 3.746953405017921e-05,
+      "loss": 0.3219,
+      "step": 50460
+    },
+    {
+      "epoch": 325.61,
+      "learning_rate": 3.746594982078853e-05,
+      "loss": 0.2996,
+      "step": 50470
+    },
+    {
+      "epoch": 325.68,
+      "learning_rate": 3.7462365591397855e-05,
+      "loss": 0.3266,
+      "step": 50480
+    },
+    {
+      "epoch": 325.74,
+      "learning_rate": 3.745878136200717e-05,
+      "loss": 0.2924,
+      "step": 50490
+    },
+    {
+      "epoch": 325.8,
+      "learning_rate": 3.7455197132616484e-05,
+      "loss": 0.3304,
+      "step": 50500
+    },
+    {
+      "epoch": 325.87,
+      "learning_rate": 3.745161290322581e-05,
+      "loss": 0.3023,
+      "step": 50510
+    },
+    {
+      "epoch": 325.93,
+      "learning_rate": 3.744802867383513e-05,
+      "loss": 0.3167,
+      "step": 50520
+    },
+    {
+      "epoch": 326.0,
+      "learning_rate": 3.7444444444444446e-05,
+      "loss": 0.3569,
+      "step": 50530
+    },
+    {
+      "epoch": 326.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0229547023773193,
+      "eval_runtime": 0.9117,
+      "eval_samples_per_second": 220.456,
+      "eval_steps_per_second": 7.678,
+      "step": 50530
+    },
+    {
+      "epoch": 326.06,
+      "learning_rate": 3.7440860215053764e-05,
+      "loss": 0.2791,
+      "step": 50540
+    },
+    {
+      "epoch": 326.13,
+      "learning_rate": 3.743727598566308e-05,
+      "loss": 0.2856,
+      "step": 50550
+    },
+    {
+      "epoch": 326.19,
+      "learning_rate": 3.743369175627241e-05,
+      "loss": 0.3487,
+      "step": 50560
+    },
+    {
+      "epoch": 326.26,
+      "learning_rate": 3.7430107526881725e-05,
+      "loss": 0.3222,
+      "step": 50570
+    },
+    {
+      "epoch": 326.32,
+      "learning_rate": 3.742652329749104e-05,
+      "loss": 0.3556,
+      "step": 50580
+    },
+    {
+      "epoch": 326.39,
+      "learning_rate": 3.742293906810036e-05,
+      "loss": 0.3209,
+      "step": 50590
+    },
+    {
+      "epoch": 326.45,
+      "learning_rate": 3.741935483870968e-05,
+      "loss": 0.3147,
+      "step": 50600
+    },
+    {
+      "epoch": 326.51,
+      "learning_rate": 3.7415770609319e-05,
+      "loss": 0.2518,
+      "step": 50610
+    },
+    {
+      "epoch": 326.58,
+      "learning_rate": 3.7412186379928316e-05,
+      "loss": 0.2781,
+      "step": 50620
+    },
+    {
+      "epoch": 326.64,
+      "learning_rate": 3.7408602150537635e-05,
+      "loss": 0.2618,
+      "step": 50630
+    },
+    {
+      "epoch": 326.71,
+      "learning_rate": 3.740501792114696e-05,
+      "loss": 0.3108,
+      "step": 50640
+    },
+    {
+      "epoch": 326.77,
+      "learning_rate": 3.740143369175628e-05,
+      "loss": 0.3497,
+      "step": 50650
+    },
+    {
+      "epoch": 326.84,
+      "learning_rate": 3.739784946236559e-05,
+      "loss": 0.3198,
+      "step": 50660
+    },
+    {
+      "epoch": 326.9,
+      "learning_rate": 3.7394265232974914e-05,
+      "loss": 0.2727,
+      "step": 50670
+    },
+    {
+      "epoch": 326.96,
+      "learning_rate": 3.739068100358423e-05,
+      "loss": 0.3236,
+      "step": 50680
+    },
+    {
+      "epoch": 327.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0764998197555542,
+      "eval_runtime": 0.8711,
+      "eval_samples_per_second": 230.748,
+      "eval_steps_per_second": 8.036,
+      "step": 50685
+    },
+    {
+      "epoch": 327.03,
+      "learning_rate": 3.738709677419355e-05,
+      "loss": 0.3124,
+      "step": 50690
+    },
+    {
+      "epoch": 327.1,
+      "learning_rate": 3.738351254480287e-05,
+      "loss": 0.3013,
+      "step": 50700
+    },
+    {
+      "epoch": 327.16,
+      "learning_rate": 3.737992831541219e-05,
+      "loss": 0.3045,
+      "step": 50710
+    },
+    {
+      "epoch": 327.23,
+      "learning_rate": 3.7376344086021505e-05,
+      "loss": 0.3367,
+      "step": 50720
+    },
+    {
+      "epoch": 327.29,
+      "learning_rate": 3.7372759856630823e-05,
+      "loss": 0.3284,
+      "step": 50730
+    },
+    {
+      "epoch": 327.35,
+      "learning_rate": 3.736917562724014e-05,
+      "loss": 0.3461,
+      "step": 50740
+    },
+    {
+      "epoch": 327.42,
+      "learning_rate": 3.736559139784947e-05,
+      "loss": 0.301,
+      "step": 50750
+    },
+    {
+      "epoch": 327.48,
+      "learning_rate": 3.7362007168458785e-05,
+      "loss": 0.3596,
+      "step": 50760
+    },
+    {
+      "epoch": 327.55,
+      "learning_rate": 3.7358422939068096e-05,
+      "loss": 0.3697,
+      "step": 50770
+    },
+    {
+      "epoch": 327.61,
+      "learning_rate": 3.735483870967742e-05,
+      "loss": 0.302,
+      "step": 50780
+    },
+    {
+      "epoch": 327.68,
+      "learning_rate": 3.735125448028674e-05,
+      "loss": 0.2907,
+      "step": 50790
+    },
+    {
+      "epoch": 327.74,
+      "learning_rate": 3.734767025089606e-05,
+      "loss": 0.33,
+      "step": 50800
+    },
+    {
+      "epoch": 327.8,
+      "learning_rate": 3.7344086021505376e-05,
+      "loss": 0.2711,
+      "step": 50810
+    },
+    {
+      "epoch": 327.87,
+      "learning_rate": 3.7340501792114694e-05,
+      "loss": 0.2805,
+      "step": 50820
+    },
+    {
+      "epoch": 327.93,
+      "learning_rate": 3.733691756272402e-05,
+      "loss": 0.2946,
+      "step": 50830
+    },
+    {
+      "epoch": 328.0,
+      "learning_rate": 3.733333333333334e-05,
+      "loss": 0.3121,
+      "step": 50840
+    },
+    {
+      "epoch": 328.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.040176272392273,
+      "eval_runtime": 0.8504,
+      "eval_samples_per_second": 236.363,
+      "eval_steps_per_second": 8.232,
+      "step": 50840
+    },
+    {
+      "epoch": 328.06,
+      "learning_rate": 3.732974910394265e-05,
+      "loss": 0.3285,
+      "step": 50850
+    },
+    {
+      "epoch": 328.13,
+      "learning_rate": 3.7326164874551974e-05,
+      "loss": 0.3161,
+      "step": 50860
+    },
+    {
+      "epoch": 328.19,
+      "learning_rate": 3.732258064516129e-05,
+      "loss": 0.2778,
+      "step": 50870
+    },
+    {
+      "epoch": 328.26,
+      "learning_rate": 3.731899641577061e-05,
+      "loss": 0.3523,
+      "step": 50880
+    },
+    {
+      "epoch": 328.32,
+      "learning_rate": 3.731541218637993e-05,
+      "loss": 0.3298,
+      "step": 50890
+    },
+    {
+      "epoch": 328.39,
+      "learning_rate": 3.7311827956989247e-05,
+      "loss": 0.3176,
+      "step": 50900
+    },
+    {
+      "epoch": 328.45,
+      "learning_rate": 3.730824372759857e-05,
+      "loss": 0.2539,
+      "step": 50910
+    },
+    {
+      "epoch": 328.51,
+      "learning_rate": 3.730465949820789e-05,
+      "loss": 0.3358,
+      "step": 50920
+    },
+    {
+      "epoch": 328.58,
+      "learning_rate": 3.73010752688172e-05,
+      "loss": 0.2704,
+      "step": 50930
+    },
+    {
+      "epoch": 328.64,
+      "learning_rate": 3.7297491039426526e-05,
+      "loss": 0.2975,
+      "step": 50940
+    },
+    {
+      "epoch": 328.71,
+      "learning_rate": 3.7293906810035844e-05,
+      "loss": 0.2768,
+      "step": 50950
+    },
+    {
+      "epoch": 328.77,
+      "learning_rate": 3.729032258064516e-05,
+      "loss": 0.2795,
+      "step": 50960
+    },
+    {
+      "epoch": 328.84,
+      "learning_rate": 3.728673835125448e-05,
+      "loss": 0.2824,
+      "step": 50970
+    },
+    {
+      "epoch": 328.9,
+      "learning_rate": 3.72831541218638e-05,
+      "loss": 0.3439,
+      "step": 50980
+    },
+    {
+      "epoch": 328.96,
+      "learning_rate": 3.7279569892473124e-05,
+      "loss": 0.2945,
+      "step": 50990
+    },
+    {
+      "epoch": 329.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0734318494796753,
+      "eval_runtime": 0.8578,
+      "eval_samples_per_second": 234.33,
+      "eval_steps_per_second": 8.161,
+      "step": 50995
+    },
+    {
+      "epoch": 329.03,
+      "learning_rate": 3.727598566308244e-05,
+      "loss": 0.3265,
+      "step": 51000
+    },
+    {
+      "epoch": 329.1,
+      "learning_rate": 3.7272401433691754e-05,
+      "loss": 0.3196,
+      "step": 51010
+    },
+    {
+      "epoch": 329.16,
+      "learning_rate": 3.726881720430108e-05,
+      "loss": 0.3303,
+      "step": 51020
+    },
+    {
+      "epoch": 329.23,
+      "learning_rate": 3.72652329749104e-05,
+      "loss": 0.3257,
+      "step": 51030
+    },
+    {
+      "epoch": 329.29,
+      "learning_rate": 3.7261648745519715e-05,
+      "loss": 0.3248,
+      "step": 51040
+    },
+    {
+      "epoch": 329.35,
+      "learning_rate": 3.725806451612903e-05,
+      "loss": 0.3061,
+      "step": 51050
+    },
+    {
+      "epoch": 329.42,
+      "learning_rate": 3.725448028673835e-05,
+      "loss": 0.313,
+      "step": 51060
+    },
+    {
+      "epoch": 329.48,
+      "learning_rate": 3.725089605734767e-05,
+      "loss": 0.2759,
+      "step": 51070
+    },
+    {
+      "epoch": 329.55,
+      "learning_rate": 3.7247311827956995e-05,
+      "loss": 0.298,
+      "step": 51080
+    },
+    {
+      "epoch": 329.61,
+      "learning_rate": 3.7243727598566306e-05,
+      "loss": 0.3311,
+      "step": 51090
+    },
+    {
+      "epoch": 329.68,
+      "learning_rate": 3.724014336917563e-05,
+      "loss": 0.3393,
+      "step": 51100
+    },
+    {
+      "epoch": 329.74,
+      "learning_rate": 3.723655913978495e-05,
+      "loss": 0.266,
+      "step": 51110
+    },
+    {
+      "epoch": 329.8,
+      "learning_rate": 3.723297491039427e-05,
+      "loss": 0.3224,
+      "step": 51120
+    },
+    {
+      "epoch": 329.87,
+      "learning_rate": 3.7229390681003586e-05,
+      "loss": 0.3052,
+      "step": 51130
+    },
+    {
+      "epoch": 329.93,
+      "learning_rate": 3.7225806451612904e-05,
+      "loss": 0.3543,
+      "step": 51140
+    },
+    {
+      "epoch": 330.0,
+      "learning_rate": 3.722222222222222e-05,
+      "loss": 0.3016,
+      "step": 51150
+    },
+    {
+      "epoch": 330.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.075447678565979,
+      "eval_runtime": 0.8772,
+      "eval_samples_per_second": 229.13,
+      "eval_steps_per_second": 7.98,
+      "step": 51150
+    },
+    {
+      "epoch": 330.06,
+      "learning_rate": 3.721863799283155e-05,
+      "loss": 0.3284,
+      "step": 51160
+    },
+    {
+      "epoch": 330.13,
+      "learning_rate": 3.721505376344086e-05,
+      "loss": 0.3017,
+      "step": 51170
+    },
+    {
+      "epoch": 330.19,
+      "learning_rate": 3.721146953405018e-05,
+      "loss": 0.3423,
+      "step": 51180
+    },
+    {
+      "epoch": 330.26,
+      "learning_rate": 3.72078853046595e-05,
+      "loss": 0.3183,
+      "step": 51190
+    },
+    {
+      "epoch": 330.32,
+      "learning_rate": 3.720430107526882e-05,
+      "loss": 0.296,
+      "step": 51200
+    },
+    {
+      "epoch": 330.39,
+      "learning_rate": 3.720071684587814e-05,
+      "loss": 0.2747,
+      "step": 51210
+    },
+    {
+      "epoch": 330.45,
+      "learning_rate": 3.7197132616487456e-05,
+      "loss": 0.3707,
+      "step": 51220
+    },
+    {
+      "epoch": 330.51,
+      "learning_rate": 3.7193548387096774e-05,
+      "loss": 0.298,
+      "step": 51230
+    },
+    {
+      "epoch": 330.58,
+      "learning_rate": 3.71899641577061e-05,
+      "loss": 0.254,
+      "step": 51240
+    },
+    {
+      "epoch": 330.64,
+      "learning_rate": 3.718637992831541e-05,
+      "loss": 0.2815,
+      "step": 51250
+    },
+    {
+      "epoch": 330.71,
+      "learning_rate": 3.7182795698924736e-05,
+      "loss": 0.29,
+      "step": 51260
+    },
+    {
+      "epoch": 330.77,
+      "learning_rate": 3.7179211469534054e-05,
+      "loss": 0.3095,
+      "step": 51270
+    },
+    {
+      "epoch": 330.84,
+      "learning_rate": 3.717562724014337e-05,
+      "loss": 0.3012,
+      "step": 51280
+    },
+    {
+      "epoch": 330.9,
+      "learning_rate": 3.717204301075269e-05,
+      "loss": 0.3453,
+      "step": 51290
+    },
+    {
+      "epoch": 330.96,
+      "learning_rate": 3.716845878136201e-05,
+      "loss": 0.3143,
+      "step": 51300
+    },
+    {
+      "epoch": 331.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9996315240859985,
+      "eval_runtime": 0.8686,
+      "eval_samples_per_second": 231.397,
+      "eval_steps_per_second": 8.059,
+      "step": 51305
+    },
+    {
+      "epoch": 331.03,
+      "learning_rate": 3.716487455197133e-05,
+      "loss": 0.3135,
+      "step": 51310
+    },
+    {
+      "epoch": 331.1,
+      "learning_rate": 3.716129032258065e-05,
+      "loss": 0.2639,
+      "step": 51320
+    },
+    {
+      "epoch": 331.16,
+      "learning_rate": 3.715770609318996e-05,
+      "loss": 0.2932,
+      "step": 51330
+    },
+    {
+      "epoch": 331.23,
+      "learning_rate": 3.715412186379929e-05,
+      "loss": 0.3543,
+      "step": 51340
+    },
+    {
+      "epoch": 331.29,
+      "learning_rate": 3.7150537634408606e-05,
+      "loss": 0.2943,
+      "step": 51350
+    },
+    {
+      "epoch": 331.35,
+      "learning_rate": 3.714695340501792e-05,
+      "loss": 0.2905,
+      "step": 51360
+    },
+    {
+      "epoch": 331.42,
+      "learning_rate": 3.714336917562724e-05,
+      "loss": 0.295,
+      "step": 51370
+    },
+    {
+      "epoch": 331.48,
+      "learning_rate": 3.713978494623656e-05,
+      "loss": 0.2446,
+      "step": 51380
+    },
+    {
+      "epoch": 331.55,
+      "learning_rate": 3.713620071684588e-05,
+      "loss": 0.3144,
+      "step": 51390
+    },
+    {
+      "epoch": 331.61,
+      "learning_rate": 3.71326164874552e-05,
+      "loss": 0.3086,
+      "step": 51400
+    },
+    {
+      "epoch": 331.68,
+      "learning_rate": 3.7129032258064516e-05,
+      "loss": 0.2677,
+      "step": 51410
+    },
+    {
+      "epoch": 331.74,
+      "learning_rate": 3.7125448028673834e-05,
+      "loss": 0.2641,
+      "step": 51420
+    },
+    {
+      "epoch": 331.8,
+      "learning_rate": 3.712186379928316e-05,
+      "loss": 0.2698,
+      "step": 51430
+    },
+    {
+      "epoch": 331.87,
+      "learning_rate": 3.711827956989247e-05,
+      "loss": 0.304,
+      "step": 51440
+    },
+    {
+      "epoch": 331.93,
+      "learning_rate": 3.7114695340501795e-05,
+      "loss": 0.3435,
+      "step": 51450
+    },
+    {
+      "epoch": 332.0,
+      "learning_rate": 3.7111111111111113e-05,
+      "loss": 0.2722,
+      "step": 51460
+    },
+    {
+      "epoch": 332.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9895583391189575,
+      "eval_runtime": 0.8715,
+      "eval_samples_per_second": 230.646,
+      "eval_steps_per_second": 8.032,
+      "step": 51460
+    },
+    {
+      "epoch": 332.06,
+      "learning_rate": 3.710752688172043e-05,
+      "loss": 0.2813,
+      "step": 51470
+    },
+    {
+      "epoch": 332.13,
+      "learning_rate": 3.710394265232975e-05,
+      "loss": 0.293,
+      "step": 51480
+    },
+    {
+      "epoch": 332.19,
+      "learning_rate": 3.710035842293907e-05,
+      "loss": 0.3081,
+      "step": 51490
+    },
+    {
+      "epoch": 332.26,
+      "learning_rate": 3.7096774193548386e-05,
+      "loss": 0.305,
+      "step": 51500
+    },
+    {
+      "epoch": 332.32,
+      "learning_rate": 3.709318996415771e-05,
+      "loss": 0.3364,
+      "step": 51510
+    },
+    {
+      "epoch": 332.39,
+      "learning_rate": 3.708960573476702e-05,
+      "loss": 0.3429,
+      "step": 51520
+    },
+    {
+      "epoch": 332.45,
+      "learning_rate": 3.708602150537635e-05,
+      "loss": 0.3351,
+      "step": 51530
+    },
+    {
+      "epoch": 332.51,
+      "learning_rate": 3.7082437275985666e-05,
+      "loss": 0.3041,
+      "step": 51540
+    },
+    {
+      "epoch": 332.58,
+      "learning_rate": 3.7078853046594984e-05,
+      "loss": 0.3039,
+      "step": 51550
+    },
+    {
+      "epoch": 332.64,
+      "learning_rate": 3.70752688172043e-05,
+      "loss": 0.2366,
+      "step": 51560
+    },
+    {
+      "epoch": 332.71,
+      "learning_rate": 3.707168458781362e-05,
+      "loss": 0.3359,
+      "step": 51570
+    },
+    {
+      "epoch": 332.77,
+      "learning_rate": 3.706810035842294e-05,
+      "loss": 0.2877,
+      "step": 51580
+    },
+    {
+      "epoch": 332.84,
+      "learning_rate": 3.7064516129032264e-05,
+      "loss": 0.2584,
+      "step": 51590
+    },
+    {
+      "epoch": 332.9,
+      "learning_rate": 3.7060931899641575e-05,
+      "loss": 0.2982,
+      "step": 51600
+    },
+    {
+      "epoch": 332.96,
+      "learning_rate": 3.70573476702509e-05,
+      "loss": 0.3031,
+      "step": 51610
+    },
+    {
+      "epoch": 333.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0519287586212158,
+      "eval_runtime": 0.8494,
+      "eval_samples_per_second": 236.642,
+      "eval_steps_per_second": 8.241,
+      "step": 51615
+    },
+    {
+      "epoch": 333.03,
+      "learning_rate": 3.705376344086022e-05,
+      "loss": 0.2975,
+      "step": 51620
+    },
+    {
+      "epoch": 333.1,
+      "learning_rate": 3.7050179211469536e-05,
+      "loss": 0.3268,
+      "step": 51630
+    },
+    {
+      "epoch": 333.16,
+      "learning_rate": 3.7046594982078855e-05,
+      "loss": 0.2962,
+      "step": 51640
+    },
+    {
+      "epoch": 333.23,
+      "learning_rate": 3.704301075268817e-05,
+      "loss": 0.2922,
+      "step": 51650
+    },
+    {
+      "epoch": 333.29,
+      "learning_rate": 3.703942652329749e-05,
+      "loss": 0.3116,
+      "step": 51660
+    },
+    {
+      "epoch": 333.35,
+      "learning_rate": 3.7035842293906816e-05,
+      "loss": 0.3366,
+      "step": 51670
+    },
+    {
+      "epoch": 333.42,
+      "learning_rate": 3.703225806451613e-05,
+      "loss": 0.2799,
+      "step": 51680
+    },
+    {
+      "epoch": 333.48,
+      "learning_rate": 3.702867383512545e-05,
+      "loss": 0.2573,
+      "step": 51690
+    },
+    {
+      "epoch": 333.55,
+      "learning_rate": 3.702508960573477e-05,
+      "loss": 0.3294,
+      "step": 51700
+    },
+    {
+      "epoch": 333.61,
+      "learning_rate": 3.702150537634409e-05,
+      "loss": 0.2848,
+      "step": 51710
+    },
+    {
+      "epoch": 333.68,
+      "learning_rate": 3.701792114695341e-05,
+      "loss": 0.3192,
+      "step": 51720
+    },
+    {
+      "epoch": 333.74,
+      "learning_rate": 3.7014336917562725e-05,
+      "loss": 0.3207,
+      "step": 51730
+    },
+    {
+      "epoch": 333.8,
+      "learning_rate": 3.7010752688172043e-05,
+      "loss": 0.2949,
+      "step": 51740
+    },
+    {
+      "epoch": 333.87,
+      "learning_rate": 3.700716845878137e-05,
+      "loss": 0.3103,
+      "step": 51750
+    },
+    {
+      "epoch": 333.93,
+      "learning_rate": 3.700358422939068e-05,
+      "loss": 0.2751,
+      "step": 51760
+    },
+    {
+      "epoch": 334.0,
+      "learning_rate": 3.7e-05,
+      "loss": 0.3377,
+      "step": 51770
+    },
+    {
+      "epoch": 334.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.001418113708496,
+      "eval_runtime": 0.8458,
+      "eval_samples_per_second": 237.632,
+      "eval_steps_per_second": 8.276,
+      "step": 51770
+    },
+    {
+      "epoch": 334.06,
+      "learning_rate": 3.699641577060932e-05,
+      "loss": 0.2757,
+      "step": 51780
+    },
+    {
+      "epoch": 334.13,
+      "learning_rate": 3.699283154121864e-05,
+      "loss": 0.2966,
+      "step": 51790
+    },
+    {
+      "epoch": 334.19,
+      "learning_rate": 3.698924731182796e-05,
+      "loss": 0.289,
+      "step": 51800
+    },
+    {
+      "epoch": 334.26,
+      "learning_rate": 3.698566308243728e-05,
+      "loss": 0.2912,
+      "step": 51810
+    },
+    {
+      "epoch": 334.32,
+      "learning_rate": 3.6982078853046596e-05,
+      "loss": 0.2943,
+      "step": 51820
+    },
+    {
+      "epoch": 334.39,
+      "learning_rate": 3.697849462365592e-05,
+      "loss": 0.334,
+      "step": 51830
+    },
+    {
+      "epoch": 334.45,
+      "learning_rate": 3.697491039426523e-05,
+      "loss": 0.3172,
+      "step": 51840
+    },
+    {
+      "epoch": 334.51,
+      "learning_rate": 3.697132616487455e-05,
+      "loss": 0.2813,
+      "step": 51850
+    },
+    {
+      "epoch": 334.58,
+      "learning_rate": 3.6967741935483876e-05,
+      "loss": 0.2943,
+      "step": 51860
+    },
+    {
+      "epoch": 334.64,
+      "learning_rate": 3.6964157706093194e-05,
+      "loss": 0.3116,
+      "step": 51870
+    },
+    {
+      "epoch": 334.71,
+      "learning_rate": 3.696057347670251e-05,
+      "loss": 0.2931,
+      "step": 51880
+    },
+    {
+      "epoch": 334.77,
+      "learning_rate": 3.695698924731183e-05,
+      "loss": 0.289,
+      "step": 51890
+    },
+    {
+      "epoch": 334.84,
+      "learning_rate": 3.695340501792115e-05,
+      "loss": 0.2495,
+      "step": 51900
+    },
+    {
+      "epoch": 334.9,
+      "learning_rate": 3.694982078853047e-05,
+      "loss": 0.3026,
+      "step": 51910
+    },
+    {
+      "epoch": 334.96,
+      "learning_rate": 3.6946236559139785e-05,
+      "loss": 0.3302,
+      "step": 51920
+    },
+    {
+      "epoch": 335.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.013700246810913,
+      "eval_runtime": 0.8461,
+      "eval_samples_per_second": 237.563,
+      "eval_steps_per_second": 8.273,
+      "step": 51925
+    },
+    {
+      "epoch": 335.03,
+      "learning_rate": 3.69426523297491e-05,
+      "loss": 0.3086,
+      "step": 51930
+    },
+    {
+      "epoch": 335.1,
+      "learning_rate": 3.693906810035843e-05,
+      "loss": 0.255,
+      "step": 51940
+    },
+    {
+      "epoch": 335.16,
+      "learning_rate": 3.6935483870967746e-05,
+      "loss": 0.3266,
+      "step": 51950
+    },
+    {
+      "epoch": 335.23,
+      "learning_rate": 3.6931899641577064e-05,
+      "loss": 0.2807,
+      "step": 51960
+    },
+    {
+      "epoch": 335.29,
+      "learning_rate": 3.692831541218638e-05,
+      "loss": 0.3189,
+      "step": 51970
+    },
+    {
+      "epoch": 335.35,
+      "learning_rate": 3.69247311827957e-05,
+      "loss": 0.3096,
+      "step": 51980
+    },
+    {
+      "epoch": 335.42,
+      "learning_rate": 3.692114695340502e-05,
+      "loss": 0.2936,
+      "step": 51990
+    },
+    {
+      "epoch": 335.48,
+      "learning_rate": 3.691756272401434e-05,
+      "loss": 0.3177,
+      "step": 52000
+    },
+    {
+      "epoch": 335.55,
+      "learning_rate": 3.6913978494623655e-05,
+      "loss": 0.2873,
+      "step": 52010
+    },
+    {
+      "epoch": 335.61,
+      "learning_rate": 3.691039426523298e-05,
+      "loss": 0.2864,
+      "step": 52020
+    },
+    {
+      "epoch": 335.68,
+      "learning_rate": 3.690681003584229e-05,
+      "loss": 0.2957,
+      "step": 52030
+    },
+    {
+      "epoch": 335.74,
+      "learning_rate": 3.690322580645162e-05,
+      "loss": 0.3047,
+      "step": 52040
+    },
+    {
+      "epoch": 335.8,
+      "learning_rate": 3.6899641577060935e-05,
+      "loss": 0.2615,
+      "step": 52050
+    },
+    {
+      "epoch": 335.87,
+      "learning_rate": 3.689605734767025e-05,
+      "loss": 0.3349,
+      "step": 52060
+    },
+    {
+      "epoch": 335.93,
+      "learning_rate": 3.689247311827957e-05,
+      "loss": 0.2944,
+      "step": 52070
+    },
+    {
+      "epoch": 336.0,
+      "learning_rate": 3.688888888888889e-05,
+      "loss": 0.3047,
+      "step": 52080
+    },
+    {
+      "epoch": 336.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9742082357406616,
+      "eval_runtime": 0.8852,
+      "eval_samples_per_second": 227.061,
+      "eval_steps_per_second": 7.908,
+      "step": 52080
+    },
+    {
+      "epoch": 336.06,
+      "learning_rate": 3.688530465949821e-05,
+      "loss": 0.2941,
+      "step": 52090
+    },
+    {
+      "epoch": 336.13,
+      "learning_rate": 3.688172043010753e-05,
+      "loss": 0.3101,
+      "step": 52100
+    },
+    {
+      "epoch": 336.19,
+      "learning_rate": 3.6878136200716844e-05,
+      "loss": 0.3068,
+      "step": 52110
+    },
+    {
+      "epoch": 336.26,
+      "learning_rate": 3.687455197132616e-05,
+      "loss": 0.3243,
+      "step": 52120
+    },
+    {
+      "epoch": 336.32,
+      "learning_rate": 3.687096774193549e-05,
+      "loss": 0.3422,
+      "step": 52130
+    },
+    {
+      "epoch": 336.39,
+      "learning_rate": 3.6867383512544806e-05,
+      "loss": 0.3016,
+      "step": 52140
+    },
+    {
+      "epoch": 336.45,
+      "learning_rate": 3.6863799283154124e-05,
+      "loss": 0.3221,
+      "step": 52150
+    },
+    {
+      "epoch": 336.51,
+      "learning_rate": 3.686021505376344e-05,
+      "loss": 0.2961,
+      "step": 52160
+    },
+    {
+      "epoch": 336.58,
+      "learning_rate": 3.685663082437276e-05,
+      "loss": 0.3341,
+      "step": 52170
+    },
+    {
+      "epoch": 336.64,
+      "learning_rate": 3.6853046594982085e-05,
+      "loss": 0.3352,
+      "step": 52180
+    },
+    {
+      "epoch": 336.71,
+      "learning_rate": 3.6849462365591397e-05,
+      "loss": 0.3044,
+      "step": 52190
+    },
+    {
+      "epoch": 336.77,
+      "learning_rate": 3.6845878136200715e-05,
+      "loss": 0.2894,
+      "step": 52200
+    },
+    {
+      "epoch": 336.84,
+      "learning_rate": 3.684229390681004e-05,
+      "loss": 0.3378,
+      "step": 52210
+    },
+    {
+      "epoch": 336.9,
+      "learning_rate": 3.683870967741936e-05,
+      "loss": 0.3268,
+      "step": 52220
+    },
+    {
+      "epoch": 336.96,
+      "learning_rate": 3.6835125448028676e-05,
+      "loss": 0.2656,
+      "step": 52230
+    },
+    {
+      "epoch": 337.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.014509677886963,
+      "eval_runtime": 0.8521,
+      "eval_samples_per_second": 235.899,
+      "eval_steps_per_second": 8.215,
+      "step": 52235
+    },
+    {
+      "epoch": 337.03,
+      "learning_rate": 3.6831541218637994e-05,
+      "loss": 0.3108,
+      "step": 52240
+    },
+    {
+      "epoch": 337.1,
+      "learning_rate": 3.682795698924731e-05,
+      "loss": 0.2646,
+      "step": 52250
+    },
+    {
+      "epoch": 337.16,
+      "learning_rate": 3.682437275985664e-05,
+      "loss": 0.274,
+      "step": 52260
+    },
+    {
+      "epoch": 337.23,
+      "learning_rate": 3.682078853046595e-05,
+      "loss": 0.2572,
+      "step": 52270
+    },
+    {
+      "epoch": 337.29,
+      "learning_rate": 3.681720430107527e-05,
+      "loss": 0.2952,
+      "step": 52280
+    },
+    {
+      "epoch": 337.35,
+      "learning_rate": 3.681362007168459e-05,
+      "loss": 0.3208,
+      "step": 52290
+    },
+    {
+      "epoch": 337.42,
+      "learning_rate": 3.681003584229391e-05,
+      "loss": 0.309,
+      "step": 52300
+    },
+    {
+      "epoch": 337.48,
+      "learning_rate": 3.680645161290323e-05,
+      "loss": 0.2603,
+      "step": 52310
+    },
+    {
+      "epoch": 337.55,
+      "learning_rate": 3.680286738351255e-05,
+      "loss": 0.3073,
+      "step": 52320
+    },
+    {
+      "epoch": 337.61,
+      "learning_rate": 3.6799283154121865e-05,
+      "loss": 0.3096,
+      "step": 52330
+    },
+    {
+      "epoch": 337.68,
+      "learning_rate": 3.679569892473119e-05,
+      "loss": 0.2562,
+      "step": 52340
+    },
+    {
+      "epoch": 337.74,
+      "learning_rate": 3.67921146953405e-05,
+      "loss": 0.2877,
+      "step": 52350
+    },
+    {
+      "epoch": 337.8,
+      "learning_rate": 3.678853046594982e-05,
+      "loss": 0.3024,
+      "step": 52360
+    },
+    {
+      "epoch": 337.87,
+      "learning_rate": 3.6784946236559145e-05,
+      "loss": 0.3561,
+      "step": 52370
+    },
+    {
+      "epoch": 337.93,
+      "learning_rate": 3.678136200716846e-05,
+      "loss": 0.2722,
+      "step": 52380
+    },
+    {
+      "epoch": 338.0,
+      "learning_rate": 3.677777777777778e-05,
+      "loss": 0.2312,
+      "step": 52390
+    },
+    {
+      "epoch": 338.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0128549337387085,
+      "eval_runtime": 0.8529,
+      "eval_samples_per_second": 235.654,
+      "eval_steps_per_second": 8.207,
+      "step": 52390
+    },
+    {
+      "epoch": 338.06,
+      "learning_rate": 3.67741935483871e-05,
+      "loss": 0.2307,
+      "step": 52400
+    },
+    {
+      "epoch": 338.13,
+      "learning_rate": 3.677060931899642e-05,
+      "loss": 0.3236,
+      "step": 52410
+    },
+    {
+      "epoch": 338.19,
+      "learning_rate": 3.6767025089605736e-05,
+      "loss": 0.2594,
+      "step": 52420
+    },
+    {
+      "epoch": 338.26,
+      "learning_rate": 3.6763440860215054e-05,
+      "loss": 0.3105,
+      "step": 52430
+    },
+    {
+      "epoch": 338.32,
+      "learning_rate": 3.675985663082437e-05,
+      "loss": 0.2793,
+      "step": 52440
+    },
+    {
+      "epoch": 338.39,
+      "learning_rate": 3.67562724014337e-05,
+      "loss": 0.2824,
+      "step": 52450
+    },
+    {
+      "epoch": 338.45,
+      "learning_rate": 3.6752688172043015e-05,
+      "loss": 0.2675,
+      "step": 52460
+    },
+    {
+      "epoch": 338.51,
+      "learning_rate": 3.674910394265233e-05,
+      "loss": 0.304,
+      "step": 52470
+    },
+    {
+      "epoch": 338.58,
+      "learning_rate": 3.674551971326165e-05,
+      "loss": 0.309,
+      "step": 52480
+    },
+    {
+      "epoch": 338.64,
+      "learning_rate": 3.674193548387097e-05,
+      "loss": 0.302,
+      "step": 52490
+    },
+    {
+      "epoch": 338.71,
+      "learning_rate": 3.673835125448029e-05,
+      "loss": 0.2959,
+      "step": 52500
+    },
+    {
+      "epoch": 338.77,
+      "learning_rate": 3.6734767025089606e-05,
+      "loss": 0.3473,
+      "step": 52510
+    },
+    {
+      "epoch": 338.84,
+      "learning_rate": 3.6731182795698924e-05,
+      "loss": 0.2993,
+      "step": 52520
+    },
+    {
+      "epoch": 338.9,
+      "learning_rate": 3.672759856630825e-05,
+      "loss": 0.2615,
+      "step": 52530
+    },
+    {
+      "epoch": 338.96,
+      "learning_rate": 3.672401433691757e-05,
+      "loss": 0.3227,
+      "step": 52540
+    },
+    {
+      "epoch": 339.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9806272387504578,
+      "eval_runtime": 0.8724,
+      "eval_samples_per_second": 230.409,
+      "eval_steps_per_second": 8.024,
+      "step": 52545
+    },
+    {
+      "epoch": 339.03,
+      "learning_rate": 3.672043010752688e-05,
+      "loss": 0.3534,
+      "step": 52550
+    },
+    {
+      "epoch": 339.1,
+      "learning_rate": 3.6716845878136204e-05,
+      "loss": 0.2429,
+      "step": 52560
+    },
+    {
+      "epoch": 339.16,
+      "learning_rate": 3.671326164874552e-05,
+      "loss": 0.3279,
+      "step": 52570
+    },
+    {
+      "epoch": 339.23,
+      "learning_rate": 3.670967741935484e-05,
+      "loss": 0.3174,
+      "step": 52580
+    },
+    {
+      "epoch": 339.29,
+      "learning_rate": 3.670609318996416e-05,
+      "loss": 0.2309,
+      "step": 52590
+    },
+    {
+      "epoch": 339.35,
+      "learning_rate": 3.670250896057348e-05,
+      "loss": 0.3012,
+      "step": 52600
+    },
+    {
+      "epoch": 339.42,
+      "learning_rate": 3.66989247311828e-05,
+      "loss": 0.3185,
+      "step": 52610
+    },
+    {
+      "epoch": 339.48,
+      "learning_rate": 3.669534050179211e-05,
+      "loss": 0.3059,
+      "step": 52620
+    },
+    {
+      "epoch": 339.55,
+      "learning_rate": 3.669175627240143e-05,
+      "loss": 0.3004,
+      "step": 52630
+    },
+    {
+      "epoch": 339.61,
+      "learning_rate": 3.6688172043010756e-05,
+      "loss": 0.3196,
+      "step": 52640
+    },
+    {
+      "epoch": 339.68,
+      "learning_rate": 3.6684587813620075e-05,
+      "loss": 0.2528,
+      "step": 52650
+    },
+    {
+      "epoch": 339.74,
+      "learning_rate": 3.668100358422939e-05,
+      "loss": 0.297,
+      "step": 52660
+    },
+    {
+      "epoch": 339.8,
+      "learning_rate": 3.667741935483871e-05,
+      "loss": 0.2779,
+      "step": 52670
+    },
+    {
+      "epoch": 339.87,
+      "learning_rate": 3.667383512544803e-05,
+      "loss": 0.2832,
+      "step": 52680
+    },
+    {
+      "epoch": 339.93,
+      "learning_rate": 3.6670250896057354e-05,
+      "loss": 0.2908,
+      "step": 52690
+    },
+    {
+      "epoch": 340.0,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 0.307,
+      "step": 52700
+    },
+    {
+      "epoch": 340.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9636080265045166,
+      "eval_runtime": 0.8563,
+      "eval_samples_per_second": 234.732,
+      "eval_steps_per_second": 8.175,
+      "step": 52700
+    },
+    {
+      "epoch": 340.06,
+      "learning_rate": 3.6663082437275984e-05,
+      "loss": 0.3232,
+      "step": 52710
+    },
+    {
+      "epoch": 340.13,
+      "learning_rate": 3.665949820788531e-05,
+      "loss": 0.3272,
+      "step": 52720
+    },
+    {
+      "epoch": 340.19,
+      "learning_rate": 3.665591397849463e-05,
+      "loss": 0.2714,
+      "step": 52730
+    },
+    {
+      "epoch": 340.26,
+      "learning_rate": 3.665232974910394e-05,
+      "loss": 0.3024,
+      "step": 52740
+    },
+    {
+      "epoch": 340.32,
+      "learning_rate": 3.6648745519713264e-05,
+      "loss": 0.2654,
+      "step": 52750
+    },
+    {
+      "epoch": 340.39,
+      "learning_rate": 3.664516129032258e-05,
+      "loss": 0.3322,
+      "step": 52760
+    },
+    {
+      "epoch": 340.45,
+      "learning_rate": 3.66415770609319e-05,
+      "loss": 0.2726,
+      "step": 52770
+    },
+    {
+      "epoch": 340.51,
+      "learning_rate": 3.663799283154122e-05,
+      "loss": 0.3003,
+      "step": 52780
+    },
+    {
+      "epoch": 340.58,
+      "learning_rate": 3.6634408602150536e-05,
+      "loss": 0.3035,
+      "step": 52790
+    },
+    {
+      "epoch": 340.64,
+      "learning_rate": 3.663082437275986e-05,
+      "loss": 0.2168,
+      "step": 52800
+    },
+    {
+      "epoch": 340.71,
+      "learning_rate": 3.662724014336918e-05,
+      "loss": 0.285,
+      "step": 52810
+    },
+    {
+      "epoch": 340.77,
+      "learning_rate": 3.662365591397849e-05,
+      "loss": 0.2897,
+      "step": 52820
+    },
+    {
+      "epoch": 340.84,
+      "learning_rate": 3.6620071684587816e-05,
+      "loss": 0.3479,
+      "step": 52830
+    },
+    {
+      "epoch": 340.9,
+      "learning_rate": 3.6616487455197134e-05,
+      "loss": 0.3167,
+      "step": 52840
+    },
+    {
+      "epoch": 340.96,
+      "learning_rate": 3.661290322580645e-05,
+      "loss": 0.3093,
+      "step": 52850
+    },
+    {
+      "epoch": 341.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0412251949310303,
+      "eval_runtime": 0.8648,
+      "eval_samples_per_second": 232.415,
+      "eval_steps_per_second": 8.094,
+      "step": 52855
+    },
+    {
+      "epoch": 341.03,
+      "learning_rate": 3.660931899641577e-05,
+      "loss": 0.2987,
+      "step": 52860
+    },
+    {
+      "epoch": 341.1,
+      "learning_rate": 3.660573476702509e-05,
+      "loss": 0.2829,
+      "step": 52870
+    },
+    {
+      "epoch": 341.16,
+      "learning_rate": 3.6602150537634414e-05,
+      "loss": 0.3265,
+      "step": 52880
+    },
+    {
+      "epoch": 341.23,
+      "learning_rate": 3.659856630824373e-05,
+      "loss": 0.2874,
+      "step": 52890
+    },
+    {
+      "epoch": 341.29,
+      "learning_rate": 3.659498207885304e-05,
+      "loss": 0.3179,
+      "step": 52900
+    },
+    {
+      "epoch": 341.35,
+      "learning_rate": 3.659139784946237e-05,
+      "loss": 0.3258,
+      "step": 52910
+    },
+    {
+      "epoch": 341.42,
+      "learning_rate": 3.6587813620071687e-05,
+      "loss": 0.2693,
+      "step": 52920
+    },
+    {
+      "epoch": 341.48,
+      "learning_rate": 3.6584229390681005e-05,
+      "loss": 0.312,
+      "step": 52930
+    },
+    {
+      "epoch": 341.55,
+      "learning_rate": 3.658064516129032e-05,
+      "loss": 0.2453,
+      "step": 52940
+    },
+    {
+      "epoch": 341.61,
+      "learning_rate": 3.657706093189964e-05,
+      "loss": 0.2796,
+      "step": 52950
+    },
+    {
+      "epoch": 341.68,
+      "learning_rate": 3.6573476702508966e-05,
+      "loss": 0.2577,
+      "step": 52960
+    },
+    {
+      "epoch": 341.74,
+      "learning_rate": 3.6569892473118284e-05,
+      "loss": 0.2776,
+      "step": 52970
+    },
+    {
+      "epoch": 341.8,
+      "learning_rate": 3.6566308243727596e-05,
+      "loss": 0.3454,
+      "step": 52980
+    },
+    {
+      "epoch": 341.87,
+      "learning_rate": 3.656272401433692e-05,
+      "loss": 0.3035,
+      "step": 52990
+    },
+    {
+      "epoch": 341.93,
+      "learning_rate": 3.655913978494624e-05,
+      "loss": 0.2884,
+      "step": 53000
+    },
+    {
+      "epoch": 342.0,
+      "learning_rate": 3.655555555555556e-05,
+      "loss": 0.281,
+      "step": 53010
+    },
+    {
+      "epoch": 342.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0201414823532104,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 230.711,
+      "eval_steps_per_second": 8.035,
+      "step": 53010
+    },
+    {
+      "epoch": 342.06,
+      "learning_rate": 3.6551971326164875e-05,
+      "loss": 0.3335,
+      "step": 53020
+    },
+    {
+      "epoch": 342.13,
+      "learning_rate": 3.6548387096774194e-05,
+      "loss": 0.3159,
+      "step": 53030
+    },
+    {
+      "epoch": 342.19,
+      "learning_rate": 3.654480286738352e-05,
+      "loss": 0.3665,
+      "step": 53040
+    },
+    {
+      "epoch": 342.26,
+      "learning_rate": 3.654121863799284e-05,
+      "loss": 0.3104,
+      "step": 53050
+    },
+    {
+      "epoch": 342.32,
+      "learning_rate": 3.653763440860215e-05,
+      "loss": 0.2824,
+      "step": 53060
+    },
+    {
+      "epoch": 342.39,
+      "learning_rate": 3.653405017921147e-05,
+      "loss": 0.2912,
+      "step": 53070
+    },
+    {
+      "epoch": 342.45,
+      "learning_rate": 3.653046594982079e-05,
+      "loss": 0.3038,
+      "step": 53080
+    },
+    {
+      "epoch": 342.51,
+      "learning_rate": 3.652688172043011e-05,
+      "loss": 0.2767,
+      "step": 53090
+    },
+    {
+      "epoch": 342.58,
+      "learning_rate": 3.652329749103943e-05,
+      "loss": 0.2706,
+      "step": 53100
+    },
+    {
+      "epoch": 342.64,
+      "learning_rate": 3.6519713261648746e-05,
+      "loss": 0.2695,
+      "step": 53110
+    },
+    {
+      "epoch": 342.71,
+      "learning_rate": 3.6516129032258064e-05,
+      "loss": 0.2757,
+      "step": 53120
+    },
+    {
+      "epoch": 342.77,
+      "learning_rate": 3.651254480286739e-05,
+      "loss": 0.2788,
+      "step": 53130
+    },
+    {
+      "epoch": 342.84,
+      "learning_rate": 3.65089605734767e-05,
+      "loss": 0.3775,
+      "step": 53140
+    },
+    {
+      "epoch": 342.9,
+      "learning_rate": 3.6505376344086026e-05,
+      "loss": 0.2549,
+      "step": 53150
+    },
+    {
+      "epoch": 342.96,
+      "learning_rate": 3.6501792114695344e-05,
+      "loss": 0.2113,
+      "step": 53160
+    },
+    {
+      "epoch": 343.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9994174838066101,
+      "eval_runtime": 0.8746,
+      "eval_samples_per_second": 229.815,
+      "eval_steps_per_second": 8.003,
+      "step": 53165
+    },
+    {
+      "epoch": 343.03,
+      "learning_rate": 3.649820788530466e-05,
+      "loss": 0.3639,
+      "step": 53170
+    },
+    {
+      "epoch": 343.1,
+      "learning_rate": 3.649462365591398e-05,
+      "loss": 0.3098,
+      "step": 53180
+    },
+    {
+      "epoch": 343.16,
+      "learning_rate": 3.64910394265233e-05,
+      "loss": 0.2867,
+      "step": 53190
+    },
+    {
+      "epoch": 343.23,
+      "learning_rate": 3.648745519713262e-05,
+      "loss": 0.3162,
+      "step": 53200
+    },
+    {
+      "epoch": 343.29,
+      "learning_rate": 3.648387096774194e-05,
+      "loss": 0.2553,
+      "step": 53210
+    },
+    {
+      "epoch": 343.35,
+      "learning_rate": 3.648028673835125e-05,
+      "loss": 0.2928,
+      "step": 53220
+    },
+    {
+      "epoch": 343.42,
+      "learning_rate": 3.647670250896058e-05,
+      "loss": 0.286,
+      "step": 53230
+    },
+    {
+      "epoch": 343.48,
+      "learning_rate": 3.6473118279569896e-05,
+      "loss": 0.3082,
+      "step": 53240
+    },
+    {
+      "epoch": 343.55,
+      "learning_rate": 3.6469534050179214e-05,
+      "loss": 0.3345,
+      "step": 53250
+    },
+    {
+      "epoch": 343.61,
+      "learning_rate": 3.646594982078853e-05,
+      "loss": 0.2557,
+      "step": 53260
+    },
+    {
+      "epoch": 343.68,
+      "learning_rate": 3.646236559139785e-05,
+      "loss": 0.3229,
+      "step": 53270
+    },
+    {
+      "epoch": 343.74,
+      "learning_rate": 3.645878136200717e-05,
+      "loss": 0.3109,
+      "step": 53280
+    },
+    {
+      "epoch": 343.8,
+      "learning_rate": 3.645519713261649e-05,
+      "loss": 0.2873,
+      "step": 53290
+    },
+    {
+      "epoch": 343.87,
+      "learning_rate": 3.6451612903225805e-05,
+      "loss": 0.2892,
+      "step": 53300
+    },
+    {
+      "epoch": 343.93,
+      "learning_rate": 3.644802867383513e-05,
+      "loss": 0.3259,
+      "step": 53310
+    },
+    {
+      "epoch": 344.0,
+      "learning_rate": 3.644444444444445e-05,
+      "loss": 0.2973,
+      "step": 53320
+    },
+    {
+      "epoch": 344.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.027903437614441,
+      "eval_runtime": 0.8561,
+      "eval_samples_per_second": 234.791,
+      "eval_steps_per_second": 8.177,
+      "step": 53320
+    },
+    {
+      "epoch": 344.06,
+      "learning_rate": 3.644086021505376e-05,
+      "loss": 0.3287,
+      "step": 53330
+    },
+    {
+      "epoch": 344.13,
+      "learning_rate": 3.6437275985663085e-05,
+      "loss": 0.3189,
+      "step": 53340
+    },
+    {
+      "epoch": 344.19,
+      "learning_rate": 3.64336917562724e-05,
+      "loss": 0.2911,
+      "step": 53350
+    },
+    {
+      "epoch": 344.26,
+      "learning_rate": 3.643010752688172e-05,
+      "loss": 0.3086,
+      "step": 53360
+    },
+    {
+      "epoch": 344.32,
+      "learning_rate": 3.642652329749104e-05,
+      "loss": 0.2995,
+      "step": 53370
+    },
+    {
+      "epoch": 344.39,
+      "learning_rate": 3.642293906810036e-05,
+      "loss": 0.2621,
+      "step": 53380
+    },
+    {
+      "epoch": 344.45,
+      "learning_rate": 3.641935483870968e-05,
+      "loss": 0.2652,
+      "step": 53390
+    },
+    {
+      "epoch": 344.51,
+      "learning_rate": 3.6415770609319e-05,
+      "loss": 0.3216,
+      "step": 53400
+    },
+    {
+      "epoch": 344.58,
+      "learning_rate": 3.641218637992831e-05,
+      "loss": 0.2707,
+      "step": 53410
+    },
+    {
+      "epoch": 344.64,
+      "learning_rate": 3.640860215053764e-05,
+      "loss": 0.3311,
+      "step": 53420
+    },
+    {
+      "epoch": 344.71,
+      "learning_rate": 3.6405017921146956e-05,
+      "loss": 0.2828,
+      "step": 53430
+    },
+    {
+      "epoch": 344.77,
+      "learning_rate": 3.6401433691756274e-05,
+      "loss": 0.2556,
+      "step": 53440
+    },
+    {
+      "epoch": 344.84,
+      "learning_rate": 3.639784946236559e-05,
+      "loss": 0.2626,
+      "step": 53450
+    },
+    {
+      "epoch": 344.9,
+      "learning_rate": 3.639426523297491e-05,
+      "loss": 0.3234,
+      "step": 53460
+    },
+    {
+      "epoch": 344.96,
+      "learning_rate": 3.639068100358423e-05,
+      "loss": 0.3137,
+      "step": 53470
+    },
+    {
+      "epoch": 345.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0603790283203125,
+      "eval_runtime": 0.8558,
+      "eval_samples_per_second": 234.878,
+      "eval_steps_per_second": 8.18,
+      "step": 53475
+    },
+    {
+      "epoch": 345.03,
+      "learning_rate": 3.6387096774193553e-05,
+      "loss": 0.3179,
+      "step": 53480
+    },
+    {
+      "epoch": 345.1,
+      "learning_rate": 3.6383512544802865e-05,
+      "loss": 0.3031,
+      "step": 53490
+    },
+    {
+      "epoch": 345.16,
+      "learning_rate": 3.637992831541219e-05,
+      "loss": 0.3675,
+      "step": 53500
+    },
+    {
+      "epoch": 345.23,
+      "learning_rate": 3.637634408602151e-05,
+      "loss": 0.2756,
+      "step": 53510
+    },
+    {
+      "epoch": 345.29,
+      "learning_rate": 3.6372759856630826e-05,
+      "loss": 0.365,
+      "step": 53520
+    },
+    {
+      "epoch": 345.35,
+      "learning_rate": 3.6369175627240144e-05,
+      "loss": 0.292,
+      "step": 53530
+    },
+    {
+      "epoch": 345.42,
+      "learning_rate": 3.636559139784946e-05,
+      "loss": 0.3189,
+      "step": 53540
+    },
+    {
+      "epoch": 345.48,
+      "learning_rate": 3.636200716845878e-05,
+      "loss": 0.3013,
+      "step": 53550
+    },
+    {
+      "epoch": 345.55,
+      "learning_rate": 3.6358422939068106e-05,
+      "loss": 0.2618,
+      "step": 53560
+    },
+    {
+      "epoch": 345.61,
+      "learning_rate": 3.635483870967742e-05,
+      "loss": 0.2401,
+      "step": 53570
+    },
+    {
+      "epoch": 345.68,
+      "learning_rate": 3.635125448028674e-05,
+      "loss": 0.3124,
+      "step": 53580
+    },
+    {
+      "epoch": 345.74,
+      "learning_rate": 3.634767025089606e-05,
+      "loss": 0.2808,
+      "step": 53590
+    },
+    {
+      "epoch": 345.8,
+      "learning_rate": 3.634408602150538e-05,
+      "loss": 0.2987,
+      "step": 53600
+    },
+    {
+      "epoch": 345.87,
+      "learning_rate": 3.63405017921147e-05,
+      "loss": 0.2783,
+      "step": 53610
+    },
+    {
+      "epoch": 345.93,
+      "learning_rate": 3.6336917562724015e-05,
+      "loss": 0.3622,
+      "step": 53620
+    },
+    {
+      "epoch": 346.0,
+      "learning_rate": 3.633333333333333e-05,
+      "loss": 0.2526,
+      "step": 53630
+    },
+    {
+      "epoch": 346.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9793425798416138,
+      "eval_runtime": 0.8874,
+      "eval_samples_per_second": 226.506,
+      "eval_steps_per_second": 7.888,
+      "step": 53630
+    },
+    {
+      "epoch": 346.06,
+      "learning_rate": 3.632974910394266e-05,
+      "loss": 0.3106,
+      "step": 53640
+    },
+    {
+      "epoch": 346.13,
+      "learning_rate": 3.632616487455197e-05,
+      "loss": 0.319,
+      "step": 53650
+    },
+    {
+      "epoch": 346.19,
+      "learning_rate": 3.6322580645161295e-05,
+      "loss": 0.311,
+      "step": 53660
+    },
+    {
+      "epoch": 346.26,
+      "learning_rate": 3.631899641577061e-05,
+      "loss": 0.2708,
+      "step": 53670
+    },
+    {
+      "epoch": 346.32,
+      "learning_rate": 3.631541218637993e-05,
+      "loss": 0.3275,
+      "step": 53680
+    },
+    {
+      "epoch": 346.39,
+      "learning_rate": 3.631182795698925e-05,
+      "loss": 0.2431,
+      "step": 53690
+    },
+    {
+      "epoch": 346.45,
+      "learning_rate": 3.630824372759857e-05,
+      "loss": 0.3103,
+      "step": 53700
+    },
+    {
+      "epoch": 346.51,
+      "learning_rate": 3.6304659498207886e-05,
+      "loss": 0.3129,
+      "step": 53710
+    },
+    {
+      "epoch": 346.58,
+      "learning_rate": 3.630107526881721e-05,
+      "loss": 0.2681,
+      "step": 53720
+    },
+    {
+      "epoch": 346.64,
+      "learning_rate": 3.629749103942652e-05,
+      "loss": 0.3486,
+      "step": 53730
+    },
+    {
+      "epoch": 346.71,
+      "learning_rate": 3.629390681003584e-05,
+      "loss": 0.2865,
+      "step": 53740
+    },
+    {
+      "epoch": 346.77,
+      "learning_rate": 3.6290322580645165e-05,
+      "loss": 0.259,
+      "step": 53750
+    },
+    {
+      "epoch": 346.84,
+      "learning_rate": 3.6286738351254484e-05,
+      "loss": 0.2517,
+      "step": 53760
+    },
+    {
+      "epoch": 346.9,
+      "learning_rate": 3.62831541218638e-05,
+      "loss": 0.2574,
+      "step": 53770
+    },
+    {
+      "epoch": 346.96,
+      "learning_rate": 3.627956989247312e-05,
+      "loss": 0.3151,
+      "step": 53780
+    },
+    {
+      "epoch": 347.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9530024528503418,
+      "eval_runtime": 0.8521,
+      "eval_samples_per_second": 235.891,
+      "eval_steps_per_second": 8.215,
+      "step": 53785
+    },
+    {
+      "epoch": 347.03,
+      "learning_rate": 3.627598566308244e-05,
+      "loss": 0.2672,
+      "step": 53790
+    },
+    {
+      "epoch": 347.1,
+      "learning_rate": 3.627240143369176e-05,
+      "loss": 0.3169,
+      "step": 53800
+    },
+    {
+      "epoch": 347.16,
+      "learning_rate": 3.6268817204301075e-05,
+      "loss": 0.2978,
+      "step": 53810
+    },
+    {
+      "epoch": 347.23,
+      "learning_rate": 3.626523297491039e-05,
+      "loss": 0.3121,
+      "step": 53820
+    },
+    {
+      "epoch": 347.29,
+      "learning_rate": 3.626164874551972e-05,
+      "loss": 0.2821,
+      "step": 53830
+    },
+    {
+      "epoch": 347.35,
+      "learning_rate": 3.6258064516129036e-05,
+      "loss": 0.2278,
+      "step": 53840
+    },
+    {
+      "epoch": 347.42,
+      "learning_rate": 3.6254480286738354e-05,
+      "loss": 0.3235,
+      "step": 53850
+    },
+    {
+      "epoch": 347.48,
+      "learning_rate": 3.625089605734767e-05,
+      "loss": 0.2523,
+      "step": 53860
+    },
+    {
+      "epoch": 347.55,
+      "learning_rate": 3.624731182795699e-05,
+      "loss": 0.2997,
+      "step": 53870
+    },
+    {
+      "epoch": 347.61,
+      "learning_rate": 3.6243727598566316e-05,
+      "loss": 0.2468,
+      "step": 53880
+    },
+    {
+      "epoch": 347.68,
+      "learning_rate": 3.624014336917563e-05,
+      "loss": 0.3311,
+      "step": 53890
+    },
+    {
+      "epoch": 347.74,
+      "learning_rate": 3.6236559139784945e-05,
+      "loss": 0.2879,
+      "step": 53900
+    },
+    {
+      "epoch": 347.8,
+      "learning_rate": 3.623297491039427e-05,
+      "loss": 0.3061,
+      "step": 53910
+    },
+    {
+      "epoch": 347.87,
+      "learning_rate": 3.622939068100358e-05,
+      "loss": 0.3064,
+      "step": 53920
+    },
+    {
+      "epoch": 347.93,
+      "learning_rate": 3.6225806451612907e-05,
+      "loss": 0.3091,
+      "step": 53930
+    },
+    {
+      "epoch": 348.0,
+      "learning_rate": 3.6222222222222225e-05,
+      "loss": 0.3286,
+      "step": 53940
+    },
+    {
+      "epoch": 348.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9526683688163757,
+      "eval_runtime": 0.8625,
+      "eval_samples_per_second": 233.051,
+      "eval_steps_per_second": 8.116,
+      "step": 53940
+    },
+    {
+      "epoch": 348.06,
+      "learning_rate": 3.621863799283154e-05,
+      "loss": 0.3119,
+      "step": 53950
+    },
+    {
+      "epoch": 348.13,
+      "learning_rate": 3.621505376344086e-05,
+      "loss": 0.2977,
+      "step": 53960
+    },
+    {
+      "epoch": 348.19,
+      "learning_rate": 3.621146953405018e-05,
+      "loss": 0.3318,
+      "step": 53970
+    },
+    {
+      "epoch": 348.26,
+      "learning_rate": 3.62078853046595e-05,
+      "loss": 0.3032,
+      "step": 53980
+    },
+    {
+      "epoch": 348.32,
+      "learning_rate": 3.620430107526882e-05,
+      "loss": 0.3438,
+      "step": 53990
+    },
+    {
+      "epoch": 348.39,
+      "learning_rate": 3.6200716845878134e-05,
+      "loss": 0.2586,
+      "step": 54000
+    },
+    {
+      "epoch": 348.45,
+      "learning_rate": 3.619713261648746e-05,
+      "loss": 0.29,
+      "step": 54010
+    },
+    {
+      "epoch": 348.51,
+      "learning_rate": 3.619354838709678e-05,
+      "loss": 0.351,
+      "step": 54020
+    },
+    {
+      "epoch": 348.58,
+      "learning_rate": 3.6189964157706095e-05,
+      "loss": 0.2774,
+      "step": 54030
+    },
+    {
+      "epoch": 348.64,
+      "learning_rate": 3.6186379928315414e-05,
+      "loss": 0.2936,
+      "step": 54040
+    },
+    {
+      "epoch": 348.71,
+      "learning_rate": 3.618279569892473e-05,
+      "loss": 0.2893,
+      "step": 54050
+    },
+    {
+      "epoch": 348.77,
+      "learning_rate": 3.617921146953405e-05,
+      "loss": 0.3291,
+      "step": 54060
+    },
+    {
+      "epoch": 348.84,
+      "learning_rate": 3.6175627240143375e-05,
+      "loss": 0.3283,
+      "step": 54070
+    },
+    {
+      "epoch": 348.9,
+      "learning_rate": 3.6172043010752686e-05,
+      "loss": 0.3012,
+      "step": 54080
+    },
+    {
+      "epoch": 348.96,
+      "learning_rate": 3.6168458781362005e-05,
+      "loss": 0.2659,
+      "step": 54090
+    },
+    {
+      "epoch": 349.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9802244305610657,
+      "eval_runtime": 0.8572,
+      "eval_samples_per_second": 234.472,
+      "eval_steps_per_second": 8.166,
+      "step": 54095
+    },
+    {
+      "epoch": 349.03,
+      "learning_rate": 3.616487455197133e-05,
+      "loss": 0.2779,
+      "step": 54100
+    },
+    {
+      "epoch": 349.1,
+      "learning_rate": 3.616129032258065e-05,
+      "loss": 0.2577,
+      "step": 54110
+    },
+    {
+      "epoch": 349.16,
+      "learning_rate": 3.6157706093189966e-05,
+      "loss": 0.2577,
+      "step": 54120
+    },
+    {
+      "epoch": 349.23,
+      "learning_rate": 3.6154121863799284e-05,
+      "loss": 0.3055,
+      "step": 54130
+    },
+    {
+      "epoch": 349.29,
+      "learning_rate": 3.61505376344086e-05,
+      "loss": 0.2788,
+      "step": 54140
+    },
+    {
+      "epoch": 349.35,
+      "learning_rate": 3.614695340501793e-05,
+      "loss": 0.2622,
+      "step": 54150
+    },
+    {
+      "epoch": 349.42,
+      "learning_rate": 3.614336917562724e-05,
+      "loss": 0.3367,
+      "step": 54160
+    },
+    {
+      "epoch": 349.48,
+      "learning_rate": 3.613978494623656e-05,
+      "loss": 0.313,
+      "step": 54170
+    },
+    {
+      "epoch": 349.55,
+      "learning_rate": 3.613620071684588e-05,
+      "loss": 0.2467,
+      "step": 54180
+    },
+    {
+      "epoch": 349.61,
+      "learning_rate": 3.61326164874552e-05,
+      "loss": 0.2777,
+      "step": 54190
+    },
+    {
+      "epoch": 349.68,
+      "learning_rate": 3.612903225806452e-05,
+      "loss": 0.3179,
+      "step": 54200
+    },
+    {
+      "epoch": 349.74,
+      "learning_rate": 3.612544802867384e-05,
+      "loss": 0.2696,
+      "step": 54210
+    },
+    {
+      "epoch": 349.8,
+      "learning_rate": 3.6121863799283155e-05,
+      "loss": 0.2704,
+      "step": 54220
+    },
+    {
+      "epoch": 349.87,
+      "learning_rate": 3.611827956989248e-05,
+      "loss": 0.2829,
+      "step": 54230
+    },
+    {
+      "epoch": 349.93,
+      "learning_rate": 3.611469534050179e-05,
+      "loss": 0.2801,
+      "step": 54240
+    },
+    {
+      "epoch": 350.0,
+      "learning_rate": 3.611111111111111e-05,
+      "loss": 0.2888,
+      "step": 54250
+    },
+    {
+      "epoch": 350.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9335050582885742,
+      "eval_runtime": 0.8598,
+      "eval_samples_per_second": 233.786,
+      "eval_steps_per_second": 8.142,
+      "step": 54250
+    },
+    {
+      "epoch": 350.06,
+      "learning_rate": 3.6107526881720434e-05,
+      "loss": 0.299,
+      "step": 54260
+    },
+    {
+      "epoch": 350.13,
+      "learning_rate": 3.610394265232975e-05,
+      "loss": 0.3397,
+      "step": 54270
+    },
+    {
+      "epoch": 350.19,
+      "learning_rate": 3.610035842293907e-05,
+      "loss": 0.2226,
+      "step": 54280
+    },
+    {
+      "epoch": 350.26,
+      "learning_rate": 3.609677419354839e-05,
+      "loss": 0.2834,
+      "step": 54290
+    },
+    {
+      "epoch": 350.32,
+      "learning_rate": 3.609318996415771e-05,
+      "loss": 0.2874,
+      "step": 54300
+    },
+    {
+      "epoch": 350.39,
+      "learning_rate": 3.608960573476703e-05,
+      "loss": 0.2922,
+      "step": 54310
+    },
+    {
+      "epoch": 350.45,
+      "learning_rate": 3.6086021505376344e-05,
+      "loss": 0.2777,
+      "step": 54320
+    },
+    {
+      "epoch": 350.51,
+      "learning_rate": 3.608243727598566e-05,
+      "loss": 0.3402,
+      "step": 54330
+    },
+    {
+      "epoch": 350.58,
+      "learning_rate": 3.607885304659499e-05,
+      "loss": 0.3175,
+      "step": 54340
+    },
+    {
+      "epoch": 350.64,
+      "learning_rate": 3.6075268817204305e-05,
+      "loss": 0.3213,
+      "step": 54350
+    },
+    {
+      "epoch": 350.71,
+      "learning_rate": 3.607168458781362e-05,
+      "loss": 0.2725,
+      "step": 54360
+    },
+    {
+      "epoch": 350.77,
+      "learning_rate": 3.606810035842294e-05,
+      "loss": 0.2806,
+      "step": 54370
+    },
+    {
+      "epoch": 350.84,
+      "learning_rate": 3.606451612903226e-05,
+      "loss": 0.2595,
+      "step": 54380
+    },
+    {
+      "epoch": 350.9,
+      "learning_rate": 3.606093189964158e-05,
+      "loss": 0.2741,
+      "step": 54390
+    },
+    {
+      "epoch": 350.96,
+      "learning_rate": 3.6057347670250896e-05,
+      "loss": 0.2793,
+      "step": 54400
+    },
+    {
+      "epoch": 351.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9525732398033142,
+      "eval_runtime": 0.8552,
+      "eval_samples_per_second": 235.04,
+      "eval_steps_per_second": 8.185,
+      "step": 54405
+    },
+    {
+      "epoch": 351.03,
+      "learning_rate": 3.6053763440860214e-05,
+      "loss": 0.2847,
+      "step": 54410
+    },
+    {
+      "epoch": 351.1,
+      "learning_rate": 3.605017921146954e-05,
+      "loss": 0.3107,
+      "step": 54420
+    },
+    {
+      "epoch": 351.16,
+      "learning_rate": 3.604659498207886e-05,
+      "loss": 0.2514,
+      "step": 54430
+    },
+    {
+      "epoch": 351.23,
+      "learning_rate": 3.604301075268817e-05,
+      "loss": 0.3085,
+      "step": 54440
+    },
+    {
+      "epoch": 351.29,
+      "learning_rate": 3.6039426523297494e-05,
+      "loss": 0.2157,
+      "step": 54450
+    },
+    {
+      "epoch": 351.35,
+      "learning_rate": 3.603584229390681e-05,
+      "loss": 0.2953,
+      "step": 54460
+    },
+    {
+      "epoch": 351.42,
+      "learning_rate": 3.603225806451613e-05,
+      "loss": 0.2824,
+      "step": 54470
+    },
+    {
+      "epoch": 351.48,
+      "learning_rate": 3.602867383512545e-05,
+      "loss": 0.2792,
+      "step": 54480
+    },
+    {
+      "epoch": 351.55,
+      "learning_rate": 3.602508960573477e-05,
+      "loss": 0.2573,
+      "step": 54490
+    },
+    {
+      "epoch": 351.61,
+      "learning_rate": 3.602150537634409e-05,
+      "loss": 0.2926,
+      "step": 54500
+    },
+    {
+      "epoch": 351.68,
+      "learning_rate": 3.601792114695341e-05,
+      "loss": 0.2368,
+      "step": 54510
+    },
+    {
+      "epoch": 351.74,
+      "learning_rate": 3.601433691756272e-05,
+      "loss": 0.3141,
+      "step": 54520
+    },
+    {
+      "epoch": 351.8,
+      "learning_rate": 3.6010752688172046e-05,
+      "loss": 0.3257,
+      "step": 54530
+    },
+    {
+      "epoch": 351.87,
+      "learning_rate": 3.6007168458781365e-05,
+      "loss": 0.28,
+      "step": 54540
+    },
+    {
+      "epoch": 351.93,
+      "learning_rate": 3.600358422939068e-05,
+      "loss": 0.3015,
+      "step": 54550
+    },
+    {
+      "epoch": 352.0,
+      "learning_rate": 3.6e-05,
+      "loss": 0.3094,
+      "step": 54560
+    },
+    {
+      "epoch": 352.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9974091649055481,
+      "eval_runtime": 0.8786,
+      "eval_samples_per_second": 228.785,
+      "eval_steps_per_second": 7.968,
+      "step": 54560
+    },
+    {
+      "epoch": 352.06,
+      "learning_rate": 3.599641577060932e-05,
+      "loss": 0.3408,
+      "step": 54570
+    },
+    {
+      "epoch": 352.13,
+      "learning_rate": 3.5992831541218644e-05,
+      "loss": 0.2484,
+      "step": 54580
+    },
+    {
+      "epoch": 352.19,
+      "learning_rate": 3.5989247311827956e-05,
+      "loss": 0.305,
+      "step": 54590
+    },
+    {
+      "epoch": 352.26,
+      "learning_rate": 3.5985663082437274e-05,
+      "loss": 0.3229,
+      "step": 54600
+    },
+    {
+      "epoch": 352.32,
+      "learning_rate": 3.59820788530466e-05,
+      "loss": 0.2951,
+      "step": 54610
+    },
+    {
+      "epoch": 352.39,
+      "learning_rate": 3.597849462365592e-05,
+      "loss": 0.27,
+      "step": 54620
+    },
+    {
+      "epoch": 352.45,
+      "learning_rate": 3.5974910394265235e-05,
+      "loss": 0.2983,
+      "step": 54630
+    },
+    {
+      "epoch": 352.51,
+      "learning_rate": 3.597132616487455e-05,
+      "loss": 0.2763,
+      "step": 54640
+    },
+    {
+      "epoch": 352.58,
+      "learning_rate": 3.596774193548387e-05,
+      "loss": 0.3705,
+      "step": 54650
+    },
+    {
+      "epoch": 352.64,
+      "learning_rate": 3.5964157706093197e-05,
+      "loss": 0.3044,
+      "step": 54660
+    },
+    {
+      "epoch": 352.71,
+      "learning_rate": 3.596057347670251e-05,
+      "loss": 0.283,
+      "step": 54670
+    },
+    {
+      "epoch": 352.77,
+      "learning_rate": 3.5956989247311826e-05,
+      "loss": 0.299,
+      "step": 54680
+    },
+    {
+      "epoch": 352.84,
+      "learning_rate": 3.595340501792115e-05,
+      "loss": 0.3347,
+      "step": 54690
+    },
+    {
+      "epoch": 352.9,
+      "learning_rate": 3.594982078853047e-05,
+      "loss": 0.2673,
+      "step": 54700
+    },
+    {
+      "epoch": 352.96,
+      "learning_rate": 3.594623655913979e-05,
+      "loss": 0.2856,
+      "step": 54710
+    },
+    {
+      "epoch": 353.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9667321443557739,
+      "eval_runtime": 0.8722,
+      "eval_samples_per_second": 230.456,
+      "eval_steps_per_second": 8.026,
+      "step": 54715
+    },
+    {
+      "epoch": 353.03,
+      "learning_rate": 3.5942652329749106e-05,
+      "loss": 0.3682,
+      "step": 54720
+    },
+    {
+      "epoch": 353.1,
+      "learning_rate": 3.5939068100358424e-05,
+      "loss": 0.2836,
+      "step": 54730
+    },
+    {
+      "epoch": 353.16,
+      "learning_rate": 3.593548387096774e-05,
+      "loss": 0.2841,
+      "step": 54740
+    },
+    {
+      "epoch": 353.23,
+      "learning_rate": 3.593189964157706e-05,
+      "loss": 0.3287,
+      "step": 54750
+    },
+    {
+      "epoch": 353.29,
+      "learning_rate": 3.592831541218638e-05,
+      "loss": 0.2574,
+      "step": 54760
+    },
+    {
+      "epoch": 353.35,
+      "learning_rate": 3.5924731182795704e-05,
+      "loss": 0.3094,
+      "step": 54770
+    },
+    {
+      "epoch": 353.42,
+      "learning_rate": 3.592114695340502e-05,
+      "loss": 0.3321,
+      "step": 54780
+    },
+    {
+      "epoch": 353.48,
+      "learning_rate": 3.591756272401433e-05,
+      "loss": 0.2521,
+      "step": 54790
+    },
+    {
+      "epoch": 353.55,
+      "learning_rate": 3.591397849462366e-05,
+      "loss": 0.3501,
+      "step": 54800
+    },
+    {
+      "epoch": 353.61,
+      "learning_rate": 3.5910394265232976e-05,
+      "loss": 0.2803,
+      "step": 54810
+    },
+    {
+      "epoch": 353.68,
+      "learning_rate": 3.5906810035842295e-05,
+      "loss": 0.268,
+      "step": 54820
+    },
+    {
+      "epoch": 353.74,
+      "learning_rate": 3.590322580645161e-05,
+      "loss": 0.3105,
+      "step": 54830
+    },
+    {
+      "epoch": 353.8,
+      "learning_rate": 3.589964157706093e-05,
+      "loss": 0.2952,
+      "step": 54840
+    },
+    {
+      "epoch": 353.87,
+      "learning_rate": 3.5896057347670256e-05,
+      "loss": 0.2759,
+      "step": 54850
+    },
+    {
+      "epoch": 353.93,
+      "learning_rate": 3.5892473118279574e-05,
+      "loss": 0.2733,
+      "step": 54860
+    },
+    {
+      "epoch": 354.0,
+      "learning_rate": 3.5888888888888886e-05,
+      "loss": 0.2915,
+      "step": 54870
+    },
+    {
+      "epoch": 354.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0136024951934814,
+      "eval_runtime": 0.847,
+      "eval_samples_per_second": 237.304,
+      "eval_steps_per_second": 8.264,
+      "step": 54870
+    },
+    {
+      "epoch": 354.06,
+      "learning_rate": 3.588530465949821e-05,
+      "loss": 0.2403,
+      "step": 54880
+    },
+    {
+      "epoch": 354.13,
+      "learning_rate": 3.588172043010753e-05,
+      "loss": 0.3258,
+      "step": 54890
+    },
+    {
+      "epoch": 354.19,
+      "learning_rate": 3.587813620071685e-05,
+      "loss": 0.2695,
+      "step": 54900
+    },
+    {
+      "epoch": 354.26,
+      "learning_rate": 3.5874551971326165e-05,
+      "loss": 0.2847,
+      "step": 54910
+    },
+    {
+      "epoch": 354.32,
+      "learning_rate": 3.5870967741935483e-05,
+      "loss": 0.284,
+      "step": 54920
+    },
+    {
+      "epoch": 354.39,
+      "learning_rate": 3.586738351254481e-05,
+      "loss": 0.2494,
+      "step": 54930
+    },
+    {
+      "epoch": 354.45,
+      "learning_rate": 3.5863799283154127e-05,
+      "loss": 0.3666,
+      "step": 54940
+    },
+    {
+      "epoch": 354.51,
+      "learning_rate": 3.586021505376344e-05,
+      "loss": 0.3116,
+      "step": 54950
+    },
+    {
+      "epoch": 354.58,
+      "learning_rate": 3.585663082437276e-05,
+      "loss": 0.2832,
+      "step": 54960
+    },
+    {
+      "epoch": 354.64,
+      "learning_rate": 3.585304659498208e-05,
+      "loss": 0.332,
+      "step": 54970
+    },
+    {
+      "epoch": 354.71,
+      "learning_rate": 3.58494623655914e-05,
+      "loss": 0.3228,
+      "step": 54980
+    },
+    {
+      "epoch": 354.77,
+      "learning_rate": 3.584587813620072e-05,
+      "loss": 0.2565,
+      "step": 54990
+    },
+    {
+      "epoch": 354.84,
+      "learning_rate": 3.5842293906810036e-05,
+      "loss": 0.2836,
+      "step": 55000
+    },
+    {
+      "epoch": 354.9,
+      "learning_rate": 3.583870967741936e-05,
+      "loss": 0.2712,
+      "step": 55010
+    },
+    {
+      "epoch": 354.96,
+      "learning_rate": 3.583512544802868e-05,
+      "loss": 0.2503,
+      "step": 55020
+    },
+    {
+      "epoch": 355.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9431814551353455,
+      "eval_runtime": 0.8626,
+      "eval_samples_per_second": 233.019,
+      "eval_steps_per_second": 8.115,
+      "step": 55025
+    },
+    {
+      "epoch": 355.03,
+      "learning_rate": 3.583154121863799e-05,
+      "loss": 0.3741,
+      "step": 55030
+    },
+    {
+      "epoch": 355.1,
+      "learning_rate": 3.5827956989247315e-05,
+      "loss": 0.2886,
+      "step": 55040
+    },
+    {
+      "epoch": 355.16,
+      "learning_rate": 3.5824372759856634e-05,
+      "loss": 0.2996,
+      "step": 55050
+    },
+    {
+      "epoch": 355.23,
+      "learning_rate": 3.582078853046595e-05,
+      "loss": 0.337,
+      "step": 55060
+    },
+    {
+      "epoch": 355.29,
+      "learning_rate": 3.581720430107527e-05,
+      "loss": 0.2939,
+      "step": 55070
+    },
+    {
+      "epoch": 355.35,
+      "learning_rate": 3.581362007168459e-05,
+      "loss": 0.3193,
+      "step": 55080
+    },
+    {
+      "epoch": 355.42,
+      "learning_rate": 3.5810035842293906e-05,
+      "loss": 0.3463,
+      "step": 55090
+    },
+    {
+      "epoch": 355.48,
+      "learning_rate": 3.580645161290323e-05,
+      "loss": 0.2784,
+      "step": 55100
+    },
+    {
+      "epoch": 355.55,
+      "learning_rate": 3.580286738351254e-05,
+      "loss": 0.3003,
+      "step": 55110
+    },
+    {
+      "epoch": 355.61,
+      "learning_rate": 3.579928315412187e-05,
+      "loss": 0.2872,
+      "step": 55120
+    },
+    {
+      "epoch": 355.68,
+      "learning_rate": 3.5795698924731186e-05,
+      "loss": 0.2854,
+      "step": 55130
+    },
+    {
+      "epoch": 355.74,
+      "learning_rate": 3.5792114695340504e-05,
+      "loss": 0.3118,
+      "step": 55140
+    },
+    {
+      "epoch": 355.8,
+      "learning_rate": 3.578853046594982e-05,
+      "loss": 0.325,
+      "step": 55150
+    },
+    {
+      "epoch": 355.87,
+      "learning_rate": 3.578494623655914e-05,
+      "loss": 0.2835,
+      "step": 55160
+    },
+    {
+      "epoch": 355.93,
+      "learning_rate": 3.578136200716846e-05,
+      "loss": 0.3006,
+      "step": 55170
+    },
+    {
+      "epoch": 356.0,
+      "learning_rate": 3.577777777777778e-05,
+      "loss": 0.3435,
+      "step": 55180
+    },
+    {
+      "epoch": 356.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.992736279964447,
+      "eval_runtime": 0.8667,
+      "eval_samples_per_second": 231.917,
+      "eval_steps_per_second": 8.077,
+      "step": 55180
+    },
+    {
+      "epoch": 356.06,
+      "learning_rate": 3.5774193548387095e-05,
+      "loss": 0.3221,
+      "step": 55190
+    },
+    {
+      "epoch": 356.13,
+      "learning_rate": 3.577060931899642e-05,
+      "loss": 0.2743,
+      "step": 55200
+    },
+    {
+      "epoch": 356.19,
+      "learning_rate": 3.576702508960574e-05,
+      "loss": 0.2939,
+      "step": 55210
+    },
+    {
+      "epoch": 356.26,
+      "learning_rate": 3.576344086021505e-05,
+      "loss": 0.2692,
+      "step": 55220
+    },
+    {
+      "epoch": 356.32,
+      "learning_rate": 3.5759856630824375e-05,
+      "loss": 0.2883,
+      "step": 55230
+    },
+    {
+      "epoch": 356.39,
+      "learning_rate": 3.575627240143369e-05,
+      "loss": 0.3473,
+      "step": 55240
+    },
+    {
+      "epoch": 356.45,
+      "learning_rate": 3.575268817204301e-05,
+      "loss": 0.2648,
+      "step": 55250
+    },
+    {
+      "epoch": 356.51,
+      "learning_rate": 3.574910394265233e-05,
+      "loss": 0.2823,
+      "step": 55260
+    },
+    {
+      "epoch": 356.58,
+      "learning_rate": 3.574551971326165e-05,
+      "loss": 0.3079,
+      "step": 55270
+    },
+    {
+      "epoch": 356.64,
+      "learning_rate": 3.574193548387097e-05,
+      "loss": 0.2965,
+      "step": 55280
+    },
+    {
+      "epoch": 356.71,
+      "learning_rate": 3.573835125448029e-05,
+      "loss": 0.2969,
+      "step": 55290
+    },
+    {
+      "epoch": 356.77,
+      "learning_rate": 3.57347670250896e-05,
+      "loss": 0.2036,
+      "step": 55300
+    },
+    {
+      "epoch": 356.84,
+      "learning_rate": 3.573118279569893e-05,
+      "loss": 0.3078,
+      "step": 55310
+    },
+    {
+      "epoch": 356.9,
+      "learning_rate": 3.5727598566308245e-05,
+      "loss": 0.3307,
+      "step": 55320
+    },
+    {
+      "epoch": 356.96,
+      "learning_rate": 3.5724014336917564e-05,
+      "loss": 0.2665,
+      "step": 55330
+    },
+    {
+      "epoch": 357.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9897564053535461,
+      "eval_runtime": 0.8825,
+      "eval_samples_per_second": 227.758,
+      "eval_steps_per_second": 7.932,
+      "step": 55335
+    },
+    {
+      "epoch": 357.03,
+      "learning_rate": 3.572043010752688e-05,
+      "loss": 0.3439,
+      "step": 55340
+    },
+    {
+      "epoch": 357.1,
+      "learning_rate": 3.57168458781362e-05,
+      "loss": 0.2696,
+      "step": 55350
+    },
+    {
+      "epoch": 357.16,
+      "learning_rate": 3.5713261648745525e-05,
+      "loss": 0.2999,
+      "step": 55360
+    },
+    {
+      "epoch": 357.23,
+      "learning_rate": 3.570967741935484e-05,
+      "loss": 0.3462,
+      "step": 55370
+    },
+    {
+      "epoch": 357.29,
+      "learning_rate": 3.5706093189964155e-05,
+      "loss": 0.3185,
+      "step": 55380
+    },
+    {
+      "epoch": 357.35,
+      "learning_rate": 3.570250896057348e-05,
+      "loss": 0.3395,
+      "step": 55390
+    },
+    {
+      "epoch": 357.42,
+      "learning_rate": 3.56989247311828e-05,
+      "loss": 0.3342,
+      "step": 55400
+    },
+    {
+      "epoch": 357.48,
+      "learning_rate": 3.5695340501792116e-05,
+      "loss": 0.2873,
+      "step": 55410
+    },
+    {
+      "epoch": 357.55,
+      "learning_rate": 3.5691756272401434e-05,
+      "loss": 0.3497,
+      "step": 55420
+    },
+    {
+      "epoch": 357.61,
+      "learning_rate": 3.568817204301075e-05,
+      "loss": 0.2615,
+      "step": 55430
+    },
+    {
+      "epoch": 357.68,
+      "learning_rate": 3.568458781362007e-05,
+      "loss": 0.3053,
+      "step": 55440
+    },
+    {
+      "epoch": 357.74,
+      "learning_rate": 3.5681003584229396e-05,
+      "loss": 0.2898,
+      "step": 55450
+    },
+    {
+      "epoch": 357.8,
+      "learning_rate": 3.567741935483871e-05,
+      "loss": 0.2568,
+      "step": 55460
+    },
+    {
+      "epoch": 357.87,
+      "learning_rate": 3.567383512544803e-05,
+      "loss": 0.3317,
+      "step": 55470
+    },
+    {
+      "epoch": 357.93,
+      "learning_rate": 3.567025089605735e-05,
+      "loss": 0.3333,
+      "step": 55480
+    },
+    {
+      "epoch": 358.0,
+      "learning_rate": 3.566666666666667e-05,
+      "loss": 0.2976,
+      "step": 55490
+    },
+    {
+      "epoch": 358.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9781351685523987,
+      "eval_runtime": 0.8771,
+      "eval_samples_per_second": 229.171,
+      "eval_steps_per_second": 7.981,
+      "step": 55490
+    },
+    {
+      "epoch": 358.06,
+      "learning_rate": 3.566308243727599e-05,
+      "loss": 0.2648,
+      "step": 55500
+    },
+    {
+      "epoch": 358.13,
+      "learning_rate": 3.5659498207885305e-05,
+      "loss": 0.2838,
+      "step": 55510
+    },
+    {
+      "epoch": 358.19,
+      "learning_rate": 3.565591397849462e-05,
+      "loss": 0.2865,
+      "step": 55520
+    },
+    {
+      "epoch": 358.26,
+      "learning_rate": 3.565232974910395e-05,
+      "loss": 0.2792,
+      "step": 55530
+    },
+    {
+      "epoch": 358.32,
+      "learning_rate": 3.564874551971326e-05,
+      "loss": 0.2676,
+      "step": 55540
+    },
+    {
+      "epoch": 358.39,
+      "learning_rate": 3.5645161290322585e-05,
+      "loss": 0.2969,
+      "step": 55550
+    },
+    {
+      "epoch": 358.45,
+      "learning_rate": 3.56415770609319e-05,
+      "loss": 0.2559,
+      "step": 55560
+    },
+    {
+      "epoch": 358.51,
+      "learning_rate": 3.563799283154122e-05,
+      "loss": 0.2918,
+      "step": 55570
+    },
+    {
+      "epoch": 358.58,
+      "learning_rate": 3.563440860215054e-05,
+      "loss": 0.2648,
+      "step": 55580
+    },
+    {
+      "epoch": 358.64,
+      "learning_rate": 3.563082437275986e-05,
+      "loss": 0.3366,
+      "step": 55590
+    },
+    {
+      "epoch": 358.71,
+      "learning_rate": 3.5627240143369176e-05,
+      "loss": 0.3026,
+      "step": 55600
+    },
+    {
+      "epoch": 358.77,
+      "learning_rate": 3.56236559139785e-05,
+      "loss": 0.3036,
+      "step": 55610
+    },
+    {
+      "epoch": 358.84,
+      "learning_rate": 3.562007168458781e-05,
+      "loss": 0.266,
+      "step": 55620
+    },
+    {
+      "epoch": 358.9,
+      "learning_rate": 3.561648745519714e-05,
+      "loss": 0.2933,
+      "step": 55630
+    },
+    {
+      "epoch": 358.96,
+      "learning_rate": 3.5612903225806455e-05,
+      "loss": 0.312,
+      "step": 55640
+    },
+    {
+      "epoch": 359.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9270293116569519,
+      "eval_runtime": 0.8405,
+      "eval_samples_per_second": 239.15,
+      "eval_steps_per_second": 8.329,
+      "step": 55645
+    },
+    {
+      "epoch": 359.03,
+      "learning_rate": 3.560931899641577e-05,
+      "loss": 0.2444,
+      "step": 55650
+    },
+    {
+      "epoch": 359.1,
+      "learning_rate": 3.560573476702509e-05,
+      "loss": 0.3021,
+      "step": 55660
+    },
+    {
+      "epoch": 359.16,
+      "learning_rate": 3.560215053763441e-05,
+      "loss": 0.2379,
+      "step": 55670
+    },
+    {
+      "epoch": 359.23,
+      "learning_rate": 3.559856630824373e-05,
+      "loss": 0.2701,
+      "step": 55680
+    },
+    {
+      "epoch": 359.29,
+      "learning_rate": 3.559498207885305e-05,
+      "loss": 0.317,
+      "step": 55690
+    },
+    {
+      "epoch": 359.35,
+      "learning_rate": 3.5591397849462364e-05,
+      "loss": 0.2683,
+      "step": 55700
+    },
+    {
+      "epoch": 359.42,
+      "learning_rate": 3.558781362007169e-05,
+      "loss": 0.3078,
+      "step": 55710
+    },
+    {
+      "epoch": 359.48,
+      "learning_rate": 3.558422939068101e-05,
+      "loss": 0.3368,
+      "step": 55720
+    },
+    {
+      "epoch": 359.55,
+      "learning_rate": 3.5580645161290326e-05,
+      "loss": 0.2977,
+      "step": 55730
+    },
+    {
+      "epoch": 359.61,
+      "learning_rate": 3.5577060931899644e-05,
+      "loss": 0.2638,
+      "step": 55740
+    },
+    {
+      "epoch": 359.68,
+      "learning_rate": 3.557347670250896e-05,
+      "loss": 0.2672,
+      "step": 55750
+    },
+    {
+      "epoch": 359.74,
+      "learning_rate": 3.556989247311828e-05,
+      "loss": 0.3252,
+      "step": 55760
+    },
+    {
+      "epoch": 359.8,
+      "learning_rate": 3.5566308243727605e-05,
+      "loss": 0.2976,
+      "step": 55770
+    },
+    {
+      "epoch": 359.87,
+      "learning_rate": 3.556272401433692e-05,
+      "loss": 0.2491,
+      "step": 55780
+    },
+    {
+      "epoch": 359.93,
+      "learning_rate": 3.5559139784946235e-05,
+      "loss": 0.2894,
+      "step": 55790
+    },
+    {
+      "epoch": 360.0,
+      "learning_rate": 3.555555555555556e-05,
+      "loss": 0.27,
+      "step": 55800
+    },
+    {
+      "epoch": 360.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9621337056159973,
+      "eval_runtime": 0.851,
+      "eval_samples_per_second": 236.202,
+      "eval_steps_per_second": 8.226,
+      "step": 55800
+    },
+    {
+      "epoch": 360.06,
+      "learning_rate": 3.555197132616488e-05,
+      "loss": 0.3137,
+      "step": 55810
+    },
+    {
+      "epoch": 360.13,
+      "learning_rate": 3.5548387096774196e-05,
+      "loss": 0.299,
+      "step": 55820
+    },
+    {
+      "epoch": 360.19,
+      "learning_rate": 3.5544802867383515e-05,
+      "loss": 0.276,
+      "step": 55830
+    },
+    {
+      "epoch": 360.26,
+      "learning_rate": 3.554121863799283e-05,
+      "loss": 0.2983,
+      "step": 55840
+    },
+    {
+      "epoch": 360.32,
+      "learning_rate": 3.553763440860215e-05,
+      "loss": 0.2861,
+      "step": 55850
+    },
+    {
+      "epoch": 360.39,
+      "learning_rate": 3.553405017921147e-05,
+      "loss": 0.2568,
+      "step": 55860
+    },
+    {
+      "epoch": 360.45,
+      "learning_rate": 3.553046594982079e-05,
+      "loss": 0.3162,
+      "step": 55870
+    },
+    {
+      "epoch": 360.51,
+      "learning_rate": 3.552688172043011e-05,
+      "loss": 0.273,
+      "step": 55880
+    },
+    {
+      "epoch": 360.58,
+      "learning_rate": 3.5523297491039424e-05,
+      "loss": 0.292,
+      "step": 55890
+    },
+    {
+      "epoch": 360.64,
+      "learning_rate": 3.551971326164875e-05,
+      "loss": 0.2915,
+      "step": 55900
+    },
+    {
+      "epoch": 360.71,
+      "learning_rate": 3.551612903225807e-05,
+      "loss": 0.2858,
+      "step": 55910
+    },
+    {
+      "epoch": 360.77,
+      "learning_rate": 3.5512544802867385e-05,
+      "loss": 0.2353,
+      "step": 55920
+    },
+    {
+      "epoch": 360.84,
+      "learning_rate": 3.5508960573476703e-05,
+      "loss": 0.2678,
+      "step": 55930
+    },
+    {
+      "epoch": 360.9,
+      "learning_rate": 3.550537634408602e-05,
+      "loss": 0.3092,
+      "step": 55940
+    },
+    {
+      "epoch": 360.96,
+      "learning_rate": 3.550179211469534e-05,
+      "loss": 0.2901,
+      "step": 55950
+    },
+    {
+      "epoch": 361.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.9777515530586243,
+      "eval_runtime": 0.8523,
+      "eval_samples_per_second": 235.824,
+      "eval_steps_per_second": 8.213,
+      "step": 55955
+    },
+    {
+      "epoch": 361.03,
+      "learning_rate": 3.5498207885304665e-05,
+      "loss": 0.3591,
+      "step": 55960
+    },
+    {
+      "epoch": 361.1,
+      "learning_rate": 3.5494623655913976e-05,
+      "loss": 0.338,
+      "step": 55970
+    },
+    {
+      "epoch": 361.16,
+      "learning_rate": 3.54910394265233e-05,
+      "loss": 0.2557,
+      "step": 55980
+    },
+    {
+      "epoch": 361.23,
+      "learning_rate": 3.548745519713262e-05,
+      "loss": 0.2531,
+      "step": 55990
+    },
+    {
+      "epoch": 361.29,
+      "learning_rate": 3.548387096774194e-05,
+      "loss": 0.2966,
+      "step": 56000
+    },
+    {
+      "epoch": 361.35,
+      "learning_rate": 3.5480286738351256e-05,
+      "loss": 0.3114,
+      "step": 56010
+    },
+    {
+      "epoch": 361.42,
+      "learning_rate": 3.5476702508960574e-05,
+      "loss": 0.3114,
+      "step": 56020
+    },
+    {
+      "epoch": 361.48,
+      "learning_rate": 3.547311827956989e-05,
+      "loss": 0.273,
+      "step": 56030
+    },
+    {
+      "epoch": 361.55,
+      "learning_rate": 3.546953405017922e-05,
+      "loss": 0.3178,
+      "step": 56040
+    },
+    {
+      "epoch": 361.61,
+      "learning_rate": 3.546594982078853e-05,
+      "loss": 0.2727,
+      "step": 56050
+    },
+    {
+      "epoch": 361.68,
+      "learning_rate": 3.5462365591397854e-05,
+      "loss": 0.3194,
+      "step": 56060
+    },
+    {
+      "epoch": 361.74,
+      "learning_rate": 3.545878136200717e-05,
+      "loss": 0.2774,
+      "step": 56070
+    },
+    {
+      "epoch": 361.8,
+      "learning_rate": 3.545519713261649e-05,
+      "loss": 0.3145,
+      "step": 56080
+    },
+    {
+      "epoch": 361.87,
+      "learning_rate": 3.545161290322581e-05,
+      "loss": 0.295,
+      "step": 56090
+    },
+    {
+      "epoch": 361.93,
+      "learning_rate": 3.5448028673835126e-05,
+      "loss": 0.2739,
+      "step": 56100
+    },
+    {
+      "epoch": 362.0,
+      "learning_rate": 3.5444444444444445e-05,
+      "loss": 0.2638,
+      "step": 56110
+    },
+    {
+      "epoch": 362.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0014443397521973,
+      "eval_runtime": 0.8433,
+      "eval_samples_per_second": 238.356,
+      "eval_steps_per_second": 8.301,
+      "step": 56110
+    },
+    {
+      "epoch": 362.06,
+      "learning_rate": 3.544086021505377e-05,
+      "loss": 0.3207,
+      "step": 56120
+    },
+    {
+      "epoch": 362.13,
+      "learning_rate": 3.543727598566308e-05,
+      "loss": 0.2838,
+      "step": 56130
+    },
+    {
+      "epoch": 362.19,
+      "learning_rate": 3.54336917562724e-05,
+      "loss": 0.2941,
+      "step": 56140
+    },
+    {
+      "epoch": 362.26,
+      "learning_rate": 3.5430107526881724e-05,
+      "loss": 0.2726,
+      "step": 56150
+    },
+    {
+      "epoch": 362.32,
+      "learning_rate": 3.542652329749104e-05,
+      "loss": 0.3136,
+      "step": 56160
+    },
+    {
+      "epoch": 362.39,
+      "learning_rate": 3.542293906810036e-05,
+      "loss": 0.2842,
+      "step": 56170
+    },
+    {
+      "epoch": 362.45,
+      "learning_rate": 3.541935483870968e-05,
+      "loss": 0.3625,
+      "step": 56180
+    },
+    {
+      "epoch": 362.51,
+      "learning_rate": 3.5415770609319e-05,
+      "loss": 0.2969,
+      "step": 56190
+    },
+    {
+      "epoch": 362.58,
+      "learning_rate": 3.541218637992832e-05,
+      "loss": 0.2579,
+      "step": 56200
+    },
+    {
+      "epoch": 362.64,
+      "learning_rate": 3.5408602150537633e-05,
+      "loss": 0.2768,
+      "step": 56210
+    },
+    {
+      "epoch": 362.71,
+      "learning_rate": 3.540501792114695e-05,
+      "loss": 0.2384,
+      "step": 56220
+    },
+    {
+      "epoch": 362.77,
+      "learning_rate": 3.540143369175628e-05,
+      "loss": 0.3024,
+      "step": 56230
+    },
+    {
+      "epoch": 362.84,
+      "learning_rate": 3.5397849462365595e-05,
+      "loss": 0.2758,
+      "step": 56240
+    },
+    {
+      "epoch": 362.9,
+      "learning_rate": 3.539426523297491e-05,
+      "loss": 0.2941,
+      "step": 56250
+    },
+    {
+      "epoch": 362.96,
+      "learning_rate": 3.539068100358423e-05,
+      "loss": 0.2413,
+      "step": 56260
+    },
+    {
+      "epoch": 363.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9649718999862671,
+      "eval_runtime": 0.8518,
+      "eval_samples_per_second": 235.96,
+      "eval_steps_per_second": 8.217,
+      "step": 56265
+    },
+    {
+      "epoch": 363.03,
+      "learning_rate": 3.538709677419355e-05,
+      "loss": 0.3275,
+      "step": 56270
+    },
+    {
+      "epoch": 363.1,
+      "learning_rate": 3.5383512544802874e-05,
+      "loss": 0.2917,
+      "step": 56280
+    },
+    {
+      "epoch": 363.16,
+      "learning_rate": 3.5379928315412186e-05,
+      "loss": 0.2763,
+      "step": 56290
+    },
+    {
+      "epoch": 363.23,
+      "learning_rate": 3.5376344086021504e-05,
+      "loss": 0.3196,
+      "step": 56300
+    },
+    {
+      "epoch": 363.29,
+      "learning_rate": 3.537275985663083e-05,
+      "loss": 0.2711,
+      "step": 56310
+    },
+    {
+      "epoch": 363.35,
+      "learning_rate": 3.536917562724015e-05,
+      "loss": 0.2886,
+      "step": 56320
+    },
+    {
+      "epoch": 363.42,
+      "learning_rate": 3.5365591397849465e-05,
+      "loss": 0.3196,
+      "step": 56330
+    },
+    {
+      "epoch": 363.48,
+      "learning_rate": 3.5362007168458784e-05,
+      "loss": 0.3242,
+      "step": 56340
+    },
+    {
+      "epoch": 363.55,
+      "learning_rate": 3.53584229390681e-05,
+      "loss": 0.3295,
+      "step": 56350
+    },
+    {
+      "epoch": 363.61,
+      "learning_rate": 3.535483870967743e-05,
+      "loss": 0.3218,
+      "step": 56360
+    },
+    {
+      "epoch": 363.68,
+      "learning_rate": 3.535125448028674e-05,
+      "loss": 0.2959,
+      "step": 56370
+    },
+    {
+      "epoch": 363.74,
+      "learning_rate": 3.5347670250896057e-05,
+      "loss": 0.2873,
+      "step": 56380
+    },
+    {
+      "epoch": 363.8,
+      "learning_rate": 3.534408602150538e-05,
+      "loss": 0.2646,
+      "step": 56390
+    },
+    {
+      "epoch": 363.87,
+      "learning_rate": 3.53405017921147e-05,
+      "loss": 0.3242,
+      "step": 56400
+    },
+    {
+      "epoch": 363.93,
+      "learning_rate": 3.533691756272402e-05,
+      "loss": 0.2645,
+      "step": 56410
+    },
+    {
+      "epoch": 364.0,
+      "learning_rate": 3.5333333333333336e-05,
+      "loss": 0.2795,
+      "step": 56420
+    },
+    {
+      "epoch": 364.0,
+      "eval_accuracy": 0.7164179104477612,
+      "eval_loss": 1.0282639265060425,
+      "eval_runtime": 0.8535,
+      "eval_samples_per_second": 235.488,
+      "eval_steps_per_second": 8.201,
+      "step": 56420
+    },
+    {
+      "epoch": 364.06,
+      "learning_rate": 3.5329749103942654e-05,
+      "loss": 0.2752,
+      "step": 56430
+    },
+    {
+      "epoch": 364.13,
+      "learning_rate": 3.532616487455197e-05,
+      "loss": 0.28,
+      "step": 56440
+    },
+    {
+      "epoch": 364.19,
+      "learning_rate": 3.532258064516129e-05,
+      "loss": 0.2482,
+      "step": 56450
+    },
+    {
+      "epoch": 364.26,
+      "learning_rate": 3.531899641577061e-05,
+      "loss": 0.2648,
+      "step": 56460
+    },
+    {
+      "epoch": 364.32,
+      "learning_rate": 3.5315412186379934e-05,
+      "loss": 0.3027,
+      "step": 56470
+    },
+    {
+      "epoch": 364.39,
+      "learning_rate": 3.5311827956989245e-05,
+      "loss": 0.2418,
+      "step": 56480
+    },
+    {
+      "epoch": 364.45,
+      "learning_rate": 3.5308243727598564e-05,
+      "loss": 0.2816,
+      "step": 56490
+    },
+    {
+      "epoch": 364.51,
+      "learning_rate": 3.530465949820789e-05,
+      "loss": 0.2906,
+      "step": 56500
+    },
+    {
+      "epoch": 364.58,
+      "learning_rate": 3.530107526881721e-05,
+      "loss": 0.3107,
+      "step": 56510
+    },
+    {
+      "epoch": 364.64,
+      "learning_rate": 3.5297491039426525e-05,
+      "loss": 0.3011,
+      "step": 56520
+    },
+    {
+      "epoch": 364.71,
+      "learning_rate": 3.529390681003584e-05,
+      "loss": 0.274,
+      "step": 56530
+    },
+    {
+      "epoch": 364.77,
+      "learning_rate": 3.529032258064516e-05,
+      "loss": 0.298,
+      "step": 56540
+    },
+    {
+      "epoch": 364.84,
+      "learning_rate": 3.5286738351254486e-05,
+      "loss": 0.2778,
+      "step": 56550
+    },
+    {
+      "epoch": 364.9,
+      "learning_rate": 3.52831541218638e-05,
+      "loss": 0.3022,
+      "step": 56560
+    },
+    {
+      "epoch": 364.96,
+      "learning_rate": 3.5279569892473116e-05,
+      "loss": 0.2901,
+      "step": 56570
+    },
+    {
+      "epoch": 365.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9573914408683777,
+      "eval_runtime": 0.8733,
+      "eval_samples_per_second": 230.159,
+      "eval_steps_per_second": 8.015,
+      "step": 56575
+    },
+    {
+      "epoch": 365.03,
+      "learning_rate": 3.527598566308244e-05,
+      "loss": 0.3343,
+      "step": 56580
+    },
+    {
+      "epoch": 365.1,
+      "learning_rate": 3.527240143369176e-05,
+      "loss": 0.303,
+      "step": 56590
+    },
+    {
+      "epoch": 365.16,
+      "learning_rate": 3.526881720430108e-05,
+      "loss": 0.2526,
+      "step": 56600
+    },
+    {
+      "epoch": 365.23,
+      "learning_rate": 3.5265232974910396e-05,
+      "loss": 0.2131,
+      "step": 56610
+    },
+    {
+      "epoch": 365.29,
+      "learning_rate": 3.5261648745519714e-05,
+      "loss": 0.2908,
+      "step": 56620
+    },
+    {
+      "epoch": 365.35,
+      "learning_rate": 3.525806451612904e-05,
+      "loss": 0.2979,
+      "step": 56630
+    },
+    {
+      "epoch": 365.42,
+      "learning_rate": 3.525448028673835e-05,
+      "loss": 0.2305,
+      "step": 56640
+    },
+    {
+      "epoch": 365.48,
+      "learning_rate": 3.525089605734767e-05,
+      "loss": 0.2622,
+      "step": 56650
+    },
+    {
+      "epoch": 365.55,
+      "learning_rate": 3.524731182795699e-05,
+      "loss": 0.2821,
+      "step": 56660
+    },
+    {
+      "epoch": 365.61,
+      "learning_rate": 3.524372759856631e-05,
+      "loss": 0.2619,
+      "step": 56670
+    },
+    {
+      "epoch": 365.68,
+      "learning_rate": 3.524014336917563e-05,
+      "loss": 0.3292,
+      "step": 56680
+    },
+    {
+      "epoch": 365.74,
+      "learning_rate": 3.523655913978495e-05,
+      "loss": 0.3651,
+      "step": 56690
+    },
+    {
+      "epoch": 365.8,
+      "learning_rate": 3.5232974910394266e-05,
+      "loss": 0.2913,
+      "step": 56700
+    },
+    {
+      "epoch": 365.87,
+      "learning_rate": 3.522939068100359e-05,
+      "loss": 0.3868,
+      "step": 56710
+    },
+    {
+      "epoch": 365.93,
+      "learning_rate": 3.52258064516129e-05,
+      "loss": 0.2512,
+      "step": 56720
+    },
+    {
+      "epoch": 366.0,
+      "learning_rate": 3.522222222222222e-05,
+      "loss": 0.2695,
+      "step": 56730
+    },
+    {
+      "epoch": 366.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.8813215494155884,
+      "eval_runtime": 0.8508,
+      "eval_samples_per_second": 236.236,
+      "eval_steps_per_second": 8.227,
+      "step": 56730
+    },
+    {
+      "epoch": 366.06,
+      "learning_rate": 3.5218637992831546e-05,
+      "loss": 0.2863,
+      "step": 56740
+    },
+    {
+      "epoch": 366.13,
+      "learning_rate": 3.5215053763440864e-05,
+      "loss": 0.285,
+      "step": 56750
+    },
+    {
+      "epoch": 366.19,
+      "learning_rate": 3.5211469534050175e-05,
+      "loss": 0.2335,
+      "step": 56760
+    },
+    {
+      "epoch": 366.26,
+      "learning_rate": 3.52078853046595e-05,
+      "loss": 0.3028,
+      "step": 56770
+    },
+    {
+      "epoch": 366.32,
+      "learning_rate": 3.520430107526882e-05,
+      "loss": 0.2711,
+      "step": 56780
+    },
+    {
+      "epoch": 366.39,
+      "learning_rate": 3.520071684587814e-05,
+      "loss": 0.2944,
+      "step": 56790
+    },
+    {
+      "epoch": 366.45,
+      "learning_rate": 3.5197132616487455e-05,
+      "loss": 0.3166,
+      "step": 56800
+    },
+    {
+      "epoch": 366.51,
+      "learning_rate": 3.519354838709677e-05,
+      "loss": 0.2529,
+      "step": 56810
+    },
+    {
+      "epoch": 366.58,
+      "learning_rate": 3.51899641577061e-05,
+      "loss": 0.26,
+      "step": 56820
+    },
+    {
+      "epoch": 366.64,
+      "learning_rate": 3.5186379928315416e-05,
+      "loss": 0.2705,
+      "step": 56830
+    },
+    {
+      "epoch": 366.71,
+      "learning_rate": 3.518279569892473e-05,
+      "loss": 0.2601,
+      "step": 56840
+    },
+    {
+      "epoch": 366.77,
+      "learning_rate": 3.517921146953405e-05,
+      "loss": 0.2566,
+      "step": 56850
+    },
+    {
+      "epoch": 366.84,
+      "learning_rate": 3.517562724014337e-05,
+      "loss": 0.2901,
+      "step": 56860
+    },
+    {
+      "epoch": 366.9,
+      "learning_rate": 3.517204301075269e-05,
+      "loss": 0.2497,
+      "step": 56870
+    },
+    {
+      "epoch": 366.96,
+      "learning_rate": 3.516845878136201e-05,
+      "loss": 0.2949,
+      "step": 56880
+    },
+    {
+      "epoch": 367.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9488153457641602,
+      "eval_runtime": 0.8639,
+      "eval_samples_per_second": 232.678,
+      "eval_steps_per_second": 8.103,
+      "step": 56885
+    },
+    {
+      "epoch": 367.03,
+      "learning_rate": 3.5164874551971326e-05,
+      "loss": 0.3068,
+      "step": 56890
+    },
+    {
+      "epoch": 367.1,
+      "learning_rate": 3.516129032258065e-05,
+      "loss": 0.2767,
+      "step": 56900
+    },
+    {
+      "epoch": 367.16,
+      "learning_rate": 3.515770609318997e-05,
+      "loss": 0.2561,
+      "step": 56910
+    },
+    {
+      "epoch": 367.23,
+      "learning_rate": 3.515412186379928e-05,
+      "loss": 0.2718,
+      "step": 56920
+    },
+    {
+      "epoch": 367.29,
+      "learning_rate": 3.5150537634408605e-05,
+      "loss": 0.3179,
+      "step": 56930
+    },
+    {
+      "epoch": 367.35,
+      "learning_rate": 3.5146953405017923e-05,
+      "loss": 0.2986,
+      "step": 56940
+    },
+    {
+      "epoch": 367.42,
+      "learning_rate": 3.514336917562724e-05,
+      "loss": 0.2861,
+      "step": 56950
+    },
+    {
+      "epoch": 367.48,
+      "learning_rate": 3.513978494623656e-05,
+      "loss": 0.2897,
+      "step": 56960
+    },
+    {
+      "epoch": 367.55,
+      "learning_rate": 3.513620071684588e-05,
+      "loss": 0.2889,
+      "step": 56970
+    },
+    {
+      "epoch": 367.61,
+      "learning_rate": 3.51326164874552e-05,
+      "loss": 0.2855,
+      "step": 56980
+    },
+    {
+      "epoch": 367.68,
+      "learning_rate": 3.512903225806452e-05,
+      "loss": 0.3283,
+      "step": 56990
+    },
+    {
+      "epoch": 367.74,
+      "learning_rate": 3.512544802867383e-05,
+      "loss": 0.2636,
+      "step": 57000
+    },
+    {
+      "epoch": 367.8,
+      "learning_rate": 3.512186379928316e-05,
+      "loss": 0.2686,
+      "step": 57010
+    },
+    {
+      "epoch": 367.87,
+      "learning_rate": 3.5118279569892476e-05,
+      "loss": 0.2878,
+      "step": 57020
+    },
+    {
+      "epoch": 367.93,
+      "learning_rate": 3.5114695340501794e-05,
+      "loss": 0.3114,
+      "step": 57030
+    },
+    {
+      "epoch": 368.0,
+      "learning_rate": 3.511111111111111e-05,
+      "loss": 0.3025,
+      "step": 57040
+    },
+    {
+      "epoch": 368.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9307913780212402,
+      "eval_runtime": 0.8494,
+      "eval_samples_per_second": 236.625,
+      "eval_steps_per_second": 8.241,
+      "step": 57040
+    },
+    {
+      "epoch": 368.06,
+      "learning_rate": 3.510752688172043e-05,
+      "loss": 0.2943,
+      "step": 57050
+    },
+    {
+      "epoch": 368.13,
+      "learning_rate": 3.5103942652329755e-05,
+      "loss": 0.2627,
+      "step": 57060
+    },
+    {
+      "epoch": 368.19,
+      "learning_rate": 3.5100358422939074e-05,
+      "loss": 0.2479,
+      "step": 57070
+    },
+    {
+      "epoch": 368.26,
+      "learning_rate": 3.5096774193548385e-05,
+      "loss": 0.2561,
+      "step": 57080
+    },
+    {
+      "epoch": 368.32,
+      "learning_rate": 3.509318996415771e-05,
+      "loss": 0.2867,
+      "step": 57090
+    },
+    {
+      "epoch": 368.39,
+      "learning_rate": 3.508960573476703e-05,
+      "loss": 0.3399,
+      "step": 57100
+    },
+    {
+      "epoch": 368.45,
+      "learning_rate": 3.508602150537634e-05,
+      "loss": 0.2496,
+      "step": 57110
+    },
+    {
+      "epoch": 368.51,
+      "learning_rate": 3.5082437275985665e-05,
+      "loss": 0.4027,
+      "step": 57120
+    },
+    {
+      "epoch": 368.58,
+      "learning_rate": 3.507885304659498e-05,
+      "loss": 0.3128,
+      "step": 57130
+    },
+    {
+      "epoch": 368.64,
+      "learning_rate": 3.50752688172043e-05,
+      "loss": 0.3061,
+      "step": 57140
+    },
+    {
+      "epoch": 368.71,
+      "learning_rate": 3.507168458781362e-05,
+      "loss": 0.251,
+      "step": 57150
+    },
+    {
+      "epoch": 368.77,
+      "learning_rate": 3.506810035842294e-05,
+      "loss": 0.2502,
+      "step": 57160
+    },
+    {
+      "epoch": 368.84,
+      "learning_rate": 3.506451612903226e-05,
+      "loss": 0.2705,
+      "step": 57170
+    },
+    {
+      "epoch": 368.9,
+      "learning_rate": 3.506093189964158e-05,
+      "loss": 0.3087,
+      "step": 57180
+    },
+    {
+      "epoch": 368.96,
+      "learning_rate": 3.505734767025089e-05,
+      "loss": 0.3177,
+      "step": 57190
+    },
+    {
+      "epoch": 369.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.972366988658905,
+      "eval_runtime": 0.8438,
+      "eval_samples_per_second": 238.217,
+      "eval_steps_per_second": 8.296,
+      "step": 57195
+    },
+    {
+      "epoch": 369.03,
+      "learning_rate": 3.505376344086022e-05,
+      "loss": 0.3222,
+      "step": 57200
+    },
+    {
+      "epoch": 369.1,
+      "learning_rate": 3.5050179211469535e-05,
+      "loss": 0.2946,
+      "step": 57210
+    },
+    {
+      "epoch": 369.16,
+      "learning_rate": 3.5046594982078853e-05,
+      "loss": 0.3099,
+      "step": 57220
+    },
+    {
+      "epoch": 369.23,
+      "learning_rate": 3.504301075268817e-05,
+      "loss": 0.2892,
+      "step": 57230
+    },
+    {
+      "epoch": 369.29,
+      "learning_rate": 3.503942652329749e-05,
+      "loss": 0.3057,
+      "step": 57240
+    },
+    {
+      "epoch": 369.35,
+      "learning_rate": 3.5035842293906815e-05,
+      "loss": 0.2689,
+      "step": 57250
+    },
+    {
+      "epoch": 369.42,
+      "learning_rate": 3.503225806451613e-05,
+      "loss": 0.3056,
+      "step": 57260
+    },
+    {
+      "epoch": 369.48,
+      "learning_rate": 3.5028673835125445e-05,
+      "loss": 0.2413,
+      "step": 57270
+    },
+    {
+      "epoch": 369.55,
+      "learning_rate": 3.502508960573477e-05,
+      "loss": 0.2731,
+      "step": 57280
+    },
+    {
+      "epoch": 369.61,
+      "learning_rate": 3.502150537634409e-05,
+      "loss": 0.2893,
+      "step": 57290
+    },
+    {
+      "epoch": 369.68,
+      "learning_rate": 3.5017921146953406e-05,
+      "loss": 0.327,
+      "step": 57300
+    },
+    {
+      "epoch": 369.74,
+      "learning_rate": 3.5014336917562724e-05,
+      "loss": 0.3281,
+      "step": 57310
+    },
+    {
+      "epoch": 369.8,
+      "learning_rate": 3.501075268817204e-05,
+      "loss": 0.3036,
+      "step": 57320
+    },
+    {
+      "epoch": 369.87,
+      "learning_rate": 3.500716845878137e-05,
+      "loss": 0.3062,
+      "step": 57330
+    },
+    {
+      "epoch": 369.93,
+      "learning_rate": 3.5003584229390686e-05,
+      "loss": 0.2807,
+      "step": 57340
+    },
+    {
+      "epoch": 370.0,
+      "learning_rate": 3.5e-05,
+      "loss": 0.2566,
+      "step": 57350
+    },
+    {
+      "epoch": 370.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.91930091381073,
+      "eval_runtime": 0.8644,
+      "eval_samples_per_second": 232.544,
+      "eval_steps_per_second": 8.099,
+      "step": 57350
+    },
+    {
+      "epoch": 370.06,
+      "learning_rate": 3.499641577060932e-05,
+      "loss": 0.3117,
+      "step": 57360
+    },
+    {
+      "epoch": 370.13,
+      "learning_rate": 3.499283154121864e-05,
+      "loss": 0.3292,
+      "step": 57370
+    },
+    {
+      "epoch": 370.19,
+      "learning_rate": 3.498924731182796e-05,
+      "loss": 0.2655,
+      "step": 57380
+    },
+    {
+      "epoch": 370.26,
+      "learning_rate": 3.4985663082437277e-05,
+      "loss": 0.2927,
+      "step": 57390
+    },
+    {
+      "epoch": 370.32,
+      "learning_rate": 3.4982078853046595e-05,
+      "loss": 0.2543,
+      "step": 57400
+    },
+    {
+      "epoch": 370.39,
+      "learning_rate": 3.497849462365592e-05,
+      "loss": 0.3049,
+      "step": 57410
+    },
+    {
+      "epoch": 370.45,
+      "learning_rate": 3.497491039426524e-05,
+      "loss": 0.3346,
+      "step": 57420
+    },
+    {
+      "epoch": 370.51,
+      "learning_rate": 3.497132616487455e-05,
+      "loss": 0.2794,
+      "step": 57430
+    },
+    {
+      "epoch": 370.58,
+      "learning_rate": 3.4967741935483874e-05,
+      "loss": 0.3103,
+      "step": 57440
+    },
+    {
+      "epoch": 370.64,
+      "learning_rate": 3.496415770609319e-05,
+      "loss": 0.2862,
+      "step": 57450
+    },
+    {
+      "epoch": 370.71,
+      "learning_rate": 3.496057347670251e-05,
+      "loss": 0.3091,
+      "step": 57460
+    },
+    {
+      "epoch": 370.77,
+      "learning_rate": 3.495698924731183e-05,
+      "loss": 0.3099,
+      "step": 57470
+    },
+    {
+      "epoch": 370.84,
+      "learning_rate": 3.495340501792115e-05,
+      "loss": 0.2715,
+      "step": 57480
+    },
+    {
+      "epoch": 370.9,
+      "learning_rate": 3.4949820788530465e-05,
+      "loss": 0.2389,
+      "step": 57490
+    },
+    {
+      "epoch": 370.96,
+      "learning_rate": 3.494623655913979e-05,
+      "loss": 0.3012,
+      "step": 57500
+    },
+    {
+      "epoch": 371.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9895779490470886,
+      "eval_runtime": 0.8841,
+      "eval_samples_per_second": 227.363,
+      "eval_steps_per_second": 7.918,
+      "step": 57505
+    },
+    {
+      "epoch": 371.03,
+      "learning_rate": 3.49426523297491e-05,
+      "loss": 0.3208,
+      "step": 57510
+    },
+    {
+      "epoch": 371.1,
+      "learning_rate": 3.493906810035843e-05,
+      "loss": 0.2851,
+      "step": 57520
+    },
+    {
+      "epoch": 371.16,
+      "learning_rate": 3.4935483870967745e-05,
+      "loss": 0.2736,
+      "step": 57530
+    },
+    {
+      "epoch": 371.23,
+      "learning_rate": 3.493189964157706e-05,
+      "loss": 0.3153,
+      "step": 57540
+    },
+    {
+      "epoch": 371.29,
+      "learning_rate": 3.492831541218638e-05,
+      "loss": 0.2998,
+      "step": 57550
+    },
+    {
+      "epoch": 371.35,
+      "learning_rate": 3.49247311827957e-05,
+      "loss": 0.3226,
+      "step": 57560
+    },
+    {
+      "epoch": 371.42,
+      "learning_rate": 3.492114695340502e-05,
+      "loss": 0.2975,
+      "step": 57570
+    },
+    {
+      "epoch": 371.48,
+      "learning_rate": 3.491756272401434e-05,
+      "loss": 0.2656,
+      "step": 57580
+    },
+    {
+      "epoch": 371.55,
+      "learning_rate": 3.4913978494623654e-05,
+      "loss": 0.2943,
+      "step": 57590
+    },
+    {
+      "epoch": 371.61,
+      "learning_rate": 3.491039426523298e-05,
+      "loss": 0.3019,
+      "step": 57600
+    },
+    {
+      "epoch": 371.68,
+      "learning_rate": 3.49068100358423e-05,
+      "loss": 0.2947,
+      "step": 57610
+    },
+    {
+      "epoch": 371.74,
+      "learning_rate": 3.4903225806451616e-05,
+      "loss": 0.2409,
+      "step": 57620
+    },
+    {
+      "epoch": 371.8,
+      "learning_rate": 3.4899641577060934e-05,
+      "loss": 0.3263,
+      "step": 57630
+    },
+    {
+      "epoch": 371.87,
+      "learning_rate": 3.489605734767025e-05,
+      "loss": 0.2608,
+      "step": 57640
+    },
+    {
+      "epoch": 371.93,
+      "learning_rate": 3.489247311827957e-05,
+      "loss": 0.3294,
+      "step": 57650
+    },
+    {
+      "epoch": 372.0,
+      "learning_rate": 3.4888888888888895e-05,
+      "loss": 0.2617,
+      "step": 57660
+    },
+    {
+      "epoch": 372.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9749698042869568,
+      "eval_runtime": 0.8565,
+      "eval_samples_per_second": 234.672,
+      "eval_steps_per_second": 8.173,
+      "step": 57660
+    },
+    {
+      "epoch": 372.06,
+      "learning_rate": 3.4885304659498207e-05,
+      "loss": 0.2912,
+      "step": 57670
+    },
+    {
+      "epoch": 372.13,
+      "learning_rate": 3.488172043010753e-05,
+      "loss": 0.3116,
+      "step": 57680
+    },
+    {
+      "epoch": 372.19,
+      "learning_rate": 3.487813620071685e-05,
+      "loss": 0.2838,
+      "step": 57690
+    },
+    {
+      "epoch": 372.26,
+      "learning_rate": 3.487455197132617e-05,
+      "loss": 0.3073,
+      "step": 57700
+    },
+    {
+      "epoch": 372.32,
+      "learning_rate": 3.4870967741935486e-05,
+      "loss": 0.2739,
+      "step": 57710
+    },
+    {
+      "epoch": 372.39,
+      "learning_rate": 3.4867383512544804e-05,
+      "loss": 0.3075,
+      "step": 57720
+    },
+    {
+      "epoch": 372.45,
+      "learning_rate": 3.486379928315412e-05,
+      "loss": 0.2615,
+      "step": 57730
+    },
+    {
+      "epoch": 372.51,
+      "learning_rate": 3.486021505376344e-05,
+      "loss": 0.29,
+      "step": 57740
+    },
+    {
+      "epoch": 372.58,
+      "learning_rate": 3.485663082437276e-05,
+      "loss": 0.2682,
+      "step": 57750
+    },
+    {
+      "epoch": 372.64,
+      "learning_rate": 3.485304659498208e-05,
+      "loss": 0.259,
+      "step": 57760
+    },
+    {
+      "epoch": 372.71,
+      "learning_rate": 3.48494623655914e-05,
+      "loss": 0.2947,
+      "step": 57770
+    },
+    {
+      "epoch": 372.77,
+      "learning_rate": 3.4845878136200714e-05,
+      "loss": 0.2961,
+      "step": 57780
+    },
+    {
+      "epoch": 372.84,
+      "learning_rate": 3.484229390681004e-05,
+      "loss": 0.3269,
+      "step": 57790
+    },
+    {
+      "epoch": 372.9,
+      "learning_rate": 3.483870967741936e-05,
+      "loss": 0.2896,
+      "step": 57800
+    },
+    {
+      "epoch": 372.96,
+      "learning_rate": 3.4835125448028675e-05,
+      "loss": 0.2692,
+      "step": 57810
+    },
+    {
+      "epoch": 373.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9831479787826538,
+      "eval_runtime": 0.8885,
+      "eval_samples_per_second": 226.211,
+      "eval_steps_per_second": 7.878,
+      "step": 57815
+    },
+    {
+      "epoch": 373.03,
+      "learning_rate": 3.483154121863799e-05,
+      "loss": 0.2791,
+      "step": 57820
+    },
+    {
+      "epoch": 373.1,
+      "learning_rate": 3.482795698924731e-05,
+      "loss": 0.2418,
+      "step": 57830
+    },
+    {
+      "epoch": 373.16,
+      "learning_rate": 3.482437275985663e-05,
+      "loss": 0.2833,
+      "step": 57840
+    },
+    {
+      "epoch": 373.23,
+      "learning_rate": 3.4820788530465955e-05,
+      "loss": 0.3022,
+      "step": 57850
+    },
+    {
+      "epoch": 373.29,
+      "learning_rate": 3.4817204301075266e-05,
+      "loss": 0.2818,
+      "step": 57860
+    },
+    {
+      "epoch": 373.35,
+      "learning_rate": 3.481362007168459e-05,
+      "loss": 0.336,
+      "step": 57870
+    },
+    {
+      "epoch": 373.42,
+      "learning_rate": 3.481003584229391e-05,
+      "loss": 0.266,
+      "step": 57880
+    },
+    {
+      "epoch": 373.48,
+      "learning_rate": 3.480645161290323e-05,
+      "loss": 0.3046,
+      "step": 57890
+    },
+    {
+      "epoch": 373.55,
+      "learning_rate": 3.4802867383512546e-05,
+      "loss": 0.2462,
+      "step": 57900
+    },
+    {
+      "epoch": 373.61,
+      "learning_rate": 3.4799283154121864e-05,
+      "loss": 0.2673,
+      "step": 57910
+    },
+    {
+      "epoch": 373.68,
+      "learning_rate": 3.479569892473118e-05,
+      "loss": 0.2625,
+      "step": 57920
+    },
+    {
+      "epoch": 373.74,
+      "learning_rate": 3.479211469534051e-05,
+      "loss": 0.2973,
+      "step": 57930
+    },
+    {
+      "epoch": 373.8,
+      "learning_rate": 3.478853046594982e-05,
+      "loss": 0.3129,
+      "step": 57940
+    },
+    {
+      "epoch": 373.87,
+      "learning_rate": 3.4784946236559143e-05,
+      "loss": 0.3378,
+      "step": 57950
+    },
+    {
+      "epoch": 373.93,
+      "learning_rate": 3.478136200716846e-05,
+      "loss": 0.2849,
+      "step": 57960
+    },
+    {
+      "epoch": 374.0,
+      "learning_rate": 3.477777777777778e-05,
+      "loss": 0.2643,
+      "step": 57970
+    },
+    {
+      "epoch": 374.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0188920497894287,
+      "eval_runtime": 0.8652,
+      "eval_samples_per_second": 232.307,
+      "eval_steps_per_second": 8.09,
+      "step": 57970
+    },
+    {
+      "epoch": 374.06,
+      "learning_rate": 3.47741935483871e-05,
+      "loss": 0.2846,
+      "step": 57980
+    },
+    {
+      "epoch": 374.13,
+      "learning_rate": 3.4770609318996416e-05,
+      "loss": 0.3122,
+      "step": 57990
+    },
+    {
+      "epoch": 374.19,
+      "learning_rate": 3.4767025089605734e-05,
+      "loss": 0.3138,
+      "step": 58000
+    },
+    {
+      "epoch": 374.26,
+      "learning_rate": 3.476344086021506e-05,
+      "loss": 0.3094,
+      "step": 58010
+    },
+    {
+      "epoch": 374.32,
+      "learning_rate": 3.475985663082437e-05,
+      "loss": 0.2568,
+      "step": 58020
+    },
+    {
+      "epoch": 374.39,
+      "learning_rate": 3.4756272401433696e-05,
+      "loss": 0.2932,
+      "step": 58030
+    },
+    {
+      "epoch": 374.45,
+      "learning_rate": 3.4752688172043014e-05,
+      "loss": 0.2885,
+      "step": 58040
+    },
+    {
+      "epoch": 374.51,
+      "learning_rate": 3.474910394265233e-05,
+      "loss": 0.3123,
+      "step": 58050
+    },
+    {
+      "epoch": 374.58,
+      "learning_rate": 3.474551971326165e-05,
+      "loss": 0.3091,
+      "step": 58060
+    },
+    {
+      "epoch": 374.64,
+      "learning_rate": 3.474193548387097e-05,
+      "loss": 0.3082,
+      "step": 58070
+    },
+    {
+      "epoch": 374.71,
+      "learning_rate": 3.473835125448029e-05,
+      "loss": 0.3022,
+      "step": 58080
+    },
+    {
+      "epoch": 374.77,
+      "learning_rate": 3.473476702508961e-05,
+      "loss": 0.2385,
+      "step": 58090
+    },
+    {
+      "epoch": 374.84,
+      "learning_rate": 3.473118279569892e-05,
+      "loss": 0.2346,
+      "step": 58100
+    },
+    {
+      "epoch": 374.9,
+      "learning_rate": 3.472759856630824e-05,
+      "loss": 0.293,
+      "step": 58110
+    },
+    {
+      "epoch": 374.96,
+      "learning_rate": 3.4724014336917566e-05,
+      "loss": 0.2656,
+      "step": 58120
+    },
+    {
+      "epoch": 375.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9283463358879089,
+      "eval_runtime": 0.8766,
+      "eval_samples_per_second": 229.307,
+      "eval_steps_per_second": 7.986,
+      "step": 58125
+    },
+    {
+      "epoch": 375.03,
+      "learning_rate": 3.4720430107526885e-05,
+      "loss": 0.252,
+      "step": 58130
+    },
+    {
+      "epoch": 375.1,
+      "learning_rate": 3.47168458781362e-05,
+      "loss": 0.2131,
+      "step": 58140
+    },
+    {
+      "epoch": 375.16,
+      "learning_rate": 3.471326164874552e-05,
+      "loss": 0.2903,
+      "step": 58150
+    },
+    {
+      "epoch": 375.23,
+      "learning_rate": 3.470967741935484e-05,
+      "loss": 0.2847,
+      "step": 58160
+    },
+    {
+      "epoch": 375.29,
+      "learning_rate": 3.4706093189964164e-05,
+      "loss": 0.2707,
+      "step": 58170
+    },
+    {
+      "epoch": 375.35,
+      "learning_rate": 3.4702508960573476e-05,
+      "loss": 0.3195,
+      "step": 58180
+    },
+    {
+      "epoch": 375.42,
+      "learning_rate": 3.4698924731182794e-05,
+      "loss": 0.3031,
+      "step": 58190
+    },
+    {
+      "epoch": 375.48,
+      "learning_rate": 3.469534050179212e-05,
+      "loss": 0.2789,
+      "step": 58200
+    },
+    {
+      "epoch": 375.55,
+      "learning_rate": 3.469175627240144e-05,
+      "loss": 0.2808,
+      "step": 58210
+    },
+    {
+      "epoch": 375.61,
+      "learning_rate": 3.4688172043010755e-05,
+      "loss": 0.2577,
+      "step": 58220
+    },
+    {
+      "epoch": 375.68,
+      "learning_rate": 3.4684587813620074e-05,
+      "loss": 0.2684,
+      "step": 58230
+    },
+    {
+      "epoch": 375.74,
+      "learning_rate": 3.468100358422939e-05,
+      "loss": 0.3582,
+      "step": 58240
+    },
+    {
+      "epoch": 375.8,
+      "learning_rate": 3.467741935483872e-05,
+      "loss": 0.3293,
+      "step": 58250
+    },
+    {
+      "epoch": 375.87,
+      "learning_rate": 3.467383512544803e-05,
+      "loss": 0.2927,
+      "step": 58260
+    },
+    {
+      "epoch": 375.93,
+      "learning_rate": 3.4670250896057346e-05,
+      "loss": 0.2964,
+      "step": 58270
+    },
+    {
+      "epoch": 376.0,
+      "learning_rate": 3.466666666666667e-05,
+      "loss": 0.3127,
+      "step": 58280
+    },
+    {
+      "epoch": 376.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0033504962921143,
+      "eval_runtime": 0.8811,
+      "eval_samples_per_second": 228.111,
+      "eval_steps_per_second": 7.944,
+      "step": 58280
+    },
+    {
+      "epoch": 376.06,
+      "learning_rate": 3.466308243727599e-05,
+      "loss": 0.3204,
+      "step": 58290
+    },
+    {
+      "epoch": 376.13,
+      "learning_rate": 3.465949820788531e-05,
+      "loss": 0.2762,
+      "step": 58300
+    },
+    {
+      "epoch": 376.19,
+      "learning_rate": 3.4655913978494626e-05,
+      "loss": 0.279,
+      "step": 58310
+    },
+    {
+      "epoch": 376.26,
+      "learning_rate": 3.4652329749103944e-05,
+      "loss": 0.2858,
+      "step": 58320
+    },
+    {
+      "epoch": 376.32,
+      "learning_rate": 3.464874551971327e-05,
+      "loss": 0.3224,
+      "step": 58330
+    },
+    {
+      "epoch": 376.39,
+      "learning_rate": 3.464516129032258e-05,
+      "loss": 0.2586,
+      "step": 58340
+    },
+    {
+      "epoch": 376.45,
+      "learning_rate": 3.46415770609319e-05,
+      "loss": 0.294,
+      "step": 58350
+    },
+    {
+      "epoch": 376.51,
+      "learning_rate": 3.4637992831541224e-05,
+      "loss": 0.3056,
+      "step": 58360
+    },
+    {
+      "epoch": 376.58,
+      "learning_rate": 3.463440860215054e-05,
+      "loss": 0.2963,
+      "step": 58370
+    },
+    {
+      "epoch": 376.64,
+      "learning_rate": 3.463082437275986e-05,
+      "loss": 0.2795,
+      "step": 58380
+    },
+    {
+      "epoch": 376.71,
+      "learning_rate": 3.462724014336918e-05,
+      "loss": 0.2446,
+      "step": 58390
+    },
+    {
+      "epoch": 376.77,
+      "learning_rate": 3.4623655913978497e-05,
+      "loss": 0.2814,
+      "step": 58400
+    },
+    {
+      "epoch": 376.84,
+      "learning_rate": 3.4620071684587815e-05,
+      "loss": 0.3289,
+      "step": 58410
+    },
+    {
+      "epoch": 376.9,
+      "learning_rate": 3.461648745519713e-05,
+      "loss": 0.31,
+      "step": 58420
+    },
+    {
+      "epoch": 376.96,
+      "learning_rate": 3.461290322580645e-05,
+      "loss": 0.2769,
+      "step": 58430
+    },
+    {
+      "epoch": 377.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9589763283729553,
+      "eval_runtime": 0.8512,
+      "eval_samples_per_second": 236.147,
+      "eval_steps_per_second": 8.224,
+      "step": 58435
+    },
+    {
+      "epoch": 377.03,
+      "learning_rate": 3.4609318996415776e-05,
+      "loss": 0.3217,
+      "step": 58440
+    },
+    {
+      "epoch": 377.1,
+      "learning_rate": 3.460573476702509e-05,
+      "loss": 0.3227,
+      "step": 58450
+    },
+    {
+      "epoch": 377.16,
+      "learning_rate": 3.4602150537634406e-05,
+      "loss": 0.2448,
+      "step": 58460
+    },
+    {
+      "epoch": 377.23,
+      "learning_rate": 3.459856630824373e-05,
+      "loss": 0.2737,
+      "step": 58470
+    },
+    {
+      "epoch": 377.29,
+      "learning_rate": 3.459498207885305e-05,
+      "loss": 0.3481,
+      "step": 58480
+    },
+    {
+      "epoch": 377.35,
+      "learning_rate": 3.459139784946237e-05,
+      "loss": 0.2881,
+      "step": 58490
+    },
+    {
+      "epoch": 377.42,
+      "learning_rate": 3.4587813620071685e-05,
+      "loss": 0.3072,
+      "step": 58500
+    },
+    {
+      "epoch": 377.48,
+      "learning_rate": 3.4584229390681004e-05,
+      "loss": 0.3122,
+      "step": 58510
+    },
+    {
+      "epoch": 377.55,
+      "learning_rate": 3.458064516129033e-05,
+      "loss": 0.2979,
+      "step": 58520
+    },
+    {
+      "epoch": 377.61,
+      "learning_rate": 3.457706093189964e-05,
+      "loss": 0.2448,
+      "step": 58530
+    },
+    {
+      "epoch": 377.68,
+      "learning_rate": 3.457347670250896e-05,
+      "loss": 0.2379,
+      "step": 58540
+    },
+    {
+      "epoch": 377.74,
+      "learning_rate": 3.456989247311828e-05,
+      "loss": 0.2779,
+      "step": 58550
+    },
+    {
+      "epoch": 377.8,
+      "learning_rate": 3.45663082437276e-05,
+      "loss": 0.2897,
+      "step": 58560
+    },
+    {
+      "epoch": 377.87,
+      "learning_rate": 3.456272401433692e-05,
+      "loss": 0.249,
+      "step": 58570
+    },
+    {
+      "epoch": 377.93,
+      "learning_rate": 3.455913978494624e-05,
+      "loss": 0.3029,
+      "step": 58580
+    },
+    {
+      "epoch": 378.0,
+      "learning_rate": 3.4555555555555556e-05,
+      "loss": 0.2806,
+      "step": 58590
+    },
+    {
+      "epoch": 378.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.018752932548523,
+      "eval_runtime": 0.856,
+      "eval_samples_per_second": 234.816,
+      "eval_steps_per_second": 8.178,
+      "step": 58590
+    },
+    {
+      "epoch": 378.06,
+      "learning_rate": 3.455197132616488e-05,
+      "loss": 0.2421,
+      "step": 58600
+    },
+    {
+      "epoch": 378.13,
+      "learning_rate": 3.454838709677419e-05,
+      "loss": 0.2883,
+      "step": 58610
+    },
+    {
+      "epoch": 378.19,
+      "learning_rate": 3.454480286738351e-05,
+      "loss": 0.2631,
+      "step": 58620
+    },
+    {
+      "epoch": 378.26,
+      "learning_rate": 3.4541218637992836e-05,
+      "loss": 0.3048,
+      "step": 58630
+    },
+    {
+      "epoch": 378.32,
+      "learning_rate": 3.4537634408602154e-05,
+      "loss": 0.2818,
+      "step": 58640
+    },
+    {
+      "epoch": 378.39,
+      "learning_rate": 3.453405017921147e-05,
+      "loss": 0.3148,
+      "step": 58650
+    },
+    {
+      "epoch": 378.45,
+      "learning_rate": 3.453046594982079e-05,
+      "loss": 0.2577,
+      "step": 58660
+    },
+    {
+      "epoch": 378.51,
+      "learning_rate": 3.452688172043011e-05,
+      "loss": 0.2974,
+      "step": 58670
+    },
+    {
+      "epoch": 378.58,
+      "learning_rate": 3.4523297491039433e-05,
+      "loss": 0.2742,
+      "step": 58680
+    },
+    {
+      "epoch": 378.64,
+      "learning_rate": 3.4519713261648745e-05,
+      "loss": 0.3302,
+      "step": 58690
+    },
+    {
+      "epoch": 378.71,
+      "learning_rate": 3.451612903225806e-05,
+      "loss": 0.3108,
+      "step": 58700
+    },
+    {
+      "epoch": 378.77,
+      "learning_rate": 3.451254480286739e-05,
+      "loss": 0.3095,
+      "step": 58710
+    },
+    {
+      "epoch": 378.84,
+      "learning_rate": 3.4508960573476706e-05,
+      "loss": 0.3451,
+      "step": 58720
+    },
+    {
+      "epoch": 378.9,
+      "learning_rate": 3.4505376344086024e-05,
+      "loss": 0.2845,
+      "step": 58730
+    },
+    {
+      "epoch": 378.96,
+      "learning_rate": 3.450179211469534e-05,
+      "loss": 0.2828,
+      "step": 58740
+    },
+    {
+      "epoch": 379.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0174264907836914,
+      "eval_runtime": 0.8678,
+      "eval_samples_per_second": 231.63,
+      "eval_steps_per_second": 8.067,
+      "step": 58745
+    },
+    {
+      "epoch": 379.03,
+      "learning_rate": 3.449820788530466e-05,
+      "loss": 0.2785,
+      "step": 58750
+    },
+    {
+      "epoch": 379.1,
+      "learning_rate": 3.449462365591398e-05,
+      "loss": 0.2544,
+      "step": 58760
+    },
+    {
+      "epoch": 379.16,
+      "learning_rate": 3.44910394265233e-05,
+      "loss": 0.284,
+      "step": 58770
+    },
+    {
+      "epoch": 379.23,
+      "learning_rate": 3.4487455197132615e-05,
+      "loss": 0.2722,
+      "step": 58780
+    },
+    {
+      "epoch": 379.29,
+      "learning_rate": 3.448387096774194e-05,
+      "loss": 0.2703,
+      "step": 58790
+    },
+    {
+      "epoch": 379.35,
+      "learning_rate": 3.448028673835126e-05,
+      "loss": 0.2896,
+      "step": 58800
+    },
+    {
+      "epoch": 379.42,
+      "learning_rate": 3.447670250896057e-05,
+      "loss": 0.2652,
+      "step": 58810
+    },
+    {
+      "epoch": 379.48,
+      "learning_rate": 3.4473118279569895e-05,
+      "loss": 0.3516,
+      "step": 58820
+    },
+    {
+      "epoch": 379.55,
+      "learning_rate": 3.446953405017921e-05,
+      "loss": 0.3034,
+      "step": 58830
+    },
+    {
+      "epoch": 379.61,
+      "learning_rate": 3.446594982078853e-05,
+      "loss": 0.2911,
+      "step": 58840
+    },
+    {
+      "epoch": 379.68,
+      "learning_rate": 3.446236559139785e-05,
+      "loss": 0.3092,
+      "step": 58850
+    },
+    {
+      "epoch": 379.74,
+      "learning_rate": 3.445878136200717e-05,
+      "loss": 0.3659,
+      "step": 58860
+    },
+    {
+      "epoch": 379.8,
+      "learning_rate": 3.445519713261649e-05,
+      "loss": 0.2338,
+      "step": 58870
+    },
+    {
+      "epoch": 379.87,
+      "learning_rate": 3.445161290322581e-05,
+      "loss": 0.2839,
+      "step": 58880
+    },
+    {
+      "epoch": 379.93,
+      "learning_rate": 3.444802867383512e-05,
+      "loss": 0.3485,
+      "step": 58890
+    },
+    {
+      "epoch": 380.0,
+      "learning_rate": 3.444444444444445e-05,
+      "loss": 0.2716,
+      "step": 58900
+    },
+    {
+      "epoch": 380.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.063020944595337,
+      "eval_runtime": 0.8725,
+      "eval_samples_per_second": 230.362,
+      "eval_steps_per_second": 8.023,
+      "step": 58900
+    },
+    {
+      "epoch": 380.06,
+      "learning_rate": 3.4440860215053766e-05,
+      "loss": 0.2831,
+      "step": 58910
+    },
+    {
+      "epoch": 380.13,
+      "learning_rate": 3.4437275985663084e-05,
+      "loss": 0.2522,
+      "step": 58920
+    },
+    {
+      "epoch": 380.19,
+      "learning_rate": 3.44336917562724e-05,
+      "loss": 0.2763,
+      "step": 58930
+    },
+    {
+      "epoch": 380.26,
+      "learning_rate": 3.443010752688172e-05,
+      "loss": 0.3219,
+      "step": 58940
+    },
+    {
+      "epoch": 380.32,
+      "learning_rate": 3.4426523297491045e-05,
+      "loss": 0.2648,
+      "step": 58950
+    },
+    {
+      "epoch": 380.39,
+      "learning_rate": 3.4422939068100363e-05,
+      "loss": 0.3379,
+      "step": 58960
+    },
+    {
+      "epoch": 380.45,
+      "learning_rate": 3.4419354838709675e-05,
+      "loss": 0.3186,
+      "step": 58970
+    },
+    {
+      "epoch": 380.51,
+      "learning_rate": 3.4415770609319e-05,
+      "loss": 0.2465,
+      "step": 58980
+    },
+    {
+      "epoch": 380.58,
+      "learning_rate": 3.441218637992832e-05,
+      "loss": 0.2938,
+      "step": 58990
+    },
+    {
+      "epoch": 380.64,
+      "learning_rate": 3.4408602150537636e-05,
+      "loss": 0.283,
+      "step": 59000
+    },
+    {
+      "epoch": 380.71,
+      "learning_rate": 3.4405017921146954e-05,
+      "loss": 0.2968,
+      "step": 59010
+    },
+    {
+      "epoch": 380.77,
+      "learning_rate": 3.440143369175627e-05,
+      "loss": 0.2951,
+      "step": 59020
+    },
+    {
+      "epoch": 380.84,
+      "learning_rate": 3.43978494623656e-05,
+      "loss": 0.2687,
+      "step": 59030
+    },
+    {
+      "epoch": 380.9,
+      "learning_rate": 3.439426523297491e-05,
+      "loss": 0.2995,
+      "step": 59040
+    },
+    {
+      "epoch": 380.96,
+      "learning_rate": 3.439068100358423e-05,
+      "loss": 0.2509,
+      "step": 59050
+    },
+    {
+      "epoch": 381.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0490702390670776,
+      "eval_runtime": 0.8469,
+      "eval_samples_per_second": 237.331,
+      "eval_steps_per_second": 8.265,
+      "step": 59055
+    },
+    {
+      "epoch": 381.03,
+      "learning_rate": 3.438709677419355e-05,
+      "loss": 0.2854,
+      "step": 59060
+    },
+    {
+      "epoch": 381.1,
+      "learning_rate": 3.438351254480287e-05,
+      "loss": 0.2663,
+      "step": 59070
+    },
+    {
+      "epoch": 381.16,
+      "learning_rate": 3.437992831541219e-05,
+      "loss": 0.2744,
+      "step": 59080
+    },
+    {
+      "epoch": 381.23,
+      "learning_rate": 3.437634408602151e-05,
+      "loss": 0.2294,
+      "step": 59090
+    },
+    {
+      "epoch": 381.29,
+      "learning_rate": 3.4372759856630825e-05,
+      "loss": 0.2066,
+      "step": 59100
+    },
+    {
+      "epoch": 381.35,
+      "learning_rate": 3.436917562724014e-05,
+      "loss": 0.2562,
+      "step": 59110
+    },
+    {
+      "epoch": 381.42,
+      "learning_rate": 3.436559139784946e-05,
+      "loss": 0.2554,
+      "step": 59120
+    },
+    {
+      "epoch": 381.48,
+      "learning_rate": 3.436200716845878e-05,
+      "loss": 0.2974,
+      "step": 59130
+    },
+    {
+      "epoch": 381.55,
+      "learning_rate": 3.4358422939068105e-05,
+      "loss": 0.2793,
+      "step": 59140
+    },
+    {
+      "epoch": 381.61,
+      "learning_rate": 3.435483870967742e-05,
+      "loss": 0.2644,
+      "step": 59150
+    },
+    {
+      "epoch": 381.68,
+      "learning_rate": 3.4351254480286734e-05,
+      "loss": 0.2617,
+      "step": 59160
+    },
+    {
+      "epoch": 381.74,
+      "learning_rate": 3.434767025089606e-05,
+      "loss": 0.3126,
+      "step": 59170
+    },
+    {
+      "epoch": 381.8,
+      "learning_rate": 3.434408602150538e-05,
+      "loss": 0.2783,
+      "step": 59180
+    },
+    {
+      "epoch": 381.87,
+      "learning_rate": 3.4340501792114696e-05,
+      "loss": 0.2469,
+      "step": 59190
+    },
+    {
+      "epoch": 381.93,
+      "learning_rate": 3.4336917562724014e-05,
+      "loss": 0.2688,
+      "step": 59200
+    },
+    {
+      "epoch": 382.0,
+      "learning_rate": 3.433333333333333e-05,
+      "loss": 0.3307,
+      "step": 59210
+    },
+    {
+      "epoch": 382.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 1.070288062095642,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.915,
+      "eval_steps_per_second": 7.937,
+      "step": 59210
+    },
+    {
+      "epoch": 382.06,
+      "learning_rate": 3.432974910394266e-05,
+      "loss": 0.3056,
+      "step": 59220
+    },
+    {
+      "epoch": 382.13,
+      "learning_rate": 3.4326164874551975e-05,
+      "loss": 0.2543,
+      "step": 59230
+    },
+    {
+      "epoch": 382.19,
+      "learning_rate": 3.432258064516129e-05,
+      "loss": 0.2527,
+      "step": 59240
+    },
+    {
+      "epoch": 382.26,
+      "learning_rate": 3.431899641577061e-05,
+      "loss": 0.2935,
+      "step": 59250
+    },
+    {
+      "epoch": 382.32,
+      "learning_rate": 3.431541218637993e-05,
+      "loss": 0.2809,
+      "step": 59260
+    },
+    {
+      "epoch": 382.39,
+      "learning_rate": 3.431182795698925e-05,
+      "loss": 0.2411,
+      "step": 59270
+    },
+    {
+      "epoch": 382.45,
+      "learning_rate": 3.4308243727598566e-05,
+      "loss": 0.2522,
+      "step": 59280
+    },
+    {
+      "epoch": 382.51,
+      "learning_rate": 3.4304659498207885e-05,
+      "loss": 0.3524,
+      "step": 59290
+    },
+    {
+      "epoch": 382.58,
+      "learning_rate": 3.430107526881721e-05,
+      "loss": 0.3148,
+      "step": 59300
+    },
+    {
+      "epoch": 382.64,
+      "learning_rate": 3.429749103942653e-05,
+      "loss": 0.2872,
+      "step": 59310
+    },
+    {
+      "epoch": 382.71,
+      "learning_rate": 3.429390681003584e-05,
+      "loss": 0.2974,
+      "step": 59320
+    },
+    {
+      "epoch": 382.77,
+      "learning_rate": 3.4290322580645164e-05,
+      "loss": 0.3382,
+      "step": 59330
+    },
+    {
+      "epoch": 382.84,
+      "learning_rate": 3.428673835125448e-05,
+      "loss": 0.3316,
+      "step": 59340
+    },
+    {
+      "epoch": 382.9,
+      "learning_rate": 3.42831541218638e-05,
+      "loss": 0.2641,
+      "step": 59350
+    },
+    {
+      "epoch": 382.96,
+      "learning_rate": 3.427956989247312e-05,
+      "loss": 0.2682,
+      "step": 59360
+    },
+    {
+      "epoch": 383.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9822870492935181,
+      "eval_runtime": 0.8668,
+      "eval_samples_per_second": 231.876,
+      "eval_steps_per_second": 8.075,
+      "step": 59365
+    },
+    {
+      "epoch": 383.03,
+      "learning_rate": 3.427598566308244e-05,
+      "loss": 0.2869,
+      "step": 59370
+    },
+    {
+      "epoch": 383.1,
+      "learning_rate": 3.427240143369176e-05,
+      "loss": 0.2648,
+      "step": 59380
+    },
+    {
+      "epoch": 383.16,
+      "learning_rate": 3.426881720430108e-05,
+      "loss": 0.3028,
+      "step": 59390
+    },
+    {
+      "epoch": 383.23,
+      "learning_rate": 3.426523297491039e-05,
+      "loss": 0.252,
+      "step": 59400
+    },
+    {
+      "epoch": 383.29,
+      "learning_rate": 3.4261648745519717e-05,
+      "loss": 0.2787,
+      "step": 59410
+    },
+    {
+      "epoch": 383.35,
+      "learning_rate": 3.4258064516129035e-05,
+      "loss": 0.3102,
+      "step": 59420
+    },
+    {
+      "epoch": 383.42,
+      "learning_rate": 3.425448028673835e-05,
+      "loss": 0.2708,
+      "step": 59430
+    },
+    {
+      "epoch": 383.48,
+      "learning_rate": 3.425089605734767e-05,
+      "loss": 0.27,
+      "step": 59440
+    },
+    {
+      "epoch": 383.55,
+      "learning_rate": 3.424731182795699e-05,
+      "loss": 0.2736,
+      "step": 59450
+    },
+    {
+      "epoch": 383.61,
+      "learning_rate": 3.424372759856631e-05,
+      "loss": 0.2953,
+      "step": 59460
+    },
+    {
+      "epoch": 383.68,
+      "learning_rate": 3.424014336917563e-05,
+      "loss": 0.2875,
+      "step": 59470
+    },
+    {
+      "epoch": 383.74,
+      "learning_rate": 3.4236559139784944e-05,
+      "loss": 0.265,
+      "step": 59480
+    },
+    {
+      "epoch": 383.8,
+      "learning_rate": 3.423297491039427e-05,
+      "loss": 0.3491,
+      "step": 59490
+    },
+    {
+      "epoch": 383.87,
+      "learning_rate": 3.422939068100359e-05,
+      "loss": 0.214,
+      "step": 59500
+    },
+    {
+      "epoch": 383.93,
+      "learning_rate": 3.4225806451612905e-05,
+      "loss": 0.3666,
+      "step": 59510
+    },
+    {
+      "epoch": 384.0,
+      "learning_rate": 3.4222222222222224e-05,
+      "loss": 0.222,
+      "step": 59520
+    },
+    {
+      "epoch": 384.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9881067872047424,
+      "eval_runtime": 0.8672,
+      "eval_samples_per_second": 231.791,
+      "eval_steps_per_second": 8.072,
+      "step": 59520
+    },
+    {
+      "epoch": 384.06,
+      "learning_rate": 3.421863799283154e-05,
+      "loss": 0.2494,
+      "step": 59530
+    },
+    {
+      "epoch": 384.13,
+      "learning_rate": 3.421505376344086e-05,
+      "loss": 0.2957,
+      "step": 59540
+    },
+    {
+      "epoch": 384.19,
+      "learning_rate": 3.4211469534050185e-05,
+      "loss": 0.2491,
+      "step": 59550
+    },
+    {
+      "epoch": 384.26,
+      "learning_rate": 3.4207885304659496e-05,
+      "loss": 0.304,
+      "step": 59560
+    },
+    {
+      "epoch": 384.32,
+      "learning_rate": 3.420430107526882e-05,
+      "loss": 0.2695,
+      "step": 59570
+    },
+    {
+      "epoch": 384.39,
+      "learning_rate": 3.420071684587814e-05,
+      "loss": 0.3031,
+      "step": 59580
+    },
+    {
+      "epoch": 384.45,
+      "learning_rate": 3.419713261648746e-05,
+      "loss": 0.2869,
+      "step": 59590
+    },
+    {
+      "epoch": 384.51,
+      "learning_rate": 3.4193548387096776e-05,
+      "loss": 0.2964,
+      "step": 59600
+    },
+    {
+      "epoch": 384.58,
+      "learning_rate": 3.4189964157706094e-05,
+      "loss": 0.2788,
+      "step": 59610
+    },
+    {
+      "epoch": 384.64,
+      "learning_rate": 3.418637992831541e-05,
+      "loss": 0.2733,
+      "step": 59620
+    },
+    {
+      "epoch": 384.71,
+      "learning_rate": 3.418279569892474e-05,
+      "loss": 0.2696,
+      "step": 59630
+    },
+    {
+      "epoch": 384.77,
+      "learning_rate": 3.417921146953405e-05,
+      "loss": 0.2479,
+      "step": 59640
+    },
+    {
+      "epoch": 384.84,
+      "learning_rate": 3.4175627240143374e-05,
+      "loss": 0.284,
+      "step": 59650
+    },
+    {
+      "epoch": 384.9,
+      "learning_rate": 3.417204301075269e-05,
+      "loss": 0.289,
+      "step": 59660
+    },
+    {
+      "epoch": 384.96,
+      "learning_rate": 3.4168458781362003e-05,
+      "loss": 0.2789,
+      "step": 59670
+    },
+    {
+      "epoch": 385.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9584704637527466,
+      "eval_runtime": 0.8643,
+      "eval_samples_per_second": 232.564,
+      "eval_steps_per_second": 8.099,
+      "step": 59675
+    },
+    {
+      "epoch": 385.03,
+      "learning_rate": 3.416487455197133e-05,
+      "loss": 0.3333,
+      "step": 59680
+    },
+    {
+      "epoch": 385.1,
+      "learning_rate": 3.416129032258065e-05,
+      "loss": 0.2479,
+      "step": 59690
+    },
+    {
+      "epoch": 385.16,
+      "learning_rate": 3.4157706093189965e-05,
+      "loss": 0.3167,
+      "step": 59700
+    },
+    {
+      "epoch": 385.23,
+      "learning_rate": 3.415412186379928e-05,
+      "loss": 0.2347,
+      "step": 59710
+    },
+    {
+      "epoch": 385.29,
+      "learning_rate": 3.41505376344086e-05,
+      "loss": 0.2777,
+      "step": 59720
+    },
+    {
+      "epoch": 385.35,
+      "learning_rate": 3.4146953405017926e-05,
+      "loss": 0.2737,
+      "step": 59730
+    },
+    {
+      "epoch": 385.42,
+      "learning_rate": 3.4143369175627244e-05,
+      "loss": 0.2438,
+      "step": 59740
+    },
+    {
+      "epoch": 385.48,
+      "learning_rate": 3.4139784946236556e-05,
+      "loss": 0.2479,
+      "step": 59750
+    },
+    {
+      "epoch": 385.55,
+      "learning_rate": 3.413620071684588e-05,
+      "loss": 0.282,
+      "step": 59760
+    },
+    {
+      "epoch": 385.61,
+      "learning_rate": 3.41326164874552e-05,
+      "loss": 0.3211,
+      "step": 59770
+    },
+    {
+      "epoch": 385.68,
+      "learning_rate": 3.412903225806452e-05,
+      "loss": 0.3224,
+      "step": 59780
+    },
+    {
+      "epoch": 385.74,
+      "learning_rate": 3.4125448028673835e-05,
+      "loss": 0.3184,
+      "step": 59790
+    },
+    {
+      "epoch": 385.8,
+      "learning_rate": 3.4121863799283154e-05,
+      "loss": 0.2739,
+      "step": 59800
+    },
+    {
+      "epoch": 385.87,
+      "learning_rate": 3.411827956989247e-05,
+      "loss": 0.2325,
+      "step": 59810
+    },
+    {
+      "epoch": 385.93,
+      "learning_rate": 3.41146953405018e-05,
+      "loss": 0.2842,
+      "step": 59820
+    },
+    {
+      "epoch": 386.0,
+      "learning_rate": 3.411111111111111e-05,
+      "loss": 0.3209,
+      "step": 59830
+    },
+    {
+      "epoch": 386.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.961182713508606,
+      "eval_runtime": 0.8516,
+      "eval_samples_per_second": 236.02,
+      "eval_steps_per_second": 8.22,
+      "step": 59830
+    },
+    {
+      "epoch": 386.06,
+      "learning_rate": 3.410752688172043e-05,
+      "loss": 0.3471,
+      "step": 59840
+    },
+    {
+      "epoch": 386.13,
+      "learning_rate": 3.410394265232975e-05,
+      "loss": 0.282,
+      "step": 59850
+    },
+    {
+      "epoch": 386.19,
+      "learning_rate": 3.410035842293907e-05,
+      "loss": 0.2766,
+      "step": 59860
+    },
+    {
+      "epoch": 386.26,
+      "learning_rate": 3.409677419354839e-05,
+      "loss": 0.2465,
+      "step": 59870
+    },
+    {
+      "epoch": 386.32,
+      "learning_rate": 3.4093189964157706e-05,
+      "loss": 0.3112,
+      "step": 59880
+    },
+    {
+      "epoch": 386.39,
+      "learning_rate": 3.4089605734767024e-05,
+      "loss": 0.2691,
+      "step": 59890
+    },
+    {
+      "epoch": 386.45,
+      "learning_rate": 3.408602150537635e-05,
+      "loss": 0.2683,
+      "step": 59900
+    },
+    {
+      "epoch": 386.51,
+      "learning_rate": 3.408243727598566e-05,
+      "loss": 0.3071,
+      "step": 59910
+    },
+    {
+      "epoch": 386.58,
+      "learning_rate": 3.4078853046594986e-05,
+      "loss": 0.2564,
+      "step": 59920
+    },
+    {
+      "epoch": 386.64,
+      "learning_rate": 3.4075268817204304e-05,
+      "loss": 0.2974,
+      "step": 59930
+    },
+    {
+      "epoch": 386.71,
+      "learning_rate": 3.407168458781362e-05,
+      "loss": 0.254,
+      "step": 59940
+    },
+    {
+      "epoch": 386.77,
+      "learning_rate": 3.406810035842294e-05,
+      "loss": 0.3234,
+      "step": 59950
+    },
+    {
+      "epoch": 386.84,
+      "learning_rate": 3.406451612903226e-05,
+      "loss": 0.2942,
+      "step": 59960
+    },
+    {
+      "epoch": 386.9,
+      "learning_rate": 3.406093189964158e-05,
+      "loss": 0.3468,
+      "step": 59970
+    },
+    {
+      "epoch": 386.96,
+      "learning_rate": 3.40573476702509e-05,
+      "loss": 0.2782,
+      "step": 59980
+    },
+    {
+      "epoch": 387.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9338664412498474,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 233.137,
+      "eval_steps_per_second": 8.119,
+      "step": 59985
+    },
+    {
+      "epoch": 387.03,
+      "learning_rate": 3.405376344086021e-05,
+      "loss": 0.3463,
+      "step": 59990
+    },
+    {
+      "epoch": 387.1,
+      "learning_rate": 3.405017921146954e-05,
+      "loss": 0.3036,
+      "step": 60000
+    },
+    {
+      "epoch": 387.16,
+      "learning_rate": 3.4046594982078856e-05,
+      "loss": 0.276,
+      "step": 60010
+    },
+    {
+      "epoch": 387.23,
+      "learning_rate": 3.4043010752688174e-05,
+      "loss": 0.3334,
+      "step": 60020
+    },
+    {
+      "epoch": 387.29,
+      "learning_rate": 3.403942652329749e-05,
+      "loss": 0.2414,
+      "step": 60030
+    },
+    {
+      "epoch": 387.35,
+      "learning_rate": 3.403584229390681e-05,
+      "loss": 0.3031,
+      "step": 60040
+    },
+    {
+      "epoch": 387.42,
+      "learning_rate": 3.403225806451613e-05,
+      "loss": 0.2805,
+      "step": 60050
+    },
+    {
+      "epoch": 387.48,
+      "learning_rate": 3.4028673835125454e-05,
+      "loss": 0.3284,
+      "step": 60060
+    },
+    {
+      "epoch": 387.55,
+      "learning_rate": 3.4025089605734766e-05,
+      "loss": 0.2628,
+      "step": 60070
+    },
+    {
+      "epoch": 387.61,
+      "learning_rate": 3.402150537634409e-05,
+      "loss": 0.2372,
+      "step": 60080
+    },
+    {
+      "epoch": 387.68,
+      "learning_rate": 3.401792114695341e-05,
+      "loss": 0.2607,
+      "step": 60090
+    },
+    {
+      "epoch": 387.74,
+      "learning_rate": 3.401433691756273e-05,
+      "loss": 0.2831,
+      "step": 60100
+    },
+    {
+      "epoch": 387.8,
+      "learning_rate": 3.4010752688172045e-05,
+      "loss": 0.2731,
+      "step": 60110
+    },
+    {
+      "epoch": 387.87,
+      "learning_rate": 3.400716845878136e-05,
+      "loss": 0.2766,
+      "step": 60120
+    },
+    {
+      "epoch": 387.93,
+      "learning_rate": 3.400358422939068e-05,
+      "loss": 0.2676,
+      "step": 60130
+    },
+    {
+      "epoch": 388.0,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.3016,
+      "step": 60140
+    },
+    {
+      "epoch": 388.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9429126977920532,
+      "eval_runtime": 0.865,
+      "eval_samples_per_second": 232.372,
+      "eval_steps_per_second": 8.093,
+      "step": 60140
+    },
+    {
+      "epoch": 388.06,
+      "learning_rate": 3.399641577060932e-05,
+      "loss": 0.2855,
+      "step": 60150
+    },
+    {
+      "epoch": 388.13,
+      "learning_rate": 3.3992831541218636e-05,
+      "loss": 0.2533,
+      "step": 60160
+    },
+    {
+      "epoch": 388.19,
+      "learning_rate": 3.398924731182796e-05,
+      "loss": 0.2873,
+      "step": 60170
+    },
+    {
+      "epoch": 388.26,
+      "learning_rate": 3.398566308243728e-05,
+      "loss": 0.3298,
+      "step": 60180
+    },
+    {
+      "epoch": 388.32,
+      "learning_rate": 3.39820788530466e-05,
+      "loss": 0.2666,
+      "step": 60190
+    },
+    {
+      "epoch": 388.39,
+      "learning_rate": 3.3978494623655916e-05,
+      "loss": 0.2986,
+      "step": 60200
+    },
+    {
+      "epoch": 388.45,
+      "learning_rate": 3.3974910394265234e-05,
+      "loss": 0.2845,
+      "step": 60210
+    },
+    {
+      "epoch": 388.51,
+      "learning_rate": 3.397132616487456e-05,
+      "loss": 0.2873,
+      "step": 60220
+    },
+    {
+      "epoch": 388.58,
+      "learning_rate": 3.396774193548387e-05,
+      "loss": 0.3158,
+      "step": 60230
+    },
+    {
+      "epoch": 388.64,
+      "learning_rate": 3.396415770609319e-05,
+      "loss": 0.3142,
+      "step": 60240
+    },
+    {
+      "epoch": 388.71,
+      "learning_rate": 3.3960573476702514e-05,
+      "loss": 0.2991,
+      "step": 60250
+    },
+    {
+      "epoch": 388.77,
+      "learning_rate": 3.395698924731183e-05,
+      "loss": 0.3128,
+      "step": 60260
+    },
+    {
+      "epoch": 388.84,
+      "learning_rate": 3.395340501792115e-05,
+      "loss": 0.2683,
+      "step": 60270
+    },
+    {
+      "epoch": 388.9,
+      "learning_rate": 3.394982078853047e-05,
+      "loss": 0.2651,
+      "step": 60280
+    },
+    {
+      "epoch": 388.96,
+      "learning_rate": 3.3946236559139786e-05,
+      "loss": 0.2413,
+      "step": 60290
+    },
+    {
+      "epoch": 389.0,
+      "eval_accuracy": 0.8009950248756219,
+      "eval_loss": 0.9210174679756165,
+      "eval_runtime": 0.8627,
+      "eval_samples_per_second": 232.991,
+      "eval_steps_per_second": 8.114,
+      "step": 60295
+    },
+    {
+      "epoch": 389.03,
+      "learning_rate": 3.3942652329749105e-05,
+      "loss": 0.2974,
+      "step": 60300
+    },
+    {
+      "epoch": 389.1,
+      "learning_rate": 3.393906810035842e-05,
+      "loss": 0.2795,
+      "step": 60310
+    },
+    {
+      "epoch": 389.16,
+      "learning_rate": 3.393548387096774e-05,
+      "loss": 0.2891,
+      "step": 60320
+    },
+    {
+      "epoch": 389.23,
+      "learning_rate": 3.3931899641577066e-05,
+      "loss": 0.2307,
+      "step": 60330
+    },
+    {
+      "epoch": 389.29,
+      "learning_rate": 3.392831541218638e-05,
+      "loss": 0.3082,
+      "step": 60340
+    },
+    {
+      "epoch": 389.35,
+      "learning_rate": 3.39247311827957e-05,
+      "loss": 0.2602,
+      "step": 60350
+    },
+    {
+      "epoch": 389.42,
+      "learning_rate": 3.392114695340502e-05,
+      "loss": 0.329,
+      "step": 60360
+    },
+    {
+      "epoch": 389.48,
+      "learning_rate": 3.391756272401434e-05,
+      "loss": 0.3434,
+      "step": 60370
+    },
+    {
+      "epoch": 389.55,
+      "learning_rate": 3.391397849462366e-05,
+      "loss": 0.3121,
+      "step": 60380
+    },
+    {
+      "epoch": 389.61,
+      "learning_rate": 3.3910394265232975e-05,
+      "loss": 0.2892,
+      "step": 60390
+    },
+    {
+      "epoch": 389.68,
+      "learning_rate": 3.3906810035842293e-05,
+      "loss": 0.2942,
+      "step": 60400
+    },
+    {
+      "epoch": 389.74,
+      "learning_rate": 3.390322580645162e-05,
+      "loss": 0.244,
+      "step": 60410
+    },
+    {
+      "epoch": 389.8,
+      "learning_rate": 3.389964157706093e-05,
+      "loss": 0.2946,
+      "step": 60420
+    },
+    {
+      "epoch": 389.87,
+      "learning_rate": 3.3896057347670255e-05,
+      "loss": 0.2351,
+      "step": 60430
+    },
+    {
+      "epoch": 389.93,
+      "learning_rate": 3.389247311827957e-05,
+      "loss": 0.2855,
+      "step": 60440
+    },
+    {
+      "epoch": 390.0,
+      "learning_rate": 3.388888888888889e-05,
+      "loss": 0.2954,
+      "step": 60450
+    },
+    {
+      "epoch": 390.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9753162860870361,
+      "eval_runtime": 0.8488,
+      "eval_samples_per_second": 236.813,
+      "eval_steps_per_second": 8.247,
+      "step": 60450
+    },
+    {
+      "epoch": 390.06,
+      "learning_rate": 3.388530465949821e-05,
+      "loss": 0.3479,
+      "step": 60460
+    },
+    {
+      "epoch": 390.13,
+      "learning_rate": 3.388172043010753e-05,
+      "loss": 0.3108,
+      "step": 60470
+    },
+    {
+      "epoch": 390.19,
+      "learning_rate": 3.3878136200716846e-05,
+      "loss": 0.3091,
+      "step": 60480
+    },
+    {
+      "epoch": 390.26,
+      "learning_rate": 3.387455197132617e-05,
+      "loss": 0.3125,
+      "step": 60490
+    },
+    {
+      "epoch": 390.32,
+      "learning_rate": 3.387096774193548e-05,
+      "loss": 0.2831,
+      "step": 60500
+    },
+    {
+      "epoch": 390.39,
+      "learning_rate": 3.38673835125448e-05,
+      "loss": 0.3503,
+      "step": 60510
+    },
+    {
+      "epoch": 390.45,
+      "learning_rate": 3.3863799283154125e-05,
+      "loss": 0.3077,
+      "step": 60520
+    },
+    {
+      "epoch": 390.51,
+      "learning_rate": 3.3860215053763444e-05,
+      "loss": 0.269,
+      "step": 60530
+    },
+    {
+      "epoch": 390.58,
+      "learning_rate": 3.385663082437276e-05,
+      "loss": 0.2698,
+      "step": 60540
+    },
+    {
+      "epoch": 390.64,
+      "learning_rate": 3.385304659498208e-05,
+      "loss": 0.2904,
+      "step": 60550
+    },
+    {
+      "epoch": 390.71,
+      "learning_rate": 3.38494623655914e-05,
+      "loss": 0.2838,
+      "step": 60560
+    },
+    {
+      "epoch": 390.77,
+      "learning_rate": 3.384587813620072e-05,
+      "loss": 0.2866,
+      "step": 60570
+    },
+    {
+      "epoch": 390.84,
+      "learning_rate": 3.3842293906810035e-05,
+      "loss": 0.3048,
+      "step": 60580
+    },
+    {
+      "epoch": 390.9,
+      "learning_rate": 3.383870967741935e-05,
+      "loss": 0.2519,
+      "step": 60590
+    },
+    {
+      "epoch": 390.96,
+      "learning_rate": 3.383512544802868e-05,
+      "loss": 0.309,
+      "step": 60600
+    },
+    {
+      "epoch": 391.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.003530740737915,
+      "eval_runtime": 0.8903,
+      "eval_samples_per_second": 225.779,
+      "eval_steps_per_second": 7.863,
+      "step": 60605
+    },
+    {
+      "epoch": 391.03,
+      "learning_rate": 3.3831541218637996e-05,
+      "loss": 0.2643,
+      "step": 60610
+    },
+    {
+      "epoch": 391.1,
+      "learning_rate": 3.3827956989247314e-05,
+      "loss": 0.3003,
+      "step": 60620
+    },
+    {
+      "epoch": 391.16,
+      "learning_rate": 3.382437275985663e-05,
+      "loss": 0.2645,
+      "step": 60630
+    },
+    {
+      "epoch": 391.23,
+      "learning_rate": 3.382078853046595e-05,
+      "loss": 0.214,
+      "step": 60640
+    },
+    {
+      "epoch": 391.29,
+      "learning_rate": 3.3817204301075276e-05,
+      "loss": 0.261,
+      "step": 60650
+    },
+    {
+      "epoch": 391.35,
+      "learning_rate": 3.381362007168459e-05,
+      "loss": 0.328,
+      "step": 60660
+    },
+    {
+      "epoch": 391.42,
+      "learning_rate": 3.3810035842293905e-05,
+      "loss": 0.2713,
+      "step": 60670
+    },
+    {
+      "epoch": 391.48,
+      "learning_rate": 3.380645161290323e-05,
+      "loss": 0.2891,
+      "step": 60680
+    },
+    {
+      "epoch": 391.55,
+      "learning_rate": 3.380286738351255e-05,
+      "loss": 0.2711,
+      "step": 60690
+    },
+    {
+      "epoch": 391.61,
+      "learning_rate": 3.379928315412187e-05,
+      "loss": 0.3485,
+      "step": 60700
+    },
+    {
+      "epoch": 391.68,
+      "learning_rate": 3.3795698924731185e-05,
+      "loss": 0.3314,
+      "step": 60710
+    },
+    {
+      "epoch": 391.74,
+      "learning_rate": 3.37921146953405e-05,
+      "loss": 0.3302,
+      "step": 60720
+    },
+    {
+      "epoch": 391.8,
+      "learning_rate": 3.378853046594983e-05,
+      "loss": 0.2527,
+      "step": 60730
+    },
+    {
+      "epoch": 391.87,
+      "learning_rate": 3.378494623655914e-05,
+      "loss": 0.3128,
+      "step": 60740
+    },
+    {
+      "epoch": 391.93,
+      "learning_rate": 3.378136200716846e-05,
+      "loss": 0.2817,
+      "step": 60750
+    },
+    {
+      "epoch": 392.0,
+      "learning_rate": 3.377777777777778e-05,
+      "loss": 0.2621,
+      "step": 60760
+    },
+    {
+      "epoch": 392.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9486048221588135,
+      "eval_runtime": 0.8718,
+      "eval_samples_per_second": 230.562,
+      "eval_steps_per_second": 8.03,
+      "step": 60760
+    },
+    {
+      "epoch": 392.06,
+      "learning_rate": 3.37741935483871e-05,
+      "loss": 0.3262,
+      "step": 60770
+    },
+    {
+      "epoch": 392.13,
+      "learning_rate": 3.377060931899641e-05,
+      "loss": 0.3102,
+      "step": 60780
+    },
+    {
+      "epoch": 392.19,
+      "learning_rate": 3.376702508960574e-05,
+      "loss": 0.286,
+      "step": 60790
+    },
+    {
+      "epoch": 392.26,
+      "learning_rate": 3.3763440860215055e-05,
+      "loss": 0.3143,
+      "step": 60800
+    },
+    {
+      "epoch": 392.32,
+      "learning_rate": 3.3759856630824374e-05,
+      "loss": 0.3101,
+      "step": 60810
+    },
+    {
+      "epoch": 392.39,
+      "learning_rate": 3.375627240143369e-05,
+      "loss": 0.2946,
+      "step": 60820
+    },
+    {
+      "epoch": 392.45,
+      "learning_rate": 3.375268817204301e-05,
+      "loss": 0.2945,
+      "step": 60830
+    },
+    {
+      "epoch": 392.51,
+      "learning_rate": 3.3749103942652335e-05,
+      "loss": 0.2505,
+      "step": 60840
+    },
+    {
+      "epoch": 392.58,
+      "learning_rate": 3.374551971326165e-05,
+      "loss": 0.2966,
+      "step": 60850
+    },
+    {
+      "epoch": 392.64,
+      "learning_rate": 3.3741935483870965e-05,
+      "loss": 0.2719,
+      "step": 60860
+    },
+    {
+      "epoch": 392.71,
+      "learning_rate": 3.373835125448029e-05,
+      "loss": 0.3219,
+      "step": 60870
+    },
+    {
+      "epoch": 392.77,
+      "learning_rate": 3.373476702508961e-05,
+      "loss": 0.2492,
+      "step": 60880
+    },
+    {
+      "epoch": 392.84,
+      "learning_rate": 3.3731182795698926e-05,
+      "loss": 0.2608,
+      "step": 60890
+    },
+    {
+      "epoch": 392.9,
+      "learning_rate": 3.3727598566308244e-05,
+      "loss": 0.2605,
+      "step": 60900
+    },
+    {
+      "epoch": 392.96,
+      "learning_rate": 3.372401433691756e-05,
+      "loss": 0.2846,
+      "step": 60910
+    },
+    {
+      "epoch": 393.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9653229117393494,
+      "eval_runtime": 0.8469,
+      "eval_samples_per_second": 237.33,
+      "eval_steps_per_second": 8.265,
+      "step": 60915
+    },
+    {
+      "epoch": 393.03,
+      "learning_rate": 3.372043010752689e-05,
+      "loss": 0.2884,
+      "step": 60920
+    },
+    {
+      "epoch": 393.1,
+      "learning_rate": 3.3716845878136206e-05,
+      "loss": 0.3018,
+      "step": 60930
+    },
+    {
+      "epoch": 393.16,
+      "learning_rate": 3.371326164874552e-05,
+      "loss": 0.3101,
+      "step": 60940
+    },
+    {
+      "epoch": 393.23,
+      "learning_rate": 3.370967741935484e-05,
+      "loss": 0.2699,
+      "step": 60950
+    },
+    {
+      "epoch": 393.29,
+      "learning_rate": 3.370609318996416e-05,
+      "loss": 0.2854,
+      "step": 60960
+    },
+    {
+      "epoch": 393.35,
+      "learning_rate": 3.370250896057348e-05,
+      "loss": 0.2842,
+      "step": 60970
+    },
+    {
+      "epoch": 393.42,
+      "learning_rate": 3.36989247311828e-05,
+      "loss": 0.2563,
+      "step": 60980
+    },
+    {
+      "epoch": 393.48,
+      "learning_rate": 3.3695340501792115e-05,
+      "loss": 0.31,
+      "step": 60990
+    },
+    {
+      "epoch": 393.55,
+      "learning_rate": 3.369175627240144e-05,
+      "loss": 0.2619,
+      "step": 61000
+    },
+    {
+      "epoch": 393.61,
+      "learning_rate": 3.368817204301075e-05,
+      "loss": 0.3024,
+      "step": 61010
+    },
+    {
+      "epoch": 393.68,
+      "learning_rate": 3.368458781362007e-05,
+      "loss": 0.2627,
+      "step": 61020
+    },
+    {
+      "epoch": 393.74,
+      "learning_rate": 3.3681003584229395e-05,
+      "loss": 0.2283,
+      "step": 61030
+    },
+    {
+      "epoch": 393.8,
+      "learning_rate": 3.367741935483871e-05,
+      "loss": 0.2719,
+      "step": 61040
+    },
+    {
+      "epoch": 393.87,
+      "learning_rate": 3.367383512544803e-05,
+      "loss": 0.2611,
+      "step": 61050
+    },
+    {
+      "epoch": 393.93,
+      "learning_rate": 3.367025089605735e-05,
+      "loss": 0.2847,
+      "step": 61060
+    },
+    {
+      "epoch": 394.0,
+      "learning_rate": 3.366666666666667e-05,
+      "loss": 0.2804,
+      "step": 61070
+    },
+    {
+      "epoch": 394.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9549418091773987,
+      "eval_runtime": 0.8727,
+      "eval_samples_per_second": 230.329,
+      "eval_steps_per_second": 8.021,
+      "step": 61070
+    },
+    {
+      "epoch": 394.06,
+      "learning_rate": 3.366308243727599e-05,
+      "loss": 0.2907,
+      "step": 61080
+    },
+    {
+      "epoch": 394.13,
+      "learning_rate": 3.3659498207885304e-05,
+      "loss": 0.3189,
+      "step": 61090
+    },
+    {
+      "epoch": 394.19,
+      "learning_rate": 3.365591397849462e-05,
+      "loss": 0.3053,
+      "step": 61100
+    },
+    {
+      "epoch": 394.26,
+      "learning_rate": 3.365232974910395e-05,
+      "loss": 0.3137,
+      "step": 61110
+    },
+    {
+      "epoch": 394.32,
+      "learning_rate": 3.3648745519713265e-05,
+      "loss": 0.2924,
+      "step": 61120
+    },
+    {
+      "epoch": 394.39,
+      "learning_rate": 3.3645161290322577e-05,
+      "loss": 0.2357,
+      "step": 61130
+    },
+    {
+      "epoch": 394.45,
+      "learning_rate": 3.36415770609319e-05,
+      "loss": 0.3369,
+      "step": 61140
+    },
+    {
+      "epoch": 394.51,
+      "learning_rate": 3.363799283154122e-05,
+      "loss": 0.2763,
+      "step": 61150
+    },
+    {
+      "epoch": 394.58,
+      "learning_rate": 3.363440860215054e-05,
+      "loss": 0.2708,
+      "step": 61160
+    },
+    {
+      "epoch": 394.64,
+      "learning_rate": 3.3630824372759856e-05,
+      "loss": 0.2571,
+      "step": 61170
+    },
+    {
+      "epoch": 394.71,
+      "learning_rate": 3.3627240143369174e-05,
+      "loss": 0.2901,
+      "step": 61180
+    },
+    {
+      "epoch": 394.77,
+      "learning_rate": 3.36236559139785e-05,
+      "loss": 0.2511,
+      "step": 61190
+    },
+    {
+      "epoch": 394.84,
+      "learning_rate": 3.362007168458782e-05,
+      "loss": 0.2882,
+      "step": 61200
+    },
+    {
+      "epoch": 394.9,
+      "learning_rate": 3.361648745519713e-05,
+      "loss": 0.2821,
+      "step": 61210
+    },
+    {
+      "epoch": 394.96,
+      "learning_rate": 3.3612903225806454e-05,
+      "loss": 0.2276,
+      "step": 61220
+    },
+    {
+      "epoch": 395.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9769439101219177,
+      "eval_runtime": 0.8563,
+      "eval_samples_per_second": 234.731,
+      "eval_steps_per_second": 8.175,
+      "step": 61225
+    },
+    {
+      "epoch": 395.03,
+      "learning_rate": 3.360931899641577e-05,
+      "loss": 0.3376,
+      "step": 61230
+    },
+    {
+      "epoch": 395.1,
+      "learning_rate": 3.360573476702509e-05,
+      "loss": 0.2964,
+      "step": 61240
+    },
+    {
+      "epoch": 395.16,
+      "learning_rate": 3.360215053763441e-05,
+      "loss": 0.2875,
+      "step": 61250
+    },
+    {
+      "epoch": 395.23,
+      "learning_rate": 3.359856630824373e-05,
+      "loss": 0.272,
+      "step": 61260
+    },
+    {
+      "epoch": 395.29,
+      "learning_rate": 3.359498207885305e-05,
+      "loss": 0.2607,
+      "step": 61270
+    },
+    {
+      "epoch": 395.35,
+      "learning_rate": 3.359139784946237e-05,
+      "loss": 0.2606,
+      "step": 61280
+    },
+    {
+      "epoch": 395.42,
+      "learning_rate": 3.358781362007168e-05,
+      "loss": 0.2684,
+      "step": 61290
+    },
+    {
+      "epoch": 395.48,
+      "learning_rate": 3.3584229390681006e-05,
+      "loss": 0.2159,
+      "step": 61300
+    },
+    {
+      "epoch": 395.55,
+      "learning_rate": 3.3580645161290325e-05,
+      "loss": 0.2618,
+      "step": 61310
+    },
+    {
+      "epoch": 395.61,
+      "learning_rate": 3.357706093189964e-05,
+      "loss": 0.2842,
+      "step": 61320
+    },
+    {
+      "epoch": 395.68,
+      "learning_rate": 3.357347670250896e-05,
+      "loss": 0.3159,
+      "step": 61330
+    },
+    {
+      "epoch": 395.74,
+      "learning_rate": 3.356989247311828e-05,
+      "loss": 0.2924,
+      "step": 61340
+    },
+    {
+      "epoch": 395.8,
+      "learning_rate": 3.3566308243727604e-05,
+      "loss": 0.2796,
+      "step": 61350
+    },
+    {
+      "epoch": 395.87,
+      "learning_rate": 3.356272401433692e-05,
+      "loss": 0.2285,
+      "step": 61360
+    },
+    {
+      "epoch": 395.93,
+      "learning_rate": 3.3559139784946234e-05,
+      "loss": 0.2577,
+      "step": 61370
+    },
+    {
+      "epoch": 396.0,
+      "learning_rate": 3.355555555555556e-05,
+      "loss": 0.2505,
+      "step": 61380
+    },
+    {
+      "epoch": 396.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9954367876052856,
+      "eval_runtime": 0.8657,
+      "eval_samples_per_second": 232.174,
+      "eval_steps_per_second": 8.086,
+      "step": 61380
+    },
+    {
+      "epoch": 396.06,
+      "learning_rate": 3.355197132616488e-05,
+      "loss": 0.3712,
+      "step": 61390
+    },
+    {
+      "epoch": 396.13,
+      "learning_rate": 3.3548387096774195e-05,
+      "loss": 0.3076,
+      "step": 61400
+    },
+    {
+      "epoch": 396.19,
+      "learning_rate": 3.3544802867383513e-05,
+      "loss": 0.306,
+      "step": 61410
+    },
+    {
+      "epoch": 396.26,
+      "learning_rate": 3.354121863799283e-05,
+      "loss": 0.2598,
+      "step": 61420
+    },
+    {
+      "epoch": 396.32,
+      "learning_rate": 3.3537634408602157e-05,
+      "loss": 0.2322,
+      "step": 61430
+    },
+    {
+      "epoch": 396.39,
+      "learning_rate": 3.3534050179211475e-05,
+      "loss": 0.29,
+      "step": 61440
+    },
+    {
+      "epoch": 396.45,
+      "learning_rate": 3.3530465949820786e-05,
+      "loss": 0.3226,
+      "step": 61450
+    },
+    {
+      "epoch": 396.51,
+      "learning_rate": 3.352688172043011e-05,
+      "loss": 0.2715,
+      "step": 61460
+    },
+    {
+      "epoch": 396.58,
+      "learning_rate": 3.352329749103943e-05,
+      "loss": 0.3452,
+      "step": 61470
+    },
+    {
+      "epoch": 396.64,
+      "learning_rate": 3.351971326164875e-05,
+      "loss": 0.2473,
+      "step": 61480
+    },
+    {
+      "epoch": 396.71,
+      "learning_rate": 3.3516129032258066e-05,
+      "loss": 0.3245,
+      "step": 61490
+    },
+    {
+      "epoch": 396.77,
+      "learning_rate": 3.3512544802867384e-05,
+      "loss": 0.25,
+      "step": 61500
+    },
+    {
+      "epoch": 396.84,
+      "learning_rate": 3.35089605734767e-05,
+      "loss": 0.2776,
+      "step": 61510
+    },
+    {
+      "epoch": 396.9,
+      "learning_rate": 3.350537634408603e-05,
+      "loss": 0.2948,
+      "step": 61520
+    },
+    {
+      "epoch": 396.96,
+      "learning_rate": 3.350179211469534e-05,
+      "loss": 0.2775,
+      "step": 61530
+    },
+    {
+      "epoch": 397.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.978253185749054,
+      "eval_runtime": 0.8814,
+      "eval_samples_per_second": 228.041,
+      "eval_steps_per_second": 7.942,
+      "step": 61535
+    },
+    {
+      "epoch": 397.03,
+      "learning_rate": 3.3498207885304664e-05,
+      "loss": 0.2822,
+      "step": 61540
+    },
+    {
+      "epoch": 397.1,
+      "learning_rate": 3.349462365591398e-05,
+      "loss": 0.2834,
+      "step": 61550
+    },
+    {
+      "epoch": 397.16,
+      "learning_rate": 3.34910394265233e-05,
+      "loss": 0.2069,
+      "step": 61560
+    },
+    {
+      "epoch": 397.23,
+      "learning_rate": 3.348745519713262e-05,
+      "loss": 0.2762,
+      "step": 61570
+    },
+    {
+      "epoch": 397.29,
+      "learning_rate": 3.3483870967741936e-05,
+      "loss": 0.2416,
+      "step": 61580
+    },
+    {
+      "epoch": 397.35,
+      "learning_rate": 3.3480286738351255e-05,
+      "loss": 0.2753,
+      "step": 61590
+    },
+    {
+      "epoch": 397.42,
+      "learning_rate": 3.347670250896057e-05,
+      "loss": 0.2474,
+      "step": 61600
+    },
+    {
+      "epoch": 397.48,
+      "learning_rate": 3.347311827956989e-05,
+      "loss": 0.2893,
+      "step": 61610
+    },
+    {
+      "epoch": 397.55,
+      "learning_rate": 3.3469534050179216e-05,
+      "loss": 0.2909,
+      "step": 61620
+    },
+    {
+      "epoch": 397.61,
+      "learning_rate": 3.3465949820788534e-05,
+      "loss": 0.2661,
+      "step": 61630
+    },
+    {
+      "epoch": 397.68,
+      "learning_rate": 3.3462365591397846e-05,
+      "loss": 0.3256,
+      "step": 61640
+    },
+    {
+      "epoch": 397.74,
+      "learning_rate": 3.345878136200717e-05,
+      "loss": 0.315,
+      "step": 61650
+    },
+    {
+      "epoch": 397.8,
+      "learning_rate": 3.345519713261649e-05,
+      "loss": 0.276,
+      "step": 61660
+    },
+    {
+      "epoch": 397.87,
+      "learning_rate": 3.345161290322581e-05,
+      "loss": 0.3275,
+      "step": 61670
+    },
+    {
+      "epoch": 397.93,
+      "learning_rate": 3.3448028673835125e-05,
+      "loss": 0.2788,
+      "step": 61680
+    },
+    {
+      "epoch": 398.0,
+      "learning_rate": 3.3444444444444443e-05,
+      "loss": 0.2494,
+      "step": 61690
+    },
+    {
+      "epoch": 398.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9821414351463318,
+      "eval_runtime": 0.8867,
+      "eval_samples_per_second": 226.686,
+      "eval_steps_per_second": 7.895,
+      "step": 61690
+    },
+    {
+      "epoch": 398.06,
+      "learning_rate": 3.344086021505377e-05,
+      "loss": 0.2945,
+      "step": 61700
+    },
+    {
+      "epoch": 398.13,
+      "learning_rate": 3.343727598566309e-05,
+      "loss": 0.2454,
+      "step": 61710
+    },
+    {
+      "epoch": 398.19,
+      "learning_rate": 3.34336917562724e-05,
+      "loss": 0.2018,
+      "step": 61720
+    },
+    {
+      "epoch": 398.26,
+      "learning_rate": 3.343010752688172e-05,
+      "loss": 0.3165,
+      "step": 61730
+    },
+    {
+      "epoch": 398.32,
+      "learning_rate": 3.342652329749104e-05,
+      "loss": 0.3193,
+      "step": 61740
+    },
+    {
+      "epoch": 398.39,
+      "learning_rate": 3.342293906810036e-05,
+      "loss": 0.2527,
+      "step": 61750
+    },
+    {
+      "epoch": 398.45,
+      "learning_rate": 3.341935483870968e-05,
+      "loss": 0.2575,
+      "step": 61760
+    },
+    {
+      "epoch": 398.51,
+      "learning_rate": 3.3415770609318996e-05,
+      "loss": 0.3103,
+      "step": 61770
+    },
+    {
+      "epoch": 398.58,
+      "learning_rate": 3.3412186379928314e-05,
+      "loss": 0.2604,
+      "step": 61780
+    },
+    {
+      "epoch": 398.64,
+      "learning_rate": 3.340860215053764e-05,
+      "loss": 0.326,
+      "step": 61790
+    },
+    {
+      "epoch": 398.71,
+      "learning_rate": 3.340501792114695e-05,
+      "loss": 0.3116,
+      "step": 61800
+    },
+    {
+      "epoch": 398.77,
+      "learning_rate": 3.3401433691756275e-05,
+      "loss": 0.2654,
+      "step": 61810
+    },
+    {
+      "epoch": 398.84,
+      "learning_rate": 3.3397849462365594e-05,
+      "loss": 0.2861,
+      "step": 61820
+    },
+    {
+      "epoch": 398.9,
+      "learning_rate": 3.339426523297491e-05,
+      "loss": 0.2873,
+      "step": 61830
+    },
+    {
+      "epoch": 398.96,
+      "learning_rate": 3.339068100358423e-05,
+      "loss": 0.2806,
+      "step": 61840
+    },
+    {
+      "epoch": 399.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.8496226668357849,
+      "eval_runtime": 0.8421,
+      "eval_samples_per_second": 238.684,
+      "eval_steps_per_second": 8.312,
+      "step": 61845
+    },
+    {
+      "epoch": 399.03,
+      "learning_rate": 3.338709677419355e-05,
+      "loss": 0.2632,
+      "step": 61850
+    },
+    {
+      "epoch": 399.1,
+      "learning_rate": 3.3383512544802867e-05,
+      "loss": 0.2965,
+      "step": 61860
+    },
+    {
+      "epoch": 399.16,
+      "learning_rate": 3.337992831541219e-05,
+      "loss": 0.3218,
+      "step": 61870
+    },
+    {
+      "epoch": 399.23,
+      "learning_rate": 3.33763440860215e-05,
+      "loss": 0.2793,
+      "step": 61880
+    },
+    {
+      "epoch": 399.29,
+      "learning_rate": 3.337275985663083e-05,
+      "loss": 0.2993,
+      "step": 61890
+    },
+    {
+      "epoch": 399.35,
+      "learning_rate": 3.3369175627240146e-05,
+      "loss": 0.2687,
+      "step": 61900
+    },
+    {
+      "epoch": 399.42,
+      "learning_rate": 3.3365591397849464e-05,
+      "loss": 0.2433,
+      "step": 61910
+    },
+    {
+      "epoch": 399.48,
+      "learning_rate": 3.336200716845878e-05,
+      "loss": 0.238,
+      "step": 61920
+    },
+    {
+      "epoch": 399.55,
+      "learning_rate": 3.33584229390681e-05,
+      "loss": 0.302,
+      "step": 61930
+    },
+    {
+      "epoch": 399.61,
+      "learning_rate": 3.335483870967742e-05,
+      "loss": 0.2459,
+      "step": 61940
+    },
+    {
+      "epoch": 399.68,
+      "learning_rate": 3.3351254480286744e-05,
+      "loss": 0.238,
+      "step": 61950
+    },
+    {
+      "epoch": 399.74,
+      "learning_rate": 3.3347670250896055e-05,
+      "loss": 0.295,
+      "step": 61960
+    },
+    {
+      "epoch": 399.8,
+      "learning_rate": 3.334408602150538e-05,
+      "loss": 0.2473,
+      "step": 61970
+    },
+    {
+      "epoch": 399.87,
+      "learning_rate": 3.33405017921147e-05,
+      "loss": 0.3058,
+      "step": 61980
+    },
+    {
+      "epoch": 399.93,
+      "learning_rate": 3.333691756272402e-05,
+      "loss": 0.2591,
+      "step": 61990
+    },
+    {
+      "epoch": 400.0,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.3155,
+      "step": 62000
+    },
+    {
+      "epoch": 400.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.967431902885437,
+      "eval_runtime": 0.8682,
+      "eval_samples_per_second": 231.51,
+      "eval_steps_per_second": 8.063,
+      "step": 62000
+    },
+    {
+      "epoch": 400.06,
+      "learning_rate": 3.332974910394265e-05,
+      "loss": 0.2691,
+      "step": 62010
+    },
+    {
+      "epoch": 400.13,
+      "learning_rate": 3.332616487455197e-05,
+      "loss": 0.3326,
+      "step": 62020
+    },
+    {
+      "epoch": 400.19,
+      "learning_rate": 3.3322580645161296e-05,
+      "loss": 0.2631,
+      "step": 62030
+    },
+    {
+      "epoch": 400.26,
+      "learning_rate": 3.331899641577061e-05,
+      "loss": 0.2761,
+      "step": 62040
+    },
+    {
+      "epoch": 400.32,
+      "learning_rate": 3.331541218637993e-05,
+      "loss": 0.3432,
+      "step": 62050
+    },
+    {
+      "epoch": 400.39,
+      "learning_rate": 3.331182795698925e-05,
+      "loss": 0.2941,
+      "step": 62060
+    },
+    {
+      "epoch": 400.45,
+      "learning_rate": 3.330824372759857e-05,
+      "loss": 0.2645,
+      "step": 62070
+    },
+    {
+      "epoch": 400.51,
+      "learning_rate": 3.330465949820789e-05,
+      "loss": 0.2415,
+      "step": 62080
+    },
+    {
+      "epoch": 400.58,
+      "learning_rate": 3.3301075268817206e-05,
+      "loss": 0.3067,
+      "step": 62090
+    },
+    {
+      "epoch": 400.64,
+      "learning_rate": 3.3297491039426524e-05,
+      "loss": 0.3097,
+      "step": 62100
+    },
+    {
+      "epoch": 400.71,
+      "learning_rate": 3.329390681003585e-05,
+      "loss": 0.328,
+      "step": 62110
+    },
+    {
+      "epoch": 400.77,
+      "learning_rate": 3.329032258064516e-05,
+      "loss": 0.2821,
+      "step": 62120
+    },
+    {
+      "epoch": 400.84,
+      "learning_rate": 3.328673835125448e-05,
+      "loss": 0.3133,
+      "step": 62130
+    },
+    {
+      "epoch": 400.9,
+      "learning_rate": 3.32831541218638e-05,
+      "loss": 0.2839,
+      "step": 62140
+    },
+    {
+      "epoch": 400.96,
+      "learning_rate": 3.327956989247312e-05,
+      "loss": 0.2908,
+      "step": 62150
+    },
+    {
+      "epoch": 401.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9093958139419556,
+      "eval_runtime": 0.8569,
+      "eval_samples_per_second": 234.556,
+      "eval_steps_per_second": 8.169,
+      "step": 62155
+    },
+    {
+      "epoch": 401.03,
+      "learning_rate": 3.327598566308244e-05,
+      "loss": 0.2361,
+      "step": 62160
+    },
+    {
+      "epoch": 401.1,
+      "learning_rate": 3.327240143369176e-05,
+      "loss": 0.2372,
+      "step": 62170
+    },
+    {
+      "epoch": 401.16,
+      "learning_rate": 3.3268817204301076e-05,
+      "loss": 0.3153,
+      "step": 62180
+    },
+    {
+      "epoch": 401.23,
+      "learning_rate": 3.32652329749104e-05,
+      "loss": 0.2768,
+      "step": 62190
+    },
+    {
+      "epoch": 401.29,
+      "learning_rate": 3.326164874551971e-05,
+      "loss": 0.3226,
+      "step": 62200
+    },
+    {
+      "epoch": 401.35,
+      "learning_rate": 3.325806451612903e-05,
+      "loss": 0.2907,
+      "step": 62210
+    },
+    {
+      "epoch": 401.42,
+      "learning_rate": 3.3254480286738356e-05,
+      "loss": 0.2874,
+      "step": 62220
+    },
+    {
+      "epoch": 401.48,
+      "learning_rate": 3.325089605734767e-05,
+      "loss": 0.273,
+      "step": 62230
+    },
+    {
+      "epoch": 401.55,
+      "learning_rate": 3.324731182795699e-05,
+      "loss": 0.2732,
+      "step": 62240
+    },
+    {
+      "epoch": 401.61,
+      "learning_rate": 3.324372759856631e-05,
+      "loss": 0.2695,
+      "step": 62250
+    },
+    {
+      "epoch": 401.68,
+      "learning_rate": 3.324014336917563e-05,
+      "loss": 0.274,
+      "step": 62260
+    },
+    {
+      "epoch": 401.74,
+      "learning_rate": 3.323655913978495e-05,
+      "loss": 0.2701,
+      "step": 62270
+    },
+    {
+      "epoch": 401.8,
+      "learning_rate": 3.3232974910394265e-05,
+      "loss": 0.2806,
+      "step": 62280
+    },
+    {
+      "epoch": 401.87,
+      "learning_rate": 3.322939068100358e-05,
+      "loss": 0.3023,
+      "step": 62290
+    },
+    {
+      "epoch": 401.93,
+      "learning_rate": 3.322580645161291e-05,
+      "loss": 0.2483,
+      "step": 62300
+    },
+    {
+      "epoch": 402.0,
+      "learning_rate": 3.322222222222222e-05,
+      "loss": 0.3078,
+      "step": 62310
+    },
+    {
+      "epoch": 402.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9292978048324585,
+      "eval_runtime": 0.8679,
+      "eval_samples_per_second": 231.604,
+      "eval_steps_per_second": 8.066,
+      "step": 62310
+    },
+    {
+      "epoch": 402.06,
+      "learning_rate": 3.3218637992831545e-05,
+      "loss": 0.3914,
+      "step": 62320
+    },
+    {
+      "epoch": 402.13,
+      "learning_rate": 3.321505376344086e-05,
+      "loss": 0.2614,
+      "step": 62330
+    },
+    {
+      "epoch": 402.19,
+      "learning_rate": 3.321146953405018e-05,
+      "loss": 0.2602,
+      "step": 62340
+    },
+    {
+      "epoch": 402.26,
+      "learning_rate": 3.32078853046595e-05,
+      "loss": 0.301,
+      "step": 62350
+    },
+    {
+      "epoch": 402.32,
+      "learning_rate": 3.320430107526882e-05,
+      "loss": 0.2924,
+      "step": 62360
+    },
+    {
+      "epoch": 402.39,
+      "learning_rate": 3.3200716845878136e-05,
+      "loss": 0.295,
+      "step": 62370
+    },
+    {
+      "epoch": 402.45,
+      "learning_rate": 3.319713261648746e-05,
+      "loss": 0.3166,
+      "step": 62380
+    },
+    {
+      "epoch": 402.51,
+      "learning_rate": 3.319354838709677e-05,
+      "loss": 0.1981,
+      "step": 62390
+    },
+    {
+      "epoch": 402.58,
+      "learning_rate": 3.31899641577061e-05,
+      "loss": 0.2734,
+      "step": 62400
+    },
+    {
+      "epoch": 402.64,
+      "learning_rate": 3.3186379928315415e-05,
+      "loss": 0.3253,
+      "step": 62410
+    },
+    {
+      "epoch": 402.71,
+      "learning_rate": 3.3182795698924733e-05,
+      "loss": 0.2836,
+      "step": 62420
+    },
+    {
+      "epoch": 402.77,
+      "learning_rate": 3.317921146953405e-05,
+      "loss": 0.2398,
+      "step": 62430
+    },
+    {
+      "epoch": 402.84,
+      "learning_rate": 3.317562724014337e-05,
+      "loss": 0.335,
+      "step": 62440
+    },
+    {
+      "epoch": 402.9,
+      "learning_rate": 3.317204301075269e-05,
+      "loss": 0.2821,
+      "step": 62450
+    },
+    {
+      "epoch": 402.96,
+      "learning_rate": 3.316845878136201e-05,
+      "loss": 0.3387,
+      "step": 62460
+    },
+    {
+      "epoch": 403.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9791634678840637,
+      "eval_runtime": 0.8747,
+      "eval_samples_per_second": 229.798,
+      "eval_steps_per_second": 8.003,
+      "step": 62465
+    },
+    {
+      "epoch": 403.03,
+      "learning_rate": 3.3164874551971324e-05,
+      "loss": 0.2915,
+      "step": 62470
+    },
+    {
+      "epoch": 403.1,
+      "learning_rate": 3.316129032258064e-05,
+      "loss": 0.2783,
+      "step": 62480
+    },
+    {
+      "epoch": 403.16,
+      "learning_rate": 3.315770609318997e-05,
+      "loss": 0.2809,
+      "step": 62490
+    },
+    {
+      "epoch": 403.23,
+      "learning_rate": 3.3154121863799286e-05,
+      "loss": 0.3473,
+      "step": 62500
+    },
+    {
+      "epoch": 403.29,
+      "learning_rate": 3.3150537634408604e-05,
+      "loss": 0.2322,
+      "step": 62510
+    },
+    {
+      "epoch": 403.35,
+      "learning_rate": 3.314695340501792e-05,
+      "loss": 0.3366,
+      "step": 62520
+    },
+    {
+      "epoch": 403.42,
+      "learning_rate": 3.314336917562724e-05,
+      "loss": 0.2425,
+      "step": 62530
+    },
+    {
+      "epoch": 403.48,
+      "learning_rate": 3.3139784946236565e-05,
+      "loss": 0.303,
+      "step": 62540
+    },
+    {
+      "epoch": 403.55,
+      "learning_rate": 3.313620071684588e-05,
+      "loss": 0.2559,
+      "step": 62550
+    },
+    {
+      "epoch": 403.61,
+      "learning_rate": 3.3132616487455195e-05,
+      "loss": 0.326,
+      "step": 62560
+    },
+    {
+      "epoch": 403.68,
+      "learning_rate": 3.312903225806452e-05,
+      "loss": 0.2747,
+      "step": 62570
+    },
+    {
+      "epoch": 403.74,
+      "learning_rate": 3.312544802867384e-05,
+      "loss": 0.2578,
+      "step": 62580
+    },
+    {
+      "epoch": 403.8,
+      "learning_rate": 3.3121863799283156e-05,
+      "loss": 0.2616,
+      "step": 62590
+    },
+    {
+      "epoch": 403.87,
+      "learning_rate": 3.3118279569892475e-05,
+      "loss": 0.2806,
+      "step": 62600
+    },
+    {
+      "epoch": 403.93,
+      "learning_rate": 3.311469534050179e-05,
+      "loss": 0.2672,
+      "step": 62610
+    },
+    {
+      "epoch": 404.0,
+      "learning_rate": 3.311111111111112e-05,
+      "loss": 0.2572,
+      "step": 62620
+    },
+    {
+      "epoch": 404.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8848350048065186,
+      "eval_runtime": 0.8442,
+      "eval_samples_per_second": 238.092,
+      "eval_steps_per_second": 8.292,
+      "step": 62620
+    },
+    {
+      "epoch": 404.06,
+      "learning_rate": 3.310752688172043e-05,
+      "loss": 0.28,
+      "step": 62630
+    },
+    {
+      "epoch": 404.13,
+      "learning_rate": 3.310394265232975e-05,
+      "loss": 0.3326,
+      "step": 62640
+    },
+    {
+      "epoch": 404.19,
+      "learning_rate": 3.310035842293907e-05,
+      "loss": 0.3042,
+      "step": 62650
+    },
+    {
+      "epoch": 404.26,
+      "learning_rate": 3.309677419354839e-05,
+      "loss": 0.2758,
+      "step": 62660
+    },
+    {
+      "epoch": 404.32,
+      "learning_rate": 3.309318996415771e-05,
+      "loss": 0.2919,
+      "step": 62670
+    },
+    {
+      "epoch": 404.39,
+      "learning_rate": 3.308960573476703e-05,
+      "loss": 0.2545,
+      "step": 62680
+    },
+    {
+      "epoch": 404.45,
+      "learning_rate": 3.3086021505376345e-05,
+      "loss": 0.2611,
+      "step": 62690
+    },
+    {
+      "epoch": 404.51,
+      "learning_rate": 3.308243727598567e-05,
+      "loss": 0.2958,
+      "step": 62700
+    },
+    {
+      "epoch": 404.58,
+      "learning_rate": 3.307885304659498e-05,
+      "loss": 0.2336,
+      "step": 62710
+    },
+    {
+      "epoch": 404.64,
+      "learning_rate": 3.30752688172043e-05,
+      "loss": 0.3075,
+      "step": 62720
+    },
+    {
+      "epoch": 404.71,
+      "learning_rate": 3.3071684587813625e-05,
+      "loss": 0.2823,
+      "step": 62730
+    },
+    {
+      "epoch": 404.77,
+      "learning_rate": 3.306810035842294e-05,
+      "loss": 0.2645,
+      "step": 62740
+    },
+    {
+      "epoch": 404.84,
+      "learning_rate": 3.306451612903226e-05,
+      "loss": 0.3209,
+      "step": 62750
+    },
+    {
+      "epoch": 404.9,
+      "learning_rate": 3.306093189964158e-05,
+      "loss": 0.2887,
+      "step": 62760
+    },
+    {
+      "epoch": 404.96,
+      "learning_rate": 3.30573476702509e-05,
+      "loss": 0.2321,
+      "step": 62770
+    },
+    {
+      "epoch": 405.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9789376258850098,
+      "eval_runtime": 0.869,
+      "eval_samples_per_second": 231.303,
+      "eval_steps_per_second": 8.055,
+      "step": 62775
+    },
+    {
+      "epoch": 405.03,
+      "learning_rate": 3.3053763440860216e-05,
+      "loss": 0.3148,
+      "step": 62780
+    },
+    {
+      "epoch": 405.1,
+      "learning_rate": 3.3050179211469534e-05,
+      "loss": 0.3363,
+      "step": 62790
+    },
+    {
+      "epoch": 405.16,
+      "learning_rate": 3.304659498207885e-05,
+      "loss": 0.2863,
+      "step": 62800
+    },
+    {
+      "epoch": 405.23,
+      "learning_rate": 3.304301075268818e-05,
+      "loss": 0.3109,
+      "step": 62810
+    },
+    {
+      "epoch": 405.29,
+      "learning_rate": 3.3039426523297495e-05,
+      "loss": 0.2399,
+      "step": 62820
+    },
+    {
+      "epoch": 405.35,
+      "learning_rate": 3.303584229390681e-05,
+      "loss": 0.2641,
+      "step": 62830
+    },
+    {
+      "epoch": 405.42,
+      "learning_rate": 3.303225806451613e-05,
+      "loss": 0.296,
+      "step": 62840
+    },
+    {
+      "epoch": 405.48,
+      "learning_rate": 3.302867383512545e-05,
+      "loss": 0.2941,
+      "step": 62850
+    },
+    {
+      "epoch": 405.55,
+      "learning_rate": 3.302508960573477e-05,
+      "loss": 0.2845,
+      "step": 62860
+    },
+    {
+      "epoch": 405.61,
+      "learning_rate": 3.3021505376344087e-05,
+      "loss": 0.2783,
+      "step": 62870
+    },
+    {
+      "epoch": 405.68,
+      "learning_rate": 3.3017921146953405e-05,
+      "loss": 0.3198,
+      "step": 62880
+    },
+    {
+      "epoch": 405.74,
+      "learning_rate": 3.301433691756273e-05,
+      "loss": 0.2872,
+      "step": 62890
+    },
+    {
+      "epoch": 405.8,
+      "learning_rate": 3.301075268817204e-05,
+      "loss": 0.2723,
+      "step": 62900
+    },
+    {
+      "epoch": 405.87,
+      "learning_rate": 3.300716845878136e-05,
+      "loss": 0.2634,
+      "step": 62910
+    },
+    {
+      "epoch": 405.93,
+      "learning_rate": 3.3003584229390684e-05,
+      "loss": 0.3093,
+      "step": 62920
+    },
+    {
+      "epoch": 406.0,
+      "learning_rate": 3.3e-05,
+      "loss": 0.2771,
+      "step": 62930
+    },
+    {
+      "epoch": 406.0,
+      "eval_accuracy": 0.8059701492537313,
+      "eval_loss": 0.9011625051498413,
+      "eval_runtime": 0.8461,
+      "eval_samples_per_second": 237.547,
+      "eval_steps_per_second": 8.273,
+      "step": 62930
+    },
+    {
+      "epoch": 406.06,
+      "learning_rate": 3.299641577060932e-05,
+      "loss": 0.3444,
+      "step": 62940
+    },
+    {
+      "epoch": 406.13,
+      "learning_rate": 3.299283154121864e-05,
+      "loss": 0.3146,
+      "step": 62950
+    },
+    {
+      "epoch": 406.19,
+      "learning_rate": 3.298924731182796e-05,
+      "loss": 0.2559,
+      "step": 62960
+    },
+    {
+      "epoch": 406.26,
+      "learning_rate": 3.298566308243728e-05,
+      "loss": 0.2754,
+      "step": 62970
+    },
+    {
+      "epoch": 406.32,
+      "learning_rate": 3.2982078853046594e-05,
+      "loss": 0.2885,
+      "step": 62980
+    },
+    {
+      "epoch": 406.39,
+      "learning_rate": 3.297849462365591e-05,
+      "loss": 0.2788,
+      "step": 62990
+    },
+    {
+      "epoch": 406.45,
+      "learning_rate": 3.297491039426524e-05,
+      "loss": 0.2702,
+      "step": 63000
+    },
+    {
+      "epoch": 406.51,
+      "learning_rate": 3.2971326164874555e-05,
+      "loss": 0.2557,
+      "step": 63010
+    },
+    {
+      "epoch": 406.58,
+      "learning_rate": 3.296774193548387e-05,
+      "loss": 0.2901,
+      "step": 63020
+    },
+    {
+      "epoch": 406.64,
+      "learning_rate": 3.296415770609319e-05,
+      "loss": 0.3147,
+      "step": 63030
+    },
+    {
+      "epoch": 406.71,
+      "learning_rate": 3.296057347670251e-05,
+      "loss": 0.278,
+      "step": 63040
+    },
+    {
+      "epoch": 406.77,
+      "learning_rate": 3.2956989247311835e-05,
+      "loss": 0.2854,
+      "step": 63050
+    },
+    {
+      "epoch": 406.84,
+      "learning_rate": 3.2953405017921146e-05,
+      "loss": 0.2765,
+      "step": 63060
+    },
+    {
+      "epoch": 406.9,
+      "learning_rate": 3.2949820788530464e-05,
+      "loss": 0.2951,
+      "step": 63070
+    },
+    {
+      "epoch": 406.96,
+      "learning_rate": 3.294623655913979e-05,
+      "loss": 0.279,
+      "step": 63080
+    },
+    {
+      "epoch": 407.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9293736815452576,
+      "eval_runtime": 0.8733,
+      "eval_samples_per_second": 230.166,
+      "eval_steps_per_second": 8.016,
+      "step": 63085
+    },
+    {
+      "epoch": 407.03,
+      "learning_rate": 3.294265232974911e-05,
+      "loss": 0.2953,
+      "step": 63090
+    },
+    {
+      "epoch": 407.1,
+      "learning_rate": 3.2939068100358426e-05,
+      "loss": 0.2795,
+      "step": 63100
+    },
+    {
+      "epoch": 407.16,
+      "learning_rate": 3.2935483870967744e-05,
+      "loss": 0.2338,
+      "step": 63110
+    },
+    {
+      "epoch": 407.23,
+      "learning_rate": 3.293189964157706e-05,
+      "loss": 0.2361,
+      "step": 63120
+    },
+    {
+      "epoch": 407.29,
+      "learning_rate": 3.292831541218638e-05,
+      "loss": 0.3071,
+      "step": 63130
+    },
+    {
+      "epoch": 407.35,
+      "learning_rate": 3.29247311827957e-05,
+      "loss": 0.2614,
+      "step": 63140
+    },
+    {
+      "epoch": 407.42,
+      "learning_rate": 3.2921146953405017e-05,
+      "loss": 0.3078,
+      "step": 63150
+    },
+    {
+      "epoch": 407.48,
+      "learning_rate": 3.291756272401434e-05,
+      "loss": 0.2675,
+      "step": 63160
+    },
+    {
+      "epoch": 407.55,
+      "learning_rate": 3.291397849462366e-05,
+      "loss": 0.3069,
+      "step": 63170
+    },
+    {
+      "epoch": 407.61,
+      "learning_rate": 3.291039426523297e-05,
+      "loss": 0.2672,
+      "step": 63180
+    },
+    {
+      "epoch": 407.68,
+      "learning_rate": 3.2906810035842296e-05,
+      "loss": 0.2769,
+      "step": 63190
+    },
+    {
+      "epoch": 407.74,
+      "learning_rate": 3.2903225806451614e-05,
+      "loss": 0.2989,
+      "step": 63200
+    },
+    {
+      "epoch": 407.8,
+      "learning_rate": 3.289964157706093e-05,
+      "loss": 0.2974,
+      "step": 63210
+    },
+    {
+      "epoch": 407.87,
+      "learning_rate": 3.289605734767025e-05,
+      "loss": 0.2354,
+      "step": 63220
+    },
+    {
+      "epoch": 407.93,
+      "learning_rate": 3.289247311827957e-05,
+      "loss": 0.2896,
+      "step": 63230
+    },
+    {
+      "epoch": 408.0,
+      "learning_rate": 3.2888888888888894e-05,
+      "loss": 0.2602,
+      "step": 63240
+    },
+    {
+      "epoch": 408.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9258705973625183,
+      "eval_runtime": 0.8627,
+      "eval_samples_per_second": 232.981,
+      "eval_steps_per_second": 8.114,
+      "step": 63240
+    },
+    {
+      "epoch": 408.06,
+      "learning_rate": 3.288530465949821e-05,
+      "loss": 0.3443,
+      "step": 63250
+    },
+    {
+      "epoch": 408.13,
+      "learning_rate": 3.2881720430107524e-05,
+      "loss": 0.3077,
+      "step": 63260
+    },
+    {
+      "epoch": 408.19,
+      "learning_rate": 3.287813620071685e-05,
+      "loss": 0.3066,
+      "step": 63270
+    },
+    {
+      "epoch": 408.26,
+      "learning_rate": 3.287455197132617e-05,
+      "loss": 0.278,
+      "step": 63280
+    },
+    {
+      "epoch": 408.32,
+      "learning_rate": 3.2870967741935485e-05,
+      "loss": 0.284,
+      "step": 63290
+    },
+    {
+      "epoch": 408.39,
+      "learning_rate": 3.28673835125448e-05,
+      "loss": 0.247,
+      "step": 63300
+    },
+    {
+      "epoch": 408.45,
+      "learning_rate": 3.286379928315412e-05,
+      "loss": 0.3318,
+      "step": 63310
+    },
+    {
+      "epoch": 408.51,
+      "learning_rate": 3.2860215053763446e-05,
+      "loss": 0.2683,
+      "step": 63320
+    },
+    {
+      "epoch": 408.58,
+      "learning_rate": 3.2856630824372765e-05,
+      "loss": 0.3043,
+      "step": 63330
+    },
+    {
+      "epoch": 408.64,
+      "learning_rate": 3.2853046594982076e-05,
+      "loss": 0.2631,
+      "step": 63340
+    },
+    {
+      "epoch": 408.71,
+      "learning_rate": 3.28494623655914e-05,
+      "loss": 0.2589,
+      "step": 63350
+    },
+    {
+      "epoch": 408.77,
+      "learning_rate": 3.284587813620072e-05,
+      "loss": 0.3359,
+      "step": 63360
+    },
+    {
+      "epoch": 408.84,
+      "learning_rate": 3.284229390681004e-05,
+      "loss": 0.251,
+      "step": 63370
+    },
+    {
+      "epoch": 408.9,
+      "learning_rate": 3.2838709677419356e-05,
+      "loss": 0.3309,
+      "step": 63380
+    },
+    {
+      "epoch": 408.96,
+      "learning_rate": 3.2835125448028674e-05,
+      "loss": 0.2851,
+      "step": 63390
+    },
+    {
+      "epoch": 409.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9009048938751221,
+      "eval_runtime": 0.8625,
+      "eval_samples_per_second": 233.039,
+      "eval_steps_per_second": 8.116,
+      "step": 63395
+    },
+    {
+      "epoch": 409.03,
+      "learning_rate": 3.2831541218638e-05,
+      "loss": 0.2925,
+      "step": 63400
+    },
+    {
+      "epoch": 409.1,
+      "learning_rate": 3.282795698924732e-05,
+      "loss": 0.2702,
+      "step": 63410
+    },
+    {
+      "epoch": 409.16,
+      "learning_rate": 3.282437275985663e-05,
+      "loss": 0.2707,
+      "step": 63420
+    },
+    {
+      "epoch": 409.23,
+      "learning_rate": 3.2820788530465953e-05,
+      "loss": 0.2259,
+      "step": 63430
+    },
+    {
+      "epoch": 409.29,
+      "learning_rate": 3.281720430107527e-05,
+      "loss": 0.3029,
+      "step": 63440
+    },
+    {
+      "epoch": 409.35,
+      "learning_rate": 3.281362007168459e-05,
+      "loss": 0.2934,
+      "step": 63450
+    },
+    {
+      "epoch": 409.42,
+      "learning_rate": 3.281003584229391e-05,
+      "loss": 0.2434,
+      "step": 63460
+    },
+    {
+      "epoch": 409.48,
+      "learning_rate": 3.2806451612903226e-05,
+      "loss": 0.3048,
+      "step": 63470
+    },
+    {
+      "epoch": 409.55,
+      "learning_rate": 3.2802867383512544e-05,
+      "loss": 0.2593,
+      "step": 63480
+    },
+    {
+      "epoch": 409.61,
+      "learning_rate": 3.279928315412187e-05,
+      "loss": 0.3115,
+      "step": 63490
+    },
+    {
+      "epoch": 409.68,
+      "learning_rate": 3.279569892473118e-05,
+      "loss": 0.2749,
+      "step": 63500
+    },
+    {
+      "epoch": 409.74,
+      "learning_rate": 3.2792114695340506e-05,
+      "loss": 0.3373,
+      "step": 63510
+    },
+    {
+      "epoch": 409.8,
+      "learning_rate": 3.2788530465949824e-05,
+      "loss": 0.3258,
+      "step": 63520
+    },
+    {
+      "epoch": 409.87,
+      "learning_rate": 3.2784946236559135e-05,
+      "loss": 0.2822,
+      "step": 63530
+    },
+    {
+      "epoch": 409.93,
+      "learning_rate": 3.278136200716846e-05,
+      "loss": 0.3317,
+      "step": 63540
+    },
+    {
+      "epoch": 410.0,
+      "learning_rate": 3.277777777777778e-05,
+      "loss": 0.255,
+      "step": 63550
+    },
+    {
+      "epoch": 410.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9668420553207397,
+      "eval_runtime": 0.849,
+      "eval_samples_per_second": 236.736,
+      "eval_steps_per_second": 8.245,
+      "step": 63550
+    },
+    {
+      "epoch": 410.06,
+      "learning_rate": 3.27741935483871e-05,
+      "loss": 0.3061,
+      "step": 63560
+    },
+    {
+      "epoch": 410.13,
+      "learning_rate": 3.2770609318996415e-05,
+      "loss": 0.2943,
+      "step": 63570
+    },
+    {
+      "epoch": 410.19,
+      "learning_rate": 3.276702508960573e-05,
+      "loss": 0.2145,
+      "step": 63580
+    },
+    {
+      "epoch": 410.26,
+      "learning_rate": 3.276344086021506e-05,
+      "loss": 0.2535,
+      "step": 63590
+    },
+    {
+      "epoch": 410.32,
+      "learning_rate": 3.2759856630824376e-05,
+      "loss": 0.2735,
+      "step": 63600
+    },
+    {
+      "epoch": 410.39,
+      "learning_rate": 3.275627240143369e-05,
+      "loss": 0.2655,
+      "step": 63610
+    },
+    {
+      "epoch": 410.45,
+      "learning_rate": 3.275268817204301e-05,
+      "loss": 0.3135,
+      "step": 63620
+    },
+    {
+      "epoch": 410.51,
+      "learning_rate": 3.274910394265233e-05,
+      "loss": 0.2848,
+      "step": 63630
+    },
+    {
+      "epoch": 410.58,
+      "learning_rate": 3.274551971326165e-05,
+      "loss": 0.2941,
+      "step": 63640
+    },
+    {
+      "epoch": 410.64,
+      "learning_rate": 3.274193548387097e-05,
+      "loss": 0.2845,
+      "step": 63650
+    },
+    {
+      "epoch": 410.71,
+      "learning_rate": 3.2738351254480286e-05,
+      "loss": 0.2847,
+      "step": 63660
+    },
+    {
+      "epoch": 410.77,
+      "learning_rate": 3.273476702508961e-05,
+      "loss": 0.2748,
+      "step": 63670
+    },
+    {
+      "epoch": 410.84,
+      "learning_rate": 3.273118279569893e-05,
+      "loss": 0.2609,
+      "step": 63680
+    },
+    {
+      "epoch": 410.9,
+      "learning_rate": 3.272759856630824e-05,
+      "loss": 0.2635,
+      "step": 63690
+    },
+    {
+      "epoch": 410.96,
+      "learning_rate": 3.2724014336917565e-05,
+      "loss": 0.3086,
+      "step": 63700
+    },
+    {
+      "epoch": 411.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.9253119230270386,
+      "eval_runtime": 0.8675,
+      "eval_samples_per_second": 231.702,
+      "eval_steps_per_second": 8.069,
+      "step": 63705
+    },
+    {
+      "epoch": 411.03,
+      "learning_rate": 3.2720430107526884e-05,
+      "loss": 0.3033,
+      "step": 63710
+    },
+    {
+      "epoch": 411.1,
+      "learning_rate": 3.27168458781362e-05,
+      "loss": 0.2724,
+      "step": 63720
+    },
+    {
+      "epoch": 411.16,
+      "learning_rate": 3.271326164874552e-05,
+      "loss": 0.3078,
+      "step": 63730
+    },
+    {
+      "epoch": 411.23,
+      "learning_rate": 3.270967741935484e-05,
+      "loss": 0.3251,
+      "step": 63740
+    },
+    {
+      "epoch": 411.29,
+      "learning_rate": 3.270609318996416e-05,
+      "loss": 0.2557,
+      "step": 63750
+    },
+    {
+      "epoch": 411.35,
+      "learning_rate": 3.270250896057348e-05,
+      "loss": 0.2709,
+      "step": 63760
+    },
+    {
+      "epoch": 411.42,
+      "learning_rate": 3.269892473118279e-05,
+      "loss": 0.2784,
+      "step": 63770
+    },
+    {
+      "epoch": 411.48,
+      "learning_rate": 3.269534050179212e-05,
+      "loss": 0.2848,
+      "step": 63780
+    },
+    {
+      "epoch": 411.55,
+      "learning_rate": 3.2691756272401436e-05,
+      "loss": 0.2107,
+      "step": 63790
+    },
+    {
+      "epoch": 411.61,
+      "learning_rate": 3.2688172043010754e-05,
+      "loss": 0.2846,
+      "step": 63800
+    },
+    {
+      "epoch": 411.68,
+      "learning_rate": 3.268458781362007e-05,
+      "loss": 0.2505,
+      "step": 63810
+    },
+    {
+      "epoch": 411.74,
+      "learning_rate": 3.268100358422939e-05,
+      "loss": 0.2811,
+      "step": 63820
+    },
+    {
+      "epoch": 411.8,
+      "learning_rate": 3.267741935483871e-05,
+      "loss": 0.2976,
+      "step": 63830
+    },
+    {
+      "epoch": 411.87,
+      "learning_rate": 3.2673835125448034e-05,
+      "loss": 0.279,
+      "step": 63840
+    },
+    {
+      "epoch": 411.93,
+      "learning_rate": 3.2670250896057345e-05,
+      "loss": 0.2623,
+      "step": 63850
+    },
+    {
+      "epoch": 412.0,
+      "learning_rate": 3.266666666666667e-05,
+      "loss": 0.2736,
+      "step": 63860
+    },
+    {
+      "epoch": 412.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9076100587844849,
+      "eval_runtime": 0.9298,
+      "eval_samples_per_second": 216.178,
+      "eval_steps_per_second": 7.529,
+      "step": 63860
+    },
+    {
+      "epoch": 412.06,
+      "learning_rate": 3.266308243727599e-05,
+      "loss": 0.2935,
+      "step": 63870
+    },
+    {
+      "epoch": 412.13,
+      "learning_rate": 3.2659498207885307e-05,
+      "loss": 0.2863,
+      "step": 63880
+    },
+    {
+      "epoch": 412.19,
+      "learning_rate": 3.2655913978494625e-05,
+      "loss": 0.2891,
+      "step": 63890
+    },
+    {
+      "epoch": 412.26,
+      "learning_rate": 3.265232974910394e-05,
+      "loss": 0.2927,
+      "step": 63900
+    },
+    {
+      "epoch": 412.32,
+      "learning_rate": 3.264874551971326e-05,
+      "loss": 0.2709,
+      "step": 63910
+    },
+    {
+      "epoch": 412.39,
+      "learning_rate": 3.2645161290322586e-05,
+      "loss": 0.2875,
+      "step": 63920
+    },
+    {
+      "epoch": 412.45,
+      "learning_rate": 3.26415770609319e-05,
+      "loss": 0.2785,
+      "step": 63930
+    },
+    {
+      "epoch": 412.51,
+      "learning_rate": 3.263799283154122e-05,
+      "loss": 0.2671,
+      "step": 63940
+    },
+    {
+      "epoch": 412.58,
+      "learning_rate": 3.263440860215054e-05,
+      "loss": 0.2527,
+      "step": 63950
+    },
+    {
+      "epoch": 412.64,
+      "learning_rate": 3.263082437275986e-05,
+      "loss": 0.2932,
+      "step": 63960
+    },
+    {
+      "epoch": 412.71,
+      "learning_rate": 3.262724014336918e-05,
+      "loss": 0.2874,
+      "step": 63970
+    },
+    {
+      "epoch": 412.77,
+      "learning_rate": 3.2623655913978495e-05,
+      "loss": 0.2807,
+      "step": 63980
+    },
+    {
+      "epoch": 412.84,
+      "learning_rate": 3.2620071684587814e-05,
+      "loss": 0.3106,
+      "step": 63990
+    },
+    {
+      "epoch": 412.9,
+      "learning_rate": 3.261648745519714e-05,
+      "loss": 0.2595,
+      "step": 64000
+    },
+    {
+      "epoch": 412.96,
+      "learning_rate": 3.261290322580645e-05,
+      "loss": 0.3069,
+      "step": 64010
+    },
+    {
+      "epoch": 413.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9553207755088806,
+      "eval_runtime": 0.8737,
+      "eval_samples_per_second": 230.059,
+      "eval_steps_per_second": 8.012,
+      "step": 64015
+    },
+    {
+      "epoch": 413.03,
+      "learning_rate": 3.2609318996415775e-05,
+      "loss": 0.2625,
+      "step": 64020
+    },
+    {
+      "epoch": 413.1,
+      "learning_rate": 3.260573476702509e-05,
+      "loss": 0.2663,
+      "step": 64030
+    },
+    {
+      "epoch": 413.16,
+      "learning_rate": 3.260215053763441e-05,
+      "loss": 0.2222,
+      "step": 64040
+    },
+    {
+      "epoch": 413.23,
+      "learning_rate": 3.259856630824373e-05,
+      "loss": 0.3058,
+      "step": 64050
+    },
+    {
+      "epoch": 413.29,
+      "learning_rate": 3.259498207885305e-05,
+      "loss": 0.3301,
+      "step": 64060
+    },
+    {
+      "epoch": 413.35,
+      "learning_rate": 3.2591397849462366e-05,
+      "loss": 0.2459,
+      "step": 64070
+    },
+    {
+      "epoch": 413.42,
+      "learning_rate": 3.258781362007169e-05,
+      "loss": 0.2235,
+      "step": 64080
+    },
+    {
+      "epoch": 413.48,
+      "learning_rate": 3.2584229390681e-05,
+      "loss": 0.2662,
+      "step": 64090
+    },
+    {
+      "epoch": 413.55,
+      "learning_rate": 3.258064516129033e-05,
+      "loss": 0.3492,
+      "step": 64100
+    },
+    {
+      "epoch": 413.61,
+      "learning_rate": 3.2577060931899646e-05,
+      "loss": 0.2536,
+      "step": 64110
+    },
+    {
+      "epoch": 413.68,
+      "learning_rate": 3.2573476702508964e-05,
+      "loss": 0.2748,
+      "step": 64120
+    },
+    {
+      "epoch": 413.74,
+      "learning_rate": 3.256989247311828e-05,
+      "loss": 0.3009,
+      "step": 64130
+    },
+    {
+      "epoch": 413.8,
+      "learning_rate": 3.25663082437276e-05,
+      "loss": 0.2497,
+      "step": 64140
+    },
+    {
+      "epoch": 413.87,
+      "learning_rate": 3.256272401433692e-05,
+      "loss": 0.3243,
+      "step": 64150
+    },
+    {
+      "epoch": 413.93,
+      "learning_rate": 3.2559139784946237e-05,
+      "loss": 0.3319,
+      "step": 64160
+    },
+    {
+      "epoch": 414.0,
+      "learning_rate": 3.2555555555555555e-05,
+      "loss": 0.3004,
+      "step": 64170
+    },
+    {
+      "epoch": 414.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9008538722991943,
+      "eval_runtime": 0.9331,
+      "eval_samples_per_second": 215.408,
+      "eval_steps_per_second": 7.502,
+      "step": 64170
+    },
+    {
+      "epoch": 414.06,
+      "learning_rate": 3.255197132616487e-05,
+      "loss": 0.364,
+      "step": 64180
+    },
+    {
+      "epoch": 414.13,
+      "learning_rate": 3.25483870967742e-05,
+      "loss": 0.279,
+      "step": 64190
+    },
+    {
+      "epoch": 414.19,
+      "learning_rate": 3.254480286738351e-05,
+      "loss": 0.2802,
+      "step": 64200
+    },
+    {
+      "epoch": 414.26,
+      "learning_rate": 3.2541218637992834e-05,
+      "loss": 0.2298,
+      "step": 64210
+    },
+    {
+      "epoch": 414.32,
+      "learning_rate": 3.253763440860215e-05,
+      "loss": 0.2529,
+      "step": 64220
+    },
+    {
+      "epoch": 414.39,
+      "learning_rate": 3.253405017921147e-05,
+      "loss": 0.2646,
+      "step": 64230
+    },
+    {
+      "epoch": 414.45,
+      "learning_rate": 3.253046594982079e-05,
+      "loss": 0.2929,
+      "step": 64240
+    },
+    {
+      "epoch": 414.51,
+      "learning_rate": 3.252688172043011e-05,
+      "loss": 0.3237,
+      "step": 64250
+    },
+    {
+      "epoch": 414.58,
+      "learning_rate": 3.2523297491039425e-05,
+      "loss": 0.298,
+      "step": 64260
+    },
+    {
+      "epoch": 414.64,
+      "learning_rate": 3.251971326164875e-05,
+      "loss": 0.2558,
+      "step": 64270
+    },
+    {
+      "epoch": 414.71,
+      "learning_rate": 3.251612903225806e-05,
+      "loss": 0.2963,
+      "step": 64280
+    },
+    {
+      "epoch": 414.77,
+      "learning_rate": 3.251254480286739e-05,
+      "loss": 0.2554,
+      "step": 64290
+    },
+    {
+      "epoch": 414.84,
+      "learning_rate": 3.2508960573476705e-05,
+      "loss": 0.2962,
+      "step": 64300
+    },
+    {
+      "epoch": 414.9,
+      "learning_rate": 3.250537634408602e-05,
+      "loss": 0.266,
+      "step": 64310
+    },
+    {
+      "epoch": 414.96,
+      "learning_rate": 3.250179211469534e-05,
+      "loss": 0.2864,
+      "step": 64320
+    },
+    {
+      "epoch": 415.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9696764349937439,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 233.128,
+      "eval_steps_per_second": 8.119,
+      "step": 64325
+    },
+    {
+      "epoch": 415.03,
+      "learning_rate": 3.249820788530466e-05,
+      "loss": 0.2953,
+      "step": 64330
+    },
+    {
+      "epoch": 415.1,
+      "learning_rate": 3.249462365591398e-05,
+      "loss": 0.2297,
+      "step": 64340
+    },
+    {
+      "epoch": 415.16,
+      "learning_rate": 3.24910394265233e-05,
+      "loss": 0.2742,
+      "step": 64350
+    },
+    {
+      "epoch": 415.23,
+      "learning_rate": 3.2487455197132614e-05,
+      "loss": 0.2691,
+      "step": 64360
+    },
+    {
+      "epoch": 415.29,
+      "learning_rate": 3.248387096774194e-05,
+      "loss": 0.2533,
+      "step": 64370
+    },
+    {
+      "epoch": 415.35,
+      "learning_rate": 3.248028673835126e-05,
+      "loss": 0.2999,
+      "step": 64380
+    },
+    {
+      "epoch": 415.42,
+      "learning_rate": 3.2476702508960576e-05,
+      "loss": 0.2561,
+      "step": 64390
+    },
+    {
+      "epoch": 415.48,
+      "learning_rate": 3.2473118279569894e-05,
+      "loss": 0.246,
+      "step": 64400
+    },
+    {
+      "epoch": 415.55,
+      "learning_rate": 3.246953405017921e-05,
+      "loss": 0.2715,
+      "step": 64410
+    },
+    {
+      "epoch": 415.61,
+      "learning_rate": 3.246594982078853e-05,
+      "loss": 0.3382,
+      "step": 64420
+    },
+    {
+      "epoch": 415.68,
+      "learning_rate": 3.2462365591397855e-05,
+      "loss": 0.2954,
+      "step": 64430
+    },
+    {
+      "epoch": 415.74,
+      "learning_rate": 3.245878136200717e-05,
+      "loss": 0.2726,
+      "step": 64440
+    },
+    {
+      "epoch": 415.8,
+      "learning_rate": 3.245519713261649e-05,
+      "loss": 0.2648,
+      "step": 64450
+    },
+    {
+      "epoch": 415.87,
+      "learning_rate": 3.245161290322581e-05,
+      "loss": 0.2635,
+      "step": 64460
+    },
+    {
+      "epoch": 415.93,
+      "learning_rate": 3.244802867383513e-05,
+      "loss": 0.3231,
+      "step": 64470
+    },
+    {
+      "epoch": 416.0,
+      "learning_rate": 3.2444444444444446e-05,
+      "loss": 0.2314,
+      "step": 64480
+    },
+    {
+      "epoch": 416.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9479461312294006,
+      "eval_runtime": 0.8935,
+      "eval_samples_per_second": 224.964,
+      "eval_steps_per_second": 7.835,
+      "step": 64480
+    },
+    {
+      "epoch": 416.06,
+      "learning_rate": 3.2440860215053764e-05,
+      "loss": 0.2629,
+      "step": 64490
+    },
+    {
+      "epoch": 416.13,
+      "learning_rate": 3.243727598566308e-05,
+      "loss": 0.3175,
+      "step": 64500
+    },
+    {
+      "epoch": 416.19,
+      "learning_rate": 3.243369175627241e-05,
+      "loss": 0.267,
+      "step": 64510
+    },
+    {
+      "epoch": 416.26,
+      "learning_rate": 3.243010752688172e-05,
+      "loss": 0.2702,
+      "step": 64520
+    },
+    {
+      "epoch": 416.32,
+      "learning_rate": 3.242652329749104e-05,
+      "loss": 0.2919,
+      "step": 64530
+    },
+    {
+      "epoch": 416.39,
+      "learning_rate": 3.242293906810036e-05,
+      "loss": 0.3126,
+      "step": 64540
+    },
+    {
+      "epoch": 416.45,
+      "learning_rate": 3.241935483870968e-05,
+      "loss": 0.2588,
+      "step": 64550
+    },
+    {
+      "epoch": 416.51,
+      "learning_rate": 3.2415770609319e-05,
+      "loss": 0.2783,
+      "step": 64560
+    },
+    {
+      "epoch": 416.58,
+      "learning_rate": 3.241218637992832e-05,
+      "loss": 0.2776,
+      "step": 64570
+    },
+    {
+      "epoch": 416.64,
+      "learning_rate": 3.2408602150537635e-05,
+      "loss": 0.2931,
+      "step": 64580
+    },
+    {
+      "epoch": 416.71,
+      "learning_rate": 3.240501792114696e-05,
+      "loss": 0.291,
+      "step": 64590
+    },
+    {
+      "epoch": 416.77,
+      "learning_rate": 3.240143369175627e-05,
+      "loss": 0.2915,
+      "step": 64600
+    },
+    {
+      "epoch": 416.84,
+      "learning_rate": 3.239784946236559e-05,
+      "loss": 0.3483,
+      "step": 64610
+    },
+    {
+      "epoch": 416.9,
+      "learning_rate": 3.2394265232974915e-05,
+      "loss": 0.3265,
+      "step": 64620
+    },
+    {
+      "epoch": 416.96,
+      "learning_rate": 3.239068100358423e-05,
+      "loss": 0.2409,
+      "step": 64630
+    },
+    {
+      "epoch": 417.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9820170402526855,
+      "eval_runtime": 0.8883,
+      "eval_samples_per_second": 226.266,
+      "eval_steps_per_second": 7.88,
+      "step": 64635
+    },
+    {
+      "epoch": 417.03,
+      "learning_rate": 3.238709677419355e-05,
+      "loss": 0.2292,
+      "step": 64640
+    },
+    {
+      "epoch": 417.1,
+      "learning_rate": 3.238351254480287e-05,
+      "loss": 0.2712,
+      "step": 64650
+    },
+    {
+      "epoch": 417.16,
+      "learning_rate": 3.237992831541219e-05,
+      "loss": 0.3151,
+      "step": 64660
+    },
+    {
+      "epoch": 417.23,
+      "learning_rate": 3.237634408602151e-05,
+      "loss": 0.2956,
+      "step": 64670
+    },
+    {
+      "epoch": 417.29,
+      "learning_rate": 3.2372759856630824e-05,
+      "loss": 0.2624,
+      "step": 64680
+    },
+    {
+      "epoch": 417.35,
+      "learning_rate": 3.236917562724014e-05,
+      "loss": 0.2427,
+      "step": 64690
+    },
+    {
+      "epoch": 417.42,
+      "learning_rate": 3.236559139784947e-05,
+      "loss": 0.2544,
+      "step": 64700
+    },
+    {
+      "epoch": 417.48,
+      "learning_rate": 3.2362007168458785e-05,
+      "loss": 0.2344,
+      "step": 64710
+    },
+    {
+      "epoch": 417.55,
+      "learning_rate": 3.2358422939068104e-05,
+      "loss": 0.2375,
+      "step": 64720
+    },
+    {
+      "epoch": 417.61,
+      "learning_rate": 3.235483870967742e-05,
+      "loss": 0.2798,
+      "step": 64730
+    },
+    {
+      "epoch": 417.68,
+      "learning_rate": 3.235125448028674e-05,
+      "loss": 0.2852,
+      "step": 64740
+    },
+    {
+      "epoch": 417.74,
+      "learning_rate": 3.2347670250896065e-05,
+      "loss": 0.28,
+      "step": 64750
+    },
+    {
+      "epoch": 417.8,
+      "learning_rate": 3.2344086021505376e-05,
+      "loss": 0.3291,
+      "step": 64760
+    },
+    {
+      "epoch": 417.87,
+      "learning_rate": 3.2340501792114695e-05,
+      "loss": 0.2698,
+      "step": 64770
+    },
+    {
+      "epoch": 417.93,
+      "learning_rate": 3.233691756272402e-05,
+      "loss": 0.272,
+      "step": 64780
+    },
+    {
+      "epoch": 418.0,
+      "learning_rate": 3.233333333333333e-05,
+      "loss": 0.2648,
+      "step": 64790
+    },
+    {
+      "epoch": 418.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9679014086723328,
+      "eval_runtime": 0.8718,
+      "eval_samples_per_second": 230.564,
+      "eval_steps_per_second": 8.03,
+      "step": 64790
+    },
+    {
+      "epoch": 418.06,
+      "learning_rate": 3.2329749103942656e-05,
+      "loss": 0.2672,
+      "step": 64800
+    },
+    {
+      "epoch": 418.13,
+      "learning_rate": 3.2326164874551974e-05,
+      "loss": 0.3237,
+      "step": 64810
+    },
+    {
+      "epoch": 418.19,
+      "learning_rate": 3.232258064516129e-05,
+      "loss": 0.2912,
+      "step": 64820
+    },
+    {
+      "epoch": 418.26,
+      "learning_rate": 3.231899641577061e-05,
+      "loss": 0.2634,
+      "step": 64830
+    },
+    {
+      "epoch": 418.32,
+      "learning_rate": 3.231541218637993e-05,
+      "loss": 0.2937,
+      "step": 64840
+    },
+    {
+      "epoch": 418.39,
+      "learning_rate": 3.231182795698925e-05,
+      "loss": 0.3208,
+      "step": 64850
+    },
+    {
+      "epoch": 418.45,
+      "learning_rate": 3.230824372759857e-05,
+      "loss": 0.3274,
+      "step": 64860
+    },
+    {
+      "epoch": 418.51,
+      "learning_rate": 3.230465949820788e-05,
+      "loss": 0.3153,
+      "step": 64870
+    },
+    {
+      "epoch": 418.58,
+      "learning_rate": 3.23010752688172e-05,
+      "loss": 0.3034,
+      "step": 64880
+    },
+    {
+      "epoch": 418.64,
+      "learning_rate": 3.2297491039426527e-05,
+      "loss": 0.2674,
+      "step": 64890
+    },
+    {
+      "epoch": 418.71,
+      "learning_rate": 3.2293906810035845e-05,
+      "loss": 0.2951,
+      "step": 64900
+    },
+    {
+      "epoch": 418.77,
+      "learning_rate": 3.229032258064516e-05,
+      "loss": 0.3054,
+      "step": 64910
+    },
+    {
+      "epoch": 418.84,
+      "learning_rate": 3.228673835125448e-05,
+      "loss": 0.3216,
+      "step": 64920
+    },
+    {
+      "epoch": 418.9,
+      "learning_rate": 3.22831541218638e-05,
+      "loss": 0.3194,
+      "step": 64930
+    },
+    {
+      "epoch": 418.96,
+      "learning_rate": 3.2279569892473124e-05,
+      "loss": 0.2543,
+      "step": 64940
+    },
+    {
+      "epoch": 419.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.0167157649993896,
+      "eval_runtime": 0.8621,
+      "eval_samples_per_second": 233.162,
+      "eval_steps_per_second": 8.12,
+      "step": 64945
+    },
+    {
+      "epoch": 419.03,
+      "learning_rate": 3.2275985663082436e-05,
+      "loss": 0.3045,
+      "step": 64950
+    },
+    {
+      "epoch": 419.1,
+      "learning_rate": 3.2272401433691754e-05,
+      "loss": 0.2851,
+      "step": 64960
+    },
+    {
+      "epoch": 419.16,
+      "learning_rate": 3.226881720430108e-05,
+      "loss": 0.2797,
+      "step": 64970
+    },
+    {
+      "epoch": 419.23,
+      "learning_rate": 3.22652329749104e-05,
+      "loss": 0.2681,
+      "step": 64980
+    },
+    {
+      "epoch": 419.29,
+      "learning_rate": 3.2261648745519715e-05,
+      "loss": 0.2395,
+      "step": 64990
+    },
+    {
+      "epoch": 419.35,
+      "learning_rate": 3.2258064516129034e-05,
+      "loss": 0.2938,
+      "step": 65000
+    },
+    {
+      "epoch": 419.42,
+      "learning_rate": 3.225448028673835e-05,
+      "loss": 0.28,
+      "step": 65010
+    },
+    {
+      "epoch": 419.48,
+      "learning_rate": 3.225089605734768e-05,
+      "loss": 0.2756,
+      "step": 65020
+    },
+    {
+      "epoch": 419.55,
+      "learning_rate": 3.224731182795699e-05,
+      "loss": 0.2923,
+      "step": 65030
+    },
+    {
+      "epoch": 419.61,
+      "learning_rate": 3.2243727598566306e-05,
+      "loss": 0.2776,
+      "step": 65040
+    },
+    {
+      "epoch": 419.68,
+      "learning_rate": 3.224014336917563e-05,
+      "loss": 0.2484,
+      "step": 65050
+    },
+    {
+      "epoch": 419.74,
+      "learning_rate": 3.223655913978495e-05,
+      "loss": 0.2598,
+      "step": 65060
+    },
+    {
+      "epoch": 419.8,
+      "learning_rate": 3.223297491039427e-05,
+      "loss": 0.2706,
+      "step": 65070
+    },
+    {
+      "epoch": 419.87,
+      "learning_rate": 3.2229390681003586e-05,
+      "loss": 0.265,
+      "step": 65080
+    },
+    {
+      "epoch": 419.93,
+      "learning_rate": 3.2225806451612904e-05,
+      "loss": 0.2665,
+      "step": 65090
+    },
+    {
+      "epoch": 420.0,
+      "learning_rate": 3.222222222222223e-05,
+      "loss": 0.2765,
+      "step": 65100
+    },
+    {
+      "epoch": 420.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9318889379501343,
+      "eval_runtime": 0.8994,
+      "eval_samples_per_second": 223.476,
+      "eval_steps_per_second": 7.783,
+      "step": 65100
+    },
+    {
+      "epoch": 420.06,
+      "learning_rate": 3.221863799283154e-05,
+      "loss": 0.2608,
+      "step": 65110
+    },
+    {
+      "epoch": 420.13,
+      "learning_rate": 3.221505376344086e-05,
+      "loss": 0.2711,
+      "step": 65120
+    },
+    {
+      "epoch": 420.19,
+      "learning_rate": 3.2211469534050184e-05,
+      "loss": 0.3204,
+      "step": 65130
+    },
+    {
+      "epoch": 420.26,
+      "learning_rate": 3.22078853046595e-05,
+      "loss": 0.2919,
+      "step": 65140
+    },
+    {
+      "epoch": 420.32,
+      "learning_rate": 3.2204301075268813e-05,
+      "loss": 0.2969,
+      "step": 65150
+    },
+    {
+      "epoch": 420.39,
+      "learning_rate": 3.220071684587814e-05,
+      "loss": 0.2671,
+      "step": 65160
+    },
+    {
+      "epoch": 420.45,
+      "learning_rate": 3.219713261648746e-05,
+      "loss": 0.2877,
+      "step": 65170
+    },
+    {
+      "epoch": 420.51,
+      "learning_rate": 3.2193548387096775e-05,
+      "loss": 0.2808,
+      "step": 65180
+    },
+    {
+      "epoch": 420.58,
+      "learning_rate": 3.218996415770609e-05,
+      "loss": 0.2811,
+      "step": 65190
+    },
+    {
+      "epoch": 420.64,
+      "learning_rate": 3.218637992831541e-05,
+      "loss": 0.2993,
+      "step": 65200
+    },
+    {
+      "epoch": 420.71,
+      "learning_rate": 3.2182795698924736e-05,
+      "loss": 0.2492,
+      "step": 65210
+    },
+    {
+      "epoch": 420.77,
+      "learning_rate": 3.2179211469534054e-05,
+      "loss": 0.2894,
+      "step": 65220
+    },
+    {
+      "epoch": 420.84,
+      "learning_rate": 3.2175627240143366e-05,
+      "loss": 0.2507,
+      "step": 65230
+    },
+    {
+      "epoch": 420.9,
+      "learning_rate": 3.217204301075269e-05,
+      "loss": 0.236,
+      "step": 65240
+    },
+    {
+      "epoch": 420.96,
+      "learning_rate": 3.216845878136201e-05,
+      "loss": 0.3191,
+      "step": 65250
+    },
+    {
+      "epoch": 421.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0078645944595337,
+      "eval_runtime": 0.8724,
+      "eval_samples_per_second": 230.387,
+      "eval_steps_per_second": 8.023,
+      "step": 65255
+    },
+    {
+      "epoch": 421.03,
+      "learning_rate": 3.216487455197133e-05,
+      "loss": 0.3144,
+      "step": 65260
+    },
+    {
+      "epoch": 421.1,
+      "learning_rate": 3.2161290322580645e-05,
+      "loss": 0.2629,
+      "step": 65270
+    },
+    {
+      "epoch": 421.16,
+      "learning_rate": 3.2157706093189964e-05,
+      "loss": 0.284,
+      "step": 65280
+    },
+    {
+      "epoch": 421.23,
+      "learning_rate": 3.215412186379929e-05,
+      "loss": 0.2717,
+      "step": 65290
+    },
+    {
+      "epoch": 421.29,
+      "learning_rate": 3.215053763440861e-05,
+      "loss": 0.2645,
+      "step": 65300
+    },
+    {
+      "epoch": 421.35,
+      "learning_rate": 3.214695340501792e-05,
+      "loss": 0.2493,
+      "step": 65310
+    },
+    {
+      "epoch": 421.42,
+      "learning_rate": 3.214336917562724e-05,
+      "loss": 0.2257,
+      "step": 65320
+    },
+    {
+      "epoch": 421.48,
+      "learning_rate": 3.213978494623656e-05,
+      "loss": 0.2888,
+      "step": 65330
+    },
+    {
+      "epoch": 421.55,
+      "learning_rate": 3.213620071684588e-05,
+      "loss": 0.2393,
+      "step": 65340
+    },
+    {
+      "epoch": 421.61,
+      "learning_rate": 3.21326164874552e-05,
+      "loss": 0.2998,
+      "step": 65350
+    },
+    {
+      "epoch": 421.68,
+      "learning_rate": 3.2129032258064516e-05,
+      "loss": 0.2746,
+      "step": 65360
+    },
+    {
+      "epoch": 421.74,
+      "learning_rate": 3.212544802867384e-05,
+      "loss": 0.2793,
+      "step": 65370
+    },
+    {
+      "epoch": 421.8,
+      "learning_rate": 3.212186379928316e-05,
+      "loss": 0.2603,
+      "step": 65380
+    },
+    {
+      "epoch": 421.87,
+      "learning_rate": 3.211827956989247e-05,
+      "loss": 0.3073,
+      "step": 65390
+    },
+    {
+      "epoch": 421.93,
+      "learning_rate": 3.2114695340501796e-05,
+      "loss": 0.2535,
+      "step": 65400
+    },
+    {
+      "epoch": 422.0,
+      "learning_rate": 3.2111111111111114e-05,
+      "loss": 0.2892,
+      "step": 65410
+    },
+    {
+      "epoch": 422.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.1275445222854614,
+      "eval_runtime": 0.8535,
+      "eval_samples_per_second": 235.5,
+      "eval_steps_per_second": 8.201,
+      "step": 65410
+    },
+    {
+      "epoch": 422.06,
+      "learning_rate": 3.210752688172043e-05,
+      "loss": 0.2631,
+      "step": 65420
+    },
+    {
+      "epoch": 422.13,
+      "learning_rate": 3.210394265232975e-05,
+      "loss": 0.2962,
+      "step": 65430
+    },
+    {
+      "epoch": 422.19,
+      "learning_rate": 3.210035842293907e-05,
+      "loss": 0.2698,
+      "step": 65440
+    },
+    {
+      "epoch": 422.26,
+      "learning_rate": 3.2096774193548393e-05,
+      "loss": 0.2635,
+      "step": 65450
+    },
+    {
+      "epoch": 422.32,
+      "learning_rate": 3.2093189964157705e-05,
+      "loss": 0.2762,
+      "step": 65460
+    },
+    {
+      "epoch": 422.39,
+      "learning_rate": 3.208960573476702e-05,
+      "loss": 0.2685,
+      "step": 65470
+    },
+    {
+      "epoch": 422.45,
+      "learning_rate": 3.208602150537635e-05,
+      "loss": 0.2629,
+      "step": 65480
+    },
+    {
+      "epoch": 422.51,
+      "learning_rate": 3.2082437275985666e-05,
+      "loss": 0.2936,
+      "step": 65490
+    },
+    {
+      "epoch": 422.58,
+      "learning_rate": 3.207885304659498e-05,
+      "loss": 0.2566,
+      "step": 65500
+    },
+    {
+      "epoch": 422.64,
+      "learning_rate": 3.20752688172043e-05,
+      "loss": 0.2766,
+      "step": 65510
+    },
+    {
+      "epoch": 422.71,
+      "learning_rate": 3.207168458781362e-05,
+      "loss": 0.2483,
+      "step": 65520
+    },
+    {
+      "epoch": 422.77,
+      "learning_rate": 3.206810035842294e-05,
+      "loss": 0.2539,
+      "step": 65530
+    },
+    {
+      "epoch": 422.84,
+      "learning_rate": 3.206451612903226e-05,
+      "loss": 0.3284,
+      "step": 65540
+    },
+    {
+      "epoch": 422.9,
+      "learning_rate": 3.2060931899641576e-05,
+      "loss": 0.2891,
+      "step": 65550
+    },
+    {
+      "epoch": 422.96,
+      "learning_rate": 3.20573476702509e-05,
+      "loss": 0.3122,
+      "step": 65560
+    },
+    {
+      "epoch": 423.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0401320457458496,
+      "eval_runtime": 0.8565,
+      "eval_samples_per_second": 234.683,
+      "eval_steps_per_second": 8.173,
+      "step": 65565
+    },
+    {
+      "epoch": 423.03,
+      "learning_rate": 3.205376344086022e-05,
+      "loss": 0.259,
+      "step": 65570
+    },
+    {
+      "epoch": 423.1,
+      "learning_rate": 3.205017921146953e-05,
+      "loss": 0.2063,
+      "step": 65580
+    },
+    {
+      "epoch": 423.16,
+      "learning_rate": 3.2046594982078855e-05,
+      "loss": 0.2528,
+      "step": 65590
+    },
+    {
+      "epoch": 423.23,
+      "learning_rate": 3.204301075268817e-05,
+      "loss": 0.2199,
+      "step": 65600
+    },
+    {
+      "epoch": 423.29,
+      "learning_rate": 3.203942652329749e-05,
+      "loss": 0.3022,
+      "step": 65610
+    },
+    {
+      "epoch": 423.35,
+      "learning_rate": 3.203584229390681e-05,
+      "loss": 0.2677,
+      "step": 65620
+    },
+    {
+      "epoch": 423.42,
+      "learning_rate": 3.203225806451613e-05,
+      "loss": 0.2815,
+      "step": 65630
+    },
+    {
+      "epoch": 423.48,
+      "learning_rate": 3.202867383512545e-05,
+      "loss": 0.2146,
+      "step": 65640
+    },
+    {
+      "epoch": 423.55,
+      "learning_rate": 3.202508960573477e-05,
+      "loss": 0.2781,
+      "step": 65650
+    },
+    {
+      "epoch": 423.61,
+      "learning_rate": 3.202150537634408e-05,
+      "loss": 0.2643,
+      "step": 65660
+    },
+    {
+      "epoch": 423.68,
+      "learning_rate": 3.201792114695341e-05,
+      "loss": 0.334,
+      "step": 65670
+    },
+    {
+      "epoch": 423.74,
+      "learning_rate": 3.2014336917562726e-05,
+      "loss": 0.2486,
+      "step": 65680
+    },
+    {
+      "epoch": 423.8,
+      "learning_rate": 3.2010752688172044e-05,
+      "loss": 0.2752,
+      "step": 65690
+    },
+    {
+      "epoch": 423.87,
+      "learning_rate": 3.200716845878136e-05,
+      "loss": 0.2711,
+      "step": 65700
+    },
+    {
+      "epoch": 423.93,
+      "learning_rate": 3.200358422939068e-05,
+      "loss": 0.3225,
+      "step": 65710
+    },
+    {
+      "epoch": 424.0,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.2607,
+      "step": 65720
+    },
+    {
+      "epoch": 424.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9745746850967407,
+      "eval_runtime": 0.8707,
+      "eval_samples_per_second": 230.854,
+      "eval_steps_per_second": 8.04,
+      "step": 65720
+    },
+    {
+      "epoch": 424.06,
+      "learning_rate": 3.1996415770609324e-05,
+      "loss": 0.2836,
+      "step": 65730
+    },
+    {
+      "epoch": 424.13,
+      "learning_rate": 3.1992831541218635e-05,
+      "loss": 0.2869,
+      "step": 65740
+    },
+    {
+      "epoch": 424.19,
+      "learning_rate": 3.198924731182796e-05,
+      "loss": 0.2533,
+      "step": 65750
+    },
+    {
+      "epoch": 424.26,
+      "learning_rate": 3.198566308243728e-05,
+      "loss": 0.3021,
+      "step": 65760
+    },
+    {
+      "epoch": 424.32,
+      "learning_rate": 3.1982078853046596e-05,
+      "loss": 0.3008,
+      "step": 65770
+    },
+    {
+      "epoch": 424.39,
+      "learning_rate": 3.1978494623655915e-05,
+      "loss": 0.2463,
+      "step": 65780
+    },
+    {
+      "epoch": 424.45,
+      "learning_rate": 3.197491039426523e-05,
+      "loss": 0.288,
+      "step": 65790
+    },
+    {
+      "epoch": 424.51,
+      "learning_rate": 3.197132616487455e-05,
+      "loss": 0.244,
+      "step": 65800
+    },
+    {
+      "epoch": 424.58,
+      "learning_rate": 3.1967741935483876e-05,
+      "loss": 0.2695,
+      "step": 65810
+    },
+    {
+      "epoch": 424.64,
+      "learning_rate": 3.196415770609319e-05,
+      "loss": 0.2685,
+      "step": 65820
+    },
+    {
+      "epoch": 424.71,
+      "learning_rate": 3.196057347670251e-05,
+      "loss": 0.2155,
+      "step": 65830
+    },
+    {
+      "epoch": 424.77,
+      "learning_rate": 3.195698924731183e-05,
+      "loss": 0.2871,
+      "step": 65840
+    },
+    {
+      "epoch": 424.84,
+      "learning_rate": 3.195340501792115e-05,
+      "loss": 0.262,
+      "step": 65850
+    },
+    {
+      "epoch": 424.9,
+      "learning_rate": 3.194982078853047e-05,
+      "loss": 0.2883,
+      "step": 65860
+    },
+    {
+      "epoch": 424.96,
+      "learning_rate": 3.1946236559139785e-05,
+      "loss": 0.2589,
+      "step": 65870
+    },
+    {
+      "epoch": 425.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0324125289916992,
+      "eval_runtime": 0.8608,
+      "eval_samples_per_second": 233.508,
+      "eval_steps_per_second": 8.132,
+      "step": 65875
+    },
+    {
+      "epoch": 425.03,
+      "learning_rate": 3.19426523297491e-05,
+      "loss": 0.2602,
+      "step": 65880
+    },
+    {
+      "epoch": 425.1,
+      "learning_rate": 3.193906810035843e-05,
+      "loss": 0.2808,
+      "step": 65890
+    },
+    {
+      "epoch": 425.16,
+      "learning_rate": 3.193548387096774e-05,
+      "loss": 0.2799,
+      "step": 65900
+    },
+    {
+      "epoch": 425.23,
+      "learning_rate": 3.1931899641577065e-05,
+      "loss": 0.2579,
+      "step": 65910
+    },
+    {
+      "epoch": 425.29,
+      "learning_rate": 3.192831541218638e-05,
+      "loss": 0.2519,
+      "step": 65920
+    },
+    {
+      "epoch": 425.35,
+      "learning_rate": 3.19247311827957e-05,
+      "loss": 0.2703,
+      "step": 65930
+    },
+    {
+      "epoch": 425.42,
+      "learning_rate": 3.192114695340502e-05,
+      "loss": 0.3305,
+      "step": 65940
+    },
+    {
+      "epoch": 425.48,
+      "learning_rate": 3.191756272401434e-05,
+      "loss": 0.2754,
+      "step": 65950
+    },
+    {
+      "epoch": 425.55,
+      "learning_rate": 3.1913978494623656e-05,
+      "loss": 0.2779,
+      "step": 65960
+    },
+    {
+      "epoch": 425.61,
+      "learning_rate": 3.191039426523298e-05,
+      "loss": 0.3071,
+      "step": 65970
+    },
+    {
+      "epoch": 425.68,
+      "learning_rate": 3.190681003584229e-05,
+      "loss": 0.3799,
+      "step": 65980
+    },
+    {
+      "epoch": 425.74,
+      "learning_rate": 3.190322580645162e-05,
+      "loss": 0.2729,
+      "step": 65990
+    },
+    {
+      "epoch": 425.8,
+      "learning_rate": 3.1899641577060935e-05,
+      "loss": 0.2611,
+      "step": 66000
+    },
+    {
+      "epoch": 425.87,
+      "learning_rate": 3.1896057347670254e-05,
+      "loss": 0.2731,
+      "step": 66010
+    },
+    {
+      "epoch": 425.93,
+      "learning_rate": 3.189247311827957e-05,
+      "loss": 0.3229,
+      "step": 66020
+    },
+    {
+      "epoch": 426.0,
+      "learning_rate": 3.188888888888889e-05,
+      "loss": 0.2919,
+      "step": 66030
+    },
+    {
+      "epoch": 426.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.074987769126892,
+      "eval_runtime": 0.8768,
+      "eval_samples_per_second": 229.24,
+      "eval_steps_per_second": 7.983,
+      "step": 66030
+    },
+    {
+      "epoch": 426.06,
+      "learning_rate": 3.188530465949821e-05,
+      "loss": 0.2632,
+      "step": 66040
+    },
+    {
+      "epoch": 426.13,
+      "learning_rate": 3.188172043010753e-05,
+      "loss": 0.2683,
+      "step": 66050
+    },
+    {
+      "epoch": 426.19,
+      "learning_rate": 3.1878136200716845e-05,
+      "loss": 0.2536,
+      "step": 66060
+    },
+    {
+      "epoch": 426.26,
+      "learning_rate": 3.187455197132617e-05,
+      "loss": 0.2722,
+      "step": 66070
+    },
+    {
+      "epoch": 426.32,
+      "learning_rate": 3.187096774193549e-05,
+      "loss": 0.2939,
+      "step": 66080
+    },
+    {
+      "epoch": 426.39,
+      "learning_rate": 3.18673835125448e-05,
+      "loss": 0.27,
+      "step": 66090
+    },
+    {
+      "epoch": 426.45,
+      "learning_rate": 3.1863799283154124e-05,
+      "loss": 0.3036,
+      "step": 66100
+    },
+    {
+      "epoch": 426.51,
+      "learning_rate": 3.186021505376344e-05,
+      "loss": 0.2975,
+      "step": 66110
+    },
+    {
+      "epoch": 426.58,
+      "learning_rate": 3.185663082437276e-05,
+      "loss": 0.2697,
+      "step": 66120
+    },
+    {
+      "epoch": 426.64,
+      "learning_rate": 3.185304659498208e-05,
+      "loss": 0.2898,
+      "step": 66130
+    },
+    {
+      "epoch": 426.71,
+      "learning_rate": 3.18494623655914e-05,
+      "loss": 0.2447,
+      "step": 66140
+    },
+    {
+      "epoch": 426.77,
+      "learning_rate": 3.1845878136200715e-05,
+      "loss": 0.2595,
+      "step": 66150
+    },
+    {
+      "epoch": 426.84,
+      "learning_rate": 3.184229390681004e-05,
+      "loss": 0.2856,
+      "step": 66160
+    },
+    {
+      "epoch": 426.9,
+      "learning_rate": 3.183870967741935e-05,
+      "loss": 0.292,
+      "step": 66170
+    },
+    {
+      "epoch": 426.96,
+      "learning_rate": 3.183512544802868e-05,
+      "loss": 0.2702,
+      "step": 66180
+    },
+    {
+      "epoch": 427.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.023012638092041,
+      "eval_runtime": 0.8885,
+      "eval_samples_per_second": 226.222,
+      "eval_steps_per_second": 7.878,
+      "step": 66185
+    },
+    {
+      "epoch": 427.03,
+      "learning_rate": 3.1831541218637995e-05,
+      "loss": 0.3059,
+      "step": 66190
+    },
+    {
+      "epoch": 427.1,
+      "learning_rate": 3.182795698924731e-05,
+      "loss": 0.3272,
+      "step": 66200
+    },
+    {
+      "epoch": 427.16,
+      "learning_rate": 3.182437275985663e-05,
+      "loss": 0.3027,
+      "step": 66210
+    },
+    {
+      "epoch": 427.23,
+      "learning_rate": 3.182078853046595e-05,
+      "loss": 0.3157,
+      "step": 66220
+    },
+    {
+      "epoch": 427.29,
+      "learning_rate": 3.181720430107527e-05,
+      "loss": 0.2386,
+      "step": 66230
+    },
+    {
+      "epoch": 427.35,
+      "learning_rate": 3.181362007168459e-05,
+      "loss": 0.2993,
+      "step": 66240
+    },
+    {
+      "epoch": 427.42,
+      "learning_rate": 3.1810035842293904e-05,
+      "loss": 0.2896,
+      "step": 66250
+    },
+    {
+      "epoch": 427.48,
+      "learning_rate": 3.180645161290323e-05,
+      "loss": 0.2959,
+      "step": 66260
+    },
+    {
+      "epoch": 427.55,
+      "learning_rate": 3.180286738351255e-05,
+      "loss": 0.2496,
+      "step": 66270
+    },
+    {
+      "epoch": 427.61,
+      "learning_rate": 3.1799283154121865e-05,
+      "loss": 0.3191,
+      "step": 66280
+    },
+    {
+      "epoch": 427.68,
+      "learning_rate": 3.1795698924731184e-05,
+      "loss": 0.2404,
+      "step": 66290
+    },
+    {
+      "epoch": 427.74,
+      "learning_rate": 3.17921146953405e-05,
+      "loss": 0.2709,
+      "step": 66300
+    },
+    {
+      "epoch": 427.8,
+      "learning_rate": 3.178853046594982e-05,
+      "loss": 0.2476,
+      "step": 66310
+    },
+    {
+      "epoch": 427.87,
+      "learning_rate": 3.1784946236559145e-05,
+      "loss": 0.2932,
+      "step": 66320
+    },
+    {
+      "epoch": 427.93,
+      "learning_rate": 3.1781362007168456e-05,
+      "loss": 0.2725,
+      "step": 66330
+    },
+    {
+      "epoch": 428.0,
+      "learning_rate": 3.177777777777778e-05,
+      "loss": 0.3,
+      "step": 66340
+    },
+    {
+      "epoch": 428.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9508754014968872,
+      "eval_runtime": 0.8528,
+      "eval_samples_per_second": 235.699,
+      "eval_steps_per_second": 8.208,
+      "step": 66340
+    },
+    {
+      "epoch": 428.06,
+      "learning_rate": 3.17741935483871e-05,
+      "loss": 0.2574,
+      "step": 66350
+    },
+    {
+      "epoch": 428.13,
+      "learning_rate": 3.177060931899642e-05,
+      "loss": 0.2248,
+      "step": 66360
+    },
+    {
+      "epoch": 428.19,
+      "learning_rate": 3.1767025089605736e-05,
+      "loss": 0.3094,
+      "step": 66370
+    },
+    {
+      "epoch": 428.26,
+      "learning_rate": 3.1763440860215054e-05,
+      "loss": 0.2589,
+      "step": 66380
+    },
+    {
+      "epoch": 428.32,
+      "learning_rate": 3.175985663082437e-05,
+      "loss": 0.2091,
+      "step": 66390
+    },
+    {
+      "epoch": 428.39,
+      "learning_rate": 3.17562724014337e-05,
+      "loss": 0.297,
+      "step": 66400
+    },
+    {
+      "epoch": 428.45,
+      "learning_rate": 3.175268817204301e-05,
+      "loss": 0.2162,
+      "step": 66410
+    },
+    {
+      "epoch": 428.51,
+      "learning_rate": 3.1749103942652334e-05,
+      "loss": 0.2523,
+      "step": 66420
+    },
+    {
+      "epoch": 428.58,
+      "learning_rate": 3.174551971326165e-05,
+      "loss": 0.2802,
+      "step": 66430
+    },
+    {
+      "epoch": 428.64,
+      "learning_rate": 3.174193548387097e-05,
+      "loss": 0.28,
+      "step": 66440
+    },
+    {
+      "epoch": 428.71,
+      "learning_rate": 3.173835125448029e-05,
+      "loss": 0.2677,
+      "step": 66450
+    },
+    {
+      "epoch": 428.77,
+      "learning_rate": 3.173476702508961e-05,
+      "loss": 0.3381,
+      "step": 66460
+    },
+    {
+      "epoch": 428.84,
+      "learning_rate": 3.1731182795698925e-05,
+      "loss": 0.2567,
+      "step": 66470
+    },
+    {
+      "epoch": 428.9,
+      "learning_rate": 3.172759856630825e-05,
+      "loss": 0.2906,
+      "step": 66480
+    },
+    {
+      "epoch": 428.96,
+      "learning_rate": 3.172401433691756e-05,
+      "loss": 0.2495,
+      "step": 66490
+    },
+    {
+      "epoch": 429.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.978017270565033,
+      "eval_runtime": 0.8897,
+      "eval_samples_per_second": 225.91,
+      "eval_steps_per_second": 7.868,
+      "step": 66495
+    },
+    {
+      "epoch": 429.03,
+      "learning_rate": 3.172043010752688e-05,
+      "loss": 0.3793,
+      "step": 66500
+    },
+    {
+      "epoch": 429.1,
+      "learning_rate": 3.1716845878136205e-05,
+      "loss": 0.2777,
+      "step": 66510
+    },
+    {
+      "epoch": 429.16,
+      "learning_rate": 3.171326164874552e-05,
+      "loss": 0.2585,
+      "step": 66520
+    },
+    {
+      "epoch": 429.23,
+      "learning_rate": 3.170967741935484e-05,
+      "loss": 0.2927,
+      "step": 66530
+    },
+    {
+      "epoch": 429.29,
+      "learning_rate": 3.170609318996416e-05,
+      "loss": 0.3054,
+      "step": 66540
+    },
+    {
+      "epoch": 429.35,
+      "learning_rate": 3.170250896057348e-05,
+      "loss": 0.2877,
+      "step": 66550
+    },
+    {
+      "epoch": 429.42,
+      "learning_rate": 3.16989247311828e-05,
+      "loss": 0.275,
+      "step": 66560
+    },
+    {
+      "epoch": 429.48,
+      "learning_rate": 3.1695340501792114e-05,
+      "loss": 0.2635,
+      "step": 66570
+    },
+    {
+      "epoch": 429.55,
+      "learning_rate": 3.169175627240143e-05,
+      "loss": 0.2703,
+      "step": 66580
+    },
+    {
+      "epoch": 429.61,
+      "learning_rate": 3.168817204301076e-05,
+      "loss": 0.2655,
+      "step": 66590
+    },
+    {
+      "epoch": 429.68,
+      "learning_rate": 3.1684587813620075e-05,
+      "loss": 0.3012,
+      "step": 66600
+    },
+    {
+      "epoch": 429.74,
+      "learning_rate": 3.168100358422939e-05,
+      "loss": 0.2793,
+      "step": 66610
+    },
+    {
+      "epoch": 429.8,
+      "learning_rate": 3.167741935483871e-05,
+      "loss": 0.2684,
+      "step": 66620
+    },
+    {
+      "epoch": 429.87,
+      "learning_rate": 3.167383512544803e-05,
+      "loss": 0.2872,
+      "step": 66630
+    },
+    {
+      "epoch": 429.93,
+      "learning_rate": 3.1670250896057355e-05,
+      "loss": 0.2856,
+      "step": 66640
+    },
+    {
+      "epoch": 430.0,
+      "learning_rate": 3.1666666666666666e-05,
+      "loss": 0.2843,
+      "step": 66650
+    },
+    {
+      "epoch": 430.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9922322034835815,
+      "eval_runtime": 0.8736,
+      "eval_samples_per_second": 230.094,
+      "eval_steps_per_second": 8.013,
+      "step": 66650
+    },
+    {
+      "epoch": 430.06,
+      "learning_rate": 3.1663082437275984e-05,
+      "loss": 0.332,
+      "step": 66660
+    },
+    {
+      "epoch": 430.13,
+      "learning_rate": 3.165949820788531e-05,
+      "loss": 0.3113,
+      "step": 66670
+    },
+    {
+      "epoch": 430.19,
+      "learning_rate": 3.165591397849463e-05,
+      "loss": 0.2814,
+      "step": 66680
+    },
+    {
+      "epoch": 430.26,
+      "learning_rate": 3.1652329749103946e-05,
+      "loss": 0.2356,
+      "step": 66690
+    },
+    {
+      "epoch": 430.32,
+      "learning_rate": 3.1648745519713264e-05,
+      "loss": 0.2941,
+      "step": 66700
+    },
+    {
+      "epoch": 430.39,
+      "learning_rate": 3.164516129032258e-05,
+      "loss": 0.25,
+      "step": 66710
+    },
+    {
+      "epoch": 430.45,
+      "learning_rate": 3.16415770609319e-05,
+      "loss": 0.2835,
+      "step": 66720
+    },
+    {
+      "epoch": 430.51,
+      "learning_rate": 3.163799283154122e-05,
+      "loss": 0.2557,
+      "step": 66730
+    },
+    {
+      "epoch": 430.58,
+      "learning_rate": 3.163440860215054e-05,
+      "loss": 0.2487,
+      "step": 66740
+    },
+    {
+      "epoch": 430.64,
+      "learning_rate": 3.163082437275986e-05,
+      "loss": 0.2535,
+      "step": 66750
+    },
+    {
+      "epoch": 430.71,
+      "learning_rate": 3.162724014336917e-05,
+      "loss": 0.2986,
+      "step": 66760
+    },
+    {
+      "epoch": 430.77,
+      "learning_rate": 3.16236559139785e-05,
+      "loss": 0.3305,
+      "step": 66770
+    },
+    {
+      "epoch": 430.84,
+      "learning_rate": 3.1620071684587816e-05,
+      "loss": 0.2626,
+      "step": 66780
+    },
+    {
+      "epoch": 430.9,
+      "learning_rate": 3.1616487455197135e-05,
+      "loss": 0.238,
+      "step": 66790
+    },
+    {
+      "epoch": 430.96,
+      "learning_rate": 3.161290322580645e-05,
+      "loss": 0.3154,
+      "step": 66800
+    },
+    {
+      "epoch": 431.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0045294761657715,
+      "eval_runtime": 0.8848,
+      "eval_samples_per_second": 227.158,
+      "eval_steps_per_second": 7.911,
+      "step": 66805
+    },
+    {
+      "epoch": 431.03,
+      "learning_rate": 3.160931899641577e-05,
+      "loss": 0.3137,
+      "step": 66810
+    },
+    {
+      "epoch": 431.1,
+      "learning_rate": 3.160573476702509e-05,
+      "loss": 0.2746,
+      "step": 66820
+    },
+    {
+      "epoch": 431.16,
+      "learning_rate": 3.1602150537634414e-05,
+      "loss": 0.3671,
+      "step": 66830
+    },
+    {
+      "epoch": 431.23,
+      "learning_rate": 3.1598566308243726e-05,
+      "loss": 0.2978,
+      "step": 66840
+    },
+    {
+      "epoch": 431.29,
+      "learning_rate": 3.1594982078853044e-05,
+      "loss": 0.3013,
+      "step": 66850
+    },
+    {
+      "epoch": 431.35,
+      "learning_rate": 3.159139784946237e-05,
+      "loss": 0.2298,
+      "step": 66860
+    },
+    {
+      "epoch": 431.42,
+      "learning_rate": 3.158781362007169e-05,
+      "loss": 0.2837,
+      "step": 66870
+    },
+    {
+      "epoch": 431.48,
+      "learning_rate": 3.1584229390681005e-05,
+      "loss": 0.2353,
+      "step": 66880
+    },
+    {
+      "epoch": 431.55,
+      "learning_rate": 3.1580645161290323e-05,
+      "loss": 0.2709,
+      "step": 66890
+    },
+    {
+      "epoch": 431.61,
+      "learning_rate": 3.157706093189964e-05,
+      "loss": 0.2574,
+      "step": 66900
+    },
+    {
+      "epoch": 431.68,
+      "learning_rate": 3.1573476702508967e-05,
+      "loss": 0.2798,
+      "step": 66910
+    },
+    {
+      "epoch": 431.74,
+      "learning_rate": 3.156989247311828e-05,
+      "loss": 0.2571,
+      "step": 66920
+    },
+    {
+      "epoch": 431.8,
+      "learning_rate": 3.1566308243727596e-05,
+      "loss": 0.242,
+      "step": 66930
+    },
+    {
+      "epoch": 431.87,
+      "learning_rate": 3.156272401433692e-05,
+      "loss": 0.274,
+      "step": 66940
+    },
+    {
+      "epoch": 431.93,
+      "learning_rate": 3.155913978494624e-05,
+      "loss": 0.2569,
+      "step": 66950
+    },
+    {
+      "epoch": 432.0,
+      "learning_rate": 3.155555555555556e-05,
+      "loss": 0.2698,
+      "step": 66960
+    },
+    {
+      "epoch": 432.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0253647565841675,
+      "eval_runtime": 0.9084,
+      "eval_samples_per_second": 221.263,
+      "eval_steps_per_second": 7.706,
+      "step": 66960
+    },
+    {
+      "epoch": 432.06,
+      "learning_rate": 3.1551971326164876e-05,
+      "loss": 0.27,
+      "step": 66970
+    },
+    {
+      "epoch": 432.13,
+      "learning_rate": 3.1548387096774194e-05,
+      "loss": 0.3084,
+      "step": 66980
+    },
+    {
+      "epoch": 432.19,
+      "learning_rate": 3.154480286738352e-05,
+      "loss": 0.2782,
+      "step": 66990
+    },
+    {
+      "epoch": 432.26,
+      "learning_rate": 3.154121863799283e-05,
+      "loss": 0.31,
+      "step": 67000
+    },
+    {
+      "epoch": 432.32,
+      "learning_rate": 3.153763440860215e-05,
+      "loss": 0.2582,
+      "step": 67010
+    },
+    {
+      "epoch": 432.39,
+      "learning_rate": 3.1534050179211474e-05,
+      "loss": 0.2879,
+      "step": 67020
+    },
+    {
+      "epoch": 432.45,
+      "learning_rate": 3.153046594982079e-05,
+      "loss": 0.2654,
+      "step": 67030
+    },
+    {
+      "epoch": 432.51,
+      "learning_rate": 3.152688172043011e-05,
+      "loss": 0.246,
+      "step": 67040
+    },
+    {
+      "epoch": 432.58,
+      "learning_rate": 3.152329749103943e-05,
+      "loss": 0.2893,
+      "step": 67050
+    },
+    {
+      "epoch": 432.64,
+      "learning_rate": 3.1519713261648746e-05,
+      "loss": 0.251,
+      "step": 67060
+    },
+    {
+      "epoch": 432.71,
+      "learning_rate": 3.151612903225807e-05,
+      "loss": 0.2985,
+      "step": 67070
+    },
+    {
+      "epoch": 432.77,
+      "learning_rate": 3.151254480286738e-05,
+      "loss": 0.338,
+      "step": 67080
+    },
+    {
+      "epoch": 432.84,
+      "learning_rate": 3.15089605734767e-05,
+      "loss": 0.2633,
+      "step": 67090
+    },
+    {
+      "epoch": 432.9,
+      "learning_rate": 3.1505376344086026e-05,
+      "loss": 0.2842,
+      "step": 67100
+    },
+    {
+      "epoch": 432.96,
+      "learning_rate": 3.1501792114695344e-05,
+      "loss": 0.2742,
+      "step": 67110
+    },
+    {
+      "epoch": 433.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.1059997081756592,
+      "eval_runtime": 0.8689,
+      "eval_samples_per_second": 231.336,
+      "eval_steps_per_second": 8.056,
+      "step": 67115
+    },
+    {
+      "epoch": 433.03,
+      "learning_rate": 3.149820788530466e-05,
+      "loss": 0.2877,
+      "step": 67120
+    },
+    {
+      "epoch": 433.1,
+      "learning_rate": 3.149462365591398e-05,
+      "loss": 0.2564,
+      "step": 67130
+    },
+    {
+      "epoch": 433.16,
+      "learning_rate": 3.14910394265233e-05,
+      "loss": 0.3222,
+      "step": 67140
+    },
+    {
+      "epoch": 433.23,
+      "learning_rate": 3.148745519713262e-05,
+      "loss": 0.2368,
+      "step": 67150
+    },
+    {
+      "epoch": 433.29,
+      "learning_rate": 3.1483870967741935e-05,
+      "loss": 0.2497,
+      "step": 67160
+    },
+    {
+      "epoch": 433.35,
+      "learning_rate": 3.1480286738351253e-05,
+      "loss": 0.3046,
+      "step": 67170
+    },
+    {
+      "epoch": 433.42,
+      "learning_rate": 3.147670250896058e-05,
+      "loss": 0.2562,
+      "step": 67180
+    },
+    {
+      "epoch": 433.48,
+      "learning_rate": 3.14731182795699e-05,
+      "loss": 0.2349,
+      "step": 67190
+    },
+    {
+      "epoch": 433.55,
+      "learning_rate": 3.146953405017921e-05,
+      "loss": 0.2745,
+      "step": 67200
+    },
+    {
+      "epoch": 433.61,
+      "learning_rate": 3.146594982078853e-05,
+      "loss": 0.3012,
+      "step": 67210
+    },
+    {
+      "epoch": 433.68,
+      "learning_rate": 3.146236559139785e-05,
+      "loss": 0.2171,
+      "step": 67220
+    },
+    {
+      "epoch": 433.74,
+      "learning_rate": 3.145878136200717e-05,
+      "loss": 0.245,
+      "step": 67230
+    },
+    {
+      "epoch": 433.8,
+      "learning_rate": 3.145519713261649e-05,
+      "loss": 0.2299,
+      "step": 67240
+    },
+    {
+      "epoch": 433.87,
+      "learning_rate": 3.1451612903225806e-05,
+      "loss": 0.2966,
+      "step": 67250
+    },
+    {
+      "epoch": 433.93,
+      "learning_rate": 3.144802867383513e-05,
+      "loss": 0.2787,
+      "step": 67260
+    },
+    {
+      "epoch": 434.0,
+      "learning_rate": 3.144444444444445e-05,
+      "loss": 0.319,
+      "step": 67270
+    },
+    {
+      "epoch": 434.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9947736263275146,
+      "eval_runtime": 0.9215,
+      "eval_samples_per_second": 218.118,
+      "eval_steps_per_second": 7.596,
+      "step": 67270
+    },
+    {
+      "epoch": 434.06,
+      "learning_rate": 3.144086021505376e-05,
+      "loss": 0.3212,
+      "step": 67280
+    },
+    {
+      "epoch": 434.13,
+      "learning_rate": 3.1437275985663085e-05,
+      "loss": 0.29,
+      "step": 67290
+    },
+    {
+      "epoch": 434.19,
+      "learning_rate": 3.1433691756272404e-05,
+      "loss": 0.234,
+      "step": 67300
+    },
+    {
+      "epoch": 434.26,
+      "learning_rate": 3.143010752688172e-05,
+      "loss": 0.2537,
+      "step": 67310
+    },
+    {
+      "epoch": 434.32,
+      "learning_rate": 3.142652329749104e-05,
+      "loss": 0.2504,
+      "step": 67320
+    },
+    {
+      "epoch": 434.39,
+      "learning_rate": 3.142293906810036e-05,
+      "loss": 0.3154,
+      "step": 67330
+    },
+    {
+      "epoch": 434.45,
+      "learning_rate": 3.141935483870968e-05,
+      "loss": 0.2638,
+      "step": 67340
+    },
+    {
+      "epoch": 434.51,
+      "learning_rate": 3.1415770609318995e-05,
+      "loss": 0.281,
+      "step": 67350
+    },
+    {
+      "epoch": 434.58,
+      "learning_rate": 3.141218637992831e-05,
+      "loss": 0.2727,
+      "step": 67360
+    },
+    {
+      "epoch": 434.64,
+      "learning_rate": 3.140860215053764e-05,
+      "loss": 0.2391,
+      "step": 67370
+    },
+    {
+      "epoch": 434.71,
+      "learning_rate": 3.1405017921146956e-05,
+      "loss": 0.2342,
+      "step": 67380
+    },
+    {
+      "epoch": 434.77,
+      "learning_rate": 3.1401433691756274e-05,
+      "loss": 0.2929,
+      "step": 67390
+    },
+    {
+      "epoch": 434.84,
+      "learning_rate": 3.139784946236559e-05,
+      "loss": 0.2805,
+      "step": 67400
+    },
+    {
+      "epoch": 434.9,
+      "learning_rate": 3.139426523297491e-05,
+      "loss": 0.256,
+      "step": 67410
+    },
+    {
+      "epoch": 434.96,
+      "learning_rate": 3.1390681003584236e-05,
+      "loss": 0.3135,
+      "step": 67420
+    },
+    {
+      "epoch": 435.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0530415773391724,
+      "eval_runtime": 0.9001,
+      "eval_samples_per_second": 223.311,
+      "eval_steps_per_second": 7.777,
+      "step": 67425
+    },
+    {
+      "epoch": 435.03,
+      "learning_rate": 3.138709677419355e-05,
+      "loss": 0.3214,
+      "step": 67430
+    },
+    {
+      "epoch": 435.1,
+      "learning_rate": 3.1383512544802865e-05,
+      "loss": 0.291,
+      "step": 67440
+    },
+    {
+      "epoch": 435.16,
+      "learning_rate": 3.137992831541219e-05,
+      "loss": 0.2708,
+      "step": 67450
+    },
+    {
+      "epoch": 435.23,
+      "learning_rate": 3.137634408602151e-05,
+      "loss": 0.2461,
+      "step": 67460
+    },
+    {
+      "epoch": 435.29,
+      "learning_rate": 3.137275985663083e-05,
+      "loss": 0.2814,
+      "step": 67470
+    },
+    {
+      "epoch": 435.35,
+      "learning_rate": 3.1369175627240145e-05,
+      "loss": 0.2612,
+      "step": 67480
+    },
+    {
+      "epoch": 435.42,
+      "learning_rate": 3.136559139784946e-05,
+      "loss": 0.247,
+      "step": 67490
+    },
+    {
+      "epoch": 435.48,
+      "learning_rate": 3.136200716845878e-05,
+      "loss": 0.305,
+      "step": 67500
+    },
+    {
+      "epoch": 435.55,
+      "learning_rate": 3.13584229390681e-05,
+      "loss": 0.2566,
+      "step": 67510
+    },
+    {
+      "epoch": 435.61,
+      "learning_rate": 3.135483870967742e-05,
+      "loss": 0.2349,
+      "step": 67520
+    },
+    {
+      "epoch": 435.68,
+      "learning_rate": 3.135125448028674e-05,
+      "loss": 0.2365,
+      "step": 67530
+    },
+    {
+      "epoch": 435.74,
+      "learning_rate": 3.134767025089606e-05,
+      "loss": 0.2276,
+      "step": 67540
+    },
+    {
+      "epoch": 435.8,
+      "learning_rate": 3.134408602150537e-05,
+      "loss": 0.2178,
+      "step": 67550
+    },
+    {
+      "epoch": 435.87,
+      "learning_rate": 3.13405017921147e-05,
+      "loss": 0.2849,
+      "step": 67560
+    },
+    {
+      "epoch": 435.93,
+      "learning_rate": 3.1336917562724016e-05,
+      "loss": 0.2116,
+      "step": 67570
+    },
+    {
+      "epoch": 436.0,
+      "learning_rate": 3.1333333333333334e-05,
+      "loss": 0.255,
+      "step": 67580
+    },
+    {
+      "epoch": 436.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9937063455581665,
+      "eval_runtime": 0.8784,
+      "eval_samples_per_second": 228.837,
+      "eval_steps_per_second": 7.969,
+      "step": 67580
+    },
+    {
+      "epoch": 436.06,
+      "learning_rate": 3.132974910394265e-05,
+      "loss": 0.289,
+      "step": 67590
+    },
+    {
+      "epoch": 436.13,
+      "learning_rate": 3.132616487455197e-05,
+      "loss": 0.3049,
+      "step": 67600
+    },
+    {
+      "epoch": 436.19,
+      "learning_rate": 3.1322580645161295e-05,
+      "loss": 0.277,
+      "step": 67610
+    },
+    {
+      "epoch": 436.26,
+      "learning_rate": 3.131899641577061e-05,
+      "loss": 0.3202,
+      "step": 67620
+    },
+    {
+      "epoch": 436.32,
+      "learning_rate": 3.1315412186379925e-05,
+      "loss": 0.2555,
+      "step": 67630
+    },
+    {
+      "epoch": 436.39,
+      "learning_rate": 3.131182795698925e-05,
+      "loss": 0.2681,
+      "step": 67640
+    },
+    {
+      "epoch": 436.45,
+      "learning_rate": 3.130824372759857e-05,
+      "loss": 0.2693,
+      "step": 67650
+    },
+    {
+      "epoch": 436.51,
+      "learning_rate": 3.1304659498207886e-05,
+      "loss": 0.2846,
+      "step": 67660
+    },
+    {
+      "epoch": 436.58,
+      "learning_rate": 3.1301075268817204e-05,
+      "loss": 0.2316,
+      "step": 67670
+    },
+    {
+      "epoch": 436.64,
+      "learning_rate": 3.129749103942652e-05,
+      "loss": 0.3051,
+      "step": 67680
+    },
+    {
+      "epoch": 436.71,
+      "learning_rate": 3.129390681003585e-05,
+      "loss": 0.2641,
+      "step": 67690
+    },
+    {
+      "epoch": 436.77,
+      "learning_rate": 3.1290322580645166e-05,
+      "loss": 0.2629,
+      "step": 67700
+    },
+    {
+      "epoch": 436.84,
+      "learning_rate": 3.128673835125448e-05,
+      "loss": 0.2318,
+      "step": 67710
+    },
+    {
+      "epoch": 436.9,
+      "learning_rate": 3.12831541218638e-05,
+      "loss": 0.2609,
+      "step": 67720
+    },
+    {
+      "epoch": 436.96,
+      "learning_rate": 3.127956989247312e-05,
+      "loss": 0.2499,
+      "step": 67730
+    },
+    {
+      "epoch": 437.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9809154272079468,
+      "eval_runtime": 0.9004,
+      "eval_samples_per_second": 223.235,
+      "eval_steps_per_second": 7.774,
+      "step": 67735
+    },
+    {
+      "epoch": 437.03,
+      "learning_rate": 3.127598566308244e-05,
+      "loss": 0.2717,
+      "step": 67740
+    },
+    {
+      "epoch": 437.1,
+      "learning_rate": 3.127240143369176e-05,
+      "loss": 0.265,
+      "step": 67750
+    },
+    {
+      "epoch": 437.16,
+      "learning_rate": 3.1268817204301075e-05,
+      "loss": 0.2781,
+      "step": 67760
+    },
+    {
+      "epoch": 437.23,
+      "learning_rate": 3.12652329749104e-05,
+      "loss": 0.2622,
+      "step": 67770
+    },
+    {
+      "epoch": 437.29,
+      "learning_rate": 3.126164874551972e-05,
+      "loss": 0.255,
+      "step": 67780
+    },
+    {
+      "epoch": 437.35,
+      "learning_rate": 3.125806451612903e-05,
+      "loss": 0.2765,
+      "step": 67790
+    },
+    {
+      "epoch": 437.42,
+      "learning_rate": 3.1254480286738355e-05,
+      "loss": 0.2579,
+      "step": 67800
+    },
+    {
+      "epoch": 437.48,
+      "learning_rate": 3.125089605734767e-05,
+      "loss": 0.2553,
+      "step": 67810
+    },
+    {
+      "epoch": 437.55,
+      "learning_rate": 3.124731182795699e-05,
+      "loss": 0.2613,
+      "step": 67820
+    },
+    {
+      "epoch": 437.61,
+      "learning_rate": 3.124372759856631e-05,
+      "loss": 0.2345,
+      "step": 67830
+    },
+    {
+      "epoch": 437.68,
+      "learning_rate": 3.124014336917563e-05,
+      "loss": 0.3029,
+      "step": 67840
+    },
+    {
+      "epoch": 437.74,
+      "learning_rate": 3.1236559139784946e-05,
+      "loss": 0.2722,
+      "step": 67850
+    },
+    {
+      "epoch": 437.8,
+      "learning_rate": 3.123297491039427e-05,
+      "loss": 0.2513,
+      "step": 67860
+    },
+    {
+      "epoch": 437.87,
+      "learning_rate": 3.122939068100358e-05,
+      "loss": 0.2756,
+      "step": 67870
+    },
+    {
+      "epoch": 437.93,
+      "learning_rate": 3.122580645161291e-05,
+      "loss": 0.2789,
+      "step": 67880
+    },
+    {
+      "epoch": 438.0,
+      "learning_rate": 3.1222222222222225e-05,
+      "loss": 0.3091,
+      "step": 67890
+    },
+    {
+      "epoch": 438.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0070457458496094,
+      "eval_runtime": 0.9778,
+      "eval_samples_per_second": 205.561,
+      "eval_steps_per_second": 7.159,
+      "step": 67890
+    },
+    {
+      "epoch": 438.06,
+      "learning_rate": 3.1218637992831543e-05,
+      "loss": 0.3242,
+      "step": 67900
+    },
+    {
+      "epoch": 438.13,
+      "learning_rate": 3.121505376344086e-05,
+      "loss": 0.2402,
+      "step": 67910
+    },
+    {
+      "epoch": 438.19,
+      "learning_rate": 3.121146953405018e-05,
+      "loss": 0.2593,
+      "step": 67920
+    },
+    {
+      "epoch": 438.26,
+      "learning_rate": 3.12078853046595e-05,
+      "loss": 0.2606,
+      "step": 67930
+    },
+    {
+      "epoch": 438.32,
+      "learning_rate": 3.120430107526882e-05,
+      "loss": 0.284,
+      "step": 67940
+    },
+    {
+      "epoch": 438.39,
+      "learning_rate": 3.1200716845878134e-05,
+      "loss": 0.2583,
+      "step": 67950
+    },
+    {
+      "epoch": 438.45,
+      "learning_rate": 3.119713261648746e-05,
+      "loss": 0.2807,
+      "step": 67960
+    },
+    {
+      "epoch": 438.51,
+      "learning_rate": 3.119354838709678e-05,
+      "loss": 0.2575,
+      "step": 67970
+    },
+    {
+      "epoch": 438.58,
+      "learning_rate": 3.118996415770609e-05,
+      "loss": 0.3019,
+      "step": 67980
+    },
+    {
+      "epoch": 438.64,
+      "learning_rate": 3.1186379928315414e-05,
+      "loss": 0.2667,
+      "step": 67990
+    },
+    {
+      "epoch": 438.71,
+      "learning_rate": 3.118279569892473e-05,
+      "loss": 0.2894,
+      "step": 68000
+    },
+    {
+      "epoch": 438.77,
+      "learning_rate": 3.117921146953405e-05,
+      "loss": 0.2856,
+      "step": 68010
+    },
+    {
+      "epoch": 438.84,
+      "learning_rate": 3.117562724014337e-05,
+      "loss": 0.2676,
+      "step": 68020
+    },
+    {
+      "epoch": 438.9,
+      "learning_rate": 3.117204301075269e-05,
+      "loss": 0.3238,
+      "step": 68030
+    },
+    {
+      "epoch": 438.96,
+      "learning_rate": 3.116845878136201e-05,
+      "loss": 0.2785,
+      "step": 68040
+    },
+    {
+      "epoch": 439.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.092349648475647,
+      "eval_runtime": 0.864,
+      "eval_samples_per_second": 232.648,
+      "eval_steps_per_second": 8.102,
+      "step": 68045
+    },
+    {
+      "epoch": 439.03,
+      "learning_rate": 3.116487455197133e-05,
+      "loss": 0.2991,
+      "step": 68050
+    },
+    {
+      "epoch": 439.1,
+      "learning_rate": 3.116129032258064e-05,
+      "loss": 0.2622,
+      "step": 68060
+    },
+    {
+      "epoch": 439.16,
+      "learning_rate": 3.1157706093189966e-05,
+      "loss": 0.2608,
+      "step": 68070
+    },
+    {
+      "epoch": 439.23,
+      "learning_rate": 3.1154121863799285e-05,
+      "loss": 0.2743,
+      "step": 68080
+    },
+    {
+      "epoch": 439.29,
+      "learning_rate": 3.11505376344086e-05,
+      "loss": 0.2532,
+      "step": 68090
+    },
+    {
+      "epoch": 439.35,
+      "learning_rate": 3.114695340501792e-05,
+      "loss": 0.2796,
+      "step": 68100
+    },
+    {
+      "epoch": 439.42,
+      "learning_rate": 3.114336917562724e-05,
+      "loss": 0.2661,
+      "step": 68110
+    },
+    {
+      "epoch": 439.48,
+      "learning_rate": 3.1139784946236564e-05,
+      "loss": 0.2844,
+      "step": 68120
+    },
+    {
+      "epoch": 439.55,
+      "learning_rate": 3.113620071684588e-05,
+      "loss": 0.2493,
+      "step": 68130
+    },
+    {
+      "epoch": 439.61,
+      "learning_rate": 3.1132616487455194e-05,
+      "loss": 0.285,
+      "step": 68140
+    },
+    {
+      "epoch": 439.68,
+      "learning_rate": 3.112903225806452e-05,
+      "loss": 0.2586,
+      "step": 68150
+    },
+    {
+      "epoch": 439.74,
+      "learning_rate": 3.112544802867384e-05,
+      "loss": 0.3259,
+      "step": 68160
+    },
+    {
+      "epoch": 439.8,
+      "learning_rate": 3.1121863799283155e-05,
+      "loss": 0.2508,
+      "step": 68170
+    },
+    {
+      "epoch": 439.87,
+      "learning_rate": 3.1118279569892473e-05,
+      "loss": 0.3064,
+      "step": 68180
+    },
+    {
+      "epoch": 439.93,
+      "learning_rate": 3.111469534050179e-05,
+      "loss": 0.2445,
+      "step": 68190
+    },
+    {
+      "epoch": 440.0,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 0.2724,
+      "step": 68200
+    },
+    {
+      "epoch": 440.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0692131519317627,
+      "eval_runtime": 0.8644,
+      "eval_samples_per_second": 232.527,
+      "eval_steps_per_second": 8.098,
+      "step": 68200
+    },
+    {
+      "epoch": 440.06,
+      "learning_rate": 3.1107526881720435e-05,
+      "loss": 0.3027,
+      "step": 68210
+    },
+    {
+      "epoch": 440.13,
+      "learning_rate": 3.1103942652329746e-05,
+      "loss": 0.2813,
+      "step": 68220
+    },
+    {
+      "epoch": 440.19,
+      "learning_rate": 3.110035842293907e-05,
+      "loss": 0.277,
+      "step": 68230
+    },
+    {
+      "epoch": 440.26,
+      "learning_rate": 3.109677419354839e-05,
+      "loss": 0.2981,
+      "step": 68240
+    },
+    {
+      "epoch": 440.32,
+      "learning_rate": 3.109318996415771e-05,
+      "loss": 0.3057,
+      "step": 68250
+    },
+    {
+      "epoch": 440.39,
+      "learning_rate": 3.1089605734767026e-05,
+      "loss": 0.2806,
+      "step": 68260
+    },
+    {
+      "epoch": 440.45,
+      "learning_rate": 3.1086021505376344e-05,
+      "loss": 0.285,
+      "step": 68270
+    },
+    {
+      "epoch": 440.51,
+      "learning_rate": 3.108243727598566e-05,
+      "loss": 0.2638,
+      "step": 68280
+    },
+    {
+      "epoch": 440.58,
+      "learning_rate": 3.107885304659499e-05,
+      "loss": 0.3232,
+      "step": 68290
+    },
+    {
+      "epoch": 440.64,
+      "learning_rate": 3.10752688172043e-05,
+      "loss": 0.2847,
+      "step": 68300
+    },
+    {
+      "epoch": 440.71,
+      "learning_rate": 3.1071684587813624e-05,
+      "loss": 0.261,
+      "step": 68310
+    },
+    {
+      "epoch": 440.77,
+      "learning_rate": 3.106810035842294e-05,
+      "loss": 0.2794,
+      "step": 68320
+    },
+    {
+      "epoch": 440.84,
+      "learning_rate": 3.106451612903226e-05,
+      "loss": 0.279,
+      "step": 68330
+    },
+    {
+      "epoch": 440.9,
+      "learning_rate": 3.106093189964158e-05,
+      "loss": 0.2807,
+      "step": 68340
+    },
+    {
+      "epoch": 440.96,
+      "learning_rate": 3.1057347670250897e-05,
+      "loss": 0.275,
+      "step": 68350
+    },
+    {
+      "epoch": 441.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.1003494262695312,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.913,
+      "eval_steps_per_second": 7.937,
+      "step": 68355
+    },
+    {
+      "epoch": 441.03,
+      "learning_rate": 3.1053763440860215e-05,
+      "loss": 0.2812,
+      "step": 68360
+    },
+    {
+      "epoch": 441.1,
+      "learning_rate": 3.105017921146954e-05,
+      "loss": 0.253,
+      "step": 68370
+    },
+    {
+      "epoch": 441.16,
+      "learning_rate": 3.104659498207885e-05,
+      "loss": 0.225,
+      "step": 68380
+    },
+    {
+      "epoch": 441.23,
+      "learning_rate": 3.1043010752688176e-05,
+      "loss": 0.2842,
+      "step": 68390
+    },
+    {
+      "epoch": 441.29,
+      "learning_rate": 3.1039426523297494e-05,
+      "loss": 0.3124,
+      "step": 68400
+    },
+    {
+      "epoch": 441.35,
+      "learning_rate": 3.103584229390681e-05,
+      "loss": 0.2667,
+      "step": 68410
+    },
+    {
+      "epoch": 441.42,
+      "learning_rate": 3.103225806451613e-05,
+      "loss": 0.2433,
+      "step": 68420
+    },
+    {
+      "epoch": 441.48,
+      "learning_rate": 3.102867383512545e-05,
+      "loss": 0.2616,
+      "step": 68430
+    },
+    {
+      "epoch": 441.55,
+      "learning_rate": 3.102508960573477e-05,
+      "loss": 0.2953,
+      "step": 68440
+    },
+    {
+      "epoch": 441.61,
+      "learning_rate": 3.102150537634409e-05,
+      "loss": 0.3006,
+      "step": 68450
+    },
+    {
+      "epoch": 441.68,
+      "learning_rate": 3.1017921146953404e-05,
+      "loss": 0.2914,
+      "step": 68460
+    },
+    {
+      "epoch": 441.74,
+      "learning_rate": 3.101433691756273e-05,
+      "loss": 0.3012,
+      "step": 68470
+    },
+    {
+      "epoch": 441.8,
+      "learning_rate": 3.101075268817205e-05,
+      "loss": 0.2844,
+      "step": 68480
+    },
+    {
+      "epoch": 441.87,
+      "learning_rate": 3.1007168458781365e-05,
+      "loss": 0.2882,
+      "step": 68490
+    },
+    {
+      "epoch": 441.93,
+      "learning_rate": 3.100358422939068e-05,
+      "loss": 0.2531,
+      "step": 68500
+    },
+    {
+      "epoch": 442.0,
+      "learning_rate": 3.1e-05,
+      "loss": 0.2614,
+      "step": 68510
+    },
+    {
+      "epoch": 442.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0606629848480225,
+      "eval_runtime": 0.899,
+      "eval_samples_per_second": 223.579,
+      "eval_steps_per_second": 7.786,
+      "step": 68510
+    },
+    {
+      "epoch": 442.06,
+      "learning_rate": 3.099641577060932e-05,
+      "loss": 0.2512,
+      "step": 68520
+    },
+    {
+      "epoch": 442.13,
+      "learning_rate": 3.0992831541218645e-05,
+      "loss": 0.2346,
+      "step": 68530
+    },
+    {
+      "epoch": 442.19,
+      "learning_rate": 3.0989247311827956e-05,
+      "loss": 0.2968,
+      "step": 68540
+    },
+    {
+      "epoch": 442.26,
+      "learning_rate": 3.0985663082437274e-05,
+      "loss": 0.2616,
+      "step": 68550
+    },
+    {
+      "epoch": 442.32,
+      "learning_rate": 3.09820788530466e-05,
+      "loss": 0.2949,
+      "step": 68560
+    },
+    {
+      "epoch": 442.39,
+      "learning_rate": 3.097849462365592e-05,
+      "loss": 0.2435,
+      "step": 68570
+    },
+    {
+      "epoch": 442.45,
+      "learning_rate": 3.0974910394265236e-05,
+      "loss": 0.2692,
+      "step": 68580
+    },
+    {
+      "epoch": 442.51,
+      "learning_rate": 3.0971326164874554e-05,
+      "loss": 0.2366,
+      "step": 68590
+    },
+    {
+      "epoch": 442.58,
+      "learning_rate": 3.096774193548387e-05,
+      "loss": 0.3336,
+      "step": 68600
+    },
+    {
+      "epoch": 442.64,
+      "learning_rate": 3.09641577060932e-05,
+      "loss": 0.3155,
+      "step": 68610
+    },
+    {
+      "epoch": 442.71,
+      "learning_rate": 3.096057347670251e-05,
+      "loss": 0.2719,
+      "step": 68620
+    },
+    {
+      "epoch": 442.77,
+      "learning_rate": 3.0956989247311827e-05,
+      "loss": 0.2566,
+      "step": 68630
+    },
+    {
+      "epoch": 442.84,
+      "learning_rate": 3.095340501792115e-05,
+      "loss": 0.2784,
+      "step": 68640
+    },
+    {
+      "epoch": 442.9,
+      "learning_rate": 3.094982078853046e-05,
+      "loss": 0.3173,
+      "step": 68650
+    },
+    {
+      "epoch": 442.96,
+      "learning_rate": 3.094623655913979e-05,
+      "loss": 0.2771,
+      "step": 68660
+    },
+    {
+      "epoch": 443.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9993885159492493,
+      "eval_runtime": 0.8771,
+      "eval_samples_per_second": 229.163,
+      "eval_steps_per_second": 7.981,
+      "step": 68665
+    },
+    {
+      "epoch": 443.03,
+      "learning_rate": 3.0942652329749106e-05,
+      "loss": 0.3592,
+      "step": 68670
+    },
+    {
+      "epoch": 443.1,
+      "learning_rate": 3.0939068100358424e-05,
+      "loss": 0.2562,
+      "step": 68680
+    },
+    {
+      "epoch": 443.16,
+      "learning_rate": 3.093548387096774e-05,
+      "loss": 0.2912,
+      "step": 68690
+    },
+    {
+      "epoch": 443.23,
+      "learning_rate": 3.093189964157706e-05,
+      "loss": 0.2838,
+      "step": 68700
+    },
+    {
+      "epoch": 443.29,
+      "learning_rate": 3.092831541218638e-05,
+      "loss": 0.2642,
+      "step": 68710
+    },
+    {
+      "epoch": 443.35,
+      "learning_rate": 3.0924731182795704e-05,
+      "loss": 0.2414,
+      "step": 68720
+    },
+    {
+      "epoch": 443.42,
+      "learning_rate": 3.0921146953405015e-05,
+      "loss": 0.2439,
+      "step": 68730
+    },
+    {
+      "epoch": 443.48,
+      "learning_rate": 3.091756272401434e-05,
+      "loss": 0.3145,
+      "step": 68740
+    },
+    {
+      "epoch": 443.55,
+      "learning_rate": 3.091397849462366e-05,
+      "loss": 0.2784,
+      "step": 68750
+    },
+    {
+      "epoch": 443.61,
+      "learning_rate": 3.091039426523298e-05,
+      "loss": 0.2601,
+      "step": 68760
+    },
+    {
+      "epoch": 443.68,
+      "learning_rate": 3.0906810035842295e-05,
+      "loss": 0.2433,
+      "step": 68770
+    },
+    {
+      "epoch": 443.74,
+      "learning_rate": 3.090322580645161e-05,
+      "loss": 0.3072,
+      "step": 68780
+    },
+    {
+      "epoch": 443.8,
+      "learning_rate": 3.089964157706093e-05,
+      "loss": 0.2218,
+      "step": 68790
+    },
+    {
+      "epoch": 443.87,
+      "learning_rate": 3.0896057347670256e-05,
+      "loss": 0.2604,
+      "step": 68800
+    },
+    {
+      "epoch": 443.93,
+      "learning_rate": 3.089247311827957e-05,
+      "loss": 0.2806,
+      "step": 68810
+    },
+    {
+      "epoch": 444.0,
+      "learning_rate": 3.088888888888889e-05,
+      "loss": 0.2368,
+      "step": 68820
+    },
+    {
+      "epoch": 444.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 1.0042835474014282,
+      "eval_runtime": 0.9202,
+      "eval_samples_per_second": 218.424,
+      "eval_steps_per_second": 7.607,
+      "step": 68820
+    },
+    {
+      "epoch": 444.06,
+      "learning_rate": 3.088530465949821e-05,
+      "loss": 0.2731,
+      "step": 68830
+    },
+    {
+      "epoch": 444.13,
+      "learning_rate": 3.088172043010753e-05,
+      "loss": 0.2794,
+      "step": 68840
+    },
+    {
+      "epoch": 444.19,
+      "learning_rate": 3.087813620071685e-05,
+      "loss": 0.2806,
+      "step": 68850
+    },
+    {
+      "epoch": 444.26,
+      "learning_rate": 3.0874551971326166e-05,
+      "loss": 0.2564,
+      "step": 68860
+    },
+    {
+      "epoch": 444.32,
+      "learning_rate": 3.0870967741935484e-05,
+      "loss": 0.286,
+      "step": 68870
+    },
+    {
+      "epoch": 444.39,
+      "learning_rate": 3.086738351254481e-05,
+      "loss": 0.3051,
+      "step": 68880
+    },
+    {
+      "epoch": 444.45,
+      "learning_rate": 3.086379928315412e-05,
+      "loss": 0.3136,
+      "step": 68890
+    },
+    {
+      "epoch": 444.51,
+      "learning_rate": 3.086021505376344e-05,
+      "loss": 0.3023,
+      "step": 68900
+    },
+    {
+      "epoch": 444.58,
+      "learning_rate": 3.0856630824372763e-05,
+      "loss": 0.2566,
+      "step": 68910
+    },
+    {
+      "epoch": 444.64,
+      "learning_rate": 3.085304659498208e-05,
+      "loss": 0.2956,
+      "step": 68920
+    },
+    {
+      "epoch": 444.71,
+      "learning_rate": 3.08494623655914e-05,
+      "loss": 0.2646,
+      "step": 68930
+    },
+    {
+      "epoch": 444.77,
+      "learning_rate": 3.084587813620072e-05,
+      "loss": 0.2317,
+      "step": 68940
+    },
+    {
+      "epoch": 444.84,
+      "learning_rate": 3.0842293906810036e-05,
+      "loss": 0.2581,
+      "step": 68950
+    },
+    {
+      "epoch": 444.9,
+      "learning_rate": 3.083870967741936e-05,
+      "loss": 0.2812,
+      "step": 68960
+    },
+    {
+      "epoch": 444.96,
+      "learning_rate": 3.083512544802867e-05,
+      "loss": 0.2874,
+      "step": 68970
+    },
+    {
+      "epoch": 445.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.042952299118042,
+      "eval_runtime": 0.8847,
+      "eval_samples_per_second": 227.2,
+      "eval_steps_per_second": 7.912,
+      "step": 68975
+    },
+    {
+      "epoch": 445.03,
+      "learning_rate": 3.083154121863799e-05,
+      "loss": 0.3543,
+      "step": 68980
+    },
+    {
+      "epoch": 445.1,
+      "learning_rate": 3.0827956989247316e-05,
+      "loss": 0.2328,
+      "step": 68990
+    },
+    {
+      "epoch": 445.16,
+      "learning_rate": 3.0824372759856634e-05,
+      "loss": 0.2268,
+      "step": 69000
+    },
+    {
+      "epoch": 445.23,
+      "learning_rate": 3.082078853046595e-05,
+      "loss": 0.2994,
+      "step": 69010
+    },
+    {
+      "epoch": 445.29,
+      "learning_rate": 3.081720430107527e-05,
+      "loss": 0.2931,
+      "step": 69020
+    },
+    {
+      "epoch": 445.35,
+      "learning_rate": 3.081362007168459e-05,
+      "loss": 0.2652,
+      "step": 69030
+    },
+    {
+      "epoch": 445.42,
+      "learning_rate": 3.0810035842293914e-05,
+      "loss": 0.3273,
+      "step": 69040
+    },
+    {
+      "epoch": 445.48,
+      "learning_rate": 3.0806451612903225e-05,
+      "loss": 0.2655,
+      "step": 69050
+    },
+    {
+      "epoch": 445.55,
+      "learning_rate": 3.080286738351254e-05,
+      "loss": 0.2241,
+      "step": 69060
+    },
+    {
+      "epoch": 445.61,
+      "learning_rate": 3.079928315412187e-05,
+      "loss": 0.2394,
+      "step": 69070
+    },
+    {
+      "epoch": 445.68,
+      "learning_rate": 3.0795698924731186e-05,
+      "loss": 0.2804,
+      "step": 69080
+    },
+    {
+      "epoch": 445.74,
+      "learning_rate": 3.0792114695340505e-05,
+      "loss": 0.2336,
+      "step": 69090
+    },
+    {
+      "epoch": 445.8,
+      "learning_rate": 3.078853046594982e-05,
+      "loss": 0.2824,
+      "step": 69100
+    },
+    {
+      "epoch": 445.87,
+      "learning_rate": 3.078494623655914e-05,
+      "loss": 0.2415,
+      "step": 69110
+    },
+    {
+      "epoch": 445.93,
+      "learning_rate": 3.0781362007168466e-05,
+      "loss": 0.2671,
+      "step": 69120
+    },
+    {
+      "epoch": 446.0,
+      "learning_rate": 3.077777777777778e-05,
+      "loss": 0.2663,
+      "step": 69130
+    },
+    {
+      "epoch": 446.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0085253715515137,
+      "eval_runtime": 0.8836,
+      "eval_samples_per_second": 227.482,
+      "eval_steps_per_second": 7.922,
+      "step": 69130
+    },
+    {
+      "epoch": 446.06,
+      "learning_rate": 3.0774193548387096e-05,
+      "loss": 0.2387,
+      "step": 69140
+    },
+    {
+      "epoch": 446.13,
+      "learning_rate": 3.077060931899642e-05,
+      "loss": 0.2464,
+      "step": 69150
+    },
+    {
+      "epoch": 446.19,
+      "learning_rate": 3.076702508960574e-05,
+      "loss": 0.37,
+      "step": 69160
+    },
+    {
+      "epoch": 446.26,
+      "learning_rate": 3.076344086021505e-05,
+      "loss": 0.295,
+      "step": 69170
+    },
+    {
+      "epoch": 446.32,
+      "learning_rate": 3.0759856630824375e-05,
+      "loss": 0.2621,
+      "step": 69180
+    },
+    {
+      "epoch": 446.39,
+      "learning_rate": 3.0756272401433693e-05,
+      "loss": 0.2991,
+      "step": 69190
+    },
+    {
+      "epoch": 446.45,
+      "learning_rate": 3.075268817204301e-05,
+      "loss": 0.3419,
+      "step": 69200
+    },
+    {
+      "epoch": 446.51,
+      "learning_rate": 3.074910394265233e-05,
+      "loss": 0.2911,
+      "step": 69210
+    },
+    {
+      "epoch": 446.58,
+      "learning_rate": 3.074551971326165e-05,
+      "loss": 0.2584,
+      "step": 69220
+    },
+    {
+      "epoch": 446.64,
+      "learning_rate": 3.074193548387097e-05,
+      "loss": 0.2839,
+      "step": 69230
+    },
+    {
+      "epoch": 446.71,
+      "learning_rate": 3.073835125448029e-05,
+      "loss": 0.2452,
+      "step": 69240
+    },
+    {
+      "epoch": 446.77,
+      "learning_rate": 3.07347670250896e-05,
+      "loss": 0.2509,
+      "step": 69250
+    },
+    {
+      "epoch": 446.84,
+      "learning_rate": 3.073118279569893e-05,
+      "loss": 0.2453,
+      "step": 69260
+    },
+    {
+      "epoch": 446.9,
+      "learning_rate": 3.0727598566308246e-05,
+      "loss": 0.2124,
+      "step": 69270
+    },
+    {
+      "epoch": 446.96,
+      "learning_rate": 3.0724014336917564e-05,
+      "loss": 0.2161,
+      "step": 69280
+    },
+    {
+      "epoch": 447.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.069527268409729,
+      "eval_runtime": 0.8518,
+      "eval_samples_per_second": 235.963,
+      "eval_steps_per_second": 8.218,
+      "step": 69285
+    },
+    {
+      "epoch": 447.03,
+      "learning_rate": 3.072043010752688e-05,
+      "loss": 0.2726,
+      "step": 69290
+    },
+    {
+      "epoch": 447.1,
+      "learning_rate": 3.07168458781362e-05,
+      "loss": 0.259,
+      "step": 69300
+    },
+    {
+      "epoch": 447.16,
+      "learning_rate": 3.0713261648745526e-05,
+      "loss": 0.2463,
+      "step": 69310
+    },
+    {
+      "epoch": 447.23,
+      "learning_rate": 3.070967741935484e-05,
+      "loss": 0.2688,
+      "step": 69320
+    },
+    {
+      "epoch": 447.29,
+      "learning_rate": 3.0706093189964155e-05,
+      "loss": 0.2841,
+      "step": 69330
+    },
+    {
+      "epoch": 447.35,
+      "learning_rate": 3.070250896057348e-05,
+      "loss": 0.3327,
+      "step": 69340
+    },
+    {
+      "epoch": 447.42,
+      "learning_rate": 3.06989247311828e-05,
+      "loss": 0.3102,
+      "step": 69350
+    },
+    {
+      "epoch": 447.48,
+      "learning_rate": 3.0695340501792117e-05,
+      "loss": 0.2007,
+      "step": 69360
+    },
+    {
+      "epoch": 447.55,
+      "learning_rate": 3.0691756272401435e-05,
+      "loss": 0.2574,
+      "step": 69370
+    },
+    {
+      "epoch": 447.61,
+      "learning_rate": 3.068817204301075e-05,
+      "loss": 0.2396,
+      "step": 69380
+    },
+    {
+      "epoch": 447.68,
+      "learning_rate": 3.068458781362008e-05,
+      "loss": 0.267,
+      "step": 69390
+    },
+    {
+      "epoch": 447.74,
+      "learning_rate": 3.068100358422939e-05,
+      "loss": 0.2657,
+      "step": 69400
+    },
+    {
+      "epoch": 447.8,
+      "learning_rate": 3.067741935483871e-05,
+      "loss": 0.2881,
+      "step": 69410
+    },
+    {
+      "epoch": 447.87,
+      "learning_rate": 3.067383512544803e-05,
+      "loss": 0.2469,
+      "step": 69420
+    },
+    {
+      "epoch": 447.93,
+      "learning_rate": 3.067025089605735e-05,
+      "loss": 0.297,
+      "step": 69430
+    },
+    {
+      "epoch": 448.0,
+      "learning_rate": 3.066666666666667e-05,
+      "loss": 0.3211,
+      "step": 69440
+    },
+    {
+      "epoch": 448.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9675971865653992,
+      "eval_runtime": 0.8719,
+      "eval_samples_per_second": 230.54,
+      "eval_steps_per_second": 8.029,
+      "step": 69440
+    },
+    {
+      "epoch": 448.06,
+      "learning_rate": 3.066308243727599e-05,
+      "loss": 0.2266,
+      "step": 69450
+    },
+    {
+      "epoch": 448.13,
+      "learning_rate": 3.0659498207885305e-05,
+      "loss": 0.2845,
+      "step": 69460
+    },
+    {
+      "epoch": 448.19,
+      "learning_rate": 3.065591397849463e-05,
+      "loss": 0.2676,
+      "step": 69470
+    },
+    {
+      "epoch": 448.26,
+      "learning_rate": 3.065232974910394e-05,
+      "loss": 0.2621,
+      "step": 69480
+    },
+    {
+      "epoch": 448.32,
+      "learning_rate": 3.064874551971326e-05,
+      "loss": 0.2591,
+      "step": 69490
+    },
+    {
+      "epoch": 448.39,
+      "learning_rate": 3.0645161290322585e-05,
+      "loss": 0.2466,
+      "step": 69500
+    },
+    {
+      "epoch": 448.45,
+      "learning_rate": 3.06415770609319e-05,
+      "loss": 0.2653,
+      "step": 69510
+    },
+    {
+      "epoch": 448.51,
+      "learning_rate": 3.0637992831541215e-05,
+      "loss": 0.2935,
+      "step": 69520
+    },
+    {
+      "epoch": 448.58,
+      "learning_rate": 3.063440860215054e-05,
+      "loss": 0.2724,
+      "step": 69530
+    },
+    {
+      "epoch": 448.64,
+      "learning_rate": 3.063082437275986e-05,
+      "loss": 0.2293,
+      "step": 69540
+    },
+    {
+      "epoch": 448.71,
+      "learning_rate": 3.0627240143369176e-05,
+      "loss": 0.2663,
+      "step": 69550
+    },
+    {
+      "epoch": 448.77,
+      "learning_rate": 3.0623655913978494e-05,
+      "loss": 0.2829,
+      "step": 69560
+    },
+    {
+      "epoch": 448.84,
+      "learning_rate": 3.062007168458781e-05,
+      "loss": 0.2391,
+      "step": 69570
+    },
+    {
+      "epoch": 448.9,
+      "learning_rate": 3.061648745519714e-05,
+      "loss": 0.2289,
+      "step": 69580
+    },
+    {
+      "epoch": 448.96,
+      "learning_rate": 3.0612903225806456e-05,
+      "loss": 0.2666,
+      "step": 69590
+    },
+    {
+      "epoch": 449.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9461472630500793,
+      "eval_runtime": 0.8707,
+      "eval_samples_per_second": 230.858,
+      "eval_steps_per_second": 8.04,
+      "step": 69595
+    },
+    {
+      "epoch": 449.03,
+      "learning_rate": 3.060931899641577e-05,
+      "loss": 0.2428,
+      "step": 69600
+    },
+    {
+      "epoch": 449.1,
+      "learning_rate": 3.060573476702509e-05,
+      "loss": 0.2317,
+      "step": 69610
+    },
+    {
+      "epoch": 449.16,
+      "learning_rate": 3.060215053763441e-05,
+      "loss": 0.2971,
+      "step": 69620
+    },
+    {
+      "epoch": 449.23,
+      "learning_rate": 3.059856630824373e-05,
+      "loss": 0.3057,
+      "step": 69630
+    },
+    {
+      "epoch": 449.29,
+      "learning_rate": 3.0594982078853047e-05,
+      "loss": 0.2661,
+      "step": 69640
+    },
+    {
+      "epoch": 449.35,
+      "learning_rate": 3.0591397849462365e-05,
+      "loss": 0.2836,
+      "step": 69650
+    },
+    {
+      "epoch": 449.42,
+      "learning_rate": 3.058781362007169e-05,
+      "loss": 0.2353,
+      "step": 69660
+    },
+    {
+      "epoch": 449.48,
+      "learning_rate": 3.058422939068101e-05,
+      "loss": 0.2901,
+      "step": 69670
+    },
+    {
+      "epoch": 449.55,
+      "learning_rate": 3.058064516129032e-05,
+      "loss": 0.2554,
+      "step": 69680
+    },
+    {
+      "epoch": 449.61,
+      "learning_rate": 3.0577060931899644e-05,
+      "loss": 0.239,
+      "step": 69690
+    },
+    {
+      "epoch": 449.68,
+      "learning_rate": 3.057347670250896e-05,
+      "loss": 0.2838,
+      "step": 69700
+    },
+    {
+      "epoch": 449.74,
+      "learning_rate": 3.056989247311828e-05,
+      "loss": 0.2739,
+      "step": 69710
+    },
+    {
+      "epoch": 449.8,
+      "learning_rate": 3.05663082437276e-05,
+      "loss": 0.2718,
+      "step": 69720
+    },
+    {
+      "epoch": 449.87,
+      "learning_rate": 3.056272401433692e-05,
+      "loss": 0.2915,
+      "step": 69730
+    },
+    {
+      "epoch": 449.93,
+      "learning_rate": 3.055913978494624e-05,
+      "loss": 0.2847,
+      "step": 69740
+    },
+    {
+      "epoch": 450.0,
+      "learning_rate": 3.055555555555556e-05,
+      "loss": 0.2566,
+      "step": 69750
+    },
+    {
+      "epoch": 450.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9784051179885864,
+      "eval_runtime": 0.8735,
+      "eval_samples_per_second": 230.108,
+      "eval_steps_per_second": 8.014,
+      "step": 69750
+    },
+    {
+      "epoch": 450.06,
+      "learning_rate": 3.055197132616487e-05,
+      "loss": 0.2929,
+      "step": 69760
+    },
+    {
+      "epoch": 450.13,
+      "learning_rate": 3.05483870967742e-05,
+      "loss": 0.2919,
+      "step": 69770
+    },
+    {
+      "epoch": 450.19,
+      "learning_rate": 3.0544802867383515e-05,
+      "loss": 0.2798,
+      "step": 69780
+    },
+    {
+      "epoch": 450.26,
+      "learning_rate": 3.054121863799283e-05,
+      "loss": 0.3057,
+      "step": 69790
+    },
+    {
+      "epoch": 450.32,
+      "learning_rate": 3.053763440860215e-05,
+      "loss": 0.3082,
+      "step": 69800
+    },
+    {
+      "epoch": 450.39,
+      "learning_rate": 3.053405017921147e-05,
+      "loss": 0.243,
+      "step": 69810
+    },
+    {
+      "epoch": 450.45,
+      "learning_rate": 3.053046594982079e-05,
+      "loss": 0.247,
+      "step": 69820
+    },
+    {
+      "epoch": 450.51,
+      "learning_rate": 3.052688172043011e-05,
+      "loss": 0.2204,
+      "step": 69830
+    },
+    {
+      "epoch": 450.58,
+      "learning_rate": 3.0523297491039424e-05,
+      "loss": 0.263,
+      "step": 69840
+    },
+    {
+      "epoch": 450.64,
+      "learning_rate": 3.051971326164875e-05,
+      "loss": 0.2613,
+      "step": 69850
+    },
+    {
+      "epoch": 450.71,
+      "learning_rate": 3.0516129032258067e-05,
+      "loss": 0.2164,
+      "step": 69860
+    },
+    {
+      "epoch": 450.77,
+      "learning_rate": 3.051254480286739e-05,
+      "loss": 0.2985,
+      "step": 69870
+    },
+    {
+      "epoch": 450.84,
+      "learning_rate": 3.0508960573476704e-05,
+      "loss": 0.202,
+      "step": 69880
+    },
+    {
+      "epoch": 450.9,
+      "learning_rate": 3.0505376344086022e-05,
+      "loss": 0.2511,
+      "step": 69890
+    },
+    {
+      "epoch": 450.96,
+      "learning_rate": 3.0501792114695344e-05,
+      "loss": 0.2469,
+      "step": 69900
+    },
+    {
+      "epoch": 451.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9609711170196533,
+      "eval_runtime": 0.9152,
+      "eval_samples_per_second": 219.619,
+      "eval_steps_per_second": 7.648,
+      "step": 69905
+    },
+    {
+      "epoch": 451.03,
+      "learning_rate": 3.049820788530466e-05,
+      "loss": 0.255,
+      "step": 69910
+    },
+    {
+      "epoch": 451.1,
+      "learning_rate": 3.049462365591398e-05,
+      "loss": 0.2513,
+      "step": 69920
+    },
+    {
+      "epoch": 451.16,
+      "learning_rate": 3.0491039426523298e-05,
+      "loss": 0.2656,
+      "step": 69930
+    },
+    {
+      "epoch": 451.23,
+      "learning_rate": 3.048745519713262e-05,
+      "loss": 0.2913,
+      "step": 69940
+    },
+    {
+      "epoch": 451.29,
+      "learning_rate": 3.0483870967741935e-05,
+      "loss": 0.2739,
+      "step": 69950
+    },
+    {
+      "epoch": 451.35,
+      "learning_rate": 3.0480286738351256e-05,
+      "loss": 0.2835,
+      "step": 69960
+    },
+    {
+      "epoch": 451.42,
+      "learning_rate": 3.0476702508960574e-05,
+      "loss": 0.3266,
+      "step": 69970
+    },
+    {
+      "epoch": 451.48,
+      "learning_rate": 3.0473118279569896e-05,
+      "loss": 0.2654,
+      "step": 69980
+    },
+    {
+      "epoch": 451.55,
+      "learning_rate": 3.046953405017921e-05,
+      "loss": 0.2065,
+      "step": 69990
+    },
+    {
+      "epoch": 451.61,
+      "learning_rate": 3.046594982078853e-05,
+      "loss": 0.3169,
+      "step": 70000
+    },
+    {
+      "epoch": 451.68,
+      "learning_rate": 3.046236559139785e-05,
+      "loss": 0.2697,
+      "step": 70010
+    },
+    {
+      "epoch": 451.74,
+      "learning_rate": 3.0458781362007172e-05,
+      "loss": 0.2768,
+      "step": 70020
+    },
+    {
+      "epoch": 451.8,
+      "learning_rate": 3.0455197132616487e-05,
+      "loss": 0.2574,
+      "step": 70030
+    },
+    {
+      "epoch": 451.87,
+      "learning_rate": 3.0451612903225805e-05,
+      "loss": 0.2889,
+      "step": 70040
+    },
+    {
+      "epoch": 451.93,
+      "learning_rate": 3.0448028673835127e-05,
+      "loss": 0.2465,
+      "step": 70050
+    },
+    {
+      "epoch": 452.0,
+      "learning_rate": 3.044444444444445e-05,
+      "loss": 0.2825,
+      "step": 70060
+    },
+    {
+      "epoch": 452.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9984765648841858,
+      "eval_runtime": 0.8799,
+      "eval_samples_per_second": 228.424,
+      "eval_steps_per_second": 7.955,
+      "step": 70060
+    },
+    {
+      "epoch": 452.06,
+      "learning_rate": 3.0440860215053763e-05,
+      "loss": 0.2678,
+      "step": 70070
+    },
+    {
+      "epoch": 452.13,
+      "learning_rate": 3.043727598566308e-05,
+      "loss": 0.2209,
+      "step": 70080
+    },
+    {
+      "epoch": 452.19,
+      "learning_rate": 3.0433691756272403e-05,
+      "loss": 0.2719,
+      "step": 70090
+    },
+    {
+      "epoch": 452.26,
+      "learning_rate": 3.0430107526881725e-05,
+      "loss": 0.235,
+      "step": 70100
+    },
+    {
+      "epoch": 452.32,
+      "learning_rate": 3.042652329749104e-05,
+      "loss": 0.2605,
+      "step": 70110
+    },
+    {
+      "epoch": 452.39,
+      "learning_rate": 3.0422939068100358e-05,
+      "loss": 0.2388,
+      "step": 70120
+    },
+    {
+      "epoch": 452.45,
+      "learning_rate": 3.041935483870968e-05,
+      "loss": 0.2924,
+      "step": 70130
+    },
+    {
+      "epoch": 452.51,
+      "learning_rate": 3.0415770609319e-05,
+      "loss": 0.288,
+      "step": 70140
+    },
+    {
+      "epoch": 452.58,
+      "learning_rate": 3.0412186379928316e-05,
+      "loss": 0.267,
+      "step": 70150
+    },
+    {
+      "epoch": 452.64,
+      "learning_rate": 3.0408602150537634e-05,
+      "loss": 0.2659,
+      "step": 70160
+    },
+    {
+      "epoch": 452.71,
+      "learning_rate": 3.0405017921146956e-05,
+      "loss": 0.2741,
+      "step": 70170
+    },
+    {
+      "epoch": 452.77,
+      "learning_rate": 3.0401433691756277e-05,
+      "loss": 0.2689,
+      "step": 70180
+    },
+    {
+      "epoch": 452.84,
+      "learning_rate": 3.0397849462365592e-05,
+      "loss": 0.2997,
+      "step": 70190
+    },
+    {
+      "epoch": 452.9,
+      "learning_rate": 3.039426523297491e-05,
+      "loss": 0.2506,
+      "step": 70200
+    },
+    {
+      "epoch": 452.96,
+      "learning_rate": 3.0390681003584232e-05,
+      "loss": 0.2897,
+      "step": 70210
+    },
+    {
+      "epoch": 453.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.004218578338623,
+      "eval_runtime": 0.8631,
+      "eval_samples_per_second": 232.89,
+      "eval_steps_per_second": 8.111,
+      "step": 70215
+    },
+    {
+      "epoch": 453.03,
+      "learning_rate": 3.0387096774193553e-05,
+      "loss": 0.2145,
+      "step": 70220
+    },
+    {
+      "epoch": 453.1,
+      "learning_rate": 3.0383512544802868e-05,
+      "loss": 0.339,
+      "step": 70230
+    },
+    {
+      "epoch": 453.16,
+      "learning_rate": 3.0379928315412186e-05,
+      "loss": 0.2696,
+      "step": 70240
+    },
+    {
+      "epoch": 453.23,
+      "learning_rate": 3.0376344086021508e-05,
+      "loss": 0.2554,
+      "step": 70250
+    },
+    {
+      "epoch": 453.29,
+      "learning_rate": 3.037275985663083e-05,
+      "loss": 0.2238,
+      "step": 70260
+    },
+    {
+      "epoch": 453.35,
+      "learning_rate": 3.0369175627240144e-05,
+      "loss": 0.2535,
+      "step": 70270
+    },
+    {
+      "epoch": 453.42,
+      "learning_rate": 3.0365591397849463e-05,
+      "loss": 0.2667,
+      "step": 70280
+    },
+    {
+      "epoch": 453.48,
+      "learning_rate": 3.0362007168458784e-05,
+      "loss": 0.2678,
+      "step": 70290
+    },
+    {
+      "epoch": 453.55,
+      "learning_rate": 3.0358422939068106e-05,
+      "loss": 0.2707,
+      "step": 70300
+    },
+    {
+      "epoch": 453.61,
+      "learning_rate": 3.035483870967742e-05,
+      "loss": 0.2644,
+      "step": 70310
+    },
+    {
+      "epoch": 453.68,
+      "learning_rate": 3.035125448028674e-05,
+      "loss": 0.2665,
+      "step": 70320
+    },
+    {
+      "epoch": 453.74,
+      "learning_rate": 3.034767025089606e-05,
+      "loss": 0.2026,
+      "step": 70330
+    },
+    {
+      "epoch": 453.8,
+      "learning_rate": 3.034408602150538e-05,
+      "loss": 0.3209,
+      "step": 70340
+    },
+    {
+      "epoch": 453.87,
+      "learning_rate": 3.0340501792114693e-05,
+      "loss": 0.2628,
+      "step": 70350
+    },
+    {
+      "epoch": 453.93,
+      "learning_rate": 3.0336917562724015e-05,
+      "loss": 0.2844,
+      "step": 70360
+    },
+    {
+      "epoch": 454.0,
+      "learning_rate": 3.0333333333333337e-05,
+      "loss": 0.2362,
+      "step": 70370
+    },
+    {
+      "epoch": 454.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9811331033706665,
+      "eval_runtime": 0.8781,
+      "eval_samples_per_second": 228.914,
+      "eval_steps_per_second": 7.972,
+      "step": 70370
+    },
+    {
+      "epoch": 454.06,
+      "learning_rate": 3.0329749103942655e-05,
+      "loss": 0.2436,
+      "step": 70380
+    },
+    {
+      "epoch": 454.13,
+      "learning_rate": 3.032616487455197e-05,
+      "loss": 0.2559,
+      "step": 70390
+    },
+    {
+      "epoch": 454.19,
+      "learning_rate": 3.032258064516129e-05,
+      "loss": 0.2768,
+      "step": 70400
+    },
+    {
+      "epoch": 454.26,
+      "learning_rate": 3.0318996415770613e-05,
+      "loss": 0.2526,
+      "step": 70410
+    },
+    {
+      "epoch": 454.32,
+      "learning_rate": 3.031541218637993e-05,
+      "loss": 0.2961,
+      "step": 70420
+    },
+    {
+      "epoch": 454.39,
+      "learning_rate": 3.0311827956989246e-05,
+      "loss": 0.2726,
+      "step": 70430
+    },
+    {
+      "epoch": 454.45,
+      "learning_rate": 3.0308243727598567e-05,
+      "loss": 0.2612,
+      "step": 70440
+    },
+    {
+      "epoch": 454.51,
+      "learning_rate": 3.030465949820789e-05,
+      "loss": 0.3285,
+      "step": 70450
+    },
+    {
+      "epoch": 454.58,
+      "learning_rate": 3.0301075268817207e-05,
+      "loss": 0.2433,
+      "step": 70460
+    },
+    {
+      "epoch": 454.64,
+      "learning_rate": 3.0297491039426522e-05,
+      "loss": 0.2469,
+      "step": 70470
+    },
+    {
+      "epoch": 454.71,
+      "learning_rate": 3.0293906810035844e-05,
+      "loss": 0.3036,
+      "step": 70480
+    },
+    {
+      "epoch": 454.77,
+      "learning_rate": 3.0290322580645165e-05,
+      "loss": 0.2529,
+      "step": 70490
+    },
+    {
+      "epoch": 454.84,
+      "learning_rate": 3.0286738351254483e-05,
+      "loss": 0.3135,
+      "step": 70500
+    },
+    {
+      "epoch": 454.9,
+      "learning_rate": 3.0283154121863798e-05,
+      "loss": 0.2833,
+      "step": 70510
+    },
+    {
+      "epoch": 454.96,
+      "learning_rate": 3.027956989247312e-05,
+      "loss": 0.221,
+      "step": 70520
+    },
+    {
+      "epoch": 455.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0014840364456177,
+      "eval_runtime": 0.8872,
+      "eval_samples_per_second": 226.546,
+      "eval_steps_per_second": 7.89,
+      "step": 70525
+    },
+    {
+      "epoch": 455.03,
+      "learning_rate": 3.027598566308244e-05,
+      "loss": 0.3008,
+      "step": 70530
+    },
+    {
+      "epoch": 455.1,
+      "learning_rate": 3.0272401433691756e-05,
+      "loss": 0.2381,
+      "step": 70540
+    },
+    {
+      "epoch": 455.16,
+      "learning_rate": 3.0268817204301074e-05,
+      "loss": 0.2786,
+      "step": 70550
+    },
+    {
+      "epoch": 455.23,
+      "learning_rate": 3.0265232974910396e-05,
+      "loss": 0.2521,
+      "step": 70560
+    },
+    {
+      "epoch": 455.29,
+      "learning_rate": 3.0261648745519718e-05,
+      "loss": 0.2606,
+      "step": 70570
+    },
+    {
+      "epoch": 455.35,
+      "learning_rate": 3.0258064516129032e-05,
+      "loss": 0.2711,
+      "step": 70580
+    },
+    {
+      "epoch": 455.42,
+      "learning_rate": 3.025448028673835e-05,
+      "loss": 0.2484,
+      "step": 70590
+    },
+    {
+      "epoch": 455.48,
+      "learning_rate": 3.0250896057347672e-05,
+      "loss": 0.2464,
+      "step": 70600
+    },
+    {
+      "epoch": 455.55,
+      "learning_rate": 3.0247311827956994e-05,
+      "loss": 0.2993,
+      "step": 70610
+    },
+    {
+      "epoch": 455.61,
+      "learning_rate": 3.024372759856631e-05,
+      "loss": 0.233,
+      "step": 70620
+    },
+    {
+      "epoch": 455.68,
+      "learning_rate": 3.0240143369175627e-05,
+      "loss": 0.2302,
+      "step": 70630
+    },
+    {
+      "epoch": 455.74,
+      "learning_rate": 3.023655913978495e-05,
+      "loss": 0.2662,
+      "step": 70640
+    },
+    {
+      "epoch": 455.8,
+      "learning_rate": 3.023297491039427e-05,
+      "loss": 0.2594,
+      "step": 70650
+    },
+    {
+      "epoch": 455.87,
+      "learning_rate": 3.0229390681003585e-05,
+      "loss": 0.3032,
+      "step": 70660
+    },
+    {
+      "epoch": 455.93,
+      "learning_rate": 3.0225806451612903e-05,
+      "loss": 0.268,
+      "step": 70670
+    },
+    {
+      "epoch": 456.0,
+      "learning_rate": 3.0222222222222225e-05,
+      "loss": 0.2966,
+      "step": 70680
+    },
+    {
+      "epoch": 456.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9985130429267883,
+      "eval_runtime": 0.9461,
+      "eval_samples_per_second": 212.45,
+      "eval_steps_per_second": 7.399,
+      "step": 70680
+    },
+    {
+      "epoch": 456.06,
+      "learning_rate": 3.0218637992831543e-05,
+      "loss": 0.2379,
+      "step": 70690
+    },
+    {
+      "epoch": 456.13,
+      "learning_rate": 3.0215053763440858e-05,
+      "loss": 0.2455,
+      "step": 70700
+    },
+    {
+      "epoch": 456.19,
+      "learning_rate": 3.021146953405018e-05,
+      "loss": 0.2527,
+      "step": 70710
+    },
+    {
+      "epoch": 456.26,
+      "learning_rate": 3.02078853046595e-05,
+      "loss": 0.2912,
+      "step": 70720
+    },
+    {
+      "epoch": 456.32,
+      "learning_rate": 3.020430107526882e-05,
+      "loss": 0.2917,
+      "step": 70730
+    },
+    {
+      "epoch": 456.39,
+      "learning_rate": 3.0200716845878134e-05,
+      "loss": 0.307,
+      "step": 70740
+    },
+    {
+      "epoch": 456.45,
+      "learning_rate": 3.0197132616487455e-05,
+      "loss": 0.2463,
+      "step": 70750
+    },
+    {
+      "epoch": 456.51,
+      "learning_rate": 3.0193548387096777e-05,
+      "loss": 0.2843,
+      "step": 70760
+    },
+    {
+      "epoch": 456.58,
+      "learning_rate": 3.0189964157706095e-05,
+      "loss": 0.3125,
+      "step": 70770
+    },
+    {
+      "epoch": 456.64,
+      "learning_rate": 3.018637992831541e-05,
+      "loss": 0.3094,
+      "step": 70780
+    },
+    {
+      "epoch": 456.71,
+      "learning_rate": 3.018279569892473e-05,
+      "loss": 0.3131,
+      "step": 70790
+    },
+    {
+      "epoch": 456.77,
+      "learning_rate": 3.0179211469534053e-05,
+      "loss": 0.263,
+      "step": 70800
+    },
+    {
+      "epoch": 456.84,
+      "learning_rate": 3.017562724014337e-05,
+      "loss": 0.3248,
+      "step": 70810
+    },
+    {
+      "epoch": 456.9,
+      "learning_rate": 3.0172043010752686e-05,
+      "loss": 0.269,
+      "step": 70820
+    },
+    {
+      "epoch": 456.96,
+      "learning_rate": 3.0168458781362008e-05,
+      "loss": 0.2642,
+      "step": 70830
+    },
+    {
+      "epoch": 457.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0560003519058228,
+      "eval_runtime": 0.8532,
+      "eval_samples_per_second": 235.587,
+      "eval_steps_per_second": 8.205,
+      "step": 70835
+    },
+    {
+      "epoch": 457.03,
+      "learning_rate": 3.016487455197133e-05,
+      "loss": 0.3007,
+      "step": 70840
+    },
+    {
+      "epoch": 457.1,
+      "learning_rate": 3.0161290322580648e-05,
+      "loss": 0.2618,
+      "step": 70850
+    },
+    {
+      "epoch": 457.16,
+      "learning_rate": 3.0157706093189962e-05,
+      "loss": 0.2857,
+      "step": 70860
+    },
+    {
+      "epoch": 457.23,
+      "learning_rate": 3.0154121863799284e-05,
+      "loss": 0.264,
+      "step": 70870
+    },
+    {
+      "epoch": 457.29,
+      "learning_rate": 3.0150537634408606e-05,
+      "loss": 0.2448,
+      "step": 70880
+    },
+    {
+      "epoch": 457.35,
+      "learning_rate": 3.0146953405017924e-05,
+      "loss": 0.2017,
+      "step": 70890
+    },
+    {
+      "epoch": 457.42,
+      "learning_rate": 3.014336917562724e-05,
+      "loss": 0.2534,
+      "step": 70900
+    },
+    {
+      "epoch": 457.48,
+      "learning_rate": 3.013978494623656e-05,
+      "loss": 0.2984,
+      "step": 70910
+    },
+    {
+      "epoch": 457.55,
+      "learning_rate": 3.0136200716845882e-05,
+      "loss": 0.2834,
+      "step": 70920
+    },
+    {
+      "epoch": 457.61,
+      "learning_rate": 3.01326164874552e-05,
+      "loss": 0.2229,
+      "step": 70930
+    },
+    {
+      "epoch": 457.68,
+      "learning_rate": 3.0129032258064515e-05,
+      "loss": 0.2823,
+      "step": 70940
+    },
+    {
+      "epoch": 457.74,
+      "learning_rate": 3.0125448028673836e-05,
+      "loss": 0.2592,
+      "step": 70950
+    },
+    {
+      "epoch": 457.8,
+      "learning_rate": 3.0121863799283158e-05,
+      "loss": 0.267,
+      "step": 70960
+    },
+    {
+      "epoch": 457.87,
+      "learning_rate": 3.0118279569892476e-05,
+      "loss": 0.2673,
+      "step": 70970
+    },
+    {
+      "epoch": 457.93,
+      "learning_rate": 3.011469534050179e-05,
+      "loss": 0.2097,
+      "step": 70980
+    },
+    {
+      "epoch": 458.0,
+      "learning_rate": 3.0111111111111113e-05,
+      "loss": 0.2271,
+      "step": 70990
+    },
+    {
+      "epoch": 458.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0303505659103394,
+      "eval_runtime": 0.8579,
+      "eval_samples_per_second": 234.291,
+      "eval_steps_per_second": 8.159,
+      "step": 70990
+    },
+    {
+      "epoch": 458.06,
+      "learning_rate": 3.010752688172043e-05,
+      "loss": 0.287,
+      "step": 71000
+    },
+    {
+      "epoch": 458.13,
+      "learning_rate": 3.0103942652329752e-05,
+      "loss": 0.2749,
+      "step": 71010
+    },
+    {
+      "epoch": 458.19,
+      "learning_rate": 3.0100358422939067e-05,
+      "loss": 0.243,
+      "step": 71020
+    },
+    {
+      "epoch": 458.26,
+      "learning_rate": 3.009677419354839e-05,
+      "loss": 0.1958,
+      "step": 71030
+    },
+    {
+      "epoch": 458.32,
+      "learning_rate": 3.0093189964157707e-05,
+      "loss": 0.289,
+      "step": 71040
+    },
+    {
+      "epoch": 458.39,
+      "learning_rate": 3.008960573476703e-05,
+      "loss": 0.2553,
+      "step": 71050
+    },
+    {
+      "epoch": 458.45,
+      "learning_rate": 3.0086021505376344e-05,
+      "loss": 0.2797,
+      "step": 71060
+    },
+    {
+      "epoch": 458.51,
+      "learning_rate": 3.0082437275985665e-05,
+      "loss": 0.3164,
+      "step": 71070
+    },
+    {
+      "epoch": 458.58,
+      "learning_rate": 3.0078853046594983e-05,
+      "loss": 0.2359,
+      "step": 71080
+    },
+    {
+      "epoch": 458.64,
+      "learning_rate": 3.0075268817204305e-05,
+      "loss": 0.2318,
+      "step": 71090
+    },
+    {
+      "epoch": 458.71,
+      "learning_rate": 3.007168458781362e-05,
+      "loss": 0.3032,
+      "step": 71100
+    },
+    {
+      "epoch": 458.77,
+      "learning_rate": 3.006810035842294e-05,
+      "loss": 0.2803,
+      "step": 71110
+    },
+    {
+      "epoch": 458.84,
+      "learning_rate": 3.006451612903226e-05,
+      "loss": 0.2345,
+      "step": 71120
+    },
+    {
+      "epoch": 458.9,
+      "learning_rate": 3.006093189964158e-05,
+      "loss": 0.2237,
+      "step": 71130
+    },
+    {
+      "epoch": 458.96,
+      "learning_rate": 3.0057347670250896e-05,
+      "loss": 0.2871,
+      "step": 71140
+    },
+    {
+      "epoch": 459.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0907621383666992,
+      "eval_runtime": 0.8827,
+      "eval_samples_per_second": 227.72,
+      "eval_steps_per_second": 7.931,
+      "step": 71145
+    },
+    {
+      "epoch": 459.03,
+      "learning_rate": 3.0053763440860218e-05,
+      "loss": 0.2843,
+      "step": 71150
+    },
+    {
+      "epoch": 459.1,
+      "learning_rate": 3.0050179211469536e-05,
+      "loss": 0.2575,
+      "step": 71160
+    },
+    {
+      "epoch": 459.16,
+      "learning_rate": 3.0046594982078857e-05,
+      "loss": 0.2692,
+      "step": 71170
+    },
+    {
+      "epoch": 459.23,
+      "learning_rate": 3.0043010752688172e-05,
+      "loss": 0.2971,
+      "step": 71180
+    },
+    {
+      "epoch": 459.29,
+      "learning_rate": 3.0039426523297494e-05,
+      "loss": 0.2534,
+      "step": 71190
+    },
+    {
+      "epoch": 459.35,
+      "learning_rate": 3.0035842293906812e-05,
+      "loss": 0.2563,
+      "step": 71200
+    },
+    {
+      "epoch": 459.42,
+      "learning_rate": 3.0032258064516127e-05,
+      "loss": 0.2652,
+      "step": 71210
+    },
+    {
+      "epoch": 459.48,
+      "learning_rate": 3.002867383512545e-05,
+      "loss": 0.2665,
+      "step": 71220
+    },
+    {
+      "epoch": 459.55,
+      "learning_rate": 3.002508960573477e-05,
+      "loss": 0.2995,
+      "step": 71230
+    },
+    {
+      "epoch": 459.61,
+      "learning_rate": 3.0021505376344088e-05,
+      "loss": 0.258,
+      "step": 71240
+    },
+    {
+      "epoch": 459.68,
+      "learning_rate": 3.0017921146953403e-05,
+      "loss": 0.2418,
+      "step": 71250
+    },
+    {
+      "epoch": 459.74,
+      "learning_rate": 3.0014336917562725e-05,
+      "loss": 0.2641,
+      "step": 71260
+    },
+    {
+      "epoch": 459.8,
+      "learning_rate": 3.0010752688172046e-05,
+      "loss": 0.2461,
+      "step": 71270
+    },
+    {
+      "epoch": 459.87,
+      "learning_rate": 3.0007168458781364e-05,
+      "loss": 0.2612,
+      "step": 71280
+    },
+    {
+      "epoch": 459.93,
+      "learning_rate": 3.000358422939068e-05,
+      "loss": 0.2759,
+      "step": 71290
+    },
+    {
+      "epoch": 460.0,
+      "learning_rate": 3e-05,
+      "loss": 0.2897,
+      "step": 71300
+    },
+    {
+      "epoch": 460.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0374939441680908,
+      "eval_runtime": 0.8482,
+      "eval_samples_per_second": 236.976,
+      "eval_steps_per_second": 8.253,
+      "step": 71300
+    },
+    {
+      "epoch": 460.06,
+      "learning_rate": 2.9996415770609322e-05,
+      "loss": 0.2879,
+      "step": 71310
+    },
+    {
+      "epoch": 460.13,
+      "learning_rate": 2.999283154121864e-05,
+      "loss": 0.3043,
+      "step": 71320
+    },
+    {
+      "epoch": 460.19,
+      "learning_rate": 2.9989247311827955e-05,
+      "loss": 0.2178,
+      "step": 71330
+    },
+    {
+      "epoch": 460.26,
+      "learning_rate": 2.9985663082437277e-05,
+      "loss": 0.2795,
+      "step": 71340
+    },
+    {
+      "epoch": 460.32,
+      "learning_rate": 2.9982078853046595e-05,
+      "loss": 0.251,
+      "step": 71350
+    },
+    {
+      "epoch": 460.39,
+      "learning_rate": 2.9978494623655917e-05,
+      "loss": 0.2644,
+      "step": 71360
+    },
+    {
+      "epoch": 460.45,
+      "learning_rate": 2.997491039426523e-05,
+      "loss": 0.2857,
+      "step": 71370
+    },
+    {
+      "epoch": 460.51,
+      "learning_rate": 2.9971326164874553e-05,
+      "loss": 0.2752,
+      "step": 71380
+    },
+    {
+      "epoch": 460.58,
+      "learning_rate": 2.996774193548387e-05,
+      "loss": 0.2629,
+      "step": 71390
+    },
+    {
+      "epoch": 460.64,
+      "learning_rate": 2.9964157706093193e-05,
+      "loss": 0.2843,
+      "step": 71400
+    },
+    {
+      "epoch": 460.71,
+      "learning_rate": 2.9960573476702508e-05,
+      "loss": 0.3097,
+      "step": 71410
+    },
+    {
+      "epoch": 460.77,
+      "learning_rate": 2.995698924731183e-05,
+      "loss": 0.2279,
+      "step": 71420
+    },
+    {
+      "epoch": 460.84,
+      "learning_rate": 2.9953405017921148e-05,
+      "loss": 0.2382,
+      "step": 71430
+    },
+    {
+      "epoch": 460.9,
+      "learning_rate": 2.994982078853047e-05,
+      "loss": 0.268,
+      "step": 71440
+    },
+    {
+      "epoch": 460.96,
+      "learning_rate": 2.9946236559139784e-05,
+      "loss": 0.253,
+      "step": 71450
+    },
+    {
+      "epoch": 461.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0764497518539429,
+      "eval_runtime": 0.8862,
+      "eval_samples_per_second": 226.813,
+      "eval_steps_per_second": 7.899,
+      "step": 71455
+    },
+    {
+      "epoch": 461.03,
+      "learning_rate": 2.9942652329749106e-05,
+      "loss": 0.293,
+      "step": 71460
+    },
+    {
+      "epoch": 461.1,
+      "learning_rate": 2.9939068100358424e-05,
+      "loss": 0.2143,
+      "step": 71470
+    },
+    {
+      "epoch": 461.16,
+      "learning_rate": 2.9935483870967745e-05,
+      "loss": 0.2439,
+      "step": 71480
+    },
+    {
+      "epoch": 461.23,
+      "learning_rate": 2.993189964157706e-05,
+      "loss": 0.2874,
+      "step": 71490
+    },
+    {
+      "epoch": 461.29,
+      "learning_rate": 2.9928315412186382e-05,
+      "loss": 0.2585,
+      "step": 71500
+    },
+    {
+      "epoch": 461.35,
+      "learning_rate": 2.99247311827957e-05,
+      "loss": 0.2449,
+      "step": 71510
+    },
+    {
+      "epoch": 461.42,
+      "learning_rate": 2.992114695340502e-05,
+      "loss": 0.3055,
+      "step": 71520
+    },
+    {
+      "epoch": 461.48,
+      "learning_rate": 2.9917562724014336e-05,
+      "loss": 0.2621,
+      "step": 71530
+    },
+    {
+      "epoch": 461.55,
+      "learning_rate": 2.9913978494623658e-05,
+      "loss": 0.2364,
+      "step": 71540
+    },
+    {
+      "epoch": 461.61,
+      "learning_rate": 2.9910394265232976e-05,
+      "loss": 0.2887,
+      "step": 71550
+    },
+    {
+      "epoch": 461.68,
+      "learning_rate": 2.9906810035842298e-05,
+      "loss": 0.255,
+      "step": 71560
+    },
+    {
+      "epoch": 461.74,
+      "learning_rate": 2.9903225806451613e-05,
+      "loss": 0.271,
+      "step": 71570
+    },
+    {
+      "epoch": 461.8,
+      "learning_rate": 2.9899641577060934e-05,
+      "loss": 0.2957,
+      "step": 71580
+    },
+    {
+      "epoch": 461.87,
+      "learning_rate": 2.9896057347670252e-05,
+      "loss": 0.2491,
+      "step": 71590
+    },
+    {
+      "epoch": 461.93,
+      "learning_rate": 2.9892473118279574e-05,
+      "loss": 0.2004,
+      "step": 71600
+    },
+    {
+      "epoch": 462.0,
+      "learning_rate": 2.988888888888889e-05,
+      "loss": 0.3055,
+      "step": 71610
+    },
+    {
+      "epoch": 462.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.003127932548523,
+      "eval_runtime": 0.8577,
+      "eval_samples_per_second": 234.36,
+      "eval_steps_per_second": 8.162,
+      "step": 71610
+    },
+    {
+      "epoch": 462.06,
+      "learning_rate": 2.988530465949821e-05,
+      "loss": 0.3123,
+      "step": 71620
+    },
+    {
+      "epoch": 462.13,
+      "learning_rate": 2.988172043010753e-05,
+      "loss": 0.2666,
+      "step": 71630
+    },
+    {
+      "epoch": 462.19,
+      "learning_rate": 2.987813620071685e-05,
+      "loss": 0.2933,
+      "step": 71640
+    },
+    {
+      "epoch": 462.26,
+      "learning_rate": 2.9874551971326165e-05,
+      "loss": 0.3059,
+      "step": 71650
+    },
+    {
+      "epoch": 462.32,
+      "learning_rate": 2.9870967741935487e-05,
+      "loss": 0.2789,
+      "step": 71660
+    },
+    {
+      "epoch": 462.39,
+      "learning_rate": 2.9867383512544805e-05,
+      "loss": 0.3188,
+      "step": 71670
+    },
+    {
+      "epoch": 462.45,
+      "learning_rate": 2.9863799283154126e-05,
+      "loss": 0.3289,
+      "step": 71680
+    },
+    {
+      "epoch": 462.51,
+      "learning_rate": 2.986021505376344e-05,
+      "loss": 0.2622,
+      "step": 71690
+    },
+    {
+      "epoch": 462.58,
+      "learning_rate": 2.985663082437276e-05,
+      "loss": 0.274,
+      "step": 71700
+    },
+    {
+      "epoch": 462.64,
+      "learning_rate": 2.985304659498208e-05,
+      "loss": 0.273,
+      "step": 71710
+    },
+    {
+      "epoch": 462.71,
+      "learning_rate": 2.9849462365591403e-05,
+      "loss": 0.2586,
+      "step": 71720
+    },
+    {
+      "epoch": 462.77,
+      "learning_rate": 2.9845878136200717e-05,
+      "loss": 0.2722,
+      "step": 71730
+    },
+    {
+      "epoch": 462.84,
+      "learning_rate": 2.9842293906810036e-05,
+      "loss": 0.2699,
+      "step": 71740
+    },
+    {
+      "epoch": 462.9,
+      "learning_rate": 2.9838709677419357e-05,
+      "loss": 0.3171,
+      "step": 71750
+    },
+    {
+      "epoch": 462.96,
+      "learning_rate": 2.983512544802868e-05,
+      "loss": 0.2577,
+      "step": 71760
+    },
+    {
+      "epoch": 463.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0312381982803345,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 233.115,
+      "eval_steps_per_second": 8.118,
+      "step": 71765
+    },
+    {
+      "epoch": 463.03,
+      "learning_rate": 2.9831541218637994e-05,
+      "loss": 0.291,
+      "step": 71770
+    },
+    {
+      "epoch": 463.1,
+      "learning_rate": 2.9827956989247312e-05,
+      "loss": 0.3166,
+      "step": 71780
+    },
+    {
+      "epoch": 463.16,
+      "learning_rate": 2.9824372759856633e-05,
+      "loss": 0.2941,
+      "step": 71790
+    },
+    {
+      "epoch": 463.23,
+      "learning_rate": 2.9820788530465955e-05,
+      "loss": 0.2453,
+      "step": 71800
+    },
+    {
+      "epoch": 463.29,
+      "learning_rate": 2.981720430107527e-05,
+      "loss": 0.2422,
+      "step": 71810
+    },
+    {
+      "epoch": 463.35,
+      "learning_rate": 2.9813620071684588e-05,
+      "loss": 0.254,
+      "step": 71820
+    },
+    {
+      "epoch": 463.42,
+      "learning_rate": 2.981003584229391e-05,
+      "loss": 0.2947,
+      "step": 71830
+    },
+    {
+      "epoch": 463.48,
+      "learning_rate": 2.9806451612903224e-05,
+      "loss": 0.2476,
+      "step": 71840
+    },
+    {
+      "epoch": 463.55,
+      "learning_rate": 2.9802867383512546e-05,
+      "loss": 0.2882,
+      "step": 71850
+    },
+    {
+      "epoch": 463.61,
+      "learning_rate": 2.9799283154121864e-05,
+      "loss": 0.2269,
+      "step": 71860
+    },
+    {
+      "epoch": 463.68,
+      "learning_rate": 2.9795698924731186e-05,
+      "loss": 0.2773,
+      "step": 71870
+    },
+    {
+      "epoch": 463.74,
+      "learning_rate": 2.97921146953405e-05,
+      "loss": 0.3338,
+      "step": 71880
+    },
+    {
+      "epoch": 463.8,
+      "learning_rate": 2.9788530465949822e-05,
+      "loss": 0.2463,
+      "step": 71890
+    },
+    {
+      "epoch": 463.87,
+      "learning_rate": 2.978494623655914e-05,
+      "loss": 0.3031,
+      "step": 71900
+    },
+    {
+      "epoch": 463.93,
+      "learning_rate": 2.9781362007168462e-05,
+      "loss": 0.2314,
+      "step": 71910
+    },
+    {
+      "epoch": 464.0,
+      "learning_rate": 2.9777777777777777e-05,
+      "loss": 0.3034,
+      "step": 71920
+    },
+    {
+      "epoch": 464.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0117812156677246,
+      "eval_runtime": 0.8491,
+      "eval_samples_per_second": 236.727,
+      "eval_steps_per_second": 8.244,
+      "step": 71920
+    },
+    {
+      "epoch": 464.06,
+      "learning_rate": 2.97741935483871e-05,
+      "loss": 0.3027,
+      "step": 71930
+    },
+    {
+      "epoch": 464.13,
+      "learning_rate": 2.9770609318996417e-05,
+      "loss": 0.2971,
+      "step": 71940
+    },
+    {
+      "epoch": 464.19,
+      "learning_rate": 2.9767025089605738e-05,
+      "loss": 0.2784,
+      "step": 71950
+    },
+    {
+      "epoch": 464.26,
+      "learning_rate": 2.9763440860215053e-05,
+      "loss": 0.3197,
+      "step": 71960
+    },
+    {
+      "epoch": 464.32,
+      "learning_rate": 2.9759856630824375e-05,
+      "loss": 0.2637,
+      "step": 71970
+    },
+    {
+      "epoch": 464.39,
+      "learning_rate": 2.9756272401433693e-05,
+      "loss": 0.263,
+      "step": 71980
+    },
+    {
+      "epoch": 464.45,
+      "learning_rate": 2.9752688172043014e-05,
+      "loss": 0.3139,
+      "step": 71990
+    },
+    {
+      "epoch": 464.51,
+      "learning_rate": 2.974910394265233e-05,
+      "loss": 0.2803,
+      "step": 72000
+    },
+    {
+      "epoch": 464.58,
+      "learning_rate": 2.9745519713261648e-05,
+      "loss": 0.26,
+      "step": 72010
+    },
+    {
+      "epoch": 464.64,
+      "learning_rate": 2.974193548387097e-05,
+      "loss": 0.2293,
+      "step": 72020
+    },
+    {
+      "epoch": 464.71,
+      "learning_rate": 2.973835125448029e-05,
+      "loss": 0.2781,
+      "step": 72030
+    },
+    {
+      "epoch": 464.77,
+      "learning_rate": 2.9734767025089606e-05,
+      "loss": 0.3037,
+      "step": 72040
+    },
+    {
+      "epoch": 464.84,
+      "learning_rate": 2.9731182795698924e-05,
+      "loss": 0.2722,
+      "step": 72050
+    },
+    {
+      "epoch": 464.9,
+      "learning_rate": 2.9727598566308245e-05,
+      "loss": 0.3107,
+      "step": 72060
+    },
+    {
+      "epoch": 464.96,
+      "learning_rate": 2.9724014336917567e-05,
+      "loss": 0.2833,
+      "step": 72070
+    },
+    {
+      "epoch": 465.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.071835994720459,
+      "eval_runtime": 0.8688,
+      "eval_samples_per_second": 231.36,
+      "eval_steps_per_second": 8.057,
+      "step": 72075
+    },
+    {
+      "epoch": 465.03,
+      "learning_rate": 2.9720430107526882e-05,
+      "loss": 0.3122,
+      "step": 72080
+    },
+    {
+      "epoch": 465.1,
+      "learning_rate": 2.97168458781362e-05,
+      "loss": 0.3173,
+      "step": 72090
+    },
+    {
+      "epoch": 465.16,
+      "learning_rate": 2.971326164874552e-05,
+      "loss": 0.2376,
+      "step": 72100
+    },
+    {
+      "epoch": 465.23,
+      "learning_rate": 2.9709677419354843e-05,
+      "loss": 0.2279,
+      "step": 72110
+    },
+    {
+      "epoch": 465.29,
+      "learning_rate": 2.9706093189964158e-05,
+      "loss": 0.2602,
+      "step": 72120
+    },
+    {
+      "epoch": 465.35,
+      "learning_rate": 2.9702508960573476e-05,
+      "loss": 0.266,
+      "step": 72130
+    },
+    {
+      "epoch": 465.42,
+      "learning_rate": 2.9698924731182798e-05,
+      "loss": 0.2522,
+      "step": 72140
+    },
+    {
+      "epoch": 465.48,
+      "learning_rate": 2.969534050179212e-05,
+      "loss": 0.2678,
+      "step": 72150
+    },
+    {
+      "epoch": 465.55,
+      "learning_rate": 2.9691756272401434e-05,
+      "loss": 0.2776,
+      "step": 72160
+    },
+    {
+      "epoch": 465.61,
+      "learning_rate": 2.9688172043010752e-05,
+      "loss": 0.2729,
+      "step": 72170
+    },
+    {
+      "epoch": 465.68,
+      "learning_rate": 2.9684587813620074e-05,
+      "loss": 0.2376,
+      "step": 72180
+    },
+    {
+      "epoch": 465.74,
+      "learning_rate": 2.9681003584229396e-05,
+      "loss": 0.2733,
+      "step": 72190
+    },
+    {
+      "epoch": 465.8,
+      "learning_rate": 2.967741935483871e-05,
+      "loss": 0.2617,
+      "step": 72200
+    },
+    {
+      "epoch": 465.87,
+      "learning_rate": 2.967383512544803e-05,
+      "loss": 0.2622,
+      "step": 72210
+    },
+    {
+      "epoch": 465.93,
+      "learning_rate": 2.967025089605735e-05,
+      "loss": 0.2766,
+      "step": 72220
+    },
+    {
+      "epoch": 466.0,
+      "learning_rate": 2.9666666666666672e-05,
+      "loss": 0.3017,
+      "step": 72230
+    },
+    {
+      "epoch": 466.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.1082024574279785,
+      "eval_runtime": 0.8838,
+      "eval_samples_per_second": 227.436,
+      "eval_steps_per_second": 7.921,
+      "step": 72230
+    },
+    {
+      "epoch": 466.06,
+      "learning_rate": 2.9663082437275987e-05,
+      "loss": 0.37,
+      "step": 72240
+    },
+    {
+      "epoch": 466.13,
+      "learning_rate": 2.9659498207885305e-05,
+      "loss": 0.2672,
+      "step": 72250
+    },
+    {
+      "epoch": 466.19,
+      "learning_rate": 2.9655913978494626e-05,
+      "loss": 0.2864,
+      "step": 72260
+    },
+    {
+      "epoch": 466.26,
+      "learning_rate": 2.9652329749103948e-05,
+      "loss": 0.2176,
+      "step": 72270
+    },
+    {
+      "epoch": 466.32,
+      "learning_rate": 2.9648745519713263e-05,
+      "loss": 0.2965,
+      "step": 72280
+    },
+    {
+      "epoch": 466.39,
+      "learning_rate": 2.964516129032258e-05,
+      "loss": 0.2831,
+      "step": 72290
+    },
+    {
+      "epoch": 466.45,
+      "learning_rate": 2.9641577060931903e-05,
+      "loss": 0.2527,
+      "step": 72300
+    },
+    {
+      "epoch": 466.51,
+      "learning_rate": 2.9637992831541224e-05,
+      "loss": 0.2657,
+      "step": 72310
+    },
+    {
+      "epoch": 466.58,
+      "learning_rate": 2.963440860215054e-05,
+      "loss": 0.2289,
+      "step": 72320
+    },
+    {
+      "epoch": 466.64,
+      "learning_rate": 2.9630824372759857e-05,
+      "loss": 0.2779,
+      "step": 72330
+    },
+    {
+      "epoch": 466.71,
+      "learning_rate": 2.962724014336918e-05,
+      "loss": 0.2953,
+      "step": 72340
+    },
+    {
+      "epoch": 466.77,
+      "learning_rate": 2.9623655913978497e-05,
+      "loss": 0.3052,
+      "step": 72350
+    },
+    {
+      "epoch": 466.84,
+      "learning_rate": 2.9620071684587812e-05,
+      "loss": 0.27,
+      "step": 72360
+    },
+    {
+      "epoch": 466.9,
+      "learning_rate": 2.9616487455197133e-05,
+      "loss": 0.2776,
+      "step": 72370
+    },
+    {
+      "epoch": 466.96,
+      "learning_rate": 2.9612903225806455e-05,
+      "loss": 0.3036,
+      "step": 72380
+    },
+    {
+      "epoch": 467.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0712283849716187,
+      "eval_runtime": 0.8587,
+      "eval_samples_per_second": 234.075,
+      "eval_steps_per_second": 8.152,
+      "step": 72385
+    },
+    {
+      "epoch": 467.03,
+      "learning_rate": 2.9609318996415773e-05,
+      "loss": 0.2391,
+      "step": 72390
+    },
+    {
+      "epoch": 467.1,
+      "learning_rate": 2.9605734767025088e-05,
+      "loss": 0.2652,
+      "step": 72400
+    },
+    {
+      "epoch": 467.16,
+      "learning_rate": 2.960215053763441e-05,
+      "loss": 0.256,
+      "step": 72410
+    },
+    {
+      "epoch": 467.23,
+      "learning_rate": 2.959856630824373e-05,
+      "loss": 0.3089,
+      "step": 72420
+    },
+    {
+      "epoch": 467.29,
+      "learning_rate": 2.959498207885305e-05,
+      "loss": 0.2469,
+      "step": 72430
+    },
+    {
+      "epoch": 467.35,
+      "learning_rate": 2.9591397849462364e-05,
+      "loss": 0.3337,
+      "step": 72440
+    },
+    {
+      "epoch": 467.42,
+      "learning_rate": 2.9587813620071686e-05,
+      "loss": 0.2241,
+      "step": 72450
+    },
+    {
+      "epoch": 467.48,
+      "learning_rate": 2.9584229390681007e-05,
+      "loss": 0.2755,
+      "step": 72460
+    },
+    {
+      "epoch": 467.55,
+      "learning_rate": 2.9580645161290322e-05,
+      "loss": 0.2635,
+      "step": 72470
+    },
+    {
+      "epoch": 467.61,
+      "learning_rate": 2.957706093189964e-05,
+      "loss": 0.23,
+      "step": 72480
+    },
+    {
+      "epoch": 467.68,
+      "learning_rate": 2.9573476702508962e-05,
+      "loss": 0.2887,
+      "step": 72490
+    },
+    {
+      "epoch": 467.74,
+      "learning_rate": 2.9569892473118284e-05,
+      "loss": 0.2756,
+      "step": 72500
+    },
+    {
+      "epoch": 467.8,
+      "learning_rate": 2.95663082437276e-05,
+      "loss": 0.2778,
+      "step": 72510
+    },
+    {
+      "epoch": 467.87,
+      "learning_rate": 2.9562724014336917e-05,
+      "loss": 0.2217,
+      "step": 72520
+    },
+    {
+      "epoch": 467.93,
+      "learning_rate": 2.9559139784946238e-05,
+      "loss": 0.2672,
+      "step": 72530
+    },
+    {
+      "epoch": 468.0,
+      "learning_rate": 2.955555555555556e-05,
+      "loss": 0.2988,
+      "step": 72540
+    },
+    {
+      "epoch": 468.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.099928617477417,
+      "eval_runtime": 0.8716,
+      "eval_samples_per_second": 230.61,
+      "eval_steps_per_second": 8.031,
+      "step": 72540
+    },
+    {
+      "epoch": 468.06,
+      "learning_rate": 2.9551971326164875e-05,
+      "loss": 0.3038,
+      "step": 72550
+    },
+    {
+      "epoch": 468.13,
+      "learning_rate": 2.9548387096774193e-05,
+      "loss": 0.2332,
+      "step": 72560
+    },
+    {
+      "epoch": 468.19,
+      "learning_rate": 2.9544802867383514e-05,
+      "loss": 0.2605,
+      "step": 72570
+    },
+    {
+      "epoch": 468.26,
+      "learning_rate": 2.9541218637992836e-05,
+      "loss": 0.2869,
+      "step": 72580
+    },
+    {
+      "epoch": 468.32,
+      "learning_rate": 2.953763440860215e-05,
+      "loss": 0.2704,
+      "step": 72590
+    },
+    {
+      "epoch": 468.39,
+      "learning_rate": 2.953405017921147e-05,
+      "loss": 0.2452,
+      "step": 72600
+    },
+    {
+      "epoch": 468.45,
+      "learning_rate": 2.953046594982079e-05,
+      "loss": 0.2962,
+      "step": 72610
+    },
+    {
+      "epoch": 468.51,
+      "learning_rate": 2.9526881720430112e-05,
+      "loss": 0.2033,
+      "step": 72620
+    },
+    {
+      "epoch": 468.58,
+      "learning_rate": 2.9523297491039427e-05,
+      "loss": 0.2659,
+      "step": 72630
+    },
+    {
+      "epoch": 468.64,
+      "learning_rate": 2.9519713261648745e-05,
+      "loss": 0.2659,
+      "step": 72640
+    },
+    {
+      "epoch": 468.71,
+      "learning_rate": 2.9516129032258067e-05,
+      "loss": 0.2516,
+      "step": 72650
+    },
+    {
+      "epoch": 468.77,
+      "learning_rate": 2.951254480286739e-05,
+      "loss": 0.2092,
+      "step": 72660
+    },
+    {
+      "epoch": 468.84,
+      "learning_rate": 2.9508960573476703e-05,
+      "loss": 0.2618,
+      "step": 72670
+    },
+    {
+      "epoch": 468.9,
+      "learning_rate": 2.950537634408602e-05,
+      "loss": 0.2504,
+      "step": 72680
+    },
+    {
+      "epoch": 468.96,
+      "learning_rate": 2.9501792114695343e-05,
+      "loss": 0.2418,
+      "step": 72690
+    },
+    {
+      "epoch": 469.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.060546875,
+      "eval_runtime": 0.8592,
+      "eval_samples_per_second": 233.932,
+      "eval_steps_per_second": 8.147,
+      "step": 72695
+    },
+    {
+      "epoch": 469.03,
+      "learning_rate": 2.949820788530466e-05,
+      "loss": 0.2627,
+      "step": 72700
+    },
+    {
+      "epoch": 469.1,
+      "learning_rate": 2.9494623655913976e-05,
+      "loss": 0.3108,
+      "step": 72710
+    },
+    {
+      "epoch": 469.16,
+      "learning_rate": 2.9491039426523298e-05,
+      "loss": 0.2678,
+      "step": 72720
+    },
+    {
+      "epoch": 469.23,
+      "learning_rate": 2.948745519713262e-05,
+      "loss": 0.2674,
+      "step": 72730
+    },
+    {
+      "epoch": 469.29,
+      "learning_rate": 2.9483870967741937e-05,
+      "loss": 0.2824,
+      "step": 72740
+    },
+    {
+      "epoch": 469.35,
+      "learning_rate": 2.9480286738351252e-05,
+      "loss": 0.2922,
+      "step": 72750
+    },
+    {
+      "epoch": 469.42,
+      "learning_rate": 2.9476702508960574e-05,
+      "loss": 0.2622,
+      "step": 72760
+    },
+    {
+      "epoch": 469.48,
+      "learning_rate": 2.9473118279569895e-05,
+      "loss": 0.2539,
+      "step": 72770
+    },
+    {
+      "epoch": 469.55,
+      "learning_rate": 2.9469534050179214e-05,
+      "loss": 0.2743,
+      "step": 72780
+    },
+    {
+      "epoch": 469.61,
+      "learning_rate": 2.946594982078853e-05,
+      "loss": 0.2712,
+      "step": 72790
+    },
+    {
+      "epoch": 469.68,
+      "learning_rate": 2.946236559139785e-05,
+      "loss": 0.2671,
+      "step": 72800
+    },
+    {
+      "epoch": 469.74,
+      "learning_rate": 2.945878136200717e-05,
+      "loss": 0.2447,
+      "step": 72810
+    },
+    {
+      "epoch": 469.8,
+      "learning_rate": 2.945519713261649e-05,
+      "loss": 0.3183,
+      "step": 72820
+    },
+    {
+      "epoch": 469.87,
+      "learning_rate": 2.9451612903225805e-05,
+      "loss": 0.2614,
+      "step": 72830
+    },
+    {
+      "epoch": 469.93,
+      "learning_rate": 2.9448028673835126e-05,
+      "loss": 0.2922,
+      "step": 72840
+    },
+    {
+      "epoch": 470.0,
+      "learning_rate": 2.9444444444444448e-05,
+      "loss": 0.2357,
+      "step": 72850
+    },
+    {
+      "epoch": 470.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0192866325378418,
+      "eval_runtime": 0.9136,
+      "eval_samples_per_second": 220.018,
+      "eval_steps_per_second": 7.662,
+      "step": 72850
+    },
+    {
+      "epoch": 470.06,
+      "learning_rate": 2.9440860215053766e-05,
+      "loss": 0.2555,
+      "step": 72860
+    },
+    {
+      "epoch": 470.13,
+      "learning_rate": 2.943727598566308e-05,
+      "loss": 0.3033,
+      "step": 72870
+    },
+    {
+      "epoch": 470.19,
+      "learning_rate": 2.9433691756272403e-05,
+      "loss": 0.259,
+      "step": 72880
+    },
+    {
+      "epoch": 470.26,
+      "learning_rate": 2.9430107526881724e-05,
+      "loss": 0.2631,
+      "step": 72890
+    },
+    {
+      "epoch": 470.32,
+      "learning_rate": 2.9426523297491042e-05,
+      "loss": 0.2625,
+      "step": 72900
+    },
+    {
+      "epoch": 470.39,
+      "learning_rate": 2.9422939068100357e-05,
+      "loss": 0.2812,
+      "step": 72910
+    },
+    {
+      "epoch": 470.45,
+      "learning_rate": 2.941935483870968e-05,
+      "loss": 0.272,
+      "step": 72920
+    },
+    {
+      "epoch": 470.51,
+      "learning_rate": 2.9415770609319e-05,
+      "loss": 0.2572,
+      "step": 72930
+    },
+    {
+      "epoch": 470.58,
+      "learning_rate": 2.941218637992832e-05,
+      "loss": 0.2883,
+      "step": 72940
+    },
+    {
+      "epoch": 470.64,
+      "learning_rate": 2.9408602150537633e-05,
+      "loss": 0.2561,
+      "step": 72950
+    },
+    {
+      "epoch": 470.71,
+      "learning_rate": 2.9405017921146955e-05,
+      "loss": 0.3336,
+      "step": 72960
+    },
+    {
+      "epoch": 470.77,
+      "learning_rate": 2.9401433691756277e-05,
+      "loss": 0.2079,
+      "step": 72970
+    },
+    {
+      "epoch": 470.84,
+      "learning_rate": 2.9397849462365595e-05,
+      "loss": 0.2699,
+      "step": 72980
+    },
+    {
+      "epoch": 470.9,
+      "learning_rate": 2.939426523297491e-05,
+      "loss": 0.2876,
+      "step": 72990
+    },
+    {
+      "epoch": 470.96,
+      "learning_rate": 2.939068100358423e-05,
+      "loss": 0.2736,
+      "step": 73000
+    },
+    {
+      "epoch": 471.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0420026779174805,
+      "eval_runtime": 0.8729,
+      "eval_samples_per_second": 230.265,
+      "eval_steps_per_second": 8.019,
+      "step": 73005
+    },
+    {
+      "epoch": 471.03,
+      "learning_rate": 2.938709677419355e-05,
+      "loss": 0.3247,
+      "step": 73010
+    },
+    {
+      "epoch": 471.1,
+      "learning_rate": 2.938351254480287e-05,
+      "loss": 0.2344,
+      "step": 73020
+    },
+    {
+      "epoch": 471.16,
+      "learning_rate": 2.9379928315412186e-05,
+      "loss": 0.2499,
+      "step": 73030
+    },
+    {
+      "epoch": 471.23,
+      "learning_rate": 2.9376344086021507e-05,
+      "loss": 0.2775,
+      "step": 73040
+    },
+    {
+      "epoch": 471.29,
+      "learning_rate": 2.9372759856630826e-05,
+      "loss": 0.2554,
+      "step": 73050
+    },
+    {
+      "epoch": 471.35,
+      "learning_rate": 2.9369175627240147e-05,
+      "loss": 0.2801,
+      "step": 73060
+    },
+    {
+      "epoch": 471.42,
+      "learning_rate": 2.9365591397849462e-05,
+      "loss": 0.2679,
+      "step": 73070
+    },
+    {
+      "epoch": 471.48,
+      "learning_rate": 2.9362007168458784e-05,
+      "loss": 0.2405,
+      "step": 73080
+    },
+    {
+      "epoch": 471.55,
+      "learning_rate": 2.9358422939068102e-05,
+      "loss": 0.2391,
+      "step": 73090
+    },
+    {
+      "epoch": 471.61,
+      "learning_rate": 2.9354838709677417e-05,
+      "loss": 0.2867,
+      "step": 73100
+    },
+    {
+      "epoch": 471.68,
+      "learning_rate": 2.9351254480286738e-05,
+      "loss": 0.2424,
+      "step": 73110
+    },
+    {
+      "epoch": 471.74,
+      "learning_rate": 2.934767025089606e-05,
+      "loss": 0.2659,
+      "step": 73120
+    },
+    {
+      "epoch": 471.8,
+      "learning_rate": 2.9344086021505378e-05,
+      "loss": 0.3095,
+      "step": 73130
+    },
+    {
+      "epoch": 471.87,
+      "learning_rate": 2.9340501792114693e-05,
+      "loss": 0.2653,
+      "step": 73140
+    },
+    {
+      "epoch": 471.93,
+      "learning_rate": 2.9336917562724014e-05,
+      "loss": 0.2653,
+      "step": 73150
+    },
+    {
+      "epoch": 472.0,
+      "learning_rate": 2.9333333333333336e-05,
+      "loss": 0.2696,
+      "step": 73160
+    },
+    {
+      "epoch": 472.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.992063581943512,
+      "eval_runtime": 0.8564,
+      "eval_samples_per_second": 234.7,
+      "eval_steps_per_second": 8.174,
+      "step": 73160
+    },
+    {
+      "epoch": 472.06,
+      "learning_rate": 2.9329749103942654e-05,
+      "loss": 0.2742,
+      "step": 73170
+    },
+    {
+      "epoch": 472.13,
+      "learning_rate": 2.932616487455197e-05,
+      "loss": 0.2457,
+      "step": 73180
+    },
+    {
+      "epoch": 472.19,
+      "learning_rate": 2.932258064516129e-05,
+      "loss": 0.3421,
+      "step": 73190
+    },
+    {
+      "epoch": 472.26,
+      "learning_rate": 2.9318996415770612e-05,
+      "loss": 0.2456,
+      "step": 73200
+    },
+    {
+      "epoch": 472.32,
+      "learning_rate": 2.931541218637993e-05,
+      "loss": 0.2627,
+      "step": 73210
+    },
+    {
+      "epoch": 472.39,
+      "learning_rate": 2.9311827956989245e-05,
+      "loss": 0.2377,
+      "step": 73220
+    },
+    {
+      "epoch": 472.45,
+      "learning_rate": 2.9308243727598567e-05,
+      "loss": 0.2189,
+      "step": 73230
+    },
+    {
+      "epoch": 472.51,
+      "learning_rate": 2.930465949820789e-05,
+      "loss": 0.2487,
+      "step": 73240
+    },
+    {
+      "epoch": 472.58,
+      "learning_rate": 2.9301075268817207e-05,
+      "loss": 0.2719,
+      "step": 73250
+    },
+    {
+      "epoch": 472.64,
+      "learning_rate": 2.929749103942652e-05,
+      "loss": 0.2671,
+      "step": 73260
+    },
+    {
+      "epoch": 472.71,
+      "learning_rate": 2.9293906810035843e-05,
+      "loss": 0.2909,
+      "step": 73270
+    },
+    {
+      "epoch": 472.77,
+      "learning_rate": 2.9290322580645165e-05,
+      "loss": 0.2795,
+      "step": 73280
+    },
+    {
+      "epoch": 472.84,
+      "learning_rate": 2.9286738351254483e-05,
+      "loss": 0.2495,
+      "step": 73290
+    },
+    {
+      "epoch": 472.9,
+      "learning_rate": 2.9283154121863798e-05,
+      "loss": 0.2499,
+      "step": 73300
+    },
+    {
+      "epoch": 472.96,
+      "learning_rate": 2.927956989247312e-05,
+      "loss": 0.2462,
+      "step": 73310
+    },
+    {
+      "epoch": 473.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0449705123901367,
+      "eval_runtime": 0.8696,
+      "eval_samples_per_second": 231.13,
+      "eval_steps_per_second": 8.049,
+      "step": 73315
+    },
+    {
+      "epoch": 473.03,
+      "learning_rate": 2.927598566308244e-05,
+      "loss": 0.3268,
+      "step": 73320
+    },
+    {
+      "epoch": 473.1,
+      "learning_rate": 2.927240143369176e-05,
+      "loss": 0.29,
+      "step": 73330
+    },
+    {
+      "epoch": 473.16,
+      "learning_rate": 2.9268817204301074e-05,
+      "loss": 0.2833,
+      "step": 73340
+    },
+    {
+      "epoch": 473.23,
+      "learning_rate": 2.9265232974910395e-05,
+      "loss": 0.2505,
+      "step": 73350
+    },
+    {
+      "epoch": 473.29,
+      "learning_rate": 2.9261648745519714e-05,
+      "loss": 0.2879,
+      "step": 73360
+    },
+    {
+      "epoch": 473.35,
+      "learning_rate": 2.9258064516129035e-05,
+      "loss": 0.2952,
+      "step": 73370
+    },
+    {
+      "epoch": 473.42,
+      "learning_rate": 2.925448028673835e-05,
+      "loss": 0.278,
+      "step": 73380
+    },
+    {
+      "epoch": 473.48,
+      "learning_rate": 2.925089605734767e-05,
+      "loss": 0.3036,
+      "step": 73390
+    },
+    {
+      "epoch": 473.55,
+      "learning_rate": 2.924731182795699e-05,
+      "loss": 0.3191,
+      "step": 73400
+    },
+    {
+      "epoch": 473.61,
+      "learning_rate": 2.924372759856631e-05,
+      "loss": 0.2461,
+      "step": 73410
+    },
+    {
+      "epoch": 473.68,
+      "learning_rate": 2.9240143369175626e-05,
+      "loss": 0.2991,
+      "step": 73420
+    },
+    {
+      "epoch": 473.74,
+      "learning_rate": 2.9236559139784948e-05,
+      "loss": 0.2047,
+      "step": 73430
+    },
+    {
+      "epoch": 473.8,
+      "learning_rate": 2.9232974910394266e-05,
+      "loss": 0.265,
+      "step": 73440
+    },
+    {
+      "epoch": 473.87,
+      "learning_rate": 2.9229390681003588e-05,
+      "loss": 0.3185,
+      "step": 73450
+    },
+    {
+      "epoch": 473.93,
+      "learning_rate": 2.9225806451612902e-05,
+      "loss": 0.2901,
+      "step": 73460
+    },
+    {
+      "epoch": 474.0,
+      "learning_rate": 2.9222222222222224e-05,
+      "loss": 0.2506,
+      "step": 73470
+    },
+    {
+      "epoch": 474.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9916217923164368,
+      "eval_runtime": 0.8767,
+      "eval_samples_per_second": 229.256,
+      "eval_steps_per_second": 7.984,
+      "step": 73470
+    },
+    {
+      "epoch": 474.06,
+      "learning_rate": 2.9218637992831542e-05,
+      "loss": 0.2629,
+      "step": 73480
+    },
+    {
+      "epoch": 474.13,
+      "learning_rate": 2.9215053763440864e-05,
+      "loss": 0.2609,
+      "step": 73490
+    },
+    {
+      "epoch": 474.19,
+      "learning_rate": 2.921146953405018e-05,
+      "loss": 0.2655,
+      "step": 73500
+    },
+    {
+      "epoch": 474.26,
+      "learning_rate": 2.92078853046595e-05,
+      "loss": 0.2706,
+      "step": 73510
+    },
+    {
+      "epoch": 474.32,
+      "learning_rate": 2.920430107526882e-05,
+      "loss": 0.2514,
+      "step": 73520
+    },
+    {
+      "epoch": 474.39,
+      "learning_rate": 2.920071684587814e-05,
+      "loss": 0.2128,
+      "step": 73530
+    },
+    {
+      "epoch": 474.45,
+      "learning_rate": 2.9197132616487455e-05,
+      "loss": 0.3162,
+      "step": 73540
+    },
+    {
+      "epoch": 474.51,
+      "learning_rate": 2.9193548387096776e-05,
+      "loss": 0.217,
+      "step": 73550
+    },
+    {
+      "epoch": 474.58,
+      "learning_rate": 2.9189964157706095e-05,
+      "loss": 0.2965,
+      "step": 73560
+    },
+    {
+      "epoch": 474.64,
+      "learning_rate": 2.9186379928315416e-05,
+      "loss": 0.2628,
+      "step": 73570
+    },
+    {
+      "epoch": 474.71,
+      "learning_rate": 2.918279569892473e-05,
+      "loss": 0.2683,
+      "step": 73580
+    },
+    {
+      "epoch": 474.77,
+      "learning_rate": 2.9179211469534053e-05,
+      "loss": 0.2758,
+      "step": 73590
+    },
+    {
+      "epoch": 474.84,
+      "learning_rate": 2.917562724014337e-05,
+      "loss": 0.2798,
+      "step": 73600
+    },
+    {
+      "epoch": 474.9,
+      "learning_rate": 2.9172043010752692e-05,
+      "loss": 0.2933,
+      "step": 73610
+    },
+    {
+      "epoch": 474.96,
+      "learning_rate": 2.9168458781362007e-05,
+      "loss": 0.2801,
+      "step": 73620
+    },
+    {
+      "epoch": 475.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 1.016597032546997,
+      "eval_runtime": 0.853,
+      "eval_samples_per_second": 235.637,
+      "eval_steps_per_second": 8.206,
+      "step": 73625
+    },
+    {
+      "epoch": 475.03,
+      "learning_rate": 2.916487455197133e-05,
+      "loss": 0.313,
+      "step": 73630
+    },
+    {
+      "epoch": 475.1,
+      "learning_rate": 2.9161290322580647e-05,
+      "loss": 0.3191,
+      "step": 73640
+    },
+    {
+      "epoch": 475.16,
+      "learning_rate": 2.915770609318997e-05,
+      "loss": 0.267,
+      "step": 73650
+    },
+    {
+      "epoch": 475.23,
+      "learning_rate": 2.9154121863799283e-05,
+      "loss": 0.2035,
+      "step": 73660
+    },
+    {
+      "epoch": 475.29,
+      "learning_rate": 2.9150537634408605e-05,
+      "loss": 0.2383,
+      "step": 73670
+    },
+    {
+      "epoch": 475.35,
+      "learning_rate": 2.9146953405017923e-05,
+      "loss": 0.3053,
+      "step": 73680
+    },
+    {
+      "epoch": 475.42,
+      "learning_rate": 2.9143369175627245e-05,
+      "loss": 0.3246,
+      "step": 73690
+    },
+    {
+      "epoch": 475.48,
+      "learning_rate": 2.913978494623656e-05,
+      "loss": 0.2095,
+      "step": 73700
+    },
+    {
+      "epoch": 475.55,
+      "learning_rate": 2.9136200716845878e-05,
+      "loss": 0.2523,
+      "step": 73710
+    },
+    {
+      "epoch": 475.61,
+      "learning_rate": 2.91326164874552e-05,
+      "loss": 0.2159,
+      "step": 73720
+    },
+    {
+      "epoch": 475.68,
+      "learning_rate": 2.912903225806452e-05,
+      "loss": 0.2666,
+      "step": 73730
+    },
+    {
+      "epoch": 475.74,
+      "learning_rate": 2.9125448028673836e-05,
+      "loss": 0.2564,
+      "step": 73740
+    },
+    {
+      "epoch": 475.8,
+      "learning_rate": 2.9121863799283154e-05,
+      "loss": 0.3143,
+      "step": 73750
+    },
+    {
+      "epoch": 475.87,
+      "learning_rate": 2.9118279569892476e-05,
+      "loss": 0.2369,
+      "step": 73760
+    },
+    {
+      "epoch": 475.93,
+      "learning_rate": 2.911469534050179e-05,
+      "loss": 0.251,
+      "step": 73770
+    },
+    {
+      "epoch": 476.0,
+      "learning_rate": 2.9111111111111112e-05,
+      "loss": 0.297,
+      "step": 73780
+    },
+    {
+      "epoch": 476.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 1.008955955505371,
+      "eval_runtime": 0.8749,
+      "eval_samples_per_second": 229.73,
+      "eval_steps_per_second": 8.001,
+      "step": 73780
+    },
+    {
+      "epoch": 476.06,
+      "learning_rate": 2.910752688172043e-05,
+      "loss": 0.2685,
+      "step": 73790
+    },
+    {
+      "epoch": 476.13,
+      "learning_rate": 2.9103942652329752e-05,
+      "loss": 0.2476,
+      "step": 73800
+    },
+    {
+      "epoch": 476.19,
+      "learning_rate": 2.9100358422939067e-05,
+      "loss": 0.2227,
+      "step": 73810
+    },
+    {
+      "epoch": 476.26,
+      "learning_rate": 2.909677419354839e-05,
+      "loss": 0.2266,
+      "step": 73820
+    },
+    {
+      "epoch": 476.32,
+      "learning_rate": 2.9093189964157707e-05,
+      "loss": 0.2337,
+      "step": 73830
+    },
+    {
+      "epoch": 476.39,
+      "learning_rate": 2.9089605734767028e-05,
+      "loss": 0.2488,
+      "step": 73840
+    },
+    {
+      "epoch": 476.45,
+      "learning_rate": 2.9086021505376343e-05,
+      "loss": 0.262,
+      "step": 73850
+    },
+    {
+      "epoch": 476.51,
+      "learning_rate": 2.9082437275985665e-05,
+      "loss": 0.2727,
+      "step": 73860
+    },
+    {
+      "epoch": 476.58,
+      "learning_rate": 2.9078853046594983e-05,
+      "loss": 0.2469,
+      "step": 73870
+    },
+    {
+      "epoch": 476.64,
+      "learning_rate": 2.9075268817204304e-05,
+      "loss": 0.3011,
+      "step": 73880
+    },
+    {
+      "epoch": 476.71,
+      "learning_rate": 2.907168458781362e-05,
+      "loss": 0.2726,
+      "step": 73890
+    },
+    {
+      "epoch": 476.77,
+      "learning_rate": 2.906810035842294e-05,
+      "loss": 0.2624,
+      "step": 73900
+    },
+    {
+      "epoch": 476.84,
+      "learning_rate": 2.906451612903226e-05,
+      "loss": 0.273,
+      "step": 73910
+    },
+    {
+      "epoch": 476.9,
+      "learning_rate": 2.906093189964158e-05,
+      "loss": 0.2813,
+      "step": 73920
+    },
+    {
+      "epoch": 476.96,
+      "learning_rate": 2.9057347670250895e-05,
+      "loss": 0.2678,
+      "step": 73930
+    },
+    {
+      "epoch": 477.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0621485710144043,
+      "eval_runtime": 0.8499,
+      "eval_samples_per_second": 236.489,
+      "eval_steps_per_second": 8.236,
+      "step": 73935
+    },
+    {
+      "epoch": 477.03,
+      "learning_rate": 2.9053763440860217e-05,
+      "loss": 0.3147,
+      "step": 73940
+    },
+    {
+      "epoch": 477.1,
+      "learning_rate": 2.9050179211469535e-05,
+      "loss": 0.2565,
+      "step": 73950
+    },
+    {
+      "epoch": 477.16,
+      "learning_rate": 2.9046594982078857e-05,
+      "loss": 0.2768,
+      "step": 73960
+    },
+    {
+      "epoch": 477.23,
+      "learning_rate": 2.904301075268817e-05,
+      "loss": 0.2942,
+      "step": 73970
+    },
+    {
+      "epoch": 477.29,
+      "learning_rate": 2.9039426523297493e-05,
+      "loss": 0.2751,
+      "step": 73980
+    },
+    {
+      "epoch": 477.35,
+      "learning_rate": 2.903584229390681e-05,
+      "loss": 0.2534,
+      "step": 73990
+    },
+    {
+      "epoch": 477.42,
+      "learning_rate": 2.9032258064516133e-05,
+      "loss": 0.2805,
+      "step": 74000
+    },
+    {
+      "epoch": 477.48,
+      "learning_rate": 2.9028673835125448e-05,
+      "loss": 0.2909,
+      "step": 74010
+    },
+    {
+      "epoch": 477.55,
+      "learning_rate": 2.902508960573477e-05,
+      "loss": 0.2402,
+      "step": 74020
+    },
+    {
+      "epoch": 477.61,
+      "learning_rate": 2.9021505376344088e-05,
+      "loss": 0.2963,
+      "step": 74030
+    },
+    {
+      "epoch": 477.68,
+      "learning_rate": 2.901792114695341e-05,
+      "loss": 0.2194,
+      "step": 74040
+    },
+    {
+      "epoch": 477.74,
+      "learning_rate": 2.9014336917562724e-05,
+      "loss": 0.2865,
+      "step": 74050
+    },
+    {
+      "epoch": 477.8,
+      "learning_rate": 2.9010752688172042e-05,
+      "loss": 0.2574,
+      "step": 74060
+    },
+    {
+      "epoch": 477.87,
+      "learning_rate": 2.9007168458781364e-05,
+      "loss": 0.3068,
+      "step": 74070
+    },
+    {
+      "epoch": 477.93,
+      "learning_rate": 2.9003584229390685e-05,
+      "loss": 0.2743,
+      "step": 74080
+    },
+    {
+      "epoch": 478.0,
+      "learning_rate": 2.9e-05,
+      "loss": 0.2643,
+      "step": 74090
+    },
+    {
+      "epoch": 478.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 1.0416908264160156,
+      "eval_runtime": 0.8589,
+      "eval_samples_per_second": 234.03,
+      "eval_steps_per_second": 8.15,
+      "step": 74090
+    },
+    {
+      "epoch": 478.06,
+      "learning_rate": 2.899641577060932e-05,
+      "loss": 0.2704,
+      "step": 74100
+    },
+    {
+      "epoch": 478.13,
+      "learning_rate": 2.899283154121864e-05,
+      "loss": 0.2712,
+      "step": 74110
+    },
+    {
+      "epoch": 478.19,
+      "learning_rate": 2.898924731182796e-05,
+      "loss": 0.2761,
+      "step": 74120
+    },
+    {
+      "epoch": 478.26,
+      "learning_rate": 2.8985663082437276e-05,
+      "loss": 0.2752,
+      "step": 74130
+    },
+    {
+      "epoch": 478.32,
+      "learning_rate": 2.8982078853046595e-05,
+      "loss": 0.2704,
+      "step": 74140
+    },
+    {
+      "epoch": 478.39,
+      "learning_rate": 2.8978494623655916e-05,
+      "loss": 0.2854,
+      "step": 74150
+    },
+    {
+      "epoch": 478.45,
+      "learning_rate": 2.8974910394265238e-05,
+      "loss": 0.2611,
+      "step": 74160
+    },
+    {
+      "epoch": 478.51,
+      "learning_rate": 2.8971326164874553e-05,
+      "loss": 0.2547,
+      "step": 74170
+    },
+    {
+      "epoch": 478.58,
+      "learning_rate": 2.896774193548387e-05,
+      "loss": 0.269,
+      "step": 74180
+    },
+    {
+      "epoch": 478.64,
+      "learning_rate": 2.8964157706093192e-05,
+      "loss": 0.2462,
+      "step": 74190
+    },
+    {
+      "epoch": 478.71,
+      "learning_rate": 2.8960573476702514e-05,
+      "loss": 0.253,
+      "step": 74200
+    },
+    {
+      "epoch": 478.77,
+      "learning_rate": 2.895698924731183e-05,
+      "loss": 0.2511,
+      "step": 74210
+    },
+    {
+      "epoch": 478.84,
+      "learning_rate": 2.8953405017921147e-05,
+      "loss": 0.1859,
+      "step": 74220
+    },
+    {
+      "epoch": 478.9,
+      "learning_rate": 2.894982078853047e-05,
+      "loss": 0.224,
+      "step": 74230
+    },
+    {
+      "epoch": 478.96,
+      "learning_rate": 2.894623655913979e-05,
+      "loss": 0.2905,
+      "step": 74240
+    },
+    {
+      "epoch": 479.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9860101342201233,
+      "eval_runtime": 0.8556,
+      "eval_samples_per_second": 234.913,
+      "eval_steps_per_second": 8.181,
+      "step": 74245
+    },
+    {
+      "epoch": 479.03,
+      "learning_rate": 2.8942652329749105e-05,
+      "loss": 0.2483,
+      "step": 74250
+    },
+    {
+      "epoch": 479.1,
+      "learning_rate": 2.8939068100358423e-05,
+      "loss": 0.2675,
+      "step": 74260
+    },
+    {
+      "epoch": 479.16,
+      "learning_rate": 2.8935483870967745e-05,
+      "loss": 0.2632,
+      "step": 74270
+    },
+    {
+      "epoch": 479.23,
+      "learning_rate": 2.8931899641577066e-05,
+      "loss": 0.2401,
+      "step": 74280
+    },
+    {
+      "epoch": 479.29,
+      "learning_rate": 2.892831541218638e-05,
+      "loss": 0.2486,
+      "step": 74290
+    },
+    {
+      "epoch": 479.35,
+      "learning_rate": 2.89247311827957e-05,
+      "loss": 0.2249,
+      "step": 74300
+    },
+    {
+      "epoch": 479.42,
+      "learning_rate": 2.892114695340502e-05,
+      "loss": 0.2645,
+      "step": 74310
+    },
+    {
+      "epoch": 479.48,
+      "learning_rate": 2.8917562724014343e-05,
+      "loss": 0.2699,
+      "step": 74320
+    },
+    {
+      "epoch": 479.55,
+      "learning_rate": 2.8913978494623657e-05,
+      "loss": 0.2443,
+      "step": 74330
+    },
+    {
+      "epoch": 479.61,
+      "learning_rate": 2.8910394265232976e-05,
+      "loss": 0.2853,
+      "step": 74340
+    },
+    {
+      "epoch": 479.68,
+      "learning_rate": 2.8906810035842297e-05,
+      "loss": 0.2876,
+      "step": 74350
+    },
+    {
+      "epoch": 479.74,
+      "learning_rate": 2.8903225806451615e-05,
+      "loss": 0.3111,
+      "step": 74360
+    },
+    {
+      "epoch": 479.8,
+      "learning_rate": 2.889964157706093e-05,
+      "loss": 0.263,
+      "step": 74370
+    },
+    {
+      "epoch": 479.87,
+      "learning_rate": 2.8896057347670252e-05,
+      "loss": 0.2434,
+      "step": 74380
+    },
+    {
+      "epoch": 479.93,
+      "learning_rate": 2.8892473118279573e-05,
+      "loss": 0.234,
+      "step": 74390
+    },
+    {
+      "epoch": 480.0,
+      "learning_rate": 2.8888888888888888e-05,
+      "loss": 0.2487,
+      "step": 74400
+    },
+    {
+      "epoch": 480.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9733458757400513,
+      "eval_runtime": 0.8844,
+      "eval_samples_per_second": 227.278,
+      "eval_steps_per_second": 7.915,
+      "step": 74400
+    },
+    {
+      "epoch": 480.06,
+      "learning_rate": 2.8885304659498206e-05,
+      "loss": 0.3641,
+      "step": 74410
+    },
+    {
+      "epoch": 480.13,
+      "learning_rate": 2.8881720430107528e-05,
+      "loss": 0.2702,
+      "step": 74420
+    },
+    {
+      "epoch": 480.19,
+      "learning_rate": 2.887813620071685e-05,
+      "loss": 0.2532,
+      "step": 74430
+    },
+    {
+      "epoch": 480.26,
+      "learning_rate": 2.8874551971326164e-05,
+      "loss": 0.2699,
+      "step": 74440
+    },
+    {
+      "epoch": 480.32,
+      "learning_rate": 2.8870967741935483e-05,
+      "loss": 0.28,
+      "step": 74450
+    },
+    {
+      "epoch": 480.39,
+      "learning_rate": 2.8867383512544804e-05,
+      "loss": 0.2639,
+      "step": 74460
+    },
+    {
+      "epoch": 480.45,
+      "learning_rate": 2.8863799283154126e-05,
+      "loss": 0.232,
+      "step": 74470
+    },
+    {
+      "epoch": 480.51,
+      "learning_rate": 2.886021505376344e-05,
+      "loss": 0.2854,
+      "step": 74480
+    },
+    {
+      "epoch": 480.58,
+      "learning_rate": 2.885663082437276e-05,
+      "loss": 0.2694,
+      "step": 74490
+    },
+    {
+      "epoch": 480.64,
+      "learning_rate": 2.885304659498208e-05,
+      "loss": 0.2546,
+      "step": 74500
+    },
+    {
+      "epoch": 480.71,
+      "learning_rate": 2.8849462365591402e-05,
+      "loss": 0.2266,
+      "step": 74510
+    },
+    {
+      "epoch": 480.77,
+      "learning_rate": 2.8845878136200717e-05,
+      "loss": 0.2886,
+      "step": 74520
+    },
+    {
+      "epoch": 480.84,
+      "learning_rate": 2.8842293906810035e-05,
+      "loss": 0.263,
+      "step": 74530
+    },
+    {
+      "epoch": 480.9,
+      "learning_rate": 2.8838709677419357e-05,
+      "loss": 0.2735,
+      "step": 74540
+    },
+    {
+      "epoch": 480.96,
+      "learning_rate": 2.8835125448028678e-05,
+      "loss": 0.2554,
+      "step": 74550
+    },
+    {
+      "epoch": 481.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9273634552955627,
+      "eval_runtime": 0.8628,
+      "eval_samples_per_second": 232.964,
+      "eval_steps_per_second": 8.113,
+      "step": 74555
+    },
+    {
+      "epoch": 481.03,
+      "learning_rate": 2.8831541218637993e-05,
+      "loss": 0.3136,
+      "step": 74560
+    },
+    {
+      "epoch": 481.1,
+      "learning_rate": 2.882795698924731e-05,
+      "loss": 0.2489,
+      "step": 74570
+    },
+    {
+      "epoch": 481.16,
+      "learning_rate": 2.8824372759856633e-05,
+      "loss": 0.2702,
+      "step": 74580
+    },
+    {
+      "epoch": 481.23,
+      "learning_rate": 2.8820788530465954e-05,
+      "loss": 0.2381,
+      "step": 74590
+    },
+    {
+      "epoch": 481.29,
+      "learning_rate": 2.881720430107527e-05,
+      "loss": 0.2727,
+      "step": 74600
+    },
+    {
+      "epoch": 481.35,
+      "learning_rate": 2.8813620071684587e-05,
+      "loss": 0.2887,
+      "step": 74610
+    },
+    {
+      "epoch": 481.42,
+      "learning_rate": 2.881003584229391e-05,
+      "loss": 0.2418,
+      "step": 74620
+    },
+    {
+      "epoch": 481.48,
+      "learning_rate": 2.880645161290323e-05,
+      "loss": 0.2917,
+      "step": 74630
+    },
+    {
+      "epoch": 481.55,
+      "learning_rate": 2.8802867383512545e-05,
+      "loss": 0.3185,
+      "step": 74640
+    },
+    {
+      "epoch": 481.61,
+      "learning_rate": 2.8799283154121864e-05,
+      "loss": 0.3337,
+      "step": 74650
+    },
+    {
+      "epoch": 481.68,
+      "learning_rate": 2.8795698924731185e-05,
+      "loss": 0.2561,
+      "step": 74660
+    },
+    {
+      "epoch": 481.74,
+      "learning_rate": 2.8792114695340507e-05,
+      "loss": 0.2256,
+      "step": 74670
+    },
+    {
+      "epoch": 481.8,
+      "learning_rate": 2.8788530465949822e-05,
+      "loss": 0.2272,
+      "step": 74680
+    },
+    {
+      "epoch": 481.87,
+      "learning_rate": 2.878494623655914e-05,
+      "loss": 0.2535,
+      "step": 74690
+    },
+    {
+      "epoch": 481.93,
+      "learning_rate": 2.878136200716846e-05,
+      "loss": 0.325,
+      "step": 74700
+    },
+    {
+      "epoch": 482.0,
+      "learning_rate": 2.877777777777778e-05,
+      "loss": 0.2607,
+      "step": 74710
+    },
+    {
+      "epoch": 482.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9427967071533203,
+      "eval_runtime": 0.8556,
+      "eval_samples_per_second": 234.911,
+      "eval_steps_per_second": 8.181,
+      "step": 74710
+    },
+    {
+      "epoch": 482.06,
+      "learning_rate": 2.8774193548387095e-05,
+      "loss": 0.321,
+      "step": 74720
+    },
+    {
+      "epoch": 482.13,
+      "learning_rate": 2.8770609318996416e-05,
+      "loss": 0.2671,
+      "step": 74730
+    },
+    {
+      "epoch": 482.19,
+      "learning_rate": 2.8767025089605738e-05,
+      "loss": 0.2981,
+      "step": 74740
+    },
+    {
+      "epoch": 482.26,
+      "learning_rate": 2.8763440860215056e-05,
+      "loss": 0.2995,
+      "step": 74750
+    },
+    {
+      "epoch": 482.32,
+      "learning_rate": 2.875985663082437e-05,
+      "loss": 0.2579,
+      "step": 74760
+    },
+    {
+      "epoch": 482.39,
+      "learning_rate": 2.8756272401433692e-05,
+      "loss": 0.2552,
+      "step": 74770
+    },
+    {
+      "epoch": 482.45,
+      "learning_rate": 2.8752688172043014e-05,
+      "loss": 0.277,
+      "step": 74780
+    },
+    {
+      "epoch": 482.51,
+      "learning_rate": 2.8749103942652332e-05,
+      "loss": 0.285,
+      "step": 74790
+    },
+    {
+      "epoch": 482.58,
+      "learning_rate": 2.8745519713261647e-05,
+      "loss": 0.2636,
+      "step": 74800
+    },
+    {
+      "epoch": 482.64,
+      "learning_rate": 2.874193548387097e-05,
+      "loss": 0.2963,
+      "step": 74810
+    },
+    {
+      "epoch": 482.71,
+      "learning_rate": 2.873835125448029e-05,
+      "loss": 0.3472,
+      "step": 74820
+    },
+    {
+      "epoch": 482.77,
+      "learning_rate": 2.873476702508961e-05,
+      "loss": 0.3177,
+      "step": 74830
+    },
+    {
+      "epoch": 482.84,
+      "learning_rate": 2.8731182795698923e-05,
+      "loss": 0.2613,
+      "step": 74840
+    },
+    {
+      "epoch": 482.9,
+      "learning_rate": 2.8727598566308245e-05,
+      "loss": 0.2281,
+      "step": 74850
+    },
+    {
+      "epoch": 482.96,
+      "learning_rate": 2.8724014336917566e-05,
+      "loss": 0.2965,
+      "step": 74860
+    },
+    {
+      "epoch": 483.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9248155355453491,
+      "eval_runtime": 0.8521,
+      "eval_samples_per_second": 235.892,
+      "eval_steps_per_second": 8.215,
+      "step": 74865
+    },
+    {
+      "epoch": 483.03,
+      "learning_rate": 2.8720430107526885e-05,
+      "loss": 0.3196,
+      "step": 74870
+    },
+    {
+      "epoch": 483.1,
+      "learning_rate": 2.87168458781362e-05,
+      "loss": 0.2153,
+      "step": 74880
+    },
+    {
+      "epoch": 483.16,
+      "learning_rate": 2.871326164874552e-05,
+      "loss": 0.2616,
+      "step": 74890
+    },
+    {
+      "epoch": 483.23,
+      "learning_rate": 2.8709677419354843e-05,
+      "loss": 0.2419,
+      "step": 74900
+    },
+    {
+      "epoch": 483.29,
+      "learning_rate": 2.870609318996416e-05,
+      "loss": 0.2373,
+      "step": 74910
+    },
+    {
+      "epoch": 483.35,
+      "learning_rate": 2.8702508960573476e-05,
+      "loss": 0.2499,
+      "step": 74920
+    },
+    {
+      "epoch": 483.42,
+      "learning_rate": 2.8698924731182797e-05,
+      "loss": 0.2647,
+      "step": 74930
+    },
+    {
+      "epoch": 483.48,
+      "learning_rate": 2.869534050179212e-05,
+      "loss": 0.2921,
+      "step": 74940
+    },
+    {
+      "epoch": 483.55,
+      "learning_rate": 2.8691756272401437e-05,
+      "loss": 0.3207,
+      "step": 74950
+    },
+    {
+      "epoch": 483.61,
+      "learning_rate": 2.8688172043010752e-05,
+      "loss": 0.308,
+      "step": 74960
+    },
+    {
+      "epoch": 483.68,
+      "learning_rate": 2.8684587813620073e-05,
+      "loss": 0.2596,
+      "step": 74970
+    },
+    {
+      "epoch": 483.74,
+      "learning_rate": 2.8681003584229395e-05,
+      "loss": 0.2225,
+      "step": 74980
+    },
+    {
+      "epoch": 483.8,
+      "learning_rate": 2.8677419354838713e-05,
+      "loss": 0.2796,
+      "step": 74990
+    },
+    {
+      "epoch": 483.87,
+      "learning_rate": 2.8673835125448028e-05,
+      "loss": 0.212,
+      "step": 75000
+    },
+    {
+      "epoch": 483.93,
+      "learning_rate": 2.867025089605735e-05,
+      "loss": 0.2405,
+      "step": 75010
+    },
+    {
+      "epoch": 484.0,
+      "learning_rate": 2.8666666666666668e-05,
+      "loss": 0.2342,
+      "step": 75020
+    },
+    {
+      "epoch": 484.0,
+      "eval_accuracy": 0.8009950248756219,
+      "eval_loss": 0.942956268787384,
+      "eval_runtime": 0.8561,
+      "eval_samples_per_second": 234.793,
+      "eval_steps_per_second": 8.177,
+      "step": 75020
+    },
+    {
+      "epoch": 484.06,
+      "learning_rate": 2.8663082437275986e-05,
+      "loss": 0.2852,
+      "step": 75030
+    },
+    {
+      "epoch": 484.13,
+      "learning_rate": 2.8659498207885304e-05,
+      "loss": 0.2436,
+      "step": 75040
+    },
+    {
+      "epoch": 484.19,
+      "learning_rate": 2.8655913978494626e-05,
+      "loss": 0.2646,
+      "step": 75050
+    },
+    {
+      "epoch": 484.26,
+      "learning_rate": 2.8652329749103944e-05,
+      "loss": 0.3022,
+      "step": 75060
+    },
+    {
+      "epoch": 484.32,
+      "learning_rate": 2.864874551971326e-05,
+      "loss": 0.2247,
+      "step": 75070
+    },
+    {
+      "epoch": 484.39,
+      "learning_rate": 2.864516129032258e-05,
+      "loss": 0.2312,
+      "step": 75080
+    },
+    {
+      "epoch": 484.45,
+      "learning_rate": 2.8641577060931902e-05,
+      "loss": 0.3103,
+      "step": 75090
+    },
+    {
+      "epoch": 484.51,
+      "learning_rate": 2.863799283154122e-05,
+      "loss": 0.2836,
+      "step": 75100
+    },
+    {
+      "epoch": 484.58,
+      "learning_rate": 2.8634408602150535e-05,
+      "loss": 0.2333,
+      "step": 75110
+    },
+    {
+      "epoch": 484.64,
+      "learning_rate": 2.8630824372759857e-05,
+      "loss": 0.2722,
+      "step": 75120
+    },
+    {
+      "epoch": 484.71,
+      "learning_rate": 2.8627240143369178e-05,
+      "loss": 0.2878,
+      "step": 75130
+    },
+    {
+      "epoch": 484.77,
+      "learning_rate": 2.8623655913978496e-05,
+      "loss": 0.302,
+      "step": 75140
+    },
+    {
+      "epoch": 484.84,
+      "learning_rate": 2.862007168458781e-05,
+      "loss": 0.2172,
+      "step": 75150
+    },
+    {
+      "epoch": 484.9,
+      "learning_rate": 2.8616487455197133e-05,
+      "loss": 0.2648,
+      "step": 75160
+    },
+    {
+      "epoch": 484.96,
+      "learning_rate": 2.8612903225806454e-05,
+      "loss": 0.2721,
+      "step": 75170
+    },
+    {
+      "epoch": 485.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9262298941612244,
+      "eval_runtime": 0.8662,
+      "eval_samples_per_second": 232.038,
+      "eval_steps_per_second": 8.081,
+      "step": 75175
+    },
+    {
+      "epoch": 485.03,
+      "learning_rate": 2.8609318996415773e-05,
+      "loss": 0.2512,
+      "step": 75180
+    },
+    {
+      "epoch": 485.1,
+      "learning_rate": 2.8605734767025087e-05,
+      "loss": 0.2383,
+      "step": 75190
+    },
+    {
+      "epoch": 485.16,
+      "learning_rate": 2.860215053763441e-05,
+      "loss": 0.2781,
+      "step": 75200
+    },
+    {
+      "epoch": 485.23,
+      "learning_rate": 2.859856630824373e-05,
+      "loss": 0.2947,
+      "step": 75210
+    },
+    {
+      "epoch": 485.29,
+      "learning_rate": 2.859498207885305e-05,
+      "loss": 0.243,
+      "step": 75220
+    },
+    {
+      "epoch": 485.35,
+      "learning_rate": 2.8591397849462364e-05,
+      "loss": 0.2655,
+      "step": 75230
+    },
+    {
+      "epoch": 485.42,
+      "learning_rate": 2.8587813620071685e-05,
+      "loss": 0.2218,
+      "step": 75240
+    },
+    {
+      "epoch": 485.48,
+      "learning_rate": 2.8584229390681007e-05,
+      "loss": 0.2935,
+      "step": 75250
+    },
+    {
+      "epoch": 485.55,
+      "learning_rate": 2.8580645161290325e-05,
+      "loss": 0.2481,
+      "step": 75260
+    },
+    {
+      "epoch": 485.61,
+      "learning_rate": 2.857706093189964e-05,
+      "loss": 0.2559,
+      "step": 75270
+    },
+    {
+      "epoch": 485.68,
+      "learning_rate": 2.857347670250896e-05,
+      "loss": 0.3021,
+      "step": 75280
+    },
+    {
+      "epoch": 485.74,
+      "learning_rate": 2.8569892473118283e-05,
+      "loss": 0.2662,
+      "step": 75290
+    },
+    {
+      "epoch": 485.8,
+      "learning_rate": 2.85663082437276e-05,
+      "loss": 0.3049,
+      "step": 75300
+    },
+    {
+      "epoch": 485.87,
+      "learning_rate": 2.8562724014336916e-05,
+      "loss": 0.2636,
+      "step": 75310
+    },
+    {
+      "epoch": 485.93,
+      "learning_rate": 2.8559139784946238e-05,
+      "loss": 0.2705,
+      "step": 75320
+    },
+    {
+      "epoch": 486.0,
+      "learning_rate": 2.855555555555556e-05,
+      "loss": 0.258,
+      "step": 75330
+    },
+    {
+      "epoch": 486.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8890553712844849,
+      "eval_runtime": 0.8809,
+      "eval_samples_per_second": 228.166,
+      "eval_steps_per_second": 7.946,
+      "step": 75330
+    },
+    {
+      "epoch": 486.06,
+      "learning_rate": 2.8551971326164877e-05,
+      "loss": 0.2713,
+      "step": 75340
+    },
+    {
+      "epoch": 486.13,
+      "learning_rate": 2.8548387096774192e-05,
+      "loss": 0.2495,
+      "step": 75350
+    },
+    {
+      "epoch": 486.19,
+      "learning_rate": 2.8544802867383514e-05,
+      "loss": 0.2623,
+      "step": 75360
+    },
+    {
+      "epoch": 486.26,
+      "learning_rate": 2.8541218637992832e-05,
+      "loss": 0.2527,
+      "step": 75370
+    },
+    {
+      "epoch": 486.32,
+      "learning_rate": 2.8537634408602154e-05,
+      "loss": 0.259,
+      "step": 75380
+    },
+    {
+      "epoch": 486.39,
+      "learning_rate": 2.853405017921147e-05,
+      "loss": 0.2611,
+      "step": 75390
+    },
+    {
+      "epoch": 486.45,
+      "learning_rate": 2.853046594982079e-05,
+      "loss": 0.2719,
+      "step": 75400
+    },
+    {
+      "epoch": 486.51,
+      "learning_rate": 2.8526881720430108e-05,
+      "loss": 0.2749,
+      "step": 75410
+    },
+    {
+      "epoch": 486.58,
+      "learning_rate": 2.852329749103943e-05,
+      "loss": 0.2382,
+      "step": 75420
+    },
+    {
+      "epoch": 486.64,
+      "learning_rate": 2.8519713261648745e-05,
+      "loss": 0.2975,
+      "step": 75430
+    },
+    {
+      "epoch": 486.71,
+      "learning_rate": 2.8516129032258066e-05,
+      "loss": 0.2908,
+      "step": 75440
+    },
+    {
+      "epoch": 486.77,
+      "learning_rate": 2.8512544802867384e-05,
+      "loss": 0.2382,
+      "step": 75450
+    },
+    {
+      "epoch": 486.84,
+      "learning_rate": 2.8508960573476706e-05,
+      "loss": 0.2829,
+      "step": 75460
+    },
+    {
+      "epoch": 486.9,
+      "learning_rate": 2.850537634408602e-05,
+      "loss": 0.3031,
+      "step": 75470
+    },
+    {
+      "epoch": 486.96,
+      "learning_rate": 2.8501792114695342e-05,
+      "loss": 0.2718,
+      "step": 75480
+    },
+    {
+      "epoch": 487.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9425584673881531,
+      "eval_runtime": 0.8509,
+      "eval_samples_per_second": 236.229,
+      "eval_steps_per_second": 8.227,
+      "step": 75485
+    },
+    {
+      "epoch": 487.03,
+      "learning_rate": 2.849820788530466e-05,
+      "loss": 0.2475,
+      "step": 75490
+    },
+    {
+      "epoch": 487.1,
+      "learning_rate": 2.8494623655913982e-05,
+      "loss": 0.2796,
+      "step": 75500
+    },
+    {
+      "epoch": 487.16,
+      "learning_rate": 2.8491039426523297e-05,
+      "loss": 0.2253,
+      "step": 75510
+    },
+    {
+      "epoch": 487.23,
+      "learning_rate": 2.848745519713262e-05,
+      "loss": 0.2425,
+      "step": 75520
+    },
+    {
+      "epoch": 487.29,
+      "learning_rate": 2.8483870967741937e-05,
+      "loss": 0.2706,
+      "step": 75530
+    },
+    {
+      "epoch": 487.35,
+      "learning_rate": 2.848028673835126e-05,
+      "loss": 0.3643,
+      "step": 75540
+    },
+    {
+      "epoch": 487.42,
+      "learning_rate": 2.8476702508960573e-05,
+      "loss": 0.2318,
+      "step": 75550
+    },
+    {
+      "epoch": 487.48,
+      "learning_rate": 2.8473118279569895e-05,
+      "loss": 0.325,
+      "step": 75560
+    },
+    {
+      "epoch": 487.55,
+      "learning_rate": 2.8469534050179213e-05,
+      "loss": 0.2495,
+      "step": 75570
+    },
+    {
+      "epoch": 487.61,
+      "learning_rate": 2.8465949820788535e-05,
+      "loss": 0.2858,
+      "step": 75580
+    },
+    {
+      "epoch": 487.68,
+      "learning_rate": 2.846236559139785e-05,
+      "loss": 0.2147,
+      "step": 75590
+    },
+    {
+      "epoch": 487.74,
+      "learning_rate": 2.845878136200717e-05,
+      "loss": 0.2443,
+      "step": 75600
+    },
+    {
+      "epoch": 487.8,
+      "learning_rate": 2.845519713261649e-05,
+      "loss": 0.2572,
+      "step": 75610
+    },
+    {
+      "epoch": 487.87,
+      "learning_rate": 2.845161290322581e-05,
+      "loss": 0.2348,
+      "step": 75620
+    },
+    {
+      "epoch": 487.93,
+      "learning_rate": 2.8448028673835126e-05,
+      "loss": 0.2947,
+      "step": 75630
+    },
+    {
+      "epoch": 488.0,
+      "learning_rate": 2.8444444444444447e-05,
+      "loss": 0.252,
+      "step": 75640
+    },
+    {
+      "epoch": 488.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9719810485839844,
+      "eval_runtime": 0.8421,
+      "eval_samples_per_second": 238.701,
+      "eval_steps_per_second": 8.313,
+      "step": 75640
+    },
+    {
+      "epoch": 488.06,
+      "learning_rate": 2.8440860215053766e-05,
+      "loss": 0.2995,
+      "step": 75650
+    },
+    {
+      "epoch": 488.13,
+      "learning_rate": 2.843727598566308e-05,
+      "loss": 0.2885,
+      "step": 75660
+    },
+    {
+      "epoch": 488.19,
+      "learning_rate": 2.8433691756272402e-05,
+      "loss": 0.2398,
+      "step": 75670
+    },
+    {
+      "epoch": 488.26,
+      "learning_rate": 2.8430107526881724e-05,
+      "loss": 0.2228,
+      "step": 75680
+    },
+    {
+      "epoch": 488.32,
+      "learning_rate": 2.8426523297491042e-05,
+      "loss": 0.2455,
+      "step": 75690
+    },
+    {
+      "epoch": 488.39,
+      "learning_rate": 2.8422939068100357e-05,
+      "loss": 0.2984,
+      "step": 75700
+    },
+    {
+      "epoch": 488.45,
+      "learning_rate": 2.8419354838709678e-05,
+      "loss": 0.2908,
+      "step": 75710
+    },
+    {
+      "epoch": 488.51,
+      "learning_rate": 2.8415770609318996e-05,
+      "loss": 0.2746,
+      "step": 75720
+    },
+    {
+      "epoch": 488.58,
+      "learning_rate": 2.8412186379928318e-05,
+      "loss": 0.2601,
+      "step": 75730
+    },
+    {
+      "epoch": 488.64,
+      "learning_rate": 2.8408602150537633e-05,
+      "loss": 0.2921,
+      "step": 75740
+    },
+    {
+      "epoch": 488.71,
+      "learning_rate": 2.8405017921146954e-05,
+      "loss": 0.2681,
+      "step": 75750
+    },
+    {
+      "epoch": 488.77,
+      "learning_rate": 2.8401433691756273e-05,
+      "loss": 0.3201,
+      "step": 75760
+    },
+    {
+      "epoch": 488.84,
+      "learning_rate": 2.8397849462365594e-05,
+      "loss": 0.2348,
+      "step": 75770
+    },
+    {
+      "epoch": 488.9,
+      "learning_rate": 2.839426523297491e-05,
+      "loss": 0.2639,
+      "step": 75780
+    },
+    {
+      "epoch": 488.96,
+      "learning_rate": 2.839068100358423e-05,
+      "loss": 0.25,
+      "step": 75790
+    },
+    {
+      "epoch": 489.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9393617510795593,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.844,
+      "eval_steps_per_second": 8.074,
+      "step": 75795
+    },
+    {
+      "epoch": 489.03,
+      "learning_rate": 2.838709677419355e-05,
+      "loss": 0.2715,
+      "step": 75800
+    },
+    {
+      "epoch": 489.1,
+      "learning_rate": 2.838351254480287e-05,
+      "loss": 0.2593,
+      "step": 75810
+    },
+    {
+      "epoch": 489.16,
+      "learning_rate": 2.8379928315412185e-05,
+      "loss": 0.2423,
+      "step": 75820
+    },
+    {
+      "epoch": 489.23,
+      "learning_rate": 2.8376344086021507e-05,
+      "loss": 0.2492,
+      "step": 75830
+    },
+    {
+      "epoch": 489.29,
+      "learning_rate": 2.8372759856630825e-05,
+      "loss": 0.294,
+      "step": 75840
+    },
+    {
+      "epoch": 489.35,
+      "learning_rate": 2.8369175627240147e-05,
+      "loss": 0.2712,
+      "step": 75850
+    },
+    {
+      "epoch": 489.42,
+      "learning_rate": 2.836559139784946e-05,
+      "loss": 0.2614,
+      "step": 75860
+    },
+    {
+      "epoch": 489.48,
+      "learning_rate": 2.8362007168458783e-05,
+      "loss": 0.2528,
+      "step": 75870
+    },
+    {
+      "epoch": 489.55,
+      "learning_rate": 2.83584229390681e-05,
+      "loss": 0.28,
+      "step": 75880
+    },
+    {
+      "epoch": 489.61,
+      "learning_rate": 2.8354838709677423e-05,
+      "loss": 0.2674,
+      "step": 75890
+    },
+    {
+      "epoch": 489.68,
+      "learning_rate": 2.8351254480286738e-05,
+      "loss": 0.319,
+      "step": 75900
+    },
+    {
+      "epoch": 489.74,
+      "learning_rate": 2.834767025089606e-05,
+      "loss": 0.3066,
+      "step": 75910
+    },
+    {
+      "epoch": 489.8,
+      "learning_rate": 2.8344086021505377e-05,
+      "loss": 0.2493,
+      "step": 75920
+    },
+    {
+      "epoch": 489.87,
+      "learning_rate": 2.83405017921147e-05,
+      "loss": 0.2527,
+      "step": 75930
+    },
+    {
+      "epoch": 489.93,
+      "learning_rate": 2.8336917562724014e-05,
+      "loss": 0.2459,
+      "step": 75940
+    },
+    {
+      "epoch": 490.0,
+      "learning_rate": 2.8333333333333335e-05,
+      "loss": 0.2735,
+      "step": 75950
+    },
+    {
+      "epoch": 490.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.8977162837982178,
+      "eval_runtime": 0.8591,
+      "eval_samples_per_second": 233.971,
+      "eval_steps_per_second": 8.148,
+      "step": 75950
+    },
+    {
+      "epoch": 490.06,
+      "learning_rate": 2.8329749103942654e-05,
+      "loss": 0.2274,
+      "step": 75960
+    },
+    {
+      "epoch": 490.13,
+      "learning_rate": 2.8326164874551975e-05,
+      "loss": 0.2134,
+      "step": 75970
+    },
+    {
+      "epoch": 490.19,
+      "learning_rate": 2.832258064516129e-05,
+      "loss": 0.2674,
+      "step": 75980
+    },
+    {
+      "epoch": 490.26,
+      "learning_rate": 2.831899641577061e-05,
+      "loss": 0.218,
+      "step": 75990
+    },
+    {
+      "epoch": 490.32,
+      "learning_rate": 2.831541218637993e-05,
+      "loss": 0.2414,
+      "step": 76000
+    },
+    {
+      "epoch": 490.39,
+      "learning_rate": 2.831182795698925e-05,
+      "loss": 0.2441,
+      "step": 76010
+    },
+    {
+      "epoch": 490.45,
+      "learning_rate": 2.8308243727598566e-05,
+      "loss": 0.2457,
+      "step": 76020
+    },
+    {
+      "epoch": 490.51,
+      "learning_rate": 2.8304659498207888e-05,
+      "loss": 0.2364,
+      "step": 76030
+    },
+    {
+      "epoch": 490.58,
+      "learning_rate": 2.8301075268817206e-05,
+      "loss": 0.2766,
+      "step": 76040
+    },
+    {
+      "epoch": 490.64,
+      "learning_rate": 2.8297491039426528e-05,
+      "loss": 0.2562,
+      "step": 76050
+    },
+    {
+      "epoch": 490.71,
+      "learning_rate": 2.8293906810035842e-05,
+      "loss": 0.2869,
+      "step": 76060
+    },
+    {
+      "epoch": 490.77,
+      "learning_rate": 2.829032258064516e-05,
+      "loss": 0.2816,
+      "step": 76070
+    },
+    {
+      "epoch": 490.84,
+      "learning_rate": 2.8286738351254482e-05,
+      "loss": 0.2097,
+      "step": 76080
+    },
+    {
+      "epoch": 490.9,
+      "learning_rate": 2.8283154121863804e-05,
+      "loss": 0.3025,
+      "step": 76090
+    },
+    {
+      "epoch": 490.96,
+      "learning_rate": 2.827956989247312e-05,
+      "loss": 0.22,
+      "step": 76100
+    },
+    {
+      "epoch": 491.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9317540526390076,
+      "eval_runtime": 0.8676,
+      "eval_samples_per_second": 231.663,
+      "eval_steps_per_second": 8.068,
+      "step": 76105
+    },
+    {
+      "epoch": 491.03,
+      "learning_rate": 2.8275985663082437e-05,
+      "loss": 0.2195,
+      "step": 76110
+    },
+    {
+      "epoch": 491.1,
+      "learning_rate": 2.827240143369176e-05,
+      "loss": 0.3036,
+      "step": 76120
+    },
+    {
+      "epoch": 491.16,
+      "learning_rate": 2.826881720430108e-05,
+      "loss": 0.2777,
+      "step": 76130
+    },
+    {
+      "epoch": 491.23,
+      "learning_rate": 2.8265232974910395e-05,
+      "loss": 0.2778,
+      "step": 76140
+    },
+    {
+      "epoch": 491.29,
+      "learning_rate": 2.8261648745519713e-05,
+      "loss": 0.2045,
+      "step": 76150
+    },
+    {
+      "epoch": 491.35,
+      "learning_rate": 2.8258064516129035e-05,
+      "loss": 0.2784,
+      "step": 76160
+    },
+    {
+      "epoch": 491.42,
+      "learning_rate": 2.8254480286738356e-05,
+      "loss": 0.29,
+      "step": 76170
+    },
+    {
+      "epoch": 491.48,
+      "learning_rate": 2.825089605734767e-05,
+      "loss": 0.3037,
+      "step": 76180
+    },
+    {
+      "epoch": 491.55,
+      "learning_rate": 2.824731182795699e-05,
+      "loss": 0.3068,
+      "step": 76190
+    },
+    {
+      "epoch": 491.61,
+      "learning_rate": 2.824372759856631e-05,
+      "loss": 0.273,
+      "step": 76200
+    },
+    {
+      "epoch": 491.68,
+      "learning_rate": 2.8240143369175632e-05,
+      "loss": 0.2737,
+      "step": 76210
+    },
+    {
+      "epoch": 491.74,
+      "learning_rate": 2.8236559139784947e-05,
+      "loss": 0.3273,
+      "step": 76220
+    },
+    {
+      "epoch": 491.8,
+      "learning_rate": 2.8232974910394265e-05,
+      "loss": 0.2451,
+      "step": 76230
+    },
+    {
+      "epoch": 491.87,
+      "learning_rate": 2.8229390681003587e-05,
+      "loss": 0.2881,
+      "step": 76240
+    },
+    {
+      "epoch": 491.93,
+      "learning_rate": 2.822580645161291e-05,
+      "loss": 0.2608,
+      "step": 76250
+    },
+    {
+      "epoch": 492.0,
+      "learning_rate": 2.8222222222222223e-05,
+      "loss": 0.3009,
+      "step": 76260
+    },
+    {
+      "epoch": 492.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9306316375732422,
+      "eval_runtime": 0.8408,
+      "eval_samples_per_second": 239.068,
+      "eval_steps_per_second": 8.326,
+      "step": 76260
+    },
+    {
+      "epoch": 492.06,
+      "learning_rate": 2.821863799283154e-05,
+      "loss": 0.2492,
+      "step": 76270
+    },
+    {
+      "epoch": 492.13,
+      "learning_rate": 2.8215053763440863e-05,
+      "loss": 0.2676,
+      "step": 76280
+    },
+    {
+      "epoch": 492.19,
+      "learning_rate": 2.8211469534050185e-05,
+      "loss": 0.2701,
+      "step": 76290
+    },
+    {
+      "epoch": 492.26,
+      "learning_rate": 2.82078853046595e-05,
+      "loss": 0.1966,
+      "step": 76300
+    },
+    {
+      "epoch": 492.32,
+      "learning_rate": 2.8204301075268818e-05,
+      "loss": 0.2225,
+      "step": 76310
+    },
+    {
+      "epoch": 492.39,
+      "learning_rate": 2.820071684587814e-05,
+      "loss": 0.2861,
+      "step": 76320
+    },
+    {
+      "epoch": 492.45,
+      "learning_rate": 2.8197132616487454e-05,
+      "loss": 0.2768,
+      "step": 76330
+    },
+    {
+      "epoch": 492.51,
+      "learning_rate": 2.8193548387096776e-05,
+      "loss": 0.2282,
+      "step": 76340
+    },
+    {
+      "epoch": 492.58,
+      "learning_rate": 2.8189964157706094e-05,
+      "loss": 0.2566,
+      "step": 76350
+    },
+    {
+      "epoch": 492.64,
+      "learning_rate": 2.8186379928315416e-05,
+      "loss": 0.2302,
+      "step": 76360
+    },
+    {
+      "epoch": 492.71,
+      "learning_rate": 2.818279569892473e-05,
+      "loss": 0.2924,
+      "step": 76370
+    },
+    {
+      "epoch": 492.77,
+      "learning_rate": 2.817921146953405e-05,
+      "loss": 0.2546,
+      "step": 76380
+    },
+    {
+      "epoch": 492.84,
+      "learning_rate": 2.817562724014337e-05,
+      "loss": 0.2738,
+      "step": 76390
+    },
+    {
+      "epoch": 492.9,
+      "learning_rate": 2.8172043010752692e-05,
+      "loss": 0.2481,
+      "step": 76400
+    },
+    {
+      "epoch": 492.96,
+      "learning_rate": 2.8168458781362007e-05,
+      "loss": 0.2884,
+      "step": 76410
+    },
+    {
+      "epoch": 493.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9090216159820557,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.469,
+      "eval_steps_per_second": 8.026,
+      "step": 76415
+    },
+    {
+      "epoch": 493.03,
+      "learning_rate": 2.8164874551971325e-05,
+      "loss": 0.3202,
+      "step": 76420
+    },
+    {
+      "epoch": 493.1,
+      "learning_rate": 2.8161290322580646e-05,
+      "loss": 0.2537,
+      "step": 76430
+    },
+    {
+      "epoch": 493.16,
+      "learning_rate": 2.8157706093189968e-05,
+      "loss": 0.2385,
+      "step": 76440
+    },
+    {
+      "epoch": 493.23,
+      "learning_rate": 2.8154121863799283e-05,
+      "loss": 0.2816,
+      "step": 76450
+    },
+    {
+      "epoch": 493.29,
+      "learning_rate": 2.81505376344086e-05,
+      "loss": 0.2445,
+      "step": 76460
+    },
+    {
+      "epoch": 493.35,
+      "learning_rate": 2.8146953405017923e-05,
+      "loss": 0.3038,
+      "step": 76470
+    },
+    {
+      "epoch": 493.42,
+      "learning_rate": 2.8143369175627244e-05,
+      "loss": 0.2499,
+      "step": 76480
+    },
+    {
+      "epoch": 493.48,
+      "learning_rate": 2.813978494623656e-05,
+      "loss": 0.2451,
+      "step": 76490
+    },
+    {
+      "epoch": 493.55,
+      "learning_rate": 2.8136200716845877e-05,
+      "loss": 0.2785,
+      "step": 76500
+    },
+    {
+      "epoch": 493.61,
+      "learning_rate": 2.81326164874552e-05,
+      "loss": 0.3053,
+      "step": 76510
+    },
+    {
+      "epoch": 493.68,
+      "learning_rate": 2.812903225806452e-05,
+      "loss": 0.3028,
+      "step": 76520
+    },
+    {
+      "epoch": 493.74,
+      "learning_rate": 2.8125448028673835e-05,
+      "loss": 0.299,
+      "step": 76530
+    },
+    {
+      "epoch": 493.8,
+      "learning_rate": 2.8121863799283154e-05,
+      "loss": 0.2826,
+      "step": 76540
+    },
+    {
+      "epoch": 493.87,
+      "learning_rate": 2.8118279569892475e-05,
+      "loss": 0.2419,
+      "step": 76550
+    },
+    {
+      "epoch": 493.93,
+      "learning_rate": 2.8114695340501797e-05,
+      "loss": 0.2407,
+      "step": 76560
+    },
+    {
+      "epoch": 494.0,
+      "learning_rate": 2.811111111111111e-05,
+      "loss": 0.2531,
+      "step": 76570
+    },
+    {
+      "epoch": 494.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.935100793838501,
+      "eval_runtime": 0.89,
+      "eval_samples_per_second": 225.834,
+      "eval_steps_per_second": 7.865,
+      "step": 76570
+    },
+    {
+      "epoch": 494.06,
+      "learning_rate": 2.810752688172043e-05,
+      "loss": 0.286,
+      "step": 76580
+    },
+    {
+      "epoch": 494.13,
+      "learning_rate": 2.810394265232975e-05,
+      "loss": 0.2444,
+      "step": 76590
+    },
+    {
+      "epoch": 494.19,
+      "learning_rate": 2.8100358422939073e-05,
+      "loss": 0.3445,
+      "step": 76600
+    },
+    {
+      "epoch": 494.26,
+      "learning_rate": 2.8096774193548388e-05,
+      "loss": 0.2558,
+      "step": 76610
+    },
+    {
+      "epoch": 494.32,
+      "learning_rate": 2.8093189964157706e-05,
+      "loss": 0.2676,
+      "step": 76620
+    },
+    {
+      "epoch": 494.39,
+      "learning_rate": 2.8089605734767028e-05,
+      "loss": 0.2054,
+      "step": 76630
+    },
+    {
+      "epoch": 494.45,
+      "learning_rate": 2.808602150537635e-05,
+      "loss": 0.2631,
+      "step": 76640
+    },
+    {
+      "epoch": 494.51,
+      "learning_rate": 2.8082437275985664e-05,
+      "loss": 0.2919,
+      "step": 76650
+    },
+    {
+      "epoch": 494.58,
+      "learning_rate": 2.8078853046594982e-05,
+      "loss": 0.2511,
+      "step": 76660
+    },
+    {
+      "epoch": 494.64,
+      "learning_rate": 2.8075268817204304e-05,
+      "loss": 0.2525,
+      "step": 76670
+    },
+    {
+      "epoch": 494.71,
+      "learning_rate": 2.8071684587813625e-05,
+      "loss": 0.2613,
+      "step": 76680
+    },
+    {
+      "epoch": 494.77,
+      "learning_rate": 2.806810035842294e-05,
+      "loss": 0.2759,
+      "step": 76690
+    },
+    {
+      "epoch": 494.84,
+      "learning_rate": 2.806451612903226e-05,
+      "loss": 0.274,
+      "step": 76700
+    },
+    {
+      "epoch": 494.9,
+      "learning_rate": 2.806093189964158e-05,
+      "loss": 0.2784,
+      "step": 76710
+    },
+    {
+      "epoch": 494.96,
+      "learning_rate": 2.8057347670250898e-05,
+      "loss": 0.2694,
+      "step": 76720
+    },
+    {
+      "epoch": 495.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9303690791130066,
+      "eval_runtime": 0.8546,
+      "eval_samples_per_second": 235.186,
+      "eval_steps_per_second": 8.191,
+      "step": 76725
+    },
+    {
+      "epoch": 495.03,
+      "learning_rate": 2.8053763440860213e-05,
+      "loss": 0.2415,
+      "step": 76730
+    },
+    {
+      "epoch": 495.1,
+      "learning_rate": 2.8050179211469535e-05,
+      "loss": 0.2416,
+      "step": 76740
+    },
+    {
+      "epoch": 495.16,
+      "learning_rate": 2.8046594982078856e-05,
+      "loss": 0.292,
+      "step": 76750
+    },
+    {
+      "epoch": 495.23,
+      "learning_rate": 2.8043010752688174e-05,
+      "loss": 0.262,
+      "step": 76760
+    },
+    {
+      "epoch": 495.29,
+      "learning_rate": 2.803942652329749e-05,
+      "loss": 0.2731,
+      "step": 76770
+    },
+    {
+      "epoch": 495.35,
+      "learning_rate": 2.803584229390681e-05,
+      "loss": 0.2674,
+      "step": 76780
+    },
+    {
+      "epoch": 495.42,
+      "learning_rate": 2.8032258064516132e-05,
+      "loss": 0.2755,
+      "step": 76790
+    },
+    {
+      "epoch": 495.48,
+      "learning_rate": 2.802867383512545e-05,
+      "loss": 0.2605,
+      "step": 76800
+    },
+    {
+      "epoch": 495.55,
+      "learning_rate": 2.8025089605734765e-05,
+      "loss": 0.2471,
+      "step": 76810
+    },
+    {
+      "epoch": 495.61,
+      "learning_rate": 2.8021505376344087e-05,
+      "loss": 0.2848,
+      "step": 76820
+    },
+    {
+      "epoch": 495.68,
+      "learning_rate": 2.801792114695341e-05,
+      "loss": 0.2976,
+      "step": 76830
+    },
+    {
+      "epoch": 495.74,
+      "learning_rate": 2.8014336917562727e-05,
+      "loss": 0.3098,
+      "step": 76840
+    },
+    {
+      "epoch": 495.8,
+      "learning_rate": 2.801075268817204e-05,
+      "loss": 0.2642,
+      "step": 76850
+    },
+    {
+      "epoch": 495.87,
+      "learning_rate": 2.8007168458781363e-05,
+      "loss": 0.2684,
+      "step": 76860
+    },
+    {
+      "epoch": 495.93,
+      "learning_rate": 2.8003584229390685e-05,
+      "loss": 0.3394,
+      "step": 76870
+    },
+    {
+      "epoch": 496.0,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.2861,
+      "step": 76880
+    },
+    {
+      "epoch": 496.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9425844550132751,
+      "eval_runtime": 0.868,
+      "eval_samples_per_second": 231.559,
+      "eval_steps_per_second": 8.064,
+      "step": 76880
+    },
+    {
+      "epoch": 496.06,
+      "learning_rate": 2.7996415770609318e-05,
+      "loss": 0.2761,
+      "step": 76890
+    },
+    {
+      "epoch": 496.13,
+      "learning_rate": 2.799283154121864e-05,
+      "loss": 0.2326,
+      "step": 76900
+    },
+    {
+      "epoch": 496.19,
+      "learning_rate": 2.798924731182796e-05,
+      "loss": 0.2672,
+      "step": 76910
+    },
+    {
+      "epoch": 496.26,
+      "learning_rate": 2.798566308243728e-05,
+      "loss": 0.2751,
+      "step": 76920
+    },
+    {
+      "epoch": 496.32,
+      "learning_rate": 2.7982078853046594e-05,
+      "loss": 0.2573,
+      "step": 76930
+    },
+    {
+      "epoch": 496.39,
+      "learning_rate": 2.7978494623655916e-05,
+      "loss": 0.2284,
+      "step": 76940
+    },
+    {
+      "epoch": 496.45,
+      "learning_rate": 2.7974910394265237e-05,
+      "loss": 0.247,
+      "step": 76950
+    },
+    {
+      "epoch": 496.51,
+      "learning_rate": 2.7971326164874552e-05,
+      "loss": 0.2602,
+      "step": 76960
+    },
+    {
+      "epoch": 496.58,
+      "learning_rate": 2.796774193548387e-05,
+      "loss": 0.2517,
+      "step": 76970
+    },
+    {
+      "epoch": 496.64,
+      "learning_rate": 2.7964157706093192e-05,
+      "loss": 0.2334,
+      "step": 76980
+    },
+    {
+      "epoch": 496.71,
+      "learning_rate": 2.7960573476702513e-05,
+      "loss": 0.2645,
+      "step": 76990
+    },
+    {
+      "epoch": 496.77,
+      "learning_rate": 2.7956989247311828e-05,
+      "loss": 0.2901,
+      "step": 77000
+    },
+    {
+      "epoch": 496.84,
+      "learning_rate": 2.7953405017921146e-05,
+      "loss": 0.2893,
+      "step": 77010
+    },
+    {
+      "epoch": 496.9,
+      "learning_rate": 2.7949820788530468e-05,
+      "loss": 0.245,
+      "step": 77020
+    },
+    {
+      "epoch": 496.96,
+      "learning_rate": 2.7946236559139786e-05,
+      "loss": 0.2793,
+      "step": 77030
+    },
+    {
+      "epoch": 497.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9734618663787842,
+      "eval_runtime": 0.8634,
+      "eval_samples_per_second": 232.812,
+      "eval_steps_per_second": 8.108,
+      "step": 77035
+    },
+    {
+      "epoch": 497.03,
+      "learning_rate": 2.7942652329749104e-05,
+      "loss": 0.3156,
+      "step": 77040
+    },
+    {
+      "epoch": 497.1,
+      "learning_rate": 2.7939068100358423e-05,
+      "loss": 0.2295,
+      "step": 77050
+    },
+    {
+      "epoch": 497.16,
+      "learning_rate": 2.7935483870967744e-05,
+      "loss": 0.3056,
+      "step": 77060
+    },
+    {
+      "epoch": 497.23,
+      "learning_rate": 2.7931899641577062e-05,
+      "loss": 0.2614,
+      "step": 77070
+    },
+    {
+      "epoch": 497.29,
+      "learning_rate": 2.7928315412186377e-05,
+      "loss": 0.2818,
+      "step": 77080
+    },
+    {
+      "epoch": 497.35,
+      "learning_rate": 2.79247311827957e-05,
+      "loss": 0.2509,
+      "step": 77090
+    },
+    {
+      "epoch": 497.42,
+      "learning_rate": 2.792114695340502e-05,
+      "loss": 0.246,
+      "step": 77100
+    },
+    {
+      "epoch": 497.48,
+      "learning_rate": 2.791756272401434e-05,
+      "loss": 0.2966,
+      "step": 77110
+    },
+    {
+      "epoch": 497.55,
+      "learning_rate": 2.7913978494623653e-05,
+      "loss": 0.2501,
+      "step": 77120
+    },
+    {
+      "epoch": 497.61,
+      "learning_rate": 2.7910394265232975e-05,
+      "loss": 0.3315,
+      "step": 77130
+    },
+    {
+      "epoch": 497.68,
+      "learning_rate": 2.7906810035842297e-05,
+      "loss": 0.2526,
+      "step": 77140
+    },
+    {
+      "epoch": 497.74,
+      "learning_rate": 2.7903225806451615e-05,
+      "loss": 0.2545,
+      "step": 77150
+    },
+    {
+      "epoch": 497.8,
+      "learning_rate": 2.789964157706093e-05,
+      "loss": 0.314,
+      "step": 77160
+    },
+    {
+      "epoch": 497.87,
+      "learning_rate": 2.789605734767025e-05,
+      "loss": 0.2269,
+      "step": 77170
+    },
+    {
+      "epoch": 497.93,
+      "learning_rate": 2.7892473118279573e-05,
+      "loss": 0.2595,
+      "step": 77180
+    },
+    {
+      "epoch": 498.0,
+      "learning_rate": 2.788888888888889e-05,
+      "loss": 0.2607,
+      "step": 77190
+    },
+    {
+      "epoch": 498.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9826405048370361,
+      "eval_runtime": 0.8658,
+      "eval_samples_per_second": 232.162,
+      "eval_steps_per_second": 8.085,
+      "step": 77190
+    },
+    {
+      "epoch": 498.06,
+      "learning_rate": 2.7885304659498206e-05,
+      "loss": 0.2759,
+      "step": 77200
+    },
+    {
+      "epoch": 498.13,
+      "learning_rate": 2.7881720430107527e-05,
+      "loss": 0.2713,
+      "step": 77210
+    },
+    {
+      "epoch": 498.19,
+      "learning_rate": 2.787813620071685e-05,
+      "loss": 0.2725,
+      "step": 77220
+    },
+    {
+      "epoch": 498.26,
+      "learning_rate": 2.7874551971326167e-05,
+      "loss": 0.2517,
+      "step": 77230
+    },
+    {
+      "epoch": 498.32,
+      "learning_rate": 2.7870967741935482e-05,
+      "loss": 0.2734,
+      "step": 77240
+    },
+    {
+      "epoch": 498.39,
+      "learning_rate": 2.7867383512544804e-05,
+      "loss": 0.2838,
+      "step": 77250
+    },
+    {
+      "epoch": 498.45,
+      "learning_rate": 2.7863799283154125e-05,
+      "loss": 0.2895,
+      "step": 77260
+    },
+    {
+      "epoch": 498.51,
+      "learning_rate": 2.7860215053763443e-05,
+      "loss": 0.2444,
+      "step": 77270
+    },
+    {
+      "epoch": 498.58,
+      "learning_rate": 2.7856630824372758e-05,
+      "loss": 0.2245,
+      "step": 77280
+    },
+    {
+      "epoch": 498.64,
+      "learning_rate": 2.785304659498208e-05,
+      "loss": 0.2527,
+      "step": 77290
+    },
+    {
+      "epoch": 498.71,
+      "learning_rate": 2.78494623655914e-05,
+      "loss": 0.2403,
+      "step": 77300
+    },
+    {
+      "epoch": 498.77,
+      "learning_rate": 2.784587813620072e-05,
+      "loss": 0.2797,
+      "step": 77310
+    },
+    {
+      "epoch": 498.84,
+      "learning_rate": 2.7842293906810034e-05,
+      "loss": 0.2366,
+      "step": 77320
+    },
+    {
+      "epoch": 498.9,
+      "learning_rate": 2.7838709677419356e-05,
+      "loss": 0.2648,
+      "step": 77330
+    },
+    {
+      "epoch": 498.96,
+      "learning_rate": 2.7835125448028678e-05,
+      "loss": 0.2893,
+      "step": 77340
+    },
+    {
+      "epoch": 499.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0050472021102905,
+      "eval_runtime": 0.8602,
+      "eval_samples_per_second": 233.66,
+      "eval_steps_per_second": 8.137,
+      "step": 77345
+    },
+    {
+      "epoch": 499.03,
+      "learning_rate": 2.7831541218637996e-05,
+      "loss": 0.2646,
+      "step": 77350
+    },
+    {
+      "epoch": 499.1,
+      "learning_rate": 2.782795698924731e-05,
+      "loss": 0.2174,
+      "step": 77360
+    },
+    {
+      "epoch": 499.16,
+      "learning_rate": 2.7824372759856632e-05,
+      "loss": 0.2736,
+      "step": 77370
+    },
+    {
+      "epoch": 499.23,
+      "learning_rate": 2.782078853046595e-05,
+      "loss": 0.2629,
+      "step": 77380
+    },
+    {
+      "epoch": 499.29,
+      "learning_rate": 2.7817204301075272e-05,
+      "loss": 0.3134,
+      "step": 77390
+    },
+    {
+      "epoch": 499.35,
+      "learning_rate": 2.7813620071684587e-05,
+      "loss": 0.2949,
+      "step": 77400
+    },
+    {
+      "epoch": 499.42,
+      "learning_rate": 2.781003584229391e-05,
+      "loss": 0.206,
+      "step": 77410
+    },
+    {
+      "epoch": 499.48,
+      "learning_rate": 2.7806451612903227e-05,
+      "loss": 0.2717,
+      "step": 77420
+    },
+    {
+      "epoch": 499.55,
+      "learning_rate": 2.7802867383512548e-05,
+      "loss": 0.2316,
+      "step": 77430
+    },
+    {
+      "epoch": 499.61,
+      "learning_rate": 2.7799283154121863e-05,
+      "loss": 0.2379,
+      "step": 77440
+    },
+    {
+      "epoch": 499.68,
+      "learning_rate": 2.7795698924731185e-05,
+      "loss": 0.3062,
+      "step": 77450
+    },
+    {
+      "epoch": 499.74,
+      "learning_rate": 2.7792114695340503e-05,
+      "loss": 0.2967,
+      "step": 77460
+    },
+    {
+      "epoch": 499.8,
+      "learning_rate": 2.7788530465949824e-05,
+      "loss": 0.2693,
+      "step": 77470
+    },
+    {
+      "epoch": 499.87,
+      "learning_rate": 2.778494623655914e-05,
+      "loss": 0.2773,
+      "step": 77480
+    },
+    {
+      "epoch": 499.93,
+      "learning_rate": 2.778136200716846e-05,
+      "loss": 0.2409,
+      "step": 77490
+    },
+    {
+      "epoch": 500.0,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 0.2689,
+      "step": 77500
+    },
+    {
+      "epoch": 500.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0201218128204346,
+      "eval_runtime": 0.8535,
+      "eval_samples_per_second": 235.513,
+      "eval_steps_per_second": 8.202,
+      "step": 77500
+    },
+    {
+      "epoch": 500.06,
+      "learning_rate": 2.77741935483871e-05,
+      "loss": 0.3092,
+      "step": 77510
+    },
+    {
+      "epoch": 500.13,
+      "learning_rate": 2.7770609318996416e-05,
+      "loss": 0.2618,
+      "step": 77520
+    },
+    {
+      "epoch": 500.19,
+      "learning_rate": 2.7767025089605737e-05,
+      "loss": 0.2849,
+      "step": 77530
+    },
+    {
+      "epoch": 500.26,
+      "learning_rate": 2.7763440860215055e-05,
+      "loss": 0.2631,
+      "step": 77540
+    },
+    {
+      "epoch": 500.32,
+      "learning_rate": 2.7759856630824377e-05,
+      "loss": 0.239,
+      "step": 77550
+    },
+    {
+      "epoch": 500.39,
+      "learning_rate": 2.7756272401433692e-05,
+      "loss": 0.2614,
+      "step": 77560
+    },
+    {
+      "epoch": 500.45,
+      "learning_rate": 2.7752688172043013e-05,
+      "loss": 0.2826,
+      "step": 77570
+    },
+    {
+      "epoch": 500.51,
+      "learning_rate": 2.774910394265233e-05,
+      "loss": 0.2476,
+      "step": 77580
+    },
+    {
+      "epoch": 500.58,
+      "learning_rate": 2.7745519713261646e-05,
+      "loss": 0.2711,
+      "step": 77590
+    },
+    {
+      "epoch": 500.64,
+      "learning_rate": 2.7741935483870968e-05,
+      "loss": 0.227,
+      "step": 77600
+    },
+    {
+      "epoch": 500.71,
+      "learning_rate": 2.773835125448029e-05,
+      "loss": 0.2333,
+      "step": 77610
+    },
+    {
+      "epoch": 500.77,
+      "learning_rate": 2.7734767025089608e-05,
+      "loss": 0.2165,
+      "step": 77620
+    },
+    {
+      "epoch": 500.84,
+      "learning_rate": 2.7731182795698923e-05,
+      "loss": 0.233,
+      "step": 77630
+    },
+    {
+      "epoch": 500.9,
+      "learning_rate": 2.7727598566308244e-05,
+      "loss": 0.2435,
+      "step": 77640
+    },
+    {
+      "epoch": 500.96,
+      "learning_rate": 2.7724014336917566e-05,
+      "loss": 0.229,
+      "step": 77650
+    },
+    {
+      "epoch": 501.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0188658237457275,
+      "eval_runtime": 0.8634,
+      "eval_samples_per_second": 232.802,
+      "eval_steps_per_second": 8.108,
+      "step": 77655
+    },
+    {
+      "epoch": 501.03,
+      "learning_rate": 2.7720430107526884e-05,
+      "loss": 0.2471,
+      "step": 77660
+    },
+    {
+      "epoch": 501.1,
+      "learning_rate": 2.77168458781362e-05,
+      "loss": 0.2522,
+      "step": 77670
+    },
+    {
+      "epoch": 501.16,
+      "learning_rate": 2.771326164874552e-05,
+      "loss": 0.232,
+      "step": 77680
+    },
+    {
+      "epoch": 501.23,
+      "learning_rate": 2.7709677419354842e-05,
+      "loss": 0.2456,
+      "step": 77690
+    },
+    {
+      "epoch": 501.29,
+      "learning_rate": 2.770609318996416e-05,
+      "loss": 0.251,
+      "step": 77700
+    },
+    {
+      "epoch": 501.35,
+      "learning_rate": 2.7702508960573475e-05,
+      "loss": 0.2345,
+      "step": 77710
+    },
+    {
+      "epoch": 501.42,
+      "learning_rate": 2.7698924731182797e-05,
+      "loss": 0.2698,
+      "step": 77720
+    },
+    {
+      "epoch": 501.48,
+      "learning_rate": 2.7695340501792115e-05,
+      "loss": 0.1972,
+      "step": 77730
+    },
+    {
+      "epoch": 501.55,
+      "learning_rate": 2.7691756272401436e-05,
+      "loss": 0.25,
+      "step": 77740
+    },
+    {
+      "epoch": 501.61,
+      "learning_rate": 2.768817204301075e-05,
+      "loss": 0.27,
+      "step": 77750
+    },
+    {
+      "epoch": 501.68,
+      "learning_rate": 2.7684587813620073e-05,
+      "loss": 0.2584,
+      "step": 77760
+    },
+    {
+      "epoch": 501.74,
+      "learning_rate": 2.768100358422939e-05,
+      "loss": 0.249,
+      "step": 77770
+    },
+    {
+      "epoch": 501.8,
+      "learning_rate": 2.7677419354838713e-05,
+      "loss": 0.2471,
+      "step": 77780
+    },
+    {
+      "epoch": 501.87,
+      "learning_rate": 2.7673835125448027e-05,
+      "loss": 0.2407,
+      "step": 77790
+    },
+    {
+      "epoch": 501.93,
+      "learning_rate": 2.767025089605735e-05,
+      "loss": 0.2719,
+      "step": 77800
+    },
+    {
+      "epoch": 502.0,
+      "learning_rate": 2.7666666666666667e-05,
+      "loss": 0.282,
+      "step": 77810
+    },
+    {
+      "epoch": 502.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 1.013315200805664,
+      "eval_runtime": 0.8741,
+      "eval_samples_per_second": 229.944,
+      "eval_steps_per_second": 8.008,
+      "step": 77810
+    },
+    {
+      "epoch": 502.06,
+      "learning_rate": 2.766308243727599e-05,
+      "loss": 0.2832,
+      "step": 77820
+    },
+    {
+      "epoch": 502.13,
+      "learning_rate": 2.7659498207885304e-05,
+      "loss": 0.2579,
+      "step": 77830
+    },
+    {
+      "epoch": 502.19,
+      "learning_rate": 2.7655913978494625e-05,
+      "loss": 0.2646,
+      "step": 77840
+    },
+    {
+      "epoch": 502.26,
+      "learning_rate": 2.7652329749103943e-05,
+      "loss": 0.246,
+      "step": 77850
+    },
+    {
+      "epoch": 502.32,
+      "learning_rate": 2.7648745519713265e-05,
+      "loss": 0.2265,
+      "step": 77860
+    },
+    {
+      "epoch": 502.39,
+      "learning_rate": 2.764516129032258e-05,
+      "loss": 0.2613,
+      "step": 77870
+    },
+    {
+      "epoch": 502.45,
+      "learning_rate": 2.76415770609319e-05,
+      "loss": 0.234,
+      "step": 77880
+    },
+    {
+      "epoch": 502.51,
+      "learning_rate": 2.763799283154122e-05,
+      "loss": 0.2874,
+      "step": 77890
+    },
+    {
+      "epoch": 502.58,
+      "learning_rate": 2.763440860215054e-05,
+      "loss": 0.2479,
+      "step": 77900
+    },
+    {
+      "epoch": 502.64,
+      "learning_rate": 2.7630824372759856e-05,
+      "loss": 0.2455,
+      "step": 77910
+    },
+    {
+      "epoch": 502.71,
+      "learning_rate": 2.7627240143369178e-05,
+      "loss": 0.2549,
+      "step": 77920
+    },
+    {
+      "epoch": 502.77,
+      "learning_rate": 2.7623655913978496e-05,
+      "loss": 0.2735,
+      "step": 77930
+    },
+    {
+      "epoch": 502.84,
+      "learning_rate": 2.7620071684587817e-05,
+      "loss": 0.285,
+      "step": 77940
+    },
+    {
+      "epoch": 502.9,
+      "learning_rate": 2.7616487455197132e-05,
+      "loss": 0.2319,
+      "step": 77950
+    },
+    {
+      "epoch": 502.96,
+      "learning_rate": 2.7612903225806454e-05,
+      "loss": 0.2371,
+      "step": 77960
+    },
+    {
+      "epoch": 503.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9993953704833984,
+      "eval_runtime": 0.8556,
+      "eval_samples_per_second": 234.919,
+      "eval_steps_per_second": 8.181,
+      "step": 77965
+    },
+    {
+      "epoch": 503.03,
+      "learning_rate": 2.7609318996415772e-05,
+      "loss": 0.2233,
+      "step": 77970
+    },
+    {
+      "epoch": 503.1,
+      "learning_rate": 2.7605734767025094e-05,
+      "loss": 0.2545,
+      "step": 77980
+    },
+    {
+      "epoch": 503.16,
+      "learning_rate": 2.760215053763441e-05,
+      "loss": 0.311,
+      "step": 77990
+    },
+    {
+      "epoch": 503.23,
+      "learning_rate": 2.759856630824373e-05,
+      "loss": 0.2383,
+      "step": 78000
+    },
+    {
+      "epoch": 503.29,
+      "learning_rate": 2.7594982078853048e-05,
+      "loss": 0.2615,
+      "step": 78010
+    },
+    {
+      "epoch": 503.35,
+      "learning_rate": 2.759139784946237e-05,
+      "loss": 0.2561,
+      "step": 78020
+    },
+    {
+      "epoch": 503.42,
+      "learning_rate": 2.7587813620071685e-05,
+      "loss": 0.2735,
+      "step": 78030
+    },
+    {
+      "epoch": 503.48,
+      "learning_rate": 2.7584229390681006e-05,
+      "loss": 0.2933,
+      "step": 78040
+    },
+    {
+      "epoch": 503.55,
+      "learning_rate": 2.7580645161290324e-05,
+      "loss": 0.2806,
+      "step": 78050
+    },
+    {
+      "epoch": 503.61,
+      "learning_rate": 2.7577060931899646e-05,
+      "loss": 0.2537,
+      "step": 78060
+    },
+    {
+      "epoch": 503.68,
+      "learning_rate": 2.757347670250896e-05,
+      "loss": 0.2628,
+      "step": 78070
+    },
+    {
+      "epoch": 503.74,
+      "learning_rate": 2.756989247311828e-05,
+      "loss": 0.268,
+      "step": 78080
+    },
+    {
+      "epoch": 503.8,
+      "learning_rate": 2.75663082437276e-05,
+      "loss": 0.2666,
+      "step": 78090
+    },
+    {
+      "epoch": 503.87,
+      "learning_rate": 2.7562724014336922e-05,
+      "loss": 0.2334,
+      "step": 78100
+    },
+    {
+      "epoch": 503.93,
+      "learning_rate": 2.7559139784946237e-05,
+      "loss": 0.2801,
+      "step": 78110
+    },
+    {
+      "epoch": 504.0,
+      "learning_rate": 2.7555555555555555e-05,
+      "loss": 0.2502,
+      "step": 78120
+    },
+    {
+      "epoch": 504.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9770559668540955,
+      "eval_runtime": 0.8383,
+      "eval_samples_per_second": 239.758,
+      "eval_steps_per_second": 8.35,
+      "step": 78120
+    },
+    {
+      "epoch": 504.06,
+      "learning_rate": 2.7551971326164877e-05,
+      "loss": 0.2433,
+      "step": 78130
+    },
+    {
+      "epoch": 504.13,
+      "learning_rate": 2.75483870967742e-05,
+      "loss": 0.2635,
+      "step": 78140
+    },
+    {
+      "epoch": 504.19,
+      "learning_rate": 2.7544802867383513e-05,
+      "loss": 0.2791,
+      "step": 78150
+    },
+    {
+      "epoch": 504.26,
+      "learning_rate": 2.754121863799283e-05,
+      "loss": 0.2316,
+      "step": 78160
+    },
+    {
+      "epoch": 504.32,
+      "learning_rate": 2.7537634408602153e-05,
+      "loss": 0.279,
+      "step": 78170
+    },
+    {
+      "epoch": 504.39,
+      "learning_rate": 2.7534050179211475e-05,
+      "loss": 0.2653,
+      "step": 78180
+    },
+    {
+      "epoch": 504.45,
+      "learning_rate": 2.753046594982079e-05,
+      "loss": 0.3007,
+      "step": 78190
+    },
+    {
+      "epoch": 504.51,
+      "learning_rate": 2.7526881720430108e-05,
+      "loss": 0.2558,
+      "step": 78200
+    },
+    {
+      "epoch": 504.58,
+      "learning_rate": 2.752329749103943e-05,
+      "loss": 0.2428,
+      "step": 78210
+    },
+    {
+      "epoch": 504.64,
+      "learning_rate": 2.7519713261648744e-05,
+      "loss": 0.2443,
+      "step": 78220
+    },
+    {
+      "epoch": 504.71,
+      "learning_rate": 2.7516129032258066e-05,
+      "loss": 0.2497,
+      "step": 78230
+    },
+    {
+      "epoch": 504.77,
+      "learning_rate": 2.7512544802867384e-05,
+      "loss": 0.2963,
+      "step": 78240
+    },
+    {
+      "epoch": 504.84,
+      "learning_rate": 2.7508960573476705e-05,
+      "loss": 0.2338,
+      "step": 78250
+    },
+    {
+      "epoch": 504.9,
+      "learning_rate": 2.750537634408602e-05,
+      "loss": 0.2369,
+      "step": 78260
+    },
+    {
+      "epoch": 504.96,
+      "learning_rate": 2.7501792114695342e-05,
+      "loss": 0.1956,
+      "step": 78270
+    },
+    {
+      "epoch": 505.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9935325384140015,
+      "eval_runtime": 0.8716,
+      "eval_samples_per_second": 230.619,
+      "eval_steps_per_second": 8.032,
+      "step": 78275
+    },
+    {
+      "epoch": 505.03,
+      "learning_rate": 2.749820788530466e-05,
+      "loss": 0.3038,
+      "step": 78280
+    },
+    {
+      "epoch": 505.1,
+      "learning_rate": 2.749462365591398e-05,
+      "loss": 0.274,
+      "step": 78290
+    },
+    {
+      "epoch": 505.16,
+      "learning_rate": 2.7491039426523296e-05,
+      "loss": 0.2508,
+      "step": 78300
+    },
+    {
+      "epoch": 505.23,
+      "learning_rate": 2.7487455197132618e-05,
+      "loss": 0.2577,
+      "step": 78310
+    },
+    {
+      "epoch": 505.29,
+      "learning_rate": 2.7483870967741936e-05,
+      "loss": 0.2878,
+      "step": 78320
+    },
+    {
+      "epoch": 505.35,
+      "learning_rate": 2.7480286738351258e-05,
+      "loss": 0.269,
+      "step": 78330
+    },
+    {
+      "epoch": 505.42,
+      "learning_rate": 2.7476702508960573e-05,
+      "loss": 0.277,
+      "step": 78340
+    },
+    {
+      "epoch": 505.48,
+      "learning_rate": 2.7473118279569894e-05,
+      "loss": 0.2399,
+      "step": 78350
+    },
+    {
+      "epoch": 505.55,
+      "learning_rate": 2.7469534050179212e-05,
+      "loss": 0.282,
+      "step": 78360
+    },
+    {
+      "epoch": 505.61,
+      "learning_rate": 2.7465949820788534e-05,
+      "loss": 0.2393,
+      "step": 78370
+    },
+    {
+      "epoch": 505.68,
+      "learning_rate": 2.746236559139785e-05,
+      "loss": 0.3177,
+      "step": 78380
+    },
+    {
+      "epoch": 505.74,
+      "learning_rate": 2.7458781362007167e-05,
+      "loss": 0.2788,
+      "step": 78390
+    },
+    {
+      "epoch": 505.8,
+      "learning_rate": 2.745519713261649e-05,
+      "loss": 0.2684,
+      "step": 78400
+    },
+    {
+      "epoch": 505.87,
+      "learning_rate": 2.745161290322581e-05,
+      "loss": 0.234,
+      "step": 78410
+    },
+    {
+      "epoch": 505.93,
+      "learning_rate": 2.7448028673835125e-05,
+      "loss": 0.2394,
+      "step": 78420
+    },
+    {
+      "epoch": 506.0,
+      "learning_rate": 2.7444444444444443e-05,
+      "loss": 0.2575,
+      "step": 78430
+    },
+    {
+      "epoch": 506.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9898877739906311,
+      "eval_runtime": 0.8572,
+      "eval_samples_per_second": 234.488,
+      "eval_steps_per_second": 8.166,
+      "step": 78430
+    },
+    {
+      "epoch": 506.06,
+      "learning_rate": 2.7440860215053765e-05,
+      "loss": 0.3037,
+      "step": 78440
+    },
+    {
+      "epoch": 506.13,
+      "learning_rate": 2.7437275985663087e-05,
+      "loss": 0.2498,
+      "step": 78450
+    },
+    {
+      "epoch": 506.19,
+      "learning_rate": 2.74336917562724e-05,
+      "loss": 0.2367,
+      "step": 78460
+    },
+    {
+      "epoch": 506.26,
+      "learning_rate": 2.743010752688172e-05,
+      "loss": 0.2492,
+      "step": 78470
+    },
+    {
+      "epoch": 506.32,
+      "learning_rate": 2.742652329749104e-05,
+      "loss": 0.1968,
+      "step": 78480
+    },
+    {
+      "epoch": 506.39,
+      "learning_rate": 2.7422939068100363e-05,
+      "loss": 0.2588,
+      "step": 78490
+    },
+    {
+      "epoch": 506.45,
+      "learning_rate": 2.7419354838709678e-05,
+      "loss": 0.2646,
+      "step": 78500
+    },
+    {
+      "epoch": 506.51,
+      "learning_rate": 2.7415770609318996e-05,
+      "loss": 0.2762,
+      "step": 78510
+    },
+    {
+      "epoch": 506.58,
+      "learning_rate": 2.7412186379928317e-05,
+      "loss": 0.2682,
+      "step": 78520
+    },
+    {
+      "epoch": 506.64,
+      "learning_rate": 2.740860215053764e-05,
+      "loss": 0.2629,
+      "step": 78530
+    },
+    {
+      "epoch": 506.71,
+      "learning_rate": 2.7405017921146954e-05,
+      "loss": 0.281,
+      "step": 78540
+    },
+    {
+      "epoch": 506.77,
+      "learning_rate": 2.7401433691756272e-05,
+      "loss": 0.2559,
+      "step": 78550
+    },
+    {
+      "epoch": 506.84,
+      "learning_rate": 2.7397849462365594e-05,
+      "loss": 0.2201,
+      "step": 78560
+    },
+    {
+      "epoch": 506.9,
+      "learning_rate": 2.7394265232974915e-05,
+      "loss": 0.2412,
+      "step": 78570
+    },
+    {
+      "epoch": 506.96,
+      "learning_rate": 2.739068100358423e-05,
+      "loss": 0.2307,
+      "step": 78580
+    },
+    {
+      "epoch": 507.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.0611976385116577,
+      "eval_runtime": 0.8738,
+      "eval_samples_per_second": 230.028,
+      "eval_steps_per_second": 8.011,
+      "step": 78585
+    },
+    {
+      "epoch": 507.03,
+      "learning_rate": 2.7387096774193548e-05,
+      "loss": 0.2495,
+      "step": 78590
+    },
+    {
+      "epoch": 507.1,
+      "learning_rate": 2.738351254480287e-05,
+      "loss": 0.2381,
+      "step": 78600
+    },
+    {
+      "epoch": 507.16,
+      "learning_rate": 2.737992831541219e-05,
+      "loss": 0.2679,
+      "step": 78610
+    },
+    {
+      "epoch": 507.23,
+      "learning_rate": 2.7376344086021506e-05,
+      "loss": 0.2476,
+      "step": 78620
+    },
+    {
+      "epoch": 507.29,
+      "learning_rate": 2.7372759856630824e-05,
+      "loss": 0.2291,
+      "step": 78630
+    },
+    {
+      "epoch": 507.35,
+      "learning_rate": 2.7369175627240146e-05,
+      "loss": 0.2318,
+      "step": 78640
+    },
+    {
+      "epoch": 507.42,
+      "learning_rate": 2.7365591397849468e-05,
+      "loss": 0.2249,
+      "step": 78650
+    },
+    {
+      "epoch": 507.48,
+      "learning_rate": 2.7362007168458782e-05,
+      "loss": 0.2533,
+      "step": 78660
+    },
+    {
+      "epoch": 507.55,
+      "learning_rate": 2.73584229390681e-05,
+      "loss": 0.2592,
+      "step": 78670
+    },
+    {
+      "epoch": 507.61,
+      "learning_rate": 2.7354838709677422e-05,
+      "loss": 0.2138,
+      "step": 78680
+    },
+    {
+      "epoch": 507.68,
+      "learning_rate": 2.7351254480286744e-05,
+      "loss": 0.33,
+      "step": 78690
+    },
+    {
+      "epoch": 507.74,
+      "learning_rate": 2.734767025089606e-05,
+      "loss": 0.2727,
+      "step": 78700
+    },
+    {
+      "epoch": 507.8,
+      "learning_rate": 2.7344086021505377e-05,
+      "loss": 0.2588,
+      "step": 78710
+    },
+    {
+      "epoch": 507.87,
+      "learning_rate": 2.73405017921147e-05,
+      "loss": 0.2796,
+      "step": 78720
+    },
+    {
+      "epoch": 507.93,
+      "learning_rate": 2.7336917562724017e-05,
+      "loss": 0.2936,
+      "step": 78730
+    },
+    {
+      "epoch": 508.0,
+      "learning_rate": 2.733333333333333e-05,
+      "loss": 0.232,
+      "step": 78740
+    },
+    {
+      "epoch": 508.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.062283992767334,
+      "eval_runtime": 0.8778,
+      "eval_samples_per_second": 228.985,
+      "eval_steps_per_second": 7.975,
+      "step": 78740
+    },
+    {
+      "epoch": 508.06,
+      "learning_rate": 2.7329749103942653e-05,
+      "loss": 0.2757,
+      "step": 78750
+    },
+    {
+      "epoch": 508.13,
+      "learning_rate": 2.7326164874551975e-05,
+      "loss": 0.3302,
+      "step": 78760
+    },
+    {
+      "epoch": 508.19,
+      "learning_rate": 2.7322580645161293e-05,
+      "loss": 0.2801,
+      "step": 78770
+    },
+    {
+      "epoch": 508.26,
+      "learning_rate": 2.7318996415770608e-05,
+      "loss": 0.2822,
+      "step": 78780
+    },
+    {
+      "epoch": 508.32,
+      "learning_rate": 2.731541218637993e-05,
+      "loss": 0.2714,
+      "step": 78790
+    },
+    {
+      "epoch": 508.39,
+      "learning_rate": 2.731182795698925e-05,
+      "loss": 0.2519,
+      "step": 78800
+    },
+    {
+      "epoch": 508.45,
+      "learning_rate": 2.730824372759857e-05,
+      "loss": 0.2554,
+      "step": 78810
+    },
+    {
+      "epoch": 508.51,
+      "learning_rate": 2.7304659498207884e-05,
+      "loss": 0.2587,
+      "step": 78820
+    },
+    {
+      "epoch": 508.58,
+      "learning_rate": 2.7301075268817205e-05,
+      "loss": 0.2353,
+      "step": 78830
+    },
+    {
+      "epoch": 508.64,
+      "learning_rate": 2.7297491039426527e-05,
+      "loss": 0.2597,
+      "step": 78840
+    },
+    {
+      "epoch": 508.71,
+      "learning_rate": 2.7293906810035845e-05,
+      "loss": 0.2112,
+      "step": 78850
+    },
+    {
+      "epoch": 508.77,
+      "learning_rate": 2.729032258064516e-05,
+      "loss": 0.2827,
+      "step": 78860
+    },
+    {
+      "epoch": 508.84,
+      "learning_rate": 2.728673835125448e-05,
+      "loss": 0.2627,
+      "step": 78870
+    },
+    {
+      "epoch": 508.9,
+      "learning_rate": 2.7283154121863803e-05,
+      "loss": 0.2466,
+      "step": 78880
+    },
+    {
+      "epoch": 508.96,
+      "learning_rate": 2.7279569892473118e-05,
+      "loss": 0.2481,
+      "step": 78890
+    },
+    {
+      "epoch": 509.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0151828527450562,
+      "eval_runtime": 0.8576,
+      "eval_samples_per_second": 234.381,
+      "eval_steps_per_second": 8.163,
+      "step": 78895
+    },
+    {
+      "epoch": 509.03,
+      "learning_rate": 2.7275985663082436e-05,
+      "loss": 0.2842,
+      "step": 78900
+    },
+    {
+      "epoch": 509.1,
+      "learning_rate": 2.7272401433691758e-05,
+      "loss": 0.2314,
+      "step": 78910
+    },
+    {
+      "epoch": 509.16,
+      "learning_rate": 2.726881720430108e-05,
+      "loss": 0.3017,
+      "step": 78920
+    },
+    {
+      "epoch": 509.23,
+      "learning_rate": 2.7265232974910394e-05,
+      "loss": 0.2396,
+      "step": 78930
+    },
+    {
+      "epoch": 509.29,
+      "learning_rate": 2.7261648745519712e-05,
+      "loss": 0.2647,
+      "step": 78940
+    },
+    {
+      "epoch": 509.35,
+      "learning_rate": 2.7258064516129034e-05,
+      "loss": 0.2615,
+      "step": 78950
+    },
+    {
+      "epoch": 509.42,
+      "learning_rate": 2.7254480286738356e-05,
+      "loss": 0.2739,
+      "step": 78960
+    },
+    {
+      "epoch": 509.48,
+      "learning_rate": 2.725089605734767e-05,
+      "loss": 0.2433,
+      "step": 78970
+    },
+    {
+      "epoch": 509.55,
+      "learning_rate": 2.724731182795699e-05,
+      "loss": 0.2439,
+      "step": 78980
+    },
+    {
+      "epoch": 509.61,
+      "learning_rate": 2.724372759856631e-05,
+      "loss": 0.2523,
+      "step": 78990
+    },
+    {
+      "epoch": 509.68,
+      "learning_rate": 2.7240143369175632e-05,
+      "loss": 0.2321,
+      "step": 79000
+    },
+    {
+      "epoch": 509.74,
+      "learning_rate": 2.7236559139784947e-05,
+      "loss": 0.2672,
+      "step": 79010
+    },
+    {
+      "epoch": 509.8,
+      "learning_rate": 2.7232974910394265e-05,
+      "loss": 0.2776,
+      "step": 79020
+    },
+    {
+      "epoch": 509.87,
+      "learning_rate": 2.7229390681003586e-05,
+      "loss": 0.2423,
+      "step": 79030
+    },
+    {
+      "epoch": 509.93,
+      "learning_rate": 2.7225806451612905e-05,
+      "loss": 0.2465,
+      "step": 79040
+    },
+    {
+      "epoch": 510.0,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 0.2984,
+      "step": 79050
+    },
+    {
+      "epoch": 510.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.005207896232605,
+      "eval_runtime": 0.872,
+      "eval_samples_per_second": 230.51,
+      "eval_steps_per_second": 8.028,
+      "step": 79050
+    },
+    {
+      "epoch": 510.06,
+      "learning_rate": 2.721863799283154e-05,
+      "loss": 0.2146,
+      "step": 79060
+    },
+    {
+      "epoch": 510.13,
+      "learning_rate": 2.7215053763440863e-05,
+      "loss": 0.1787,
+      "step": 79070
+    },
+    {
+      "epoch": 510.19,
+      "learning_rate": 2.721146953405018e-05,
+      "loss": 0.2052,
+      "step": 79080
+    },
+    {
+      "epoch": 510.26,
+      "learning_rate": 2.7207885304659496e-05,
+      "loss": 0.2352,
+      "step": 79090
+    },
+    {
+      "epoch": 510.32,
+      "learning_rate": 2.7204301075268817e-05,
+      "loss": 0.2525,
+      "step": 79100
+    },
+    {
+      "epoch": 510.39,
+      "learning_rate": 2.720071684587814e-05,
+      "loss": 0.2474,
+      "step": 79110
+    },
+    {
+      "epoch": 510.45,
+      "learning_rate": 2.7197132616487457e-05,
+      "loss": 0.2854,
+      "step": 79120
+    },
+    {
+      "epoch": 510.51,
+      "learning_rate": 2.7193548387096772e-05,
+      "loss": 0.2523,
+      "step": 79130
+    },
+    {
+      "epoch": 510.58,
+      "learning_rate": 2.7189964157706093e-05,
+      "loss": 0.269,
+      "step": 79140
+    },
+    {
+      "epoch": 510.64,
+      "learning_rate": 2.7186379928315415e-05,
+      "loss": 0.2729,
+      "step": 79150
+    },
+    {
+      "epoch": 510.71,
+      "learning_rate": 2.7182795698924733e-05,
+      "loss": 0.2866,
+      "step": 79160
+    },
+    {
+      "epoch": 510.77,
+      "learning_rate": 2.7179211469534048e-05,
+      "loss": 0.2834,
+      "step": 79170
+    },
+    {
+      "epoch": 510.84,
+      "learning_rate": 2.717562724014337e-05,
+      "loss": 0.3046,
+      "step": 79180
+    },
+    {
+      "epoch": 510.9,
+      "learning_rate": 2.717204301075269e-05,
+      "loss": 0.2046,
+      "step": 79190
+    },
+    {
+      "epoch": 510.96,
+      "learning_rate": 2.716845878136201e-05,
+      "loss": 0.2399,
+      "step": 79200
+    },
+    {
+      "epoch": 511.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0396959781646729,
+      "eval_runtime": 0.868,
+      "eval_samples_per_second": 231.571,
+      "eval_steps_per_second": 8.065,
+      "step": 79205
+    },
+    {
+      "epoch": 511.03,
+      "learning_rate": 2.7164874551971324e-05,
+      "loss": 0.2767,
+      "step": 79210
+    },
+    {
+      "epoch": 511.1,
+      "learning_rate": 2.7161290322580646e-05,
+      "loss": 0.2323,
+      "step": 79220
+    },
+    {
+      "epoch": 511.16,
+      "learning_rate": 2.7157706093189967e-05,
+      "loss": 0.2191,
+      "step": 79230
+    },
+    {
+      "epoch": 511.23,
+      "learning_rate": 2.7154121863799286e-05,
+      "loss": 0.2145,
+      "step": 79240
+    },
+    {
+      "epoch": 511.29,
+      "learning_rate": 2.71505376344086e-05,
+      "loss": 0.261,
+      "step": 79250
+    },
+    {
+      "epoch": 511.35,
+      "learning_rate": 2.7146953405017922e-05,
+      "loss": 0.27,
+      "step": 79260
+    },
+    {
+      "epoch": 511.42,
+      "learning_rate": 2.7143369175627244e-05,
+      "loss": 0.3176,
+      "step": 79270
+    },
+    {
+      "epoch": 511.48,
+      "learning_rate": 2.7139784946236562e-05,
+      "loss": 0.2829,
+      "step": 79280
+    },
+    {
+      "epoch": 511.55,
+      "learning_rate": 2.7136200716845877e-05,
+      "loss": 0.2545,
+      "step": 79290
+    },
+    {
+      "epoch": 511.61,
+      "learning_rate": 2.7132616487455198e-05,
+      "loss": 0.2318,
+      "step": 79300
+    },
+    {
+      "epoch": 511.68,
+      "learning_rate": 2.712903225806452e-05,
+      "loss": 0.2996,
+      "step": 79310
+    },
+    {
+      "epoch": 511.74,
+      "learning_rate": 2.7125448028673838e-05,
+      "loss": 0.2117,
+      "step": 79320
+    },
+    {
+      "epoch": 511.8,
+      "learning_rate": 2.7121863799283153e-05,
+      "loss": 0.2703,
+      "step": 79330
+    },
+    {
+      "epoch": 511.87,
+      "learning_rate": 2.7118279569892475e-05,
+      "loss": 0.2693,
+      "step": 79340
+    },
+    {
+      "epoch": 511.93,
+      "learning_rate": 2.7114695340501796e-05,
+      "loss": 0.2597,
+      "step": 79350
+    },
+    {
+      "epoch": 512.0,
+      "learning_rate": 2.7111111111111114e-05,
+      "loss": 0.252,
+      "step": 79360
+    },
+    {
+      "epoch": 512.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.018062949180603,
+      "eval_runtime": 0.8458,
+      "eval_samples_per_second": 237.654,
+      "eval_steps_per_second": 8.277,
+      "step": 79360
+    },
+    {
+      "epoch": 512.06,
+      "learning_rate": 2.710752688172043e-05,
+      "loss": 0.2607,
+      "step": 79370
+    },
+    {
+      "epoch": 512.13,
+      "learning_rate": 2.710394265232975e-05,
+      "loss": 0.2957,
+      "step": 79380
+    },
+    {
+      "epoch": 512.19,
+      "learning_rate": 2.710035842293907e-05,
+      "loss": 0.2857,
+      "step": 79390
+    },
+    {
+      "epoch": 512.26,
+      "learning_rate": 2.709677419354839e-05,
+      "loss": 0.2668,
+      "step": 79400
+    },
+    {
+      "epoch": 512.32,
+      "learning_rate": 2.7093189964157705e-05,
+      "loss": 0.2315,
+      "step": 79410
+    },
+    {
+      "epoch": 512.39,
+      "learning_rate": 2.7089605734767027e-05,
+      "loss": 0.2704,
+      "step": 79420
+    },
+    {
+      "epoch": 512.45,
+      "learning_rate": 2.7086021505376345e-05,
+      "loss": 0.2996,
+      "step": 79430
+    },
+    {
+      "epoch": 512.51,
+      "learning_rate": 2.7082437275985667e-05,
+      "loss": 0.2597,
+      "step": 79440
+    },
+    {
+      "epoch": 512.58,
+      "learning_rate": 2.707885304659498e-05,
+      "loss": 0.2739,
+      "step": 79450
+    },
+    {
+      "epoch": 512.64,
+      "learning_rate": 2.7075268817204303e-05,
+      "loss": 0.2455,
+      "step": 79460
+    },
+    {
+      "epoch": 512.71,
+      "learning_rate": 2.707168458781362e-05,
+      "loss": 0.2249,
+      "step": 79470
+    },
+    {
+      "epoch": 512.77,
+      "learning_rate": 2.7068100358422943e-05,
+      "loss": 0.2413,
+      "step": 79480
+    },
+    {
+      "epoch": 512.84,
+      "learning_rate": 2.7064516129032258e-05,
+      "loss": 0.2481,
+      "step": 79490
+    },
+    {
+      "epoch": 512.9,
+      "learning_rate": 2.706093189964158e-05,
+      "loss": 0.2333,
+      "step": 79500
+    },
+    {
+      "epoch": 512.96,
+      "learning_rate": 2.7057347670250898e-05,
+      "loss": 0.2259,
+      "step": 79510
+    },
+    {
+      "epoch": 513.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9844812154769897,
+      "eval_runtime": 0.8957,
+      "eval_samples_per_second": 224.393,
+      "eval_steps_per_second": 7.815,
+      "step": 79515
+    },
+    {
+      "epoch": 513.03,
+      "learning_rate": 2.7053763440860212e-05,
+      "loss": 0.2742,
+      "step": 79520
+    },
+    {
+      "epoch": 513.1,
+      "learning_rate": 2.7050179211469534e-05,
+      "loss": 0.2469,
+      "step": 79530
+    },
+    {
+      "epoch": 513.16,
+      "learning_rate": 2.7046594982078856e-05,
+      "loss": 0.2676,
+      "step": 79540
+    },
+    {
+      "epoch": 513.23,
+      "learning_rate": 2.7043010752688174e-05,
+      "loss": 0.2417,
+      "step": 79550
+    },
+    {
+      "epoch": 513.29,
+      "learning_rate": 2.703942652329749e-05,
+      "loss": 0.2256,
+      "step": 79560
+    },
+    {
+      "epoch": 513.35,
+      "learning_rate": 2.703584229390681e-05,
+      "loss": 0.2528,
+      "step": 79570
+    },
+    {
+      "epoch": 513.42,
+      "learning_rate": 2.7032258064516132e-05,
+      "loss": 0.2624,
+      "step": 79580
+    },
+    {
+      "epoch": 513.48,
+      "learning_rate": 2.702867383512545e-05,
+      "loss": 0.2627,
+      "step": 79590
+    },
+    {
+      "epoch": 513.55,
+      "learning_rate": 2.7025089605734765e-05,
+      "loss": 0.2331,
+      "step": 79600
+    },
+    {
+      "epoch": 513.61,
+      "learning_rate": 2.7021505376344086e-05,
+      "loss": 0.2517,
+      "step": 79610
+    },
+    {
+      "epoch": 513.68,
+      "learning_rate": 2.7017921146953408e-05,
+      "loss": 0.2563,
+      "step": 79620
+    },
+    {
+      "epoch": 513.74,
+      "learning_rate": 2.7014336917562726e-05,
+      "loss": 0.2338,
+      "step": 79630
+    },
+    {
+      "epoch": 513.8,
+      "learning_rate": 2.701075268817204e-05,
+      "loss": 0.2385,
+      "step": 79640
+    },
+    {
+      "epoch": 513.87,
+      "learning_rate": 2.7007168458781363e-05,
+      "loss": 0.2675,
+      "step": 79650
+    },
+    {
+      "epoch": 513.93,
+      "learning_rate": 2.7003584229390684e-05,
+      "loss": 0.2546,
+      "step": 79660
+    },
+    {
+      "epoch": 514.0,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.2768,
+      "step": 79670
+    },
+    {
+      "epoch": 514.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9594140648841858,
+      "eval_runtime": 0.8646,
+      "eval_samples_per_second": 232.466,
+      "eval_steps_per_second": 8.096,
+      "step": 79670
+    },
+    {
+      "epoch": 514.06,
+      "learning_rate": 2.6996415770609317e-05,
+      "loss": 0.2864,
+      "step": 79680
+    },
+    {
+      "epoch": 514.13,
+      "learning_rate": 2.699283154121864e-05,
+      "loss": 0.2773,
+      "step": 79690
+    },
+    {
+      "epoch": 514.19,
+      "learning_rate": 2.698924731182796e-05,
+      "loss": 0.2562,
+      "step": 79700
+    },
+    {
+      "epoch": 514.26,
+      "learning_rate": 2.698566308243728e-05,
+      "loss": 0.2485,
+      "step": 79710
+    },
+    {
+      "epoch": 514.32,
+      "learning_rate": 2.6982078853046593e-05,
+      "loss": 0.2404,
+      "step": 79720
+    },
+    {
+      "epoch": 514.39,
+      "learning_rate": 2.6978494623655915e-05,
+      "loss": 0.2365,
+      "step": 79730
+    },
+    {
+      "epoch": 514.45,
+      "learning_rate": 2.6974910394265233e-05,
+      "loss": 0.2432,
+      "step": 79740
+    },
+    {
+      "epoch": 514.51,
+      "learning_rate": 2.6971326164874555e-05,
+      "loss": 0.236,
+      "step": 79750
+    },
+    {
+      "epoch": 514.58,
+      "learning_rate": 2.696774193548387e-05,
+      "loss": 0.2168,
+      "step": 79760
+    },
+    {
+      "epoch": 514.64,
+      "learning_rate": 2.696415770609319e-05,
+      "loss": 0.2241,
+      "step": 79770
+    },
+    {
+      "epoch": 514.71,
+      "learning_rate": 2.696057347670251e-05,
+      "loss": 0.2021,
+      "step": 79780
+    },
+    {
+      "epoch": 514.77,
+      "learning_rate": 2.695698924731183e-05,
+      "loss": 0.2422,
+      "step": 79790
+    },
+    {
+      "epoch": 514.84,
+      "learning_rate": 2.6953405017921146e-05,
+      "loss": 0.233,
+      "step": 79800
+    },
+    {
+      "epoch": 514.9,
+      "learning_rate": 2.6949820788530467e-05,
+      "loss": 0.26,
+      "step": 79810
+    },
+    {
+      "epoch": 514.96,
+      "learning_rate": 2.6946236559139786e-05,
+      "loss": 0.2424,
+      "step": 79820
+    },
+    {
+      "epoch": 515.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0056157112121582,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.387,
+      "eval_steps_per_second": 8.058,
+      "step": 79825
+    },
+    {
+      "epoch": 515.03,
+      "learning_rate": 2.6942652329749107e-05,
+      "loss": 0.2663,
+      "step": 79830
+    },
+    {
+      "epoch": 515.1,
+      "learning_rate": 2.6939068100358422e-05,
+      "loss": 0.3058,
+      "step": 79840
+    },
+    {
+      "epoch": 515.16,
+      "learning_rate": 2.6935483870967744e-05,
+      "loss": 0.263,
+      "step": 79850
+    },
+    {
+      "epoch": 515.23,
+      "learning_rate": 2.6931899641577062e-05,
+      "loss": 0.2174,
+      "step": 79860
+    },
+    {
+      "epoch": 515.29,
+      "learning_rate": 2.6928315412186383e-05,
+      "loss": 0.2632,
+      "step": 79870
+    },
+    {
+      "epoch": 515.35,
+      "learning_rate": 2.6924731182795698e-05,
+      "loss": 0.2366,
+      "step": 79880
+    },
+    {
+      "epoch": 515.42,
+      "learning_rate": 2.692114695340502e-05,
+      "loss": 0.3103,
+      "step": 79890
+    },
+    {
+      "epoch": 515.48,
+      "learning_rate": 2.6917562724014338e-05,
+      "loss": 0.2848,
+      "step": 79900
+    },
+    {
+      "epoch": 515.55,
+      "learning_rate": 2.691397849462366e-05,
+      "loss": 0.2856,
+      "step": 79910
+    },
+    {
+      "epoch": 515.61,
+      "learning_rate": 2.6910394265232974e-05,
+      "loss": 0.2647,
+      "step": 79920
+    },
+    {
+      "epoch": 515.68,
+      "learning_rate": 2.6906810035842296e-05,
+      "loss": 0.2482,
+      "step": 79930
+    },
+    {
+      "epoch": 515.74,
+      "learning_rate": 2.6903225806451614e-05,
+      "loss": 0.3015,
+      "step": 79940
+    },
+    {
+      "epoch": 515.8,
+      "learning_rate": 2.6899641577060936e-05,
+      "loss": 0.2747,
+      "step": 79950
+    },
+    {
+      "epoch": 515.87,
+      "learning_rate": 2.689605734767025e-05,
+      "loss": 0.2235,
+      "step": 79960
+    },
+    {
+      "epoch": 515.93,
+      "learning_rate": 2.6892473118279572e-05,
+      "loss": 0.2355,
+      "step": 79970
+    },
+    {
+      "epoch": 516.0,
+      "learning_rate": 2.688888888888889e-05,
+      "loss": 0.326,
+      "step": 79980
+    },
+    {
+      "epoch": 516.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0243390798568726,
+      "eval_runtime": 0.8696,
+      "eval_samples_per_second": 231.144,
+      "eval_steps_per_second": 8.05,
+      "step": 79980
+    },
+    {
+      "epoch": 516.06,
+      "learning_rate": 2.6885304659498212e-05,
+      "loss": 0.2602,
+      "step": 79990
+    },
+    {
+      "epoch": 516.13,
+      "learning_rate": 2.6881720430107527e-05,
+      "loss": 0.2322,
+      "step": 80000
+    },
+    {
+      "epoch": 516.19,
+      "learning_rate": 2.687813620071685e-05,
+      "loss": 0.2953,
+      "step": 80010
+    },
+    {
+      "epoch": 516.26,
+      "learning_rate": 2.6874551971326167e-05,
+      "loss": 0.2393,
+      "step": 80020
+    },
+    {
+      "epoch": 516.32,
+      "learning_rate": 2.6870967741935488e-05,
+      "loss": 0.2385,
+      "step": 80030
+    },
+    {
+      "epoch": 516.39,
+      "learning_rate": 2.6867383512544803e-05,
+      "loss": 0.2814,
+      "step": 80040
+    },
+    {
+      "epoch": 516.45,
+      "learning_rate": 2.6863799283154125e-05,
+      "loss": 0.3105,
+      "step": 80050
+    },
+    {
+      "epoch": 516.51,
+      "learning_rate": 2.6860215053763443e-05,
+      "loss": 0.2607,
+      "step": 80060
+    },
+    {
+      "epoch": 516.58,
+      "learning_rate": 2.6856630824372764e-05,
+      "loss": 0.2368,
+      "step": 80070
+    },
+    {
+      "epoch": 516.64,
+      "learning_rate": 2.685304659498208e-05,
+      "loss": 0.2584,
+      "step": 80080
+    },
+    {
+      "epoch": 516.71,
+      "learning_rate": 2.6849462365591397e-05,
+      "loss": 0.2913,
+      "step": 80090
+    },
+    {
+      "epoch": 516.77,
+      "learning_rate": 2.684587813620072e-05,
+      "loss": 0.2382,
+      "step": 80100
+    },
+    {
+      "epoch": 516.84,
+      "learning_rate": 2.684229390681004e-05,
+      "loss": 0.2465,
+      "step": 80110
+    },
+    {
+      "epoch": 516.9,
+      "learning_rate": 2.6838709677419355e-05,
+      "loss": 0.2578,
+      "step": 80120
+    },
+    {
+      "epoch": 516.96,
+      "learning_rate": 2.6835125448028674e-05,
+      "loss": 0.2489,
+      "step": 80130
+    },
+    {
+      "epoch": 517.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9949617385864258,
+      "eval_runtime": 0.8743,
+      "eval_samples_per_second": 229.903,
+      "eval_steps_per_second": 8.007,
+      "step": 80135
+    },
+    {
+      "epoch": 517.03,
+      "learning_rate": 2.6831541218637995e-05,
+      "loss": 0.2601,
+      "step": 80140
+    },
+    {
+      "epoch": 517.1,
+      "learning_rate": 2.682795698924731e-05,
+      "loss": 0.2965,
+      "step": 80150
+    },
+    {
+      "epoch": 517.16,
+      "learning_rate": 2.682437275985663e-05,
+      "loss": 0.2718,
+      "step": 80160
+    },
+    {
+      "epoch": 517.23,
+      "learning_rate": 2.682078853046595e-05,
+      "loss": 0.2643,
+      "step": 80170
+    },
+    {
+      "epoch": 517.29,
+      "learning_rate": 2.681720430107527e-05,
+      "loss": 0.2774,
+      "step": 80180
+    },
+    {
+      "epoch": 517.35,
+      "learning_rate": 2.6813620071684586e-05,
+      "loss": 0.3198,
+      "step": 80190
+    },
+    {
+      "epoch": 517.42,
+      "learning_rate": 2.6810035842293908e-05,
+      "loss": 0.2666,
+      "step": 80200
+    },
+    {
+      "epoch": 517.48,
+      "learning_rate": 2.6806451612903226e-05,
+      "loss": 0.2637,
+      "step": 80210
+    },
+    {
+      "epoch": 517.55,
+      "learning_rate": 2.6802867383512548e-05,
+      "loss": 0.2474,
+      "step": 80220
+    },
+    {
+      "epoch": 517.61,
+      "learning_rate": 2.6799283154121863e-05,
+      "loss": 0.2911,
+      "step": 80230
+    },
+    {
+      "epoch": 517.68,
+      "learning_rate": 2.6795698924731184e-05,
+      "loss": 0.214,
+      "step": 80240
+    },
+    {
+      "epoch": 517.74,
+      "learning_rate": 2.6792114695340502e-05,
+      "loss": 0.3008,
+      "step": 80250
+    },
+    {
+      "epoch": 517.8,
+      "learning_rate": 2.6788530465949824e-05,
+      "loss": 0.246,
+      "step": 80260
+    },
+    {
+      "epoch": 517.87,
+      "learning_rate": 2.678494623655914e-05,
+      "loss": 0.2719,
+      "step": 80270
+    },
+    {
+      "epoch": 517.93,
+      "learning_rate": 2.678136200716846e-05,
+      "loss": 0.2901,
+      "step": 80280
+    },
+    {
+      "epoch": 518.0,
+      "learning_rate": 2.677777777777778e-05,
+      "loss": 0.2218,
+      "step": 80290
+    },
+    {
+      "epoch": 518.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9956569671630859,
+      "eval_runtime": 0.9007,
+      "eval_samples_per_second": 223.162,
+      "eval_steps_per_second": 7.772,
+      "step": 80290
+    },
+    {
+      "epoch": 518.06,
+      "learning_rate": 2.67741935483871e-05,
+      "loss": 0.2793,
+      "step": 80300
+    },
+    {
+      "epoch": 518.13,
+      "learning_rate": 2.6770609318996415e-05,
+      "loss": 0.3046,
+      "step": 80310
+    },
+    {
+      "epoch": 518.19,
+      "learning_rate": 2.6767025089605737e-05,
+      "loss": 0.2517,
+      "step": 80320
+    },
+    {
+      "epoch": 518.26,
+      "learning_rate": 2.6763440860215055e-05,
+      "loss": 0.2509,
+      "step": 80330
+    },
+    {
+      "epoch": 518.32,
+      "learning_rate": 2.6759856630824376e-05,
+      "loss": 0.2419,
+      "step": 80340
+    },
+    {
+      "epoch": 518.39,
+      "learning_rate": 2.675627240143369e-05,
+      "loss": 0.2781,
+      "step": 80350
+    },
+    {
+      "epoch": 518.45,
+      "learning_rate": 2.6752688172043013e-05,
+      "loss": 0.2251,
+      "step": 80360
+    },
+    {
+      "epoch": 518.51,
+      "learning_rate": 2.674910394265233e-05,
+      "loss": 0.2366,
+      "step": 80370
+    },
+    {
+      "epoch": 518.58,
+      "learning_rate": 2.6745519713261653e-05,
+      "loss": 0.2799,
+      "step": 80380
+    },
+    {
+      "epoch": 518.64,
+      "learning_rate": 2.6741935483870967e-05,
+      "loss": 0.3106,
+      "step": 80390
+    },
+    {
+      "epoch": 518.71,
+      "learning_rate": 2.6738351254480286e-05,
+      "loss": 0.2636,
+      "step": 80400
+    },
+    {
+      "epoch": 518.77,
+      "learning_rate": 2.6734767025089607e-05,
+      "loss": 0.2503,
+      "step": 80410
+    },
+    {
+      "epoch": 518.84,
+      "learning_rate": 2.673118279569893e-05,
+      "loss": 0.2488,
+      "step": 80420
+    },
+    {
+      "epoch": 518.9,
+      "learning_rate": 2.6727598566308244e-05,
+      "loss": 0.2564,
+      "step": 80430
+    },
+    {
+      "epoch": 518.96,
+      "learning_rate": 2.6724014336917562e-05,
+      "loss": 0.2754,
+      "step": 80440
+    },
+    {
+      "epoch": 519.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0420950651168823,
+      "eval_runtime": 0.8723,
+      "eval_samples_per_second": 230.432,
+      "eval_steps_per_second": 8.025,
+      "step": 80445
+    },
+    {
+      "epoch": 519.03,
+      "learning_rate": 2.6720430107526883e-05,
+      "loss": 0.2718,
+      "step": 80450
+    },
+    {
+      "epoch": 519.1,
+      "learning_rate": 2.6716845878136205e-05,
+      "loss": 0.2881,
+      "step": 80460
+    },
+    {
+      "epoch": 519.16,
+      "learning_rate": 2.671326164874552e-05,
+      "loss": 0.2378,
+      "step": 80470
+    },
+    {
+      "epoch": 519.23,
+      "learning_rate": 2.6709677419354838e-05,
+      "loss": 0.2961,
+      "step": 80480
+    },
+    {
+      "epoch": 519.29,
+      "learning_rate": 2.670609318996416e-05,
+      "loss": 0.2474,
+      "step": 80490
+    },
+    {
+      "epoch": 519.35,
+      "learning_rate": 2.670250896057348e-05,
+      "loss": 0.2113,
+      "step": 80500
+    },
+    {
+      "epoch": 519.42,
+      "learning_rate": 2.6698924731182796e-05,
+      "loss": 0.2729,
+      "step": 80510
+    },
+    {
+      "epoch": 519.48,
+      "learning_rate": 2.6695340501792114e-05,
+      "loss": 0.2463,
+      "step": 80520
+    },
+    {
+      "epoch": 519.55,
+      "learning_rate": 2.6691756272401436e-05,
+      "loss": 0.2201,
+      "step": 80530
+    },
+    {
+      "epoch": 519.61,
+      "learning_rate": 2.6688172043010757e-05,
+      "loss": 0.2453,
+      "step": 80540
+    },
+    {
+      "epoch": 519.68,
+      "learning_rate": 2.6684587813620072e-05,
+      "loss": 0.2693,
+      "step": 80550
+    },
+    {
+      "epoch": 519.74,
+      "learning_rate": 2.668100358422939e-05,
+      "loss": 0.2797,
+      "step": 80560
+    },
+    {
+      "epoch": 519.8,
+      "learning_rate": 2.6677419354838712e-05,
+      "loss": 0.2932,
+      "step": 80570
+    },
+    {
+      "epoch": 519.87,
+      "learning_rate": 2.6673835125448034e-05,
+      "loss": 0.2382,
+      "step": 80580
+    },
+    {
+      "epoch": 519.93,
+      "learning_rate": 2.667025089605735e-05,
+      "loss": 0.266,
+      "step": 80590
+    },
+    {
+      "epoch": 520.0,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 0.2168,
+      "step": 80600
+    },
+    {
+      "epoch": 520.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.980720043182373,
+      "eval_runtime": 0.8636,
+      "eval_samples_per_second": 232.757,
+      "eval_steps_per_second": 8.106,
+      "step": 80600
+    },
+    {
+      "epoch": 520.06,
+      "learning_rate": 2.6663082437275988e-05,
+      "loss": 0.2695,
+      "step": 80610
+    },
+    {
+      "epoch": 520.13,
+      "learning_rate": 2.665949820788531e-05,
+      "loss": 0.2754,
+      "step": 80620
+    },
+    {
+      "epoch": 520.19,
+      "learning_rate": 2.6655913978494625e-05,
+      "loss": 0.2532,
+      "step": 80630
+    },
+    {
+      "epoch": 520.26,
+      "learning_rate": 2.6652329749103943e-05,
+      "loss": 0.2103,
+      "step": 80640
+    },
+    {
+      "epoch": 520.32,
+      "learning_rate": 2.6648745519713264e-05,
+      "loss": 0.2354,
+      "step": 80650
+    },
+    {
+      "epoch": 520.39,
+      "learning_rate": 2.6645161290322586e-05,
+      "loss": 0.278,
+      "step": 80660
+    },
+    {
+      "epoch": 520.45,
+      "learning_rate": 2.66415770609319e-05,
+      "loss": 0.2617,
+      "step": 80670
+    },
+    {
+      "epoch": 520.51,
+      "learning_rate": 2.663799283154122e-05,
+      "loss": 0.2865,
+      "step": 80680
+    },
+    {
+      "epoch": 520.58,
+      "learning_rate": 2.663440860215054e-05,
+      "loss": 0.2608,
+      "step": 80690
+    },
+    {
+      "epoch": 520.64,
+      "learning_rate": 2.6630824372759862e-05,
+      "loss": 0.2248,
+      "step": 80700
+    },
+    {
+      "epoch": 520.71,
+      "learning_rate": 2.6627240143369177e-05,
+      "loss": 0.2905,
+      "step": 80710
+    },
+    {
+      "epoch": 520.77,
+      "learning_rate": 2.6623655913978495e-05,
+      "loss": 0.2777,
+      "step": 80720
+    },
+    {
+      "epoch": 520.84,
+      "learning_rate": 2.6620071684587817e-05,
+      "loss": 0.2758,
+      "step": 80730
+    },
+    {
+      "epoch": 520.9,
+      "learning_rate": 2.6616487455197135e-05,
+      "loss": 0.2807,
+      "step": 80740
+    },
+    {
+      "epoch": 520.96,
+      "learning_rate": 2.661290322580645e-05,
+      "loss": 0.2579,
+      "step": 80750
+    },
+    {
+      "epoch": 521.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0221185684204102,
+      "eval_runtime": 0.8715,
+      "eval_samples_per_second": 230.63,
+      "eval_steps_per_second": 8.032,
+      "step": 80755
+    },
+    {
+      "epoch": 521.03,
+      "learning_rate": 2.660931899641577e-05,
+      "loss": 0.2352,
+      "step": 80760
+    },
+    {
+      "epoch": 521.1,
+      "learning_rate": 2.6605734767025093e-05,
+      "loss": 0.2269,
+      "step": 80770
+    },
+    {
+      "epoch": 521.16,
+      "learning_rate": 2.6602150537634408e-05,
+      "loss": 0.2469,
+      "step": 80780
+    },
+    {
+      "epoch": 521.23,
+      "learning_rate": 2.6598566308243726e-05,
+      "loss": 0.2421,
+      "step": 80790
+    },
+    {
+      "epoch": 521.29,
+      "learning_rate": 2.6594982078853048e-05,
+      "loss": 0.2195,
+      "step": 80800
+    },
+    {
+      "epoch": 521.35,
+      "learning_rate": 2.659139784946237e-05,
+      "loss": 0.2813,
+      "step": 80810
+    },
+    {
+      "epoch": 521.42,
+      "learning_rate": 2.6587813620071684e-05,
+      "loss": 0.2783,
+      "step": 80820
+    },
+    {
+      "epoch": 521.48,
+      "learning_rate": 2.6584229390681002e-05,
+      "loss": 0.2215,
+      "step": 80830
+    },
+    {
+      "epoch": 521.55,
+      "learning_rate": 2.6580645161290324e-05,
+      "loss": 0.263,
+      "step": 80840
+    },
+    {
+      "epoch": 521.61,
+      "learning_rate": 2.6577060931899645e-05,
+      "loss": 0.2301,
+      "step": 80850
+    },
+    {
+      "epoch": 521.68,
+      "learning_rate": 2.657347670250896e-05,
+      "loss": 0.229,
+      "step": 80860
+    },
+    {
+      "epoch": 521.74,
+      "learning_rate": 2.656989247311828e-05,
+      "loss": 0.2739,
+      "step": 80870
+    },
+    {
+      "epoch": 521.8,
+      "learning_rate": 2.65663082437276e-05,
+      "loss": 0.2587,
+      "step": 80880
+    },
+    {
+      "epoch": 521.87,
+      "learning_rate": 2.656272401433692e-05,
+      "loss": 0.2495,
+      "step": 80890
+    },
+    {
+      "epoch": 521.93,
+      "learning_rate": 2.6559139784946236e-05,
+      "loss": 0.2858,
+      "step": 80900
+    },
+    {
+      "epoch": 522.0,
+      "learning_rate": 2.6555555555555555e-05,
+      "loss": 0.2419,
+      "step": 80910
+    },
+    {
+      "epoch": 522.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.00385582447052,
+      "eval_runtime": 0.858,
+      "eval_samples_per_second": 234.276,
+      "eval_steps_per_second": 8.159,
+      "step": 80910
+    },
+    {
+      "epoch": 522.06,
+      "learning_rate": 2.6551971326164876e-05,
+      "loss": 0.2959,
+      "step": 80920
+    },
+    {
+      "epoch": 522.13,
+      "learning_rate": 2.6548387096774198e-05,
+      "loss": 0.2563,
+      "step": 80930
+    },
+    {
+      "epoch": 522.19,
+      "learning_rate": 2.6544802867383513e-05,
+      "loss": 0.2483,
+      "step": 80940
+    },
+    {
+      "epoch": 522.26,
+      "learning_rate": 2.654121863799283e-05,
+      "loss": 0.2485,
+      "step": 80950
+    },
+    {
+      "epoch": 522.32,
+      "learning_rate": 2.6537634408602152e-05,
+      "loss": 0.2706,
+      "step": 80960
+    },
+    {
+      "epoch": 522.39,
+      "learning_rate": 2.6534050179211474e-05,
+      "loss": 0.3061,
+      "step": 80970
+    },
+    {
+      "epoch": 522.45,
+      "learning_rate": 2.653046594982079e-05,
+      "loss": 0.211,
+      "step": 80980
+    },
+    {
+      "epoch": 522.51,
+      "learning_rate": 2.6526881720430107e-05,
+      "loss": 0.2808,
+      "step": 80990
+    },
+    {
+      "epoch": 522.58,
+      "learning_rate": 2.652329749103943e-05,
+      "loss": 0.2478,
+      "step": 81000
+    },
+    {
+      "epoch": 522.64,
+      "learning_rate": 2.651971326164875e-05,
+      "loss": 0.2024,
+      "step": 81010
+    },
+    {
+      "epoch": 522.71,
+      "learning_rate": 2.6516129032258065e-05,
+      "loss": 0.2562,
+      "step": 81020
+    },
+    {
+      "epoch": 522.77,
+      "learning_rate": 2.6512544802867383e-05,
+      "loss": 0.3146,
+      "step": 81030
+    },
+    {
+      "epoch": 522.84,
+      "learning_rate": 2.6508960573476705e-05,
+      "loss": 0.2682,
+      "step": 81040
+    },
+    {
+      "epoch": 522.9,
+      "learning_rate": 2.6505376344086023e-05,
+      "loss": 0.3043,
+      "step": 81050
+    },
+    {
+      "epoch": 522.96,
+      "learning_rate": 2.650179211469534e-05,
+      "loss": 0.2834,
+      "step": 81060
+    },
+    {
+      "epoch": 523.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0385233163833618,
+      "eval_runtime": 0.8628,
+      "eval_samples_per_second": 232.975,
+      "eval_steps_per_second": 8.114,
+      "step": 81065
+    },
+    {
+      "epoch": 523.03,
+      "learning_rate": 2.649820788530466e-05,
+      "loss": 0.257,
+      "step": 81070
+    },
+    {
+      "epoch": 523.1,
+      "learning_rate": 2.649462365591398e-05,
+      "loss": 0.2633,
+      "step": 81080
+    },
+    {
+      "epoch": 523.16,
+      "learning_rate": 2.64910394265233e-05,
+      "loss": 0.2463,
+      "step": 81090
+    },
+    {
+      "epoch": 523.23,
+      "learning_rate": 2.6487455197132614e-05,
+      "loss": 0.2578,
+      "step": 81100
+    },
+    {
+      "epoch": 523.29,
+      "learning_rate": 2.6483870967741936e-05,
+      "loss": 0.2804,
+      "step": 81110
+    },
+    {
+      "epoch": 523.35,
+      "learning_rate": 2.6480286738351257e-05,
+      "loss": 0.2691,
+      "step": 81120
+    },
+    {
+      "epoch": 523.42,
+      "learning_rate": 2.6476702508960575e-05,
+      "loss": 0.2337,
+      "step": 81130
+    },
+    {
+      "epoch": 523.48,
+      "learning_rate": 2.647311827956989e-05,
+      "loss": 0.3022,
+      "step": 81140
+    },
+    {
+      "epoch": 523.55,
+      "learning_rate": 2.6469534050179212e-05,
+      "loss": 0.2679,
+      "step": 81150
+    },
+    {
+      "epoch": 523.61,
+      "learning_rate": 2.6465949820788534e-05,
+      "loss": 0.2273,
+      "step": 81160
+    },
+    {
+      "epoch": 523.68,
+      "learning_rate": 2.6462365591397852e-05,
+      "loss": 0.2874,
+      "step": 81170
+    },
+    {
+      "epoch": 523.74,
+      "learning_rate": 2.6458781362007167e-05,
+      "loss": 0.2387,
+      "step": 81180
+    },
+    {
+      "epoch": 523.8,
+      "learning_rate": 2.6455197132616488e-05,
+      "loss": 0.2554,
+      "step": 81190
+    },
+    {
+      "epoch": 523.87,
+      "learning_rate": 2.645161290322581e-05,
+      "loss": 0.2985,
+      "step": 81200
+    },
+    {
+      "epoch": 523.93,
+      "learning_rate": 2.6448028673835128e-05,
+      "loss": 0.2646,
+      "step": 81210
+    },
+    {
+      "epoch": 524.0,
+      "learning_rate": 2.6444444444444443e-05,
+      "loss": 0.2668,
+      "step": 81220
+    },
+    {
+      "epoch": 524.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0139394998550415,
+      "eval_runtime": 0.8666,
+      "eval_samples_per_second": 231.933,
+      "eval_steps_per_second": 8.077,
+      "step": 81220
+    },
+    {
+      "epoch": 524.06,
+      "learning_rate": 2.6440860215053764e-05,
+      "loss": 0.3166,
+      "step": 81230
+    },
+    {
+      "epoch": 524.13,
+      "learning_rate": 2.6437275985663086e-05,
+      "loss": 0.2576,
+      "step": 81240
+    },
+    {
+      "epoch": 524.19,
+      "learning_rate": 2.6433691756272404e-05,
+      "loss": 0.2563,
+      "step": 81250
+    },
+    {
+      "epoch": 524.26,
+      "learning_rate": 2.643010752688172e-05,
+      "loss": 0.1889,
+      "step": 81260
+    },
+    {
+      "epoch": 524.32,
+      "learning_rate": 2.642652329749104e-05,
+      "loss": 0.284,
+      "step": 81270
+    },
+    {
+      "epoch": 524.39,
+      "learning_rate": 2.6422939068100362e-05,
+      "loss": 0.2601,
+      "step": 81280
+    },
+    {
+      "epoch": 524.45,
+      "learning_rate": 2.641935483870968e-05,
+      "loss": 0.2274,
+      "step": 81290
+    },
+    {
+      "epoch": 524.51,
+      "learning_rate": 2.6415770609318995e-05,
+      "loss": 0.2478,
+      "step": 81300
+    },
+    {
+      "epoch": 524.58,
+      "learning_rate": 2.6412186379928317e-05,
+      "loss": 0.2436,
+      "step": 81310
+    },
+    {
+      "epoch": 524.64,
+      "learning_rate": 2.640860215053764e-05,
+      "loss": 0.2059,
+      "step": 81320
+    },
+    {
+      "epoch": 524.71,
+      "learning_rate": 2.6405017921146957e-05,
+      "loss": 0.2282,
+      "step": 81330
+    },
+    {
+      "epoch": 524.77,
+      "learning_rate": 2.640143369175627e-05,
+      "loss": 0.2724,
+      "step": 81340
+    },
+    {
+      "epoch": 524.84,
+      "learning_rate": 2.6397849462365593e-05,
+      "loss": 0.2918,
+      "step": 81350
+    },
+    {
+      "epoch": 524.9,
+      "learning_rate": 2.6394265232974915e-05,
+      "loss": 0.2394,
+      "step": 81360
+    },
+    {
+      "epoch": 524.96,
+      "learning_rate": 2.6390681003584233e-05,
+      "loss": 0.22,
+      "step": 81370
+    },
+    {
+      "epoch": 525.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0265573263168335,
+      "eval_runtime": 0.8726,
+      "eval_samples_per_second": 230.35,
+      "eval_steps_per_second": 8.022,
+      "step": 81375
+    },
+    {
+      "epoch": 525.03,
+      "learning_rate": 2.6387096774193548e-05,
+      "loss": 0.2706,
+      "step": 81380
+    },
+    {
+      "epoch": 525.1,
+      "learning_rate": 2.638351254480287e-05,
+      "loss": 0.2563,
+      "step": 81390
+    },
+    {
+      "epoch": 525.16,
+      "learning_rate": 2.6379928315412187e-05,
+      "loss": 0.233,
+      "step": 81400
+    },
+    {
+      "epoch": 525.23,
+      "learning_rate": 2.637634408602151e-05,
+      "loss": 0.2472,
+      "step": 81410
+    },
+    {
+      "epoch": 525.29,
+      "learning_rate": 2.6372759856630824e-05,
+      "loss": 0.2429,
+      "step": 81420
+    },
+    {
+      "epoch": 525.35,
+      "learning_rate": 2.6369175627240145e-05,
+      "loss": 0.277,
+      "step": 81430
+    },
+    {
+      "epoch": 525.42,
+      "learning_rate": 2.6365591397849464e-05,
+      "loss": 0.2885,
+      "step": 81440
+    },
+    {
+      "epoch": 525.48,
+      "learning_rate": 2.636200716845878e-05,
+      "loss": 0.2185,
+      "step": 81450
+    },
+    {
+      "epoch": 525.55,
+      "learning_rate": 2.63584229390681e-05,
+      "loss": 0.2525,
+      "step": 81460
+    },
+    {
+      "epoch": 525.61,
+      "learning_rate": 2.635483870967742e-05,
+      "loss": 0.2108,
+      "step": 81470
+    },
+    {
+      "epoch": 525.68,
+      "learning_rate": 2.635125448028674e-05,
+      "loss": 0.2451,
+      "step": 81480
+    },
+    {
+      "epoch": 525.74,
+      "learning_rate": 2.6347670250896055e-05,
+      "loss": 0.289,
+      "step": 81490
+    },
+    {
+      "epoch": 525.8,
+      "learning_rate": 2.6344086021505376e-05,
+      "loss": 0.3036,
+      "step": 81500
+    },
+    {
+      "epoch": 525.87,
+      "learning_rate": 2.6340501792114698e-05,
+      "loss": 0.2949,
+      "step": 81510
+    },
+    {
+      "epoch": 525.93,
+      "learning_rate": 2.6336917562724016e-05,
+      "loss": 0.2817,
+      "step": 81520
+    },
+    {
+      "epoch": 526.0,
+      "learning_rate": 2.633333333333333e-05,
+      "loss": 0.2466,
+      "step": 81530
+    },
+    {
+      "epoch": 526.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0535767078399658,
+      "eval_runtime": 0.8881,
+      "eval_samples_per_second": 226.318,
+      "eval_steps_per_second": 7.882,
+      "step": 81530
+    },
+    {
+      "epoch": 526.06,
+      "learning_rate": 2.6329749103942652e-05,
+      "loss": 0.2652,
+      "step": 81540
+    },
+    {
+      "epoch": 526.13,
+      "learning_rate": 2.6326164874551974e-05,
+      "loss": 0.2103,
+      "step": 81550
+    },
+    {
+      "epoch": 526.19,
+      "learning_rate": 2.6322580645161292e-05,
+      "loss": 0.2607,
+      "step": 81560
+    },
+    {
+      "epoch": 526.26,
+      "learning_rate": 2.6318996415770607e-05,
+      "loss": 0.255,
+      "step": 81570
+    },
+    {
+      "epoch": 526.32,
+      "learning_rate": 2.631541218637993e-05,
+      "loss": 0.2923,
+      "step": 81580
+    },
+    {
+      "epoch": 526.39,
+      "learning_rate": 2.631182795698925e-05,
+      "loss": 0.2896,
+      "step": 81590
+    },
+    {
+      "epoch": 526.45,
+      "learning_rate": 2.630824372759857e-05,
+      "loss": 0.2369,
+      "step": 81600
+    },
+    {
+      "epoch": 526.51,
+      "learning_rate": 2.6304659498207883e-05,
+      "loss": 0.2382,
+      "step": 81610
+    },
+    {
+      "epoch": 526.58,
+      "learning_rate": 2.6301075268817205e-05,
+      "loss": 0.2408,
+      "step": 81620
+    },
+    {
+      "epoch": 526.64,
+      "learning_rate": 2.6297491039426526e-05,
+      "loss": 0.2641,
+      "step": 81630
+    },
+    {
+      "epoch": 526.71,
+      "learning_rate": 2.6293906810035845e-05,
+      "loss": 0.2571,
+      "step": 81640
+    },
+    {
+      "epoch": 526.77,
+      "learning_rate": 2.629032258064516e-05,
+      "loss": 0.2798,
+      "step": 81650
+    },
+    {
+      "epoch": 526.84,
+      "learning_rate": 2.628673835125448e-05,
+      "loss": 0.2483,
+      "step": 81660
+    },
+    {
+      "epoch": 526.9,
+      "learning_rate": 2.6283154121863803e-05,
+      "loss": 0.2831,
+      "step": 81670
+    },
+    {
+      "epoch": 526.96,
+      "learning_rate": 2.627956989247312e-05,
+      "loss": 0.2587,
+      "step": 81680
+    },
+    {
+      "epoch": 527.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0940260887145996,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 231.253,
+      "eval_steps_per_second": 8.054,
+      "step": 81685
+    },
+    {
+      "epoch": 527.03,
+      "learning_rate": 2.6275985663082436e-05,
+      "loss": 0.2591,
+      "step": 81690
+    },
+    {
+      "epoch": 527.1,
+      "learning_rate": 2.6272401433691757e-05,
+      "loss": 0.3009,
+      "step": 81700
+    },
+    {
+      "epoch": 527.16,
+      "learning_rate": 2.626881720430108e-05,
+      "loss": 0.2706,
+      "step": 81710
+    },
+    {
+      "epoch": 527.23,
+      "learning_rate": 2.6265232974910397e-05,
+      "loss": 0.2555,
+      "step": 81720
+    },
+    {
+      "epoch": 527.29,
+      "learning_rate": 2.6261648745519712e-05,
+      "loss": 0.236,
+      "step": 81730
+    },
+    {
+      "epoch": 527.35,
+      "learning_rate": 2.6258064516129033e-05,
+      "loss": 0.2267,
+      "step": 81740
+    },
+    {
+      "epoch": 527.42,
+      "learning_rate": 2.625448028673835e-05,
+      "loss": 0.2937,
+      "step": 81750
+    },
+    {
+      "epoch": 527.48,
+      "learning_rate": 2.6250896057347673e-05,
+      "loss": 0.2101,
+      "step": 81760
+    },
+    {
+      "epoch": 527.55,
+      "learning_rate": 2.6247311827956988e-05,
+      "loss": 0.2846,
+      "step": 81770
+    },
+    {
+      "epoch": 527.61,
+      "learning_rate": 2.624372759856631e-05,
+      "loss": 0.2529,
+      "step": 81780
+    },
+    {
+      "epoch": 527.68,
+      "learning_rate": 2.6240143369175628e-05,
+      "loss": 0.2505,
+      "step": 81790
+    },
+    {
+      "epoch": 527.74,
+      "learning_rate": 2.623655913978495e-05,
+      "loss": 0.2859,
+      "step": 81800
+    },
+    {
+      "epoch": 527.8,
+      "learning_rate": 2.6232974910394264e-05,
+      "loss": 0.2707,
+      "step": 81810
+    },
+    {
+      "epoch": 527.87,
+      "learning_rate": 2.6229390681003586e-05,
+      "loss": 0.2701,
+      "step": 81820
+    },
+    {
+      "epoch": 527.93,
+      "learning_rate": 2.6225806451612904e-05,
+      "loss": 0.2817,
+      "step": 81830
+    },
+    {
+      "epoch": 528.0,
+      "learning_rate": 2.6222222222222226e-05,
+      "loss": 0.2582,
+      "step": 81840
+    },
+    {
+      "epoch": 528.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.060752511024475,
+      "eval_runtime": 0.8896,
+      "eval_samples_per_second": 225.937,
+      "eval_steps_per_second": 7.868,
+      "step": 81840
+    },
+    {
+      "epoch": 528.06,
+      "learning_rate": 2.621863799283154e-05,
+      "loss": 0.282,
+      "step": 81850
+    },
+    {
+      "epoch": 528.13,
+      "learning_rate": 2.6215053763440862e-05,
+      "loss": 0.2552,
+      "step": 81860
+    },
+    {
+      "epoch": 528.19,
+      "learning_rate": 2.621146953405018e-05,
+      "loss": 0.2175,
+      "step": 81870
+    },
+    {
+      "epoch": 528.26,
+      "learning_rate": 2.6207885304659502e-05,
+      "loss": 0.2262,
+      "step": 81880
+    },
+    {
+      "epoch": 528.32,
+      "learning_rate": 2.6204301075268817e-05,
+      "loss": 0.2669,
+      "step": 81890
+    },
+    {
+      "epoch": 528.39,
+      "learning_rate": 2.6200716845878138e-05,
+      "loss": 0.2883,
+      "step": 81900
+    },
+    {
+      "epoch": 528.45,
+      "learning_rate": 2.6197132616487456e-05,
+      "loss": 0.2414,
+      "step": 81910
+    },
+    {
+      "epoch": 528.51,
+      "learning_rate": 2.6193548387096778e-05,
+      "loss": 0.2451,
+      "step": 81920
+    },
+    {
+      "epoch": 528.58,
+      "learning_rate": 2.6189964157706093e-05,
+      "loss": 0.242,
+      "step": 81930
+    },
+    {
+      "epoch": 528.64,
+      "learning_rate": 2.6186379928315414e-05,
+      "loss": 0.2709,
+      "step": 81940
+    },
+    {
+      "epoch": 528.71,
+      "learning_rate": 2.6182795698924733e-05,
+      "loss": 0.2306,
+      "step": 81950
+    },
+    {
+      "epoch": 528.77,
+      "learning_rate": 2.6179211469534054e-05,
+      "loss": 0.2594,
+      "step": 81960
+    },
+    {
+      "epoch": 528.84,
+      "learning_rate": 2.617562724014337e-05,
+      "loss": 0.2023,
+      "step": 81970
+    },
+    {
+      "epoch": 528.9,
+      "learning_rate": 2.617204301075269e-05,
+      "loss": 0.2815,
+      "step": 81980
+    },
+    {
+      "epoch": 528.96,
+      "learning_rate": 2.616845878136201e-05,
+      "loss": 0.2195,
+      "step": 81990
+    },
+    {
+      "epoch": 529.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.096023440361023,
+      "eval_runtime": 0.8996,
+      "eval_samples_per_second": 223.423,
+      "eval_steps_per_second": 7.781,
+      "step": 81995
+    },
+    {
+      "epoch": 529.03,
+      "learning_rate": 2.616487455197133e-05,
+      "loss": 0.2179,
+      "step": 82000
+    },
+    {
+      "epoch": 529.1,
+      "learning_rate": 2.6161290322580645e-05,
+      "loss": 0.3006,
+      "step": 82010
+    },
+    {
+      "epoch": 529.16,
+      "learning_rate": 2.6157706093189967e-05,
+      "loss": 0.2635,
+      "step": 82020
+    },
+    {
+      "epoch": 529.23,
+      "learning_rate": 2.6154121863799285e-05,
+      "loss": 0.2884,
+      "step": 82030
+    },
+    {
+      "epoch": 529.29,
+      "learning_rate": 2.6150537634408607e-05,
+      "loss": 0.261,
+      "step": 82040
+    },
+    {
+      "epoch": 529.35,
+      "learning_rate": 2.614695340501792e-05,
+      "loss": 0.2528,
+      "step": 82050
+    },
+    {
+      "epoch": 529.42,
+      "learning_rate": 2.6143369175627243e-05,
+      "loss": 0.2816,
+      "step": 82060
+    },
+    {
+      "epoch": 529.48,
+      "learning_rate": 2.613978494623656e-05,
+      "loss": 0.2524,
+      "step": 82070
+    },
+    {
+      "epoch": 529.55,
+      "learning_rate": 2.6136200716845876e-05,
+      "loss": 0.2322,
+      "step": 82080
+    },
+    {
+      "epoch": 529.61,
+      "learning_rate": 2.6132616487455198e-05,
+      "loss": 0.2386,
+      "step": 82090
+    },
+    {
+      "epoch": 529.68,
+      "learning_rate": 2.6129032258064516e-05,
+      "loss": 0.2204,
+      "step": 82100
+    },
+    {
+      "epoch": 529.74,
+      "learning_rate": 2.6125448028673838e-05,
+      "loss": 0.2233,
+      "step": 82110
+    },
+    {
+      "epoch": 529.8,
+      "learning_rate": 2.6121863799283152e-05,
+      "loss": 0.2567,
+      "step": 82120
+    },
+    {
+      "epoch": 529.87,
+      "learning_rate": 2.6118279569892474e-05,
+      "loss": 0.2655,
+      "step": 82130
+    },
+    {
+      "epoch": 529.93,
+      "learning_rate": 2.6114695340501792e-05,
+      "loss": 0.2643,
+      "step": 82140
+    },
+    {
+      "epoch": 530.0,
+      "learning_rate": 2.6111111111111114e-05,
+      "loss": 0.2507,
+      "step": 82150
+    },
+    {
+      "epoch": 530.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0433999300003052,
+      "eval_runtime": 0.8634,
+      "eval_samples_per_second": 232.811,
+      "eval_steps_per_second": 8.108,
+      "step": 82150
+    },
+    {
+      "epoch": 530.06,
+      "learning_rate": 2.610752688172043e-05,
+      "loss": 0.322,
+      "step": 82160
+    },
+    {
+      "epoch": 530.13,
+      "learning_rate": 2.610394265232975e-05,
+      "loss": 0.2418,
+      "step": 82170
+    },
+    {
+      "epoch": 530.19,
+      "learning_rate": 2.610035842293907e-05,
+      "loss": 0.2782,
+      "step": 82180
+    },
+    {
+      "epoch": 530.26,
+      "learning_rate": 2.609677419354839e-05,
+      "loss": 0.2453,
+      "step": 82190
+    },
+    {
+      "epoch": 530.32,
+      "learning_rate": 2.6093189964157705e-05,
+      "loss": 0.2373,
+      "step": 82200
+    },
+    {
+      "epoch": 530.39,
+      "learning_rate": 2.6089605734767026e-05,
+      "loss": 0.3008,
+      "step": 82210
+    },
+    {
+      "epoch": 530.45,
+      "learning_rate": 2.6086021505376345e-05,
+      "loss": 0.2265,
+      "step": 82220
+    },
+    {
+      "epoch": 530.51,
+      "learning_rate": 2.6082437275985666e-05,
+      "loss": 0.3029,
+      "step": 82230
+    },
+    {
+      "epoch": 530.58,
+      "learning_rate": 2.607885304659498e-05,
+      "loss": 0.2716,
+      "step": 82240
+    },
+    {
+      "epoch": 530.64,
+      "learning_rate": 2.6075268817204303e-05,
+      "loss": 0.2567,
+      "step": 82250
+    },
+    {
+      "epoch": 530.71,
+      "learning_rate": 2.607168458781362e-05,
+      "loss": 0.2003,
+      "step": 82260
+    },
+    {
+      "epoch": 530.77,
+      "learning_rate": 2.6068100358422942e-05,
+      "loss": 0.2387,
+      "step": 82270
+    },
+    {
+      "epoch": 530.84,
+      "learning_rate": 2.6064516129032257e-05,
+      "loss": 0.2921,
+      "step": 82280
+    },
+    {
+      "epoch": 530.9,
+      "learning_rate": 2.606093189964158e-05,
+      "loss": 0.2508,
+      "step": 82290
+    },
+    {
+      "epoch": 530.96,
+      "learning_rate": 2.6057347670250897e-05,
+      "loss": 0.3006,
+      "step": 82300
+    },
+    {
+      "epoch": 531.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.1084973812103271,
+      "eval_runtime": 0.8905,
+      "eval_samples_per_second": 225.725,
+      "eval_steps_per_second": 7.861,
+      "step": 82305
+    },
+    {
+      "epoch": 531.03,
+      "learning_rate": 2.605376344086022e-05,
+      "loss": 0.2194,
+      "step": 82310
+    },
+    {
+      "epoch": 531.1,
+      "learning_rate": 2.6050179211469533e-05,
+      "loss": 0.2613,
+      "step": 82320
+    },
+    {
+      "epoch": 531.16,
+      "learning_rate": 2.6046594982078855e-05,
+      "loss": 0.2358,
+      "step": 82330
+    },
+    {
+      "epoch": 531.23,
+      "learning_rate": 2.6043010752688173e-05,
+      "loss": 0.2675,
+      "step": 82340
+    },
+    {
+      "epoch": 531.29,
+      "learning_rate": 2.6039426523297495e-05,
+      "loss": 0.2103,
+      "step": 82350
+    },
+    {
+      "epoch": 531.35,
+      "learning_rate": 2.603584229390681e-05,
+      "loss": 0.2746,
+      "step": 82360
+    },
+    {
+      "epoch": 531.42,
+      "learning_rate": 2.603225806451613e-05,
+      "loss": 0.2341,
+      "step": 82370
+    },
+    {
+      "epoch": 531.48,
+      "learning_rate": 2.602867383512545e-05,
+      "loss": 0.2868,
+      "step": 82380
+    },
+    {
+      "epoch": 531.55,
+      "learning_rate": 2.602508960573477e-05,
+      "loss": 0.2198,
+      "step": 82390
+    },
+    {
+      "epoch": 531.61,
+      "learning_rate": 2.6021505376344086e-05,
+      "loss": 0.2553,
+      "step": 82400
+    },
+    {
+      "epoch": 531.68,
+      "learning_rate": 2.6017921146953404e-05,
+      "loss": 0.2772,
+      "step": 82410
+    },
+    {
+      "epoch": 531.74,
+      "learning_rate": 2.6014336917562726e-05,
+      "loss": 0.2381,
+      "step": 82420
+    },
+    {
+      "epoch": 531.8,
+      "learning_rate": 2.6010752688172047e-05,
+      "loss": 0.2838,
+      "step": 82430
+    },
+    {
+      "epoch": 531.87,
+      "learning_rate": 2.6007168458781362e-05,
+      "loss": 0.2239,
+      "step": 82440
+    },
+    {
+      "epoch": 531.93,
+      "learning_rate": 2.600358422939068e-05,
+      "loss": 0.221,
+      "step": 82450
+    },
+    {
+      "epoch": 532.0,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.2151,
+      "step": 82460
+    },
+    {
+      "epoch": 532.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0577452182769775,
+      "eval_runtime": 0.8706,
+      "eval_samples_per_second": 230.885,
+      "eval_steps_per_second": 8.041,
+      "step": 82460
+    },
+    {
+      "epoch": 532.06,
+      "learning_rate": 2.5996415770609323e-05,
+      "loss": 0.2751,
+      "step": 82470
+    },
+    {
+      "epoch": 532.13,
+      "learning_rate": 2.5992831541218638e-05,
+      "loss": 0.2722,
+      "step": 82480
+    },
+    {
+      "epoch": 532.19,
+      "learning_rate": 2.5989247311827956e-05,
+      "loss": 0.2195,
+      "step": 82490
+    },
+    {
+      "epoch": 532.26,
+      "learning_rate": 2.5985663082437278e-05,
+      "loss": 0.2364,
+      "step": 82500
+    },
+    {
+      "epoch": 532.32,
+      "learning_rate": 2.59820788530466e-05,
+      "loss": 0.2396,
+      "step": 82510
+    },
+    {
+      "epoch": 532.39,
+      "learning_rate": 2.5978494623655914e-05,
+      "loss": 0.2414,
+      "step": 82520
+    },
+    {
+      "epoch": 532.45,
+      "learning_rate": 2.5974910394265233e-05,
+      "loss": 0.3015,
+      "step": 82530
+    },
+    {
+      "epoch": 532.51,
+      "learning_rate": 2.5971326164874554e-05,
+      "loss": 0.2537,
+      "step": 82540
+    },
+    {
+      "epoch": 532.58,
+      "learning_rate": 2.5967741935483876e-05,
+      "loss": 0.3121,
+      "step": 82550
+    },
+    {
+      "epoch": 532.64,
+      "learning_rate": 2.596415770609319e-05,
+      "loss": 0.2394,
+      "step": 82560
+    },
+    {
+      "epoch": 532.71,
+      "learning_rate": 2.596057347670251e-05,
+      "loss": 0.2365,
+      "step": 82570
+    },
+    {
+      "epoch": 532.77,
+      "learning_rate": 2.595698924731183e-05,
+      "loss": 0.2511,
+      "step": 82580
+    },
+    {
+      "epoch": 532.84,
+      "learning_rate": 2.5953405017921152e-05,
+      "loss": 0.2402,
+      "step": 82590
+    },
+    {
+      "epoch": 532.9,
+      "learning_rate": 2.5949820788530467e-05,
+      "loss": 0.2439,
+      "step": 82600
+    },
+    {
+      "epoch": 532.96,
+      "learning_rate": 2.5946236559139785e-05,
+      "loss": 0.2406,
+      "step": 82610
+    },
+    {
+      "epoch": 533.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.1022897958755493,
+      "eval_runtime": 0.8878,
+      "eval_samples_per_second": 226.402,
+      "eval_steps_per_second": 7.885,
+      "step": 82615
+    },
+    {
+      "epoch": 533.03,
+      "learning_rate": 2.5942652329749107e-05,
+      "loss": 0.2585,
+      "step": 82620
+    },
+    {
+      "epoch": 533.1,
+      "learning_rate": 2.5939068100358428e-05,
+      "loss": 0.2713,
+      "step": 82630
+    },
+    {
+      "epoch": 533.16,
+      "learning_rate": 2.5935483870967743e-05,
+      "loss": 0.2467,
+      "step": 82640
+    },
+    {
+      "epoch": 533.23,
+      "learning_rate": 2.593189964157706e-05,
+      "loss": 0.2843,
+      "step": 82650
+    },
+    {
+      "epoch": 533.29,
+      "learning_rate": 2.5928315412186383e-05,
+      "loss": 0.2294,
+      "step": 82660
+    },
+    {
+      "epoch": 533.35,
+      "learning_rate": 2.5924731182795704e-05,
+      "loss": 0.2347,
+      "step": 82670
+    },
+    {
+      "epoch": 533.42,
+      "learning_rate": 2.592114695340502e-05,
+      "loss": 0.3314,
+      "step": 82680
+    },
+    {
+      "epoch": 533.48,
+      "learning_rate": 2.5917562724014337e-05,
+      "loss": 0.2045,
+      "step": 82690
+    },
+    {
+      "epoch": 533.55,
+      "learning_rate": 2.591397849462366e-05,
+      "loss": 0.2696,
+      "step": 82700
+    },
+    {
+      "epoch": 533.61,
+      "learning_rate": 2.5910394265232974e-05,
+      "loss": 0.2643,
+      "step": 82710
+    },
+    {
+      "epoch": 533.68,
+      "learning_rate": 2.5906810035842295e-05,
+      "loss": 0.2769,
+      "step": 82720
+    },
+    {
+      "epoch": 533.74,
+      "learning_rate": 2.5903225806451614e-05,
+      "loss": 0.3151,
+      "step": 82730
+    },
+    {
+      "epoch": 533.8,
+      "learning_rate": 2.5899641577060935e-05,
+      "loss": 0.2362,
+      "step": 82740
+    },
+    {
+      "epoch": 533.87,
+      "learning_rate": 2.589605734767025e-05,
+      "loss": 0.2388,
+      "step": 82750
+    },
+    {
+      "epoch": 533.93,
+      "learning_rate": 2.5892473118279568e-05,
+      "loss": 0.222,
+      "step": 82760
+    },
+    {
+      "epoch": 534.0,
+      "learning_rate": 2.588888888888889e-05,
+      "loss": 0.2377,
+      "step": 82770
+    },
+    {
+      "epoch": 534.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0356496572494507,
+      "eval_runtime": 0.8943,
+      "eval_samples_per_second": 224.747,
+      "eval_steps_per_second": 7.827,
+      "step": 82770
+    },
+    {
+      "epoch": 534.06,
+      "learning_rate": 2.588530465949821e-05,
+      "loss": 0.2828,
+      "step": 82780
+    },
+    {
+      "epoch": 534.13,
+      "learning_rate": 2.5881720430107526e-05,
+      "loss": 0.2373,
+      "step": 82790
+    },
+    {
+      "epoch": 534.19,
+      "learning_rate": 2.5878136200716844e-05,
+      "loss": 0.3142,
+      "step": 82800
+    },
+    {
+      "epoch": 534.26,
+      "learning_rate": 2.5874551971326166e-05,
+      "loss": 0.2612,
+      "step": 82810
+    },
+    {
+      "epoch": 534.32,
+      "learning_rate": 2.5870967741935488e-05,
+      "loss": 0.2766,
+      "step": 82820
+    },
+    {
+      "epoch": 534.39,
+      "learning_rate": 2.5867383512544802e-05,
+      "loss": 0.2492,
+      "step": 82830
+    },
+    {
+      "epoch": 534.45,
+      "learning_rate": 2.586379928315412e-05,
+      "loss": 0.2607,
+      "step": 82840
+    },
+    {
+      "epoch": 534.51,
+      "learning_rate": 2.5860215053763442e-05,
+      "loss": 0.273,
+      "step": 82850
+    },
+    {
+      "epoch": 534.58,
+      "learning_rate": 2.5856630824372764e-05,
+      "loss": 0.2596,
+      "step": 82860
+    },
+    {
+      "epoch": 534.64,
+      "learning_rate": 2.585304659498208e-05,
+      "loss": 0.2562,
+      "step": 82870
+    },
+    {
+      "epoch": 534.71,
+      "learning_rate": 2.5849462365591397e-05,
+      "loss": 0.2731,
+      "step": 82880
+    },
+    {
+      "epoch": 534.77,
+      "learning_rate": 2.584587813620072e-05,
+      "loss": 0.2335,
+      "step": 82890
+    },
+    {
+      "epoch": 534.84,
+      "learning_rate": 2.584229390681004e-05,
+      "loss": 0.2142,
+      "step": 82900
+    },
+    {
+      "epoch": 534.9,
+      "learning_rate": 2.5838709677419355e-05,
+      "loss": 0.2267,
+      "step": 82910
+    },
+    {
+      "epoch": 534.96,
+      "learning_rate": 2.5835125448028673e-05,
+      "loss": 0.2954,
+      "step": 82920
+    },
+    {
+      "epoch": 535.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0586652755737305,
+      "eval_runtime": 0.8685,
+      "eval_samples_per_second": 231.424,
+      "eval_steps_per_second": 8.06,
+      "step": 82925
+    },
+    {
+      "epoch": 535.03,
+      "learning_rate": 2.5831541218637995e-05,
+      "loss": 0.3037,
+      "step": 82930
+    },
+    {
+      "epoch": 535.1,
+      "learning_rate": 2.5827956989247316e-05,
+      "loss": 0.2643,
+      "step": 82940
+    },
+    {
+      "epoch": 535.16,
+      "learning_rate": 2.582437275985663e-05,
+      "loss": 0.2626,
+      "step": 82950
+    },
+    {
+      "epoch": 535.23,
+      "learning_rate": 2.582078853046595e-05,
+      "loss": 0.3102,
+      "step": 82960
+    },
+    {
+      "epoch": 535.29,
+      "learning_rate": 2.581720430107527e-05,
+      "loss": 0.2857,
+      "step": 82970
+    },
+    {
+      "epoch": 535.35,
+      "learning_rate": 2.5813620071684592e-05,
+      "loss": 0.2919,
+      "step": 82980
+    },
+    {
+      "epoch": 535.42,
+      "learning_rate": 2.5810035842293907e-05,
+      "loss": 0.2851,
+      "step": 82990
+    },
+    {
+      "epoch": 535.48,
+      "learning_rate": 2.5806451612903226e-05,
+      "loss": 0.2785,
+      "step": 83000
+    },
+    {
+      "epoch": 535.55,
+      "learning_rate": 2.5802867383512547e-05,
+      "loss": 0.2254,
+      "step": 83010
+    },
+    {
+      "epoch": 535.61,
+      "learning_rate": 2.579928315412187e-05,
+      "loss": 0.2573,
+      "step": 83020
+    },
+    {
+      "epoch": 535.68,
+      "learning_rate": 2.5795698924731184e-05,
+      "loss": 0.2869,
+      "step": 83030
+    },
+    {
+      "epoch": 535.74,
+      "learning_rate": 2.5792114695340502e-05,
+      "loss": 0.3013,
+      "step": 83040
+    },
+    {
+      "epoch": 535.8,
+      "learning_rate": 2.5788530465949823e-05,
+      "loss": 0.2511,
+      "step": 83050
+    },
+    {
+      "epoch": 535.87,
+      "learning_rate": 2.578494623655914e-05,
+      "loss": 0.2224,
+      "step": 83060
+    },
+    {
+      "epoch": 535.93,
+      "learning_rate": 2.578136200716846e-05,
+      "loss": 0.2409,
+      "step": 83070
+    },
+    {
+      "epoch": 536.0,
+      "learning_rate": 2.5777777777777778e-05,
+      "loss": 0.2298,
+      "step": 83080
+    },
+    {
+      "epoch": 536.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0553405284881592,
+      "eval_runtime": 0.8878,
+      "eval_samples_per_second": 226.405,
+      "eval_steps_per_second": 7.885,
+      "step": 83080
+    },
+    {
+      "epoch": 536.06,
+      "learning_rate": 2.57741935483871e-05,
+      "loss": 0.2767,
+      "step": 83090
+    },
+    {
+      "epoch": 536.13,
+      "learning_rate": 2.5770609318996418e-05,
+      "loss": 0.272,
+      "step": 83100
+    },
+    {
+      "epoch": 536.19,
+      "learning_rate": 2.5767025089605733e-05,
+      "loss": 0.2644,
+      "step": 83110
+    },
+    {
+      "epoch": 536.26,
+      "learning_rate": 2.5763440860215054e-05,
+      "loss": 0.2854,
+      "step": 83120
+    },
+    {
+      "epoch": 536.32,
+      "learning_rate": 2.5759856630824376e-05,
+      "loss": 0.2392,
+      "step": 83130
+    },
+    {
+      "epoch": 536.39,
+      "learning_rate": 2.5756272401433694e-05,
+      "loss": 0.2931,
+      "step": 83140
+    },
+    {
+      "epoch": 536.45,
+      "learning_rate": 2.575268817204301e-05,
+      "loss": 0.2521,
+      "step": 83150
+    },
+    {
+      "epoch": 536.51,
+      "learning_rate": 2.574910394265233e-05,
+      "loss": 0.2509,
+      "step": 83160
+    },
+    {
+      "epoch": 536.58,
+      "learning_rate": 2.5745519713261652e-05,
+      "loss": 0.2804,
+      "step": 83170
+    },
+    {
+      "epoch": 536.64,
+      "learning_rate": 2.574193548387097e-05,
+      "loss": 0.2231,
+      "step": 83180
+    },
+    {
+      "epoch": 536.71,
+      "learning_rate": 2.5738351254480285e-05,
+      "loss": 0.2297,
+      "step": 83190
+    },
+    {
+      "epoch": 536.77,
+      "learning_rate": 2.5734767025089607e-05,
+      "loss": 0.2509,
+      "step": 83200
+    },
+    {
+      "epoch": 536.84,
+      "learning_rate": 2.5731182795698928e-05,
+      "loss": 0.2213,
+      "step": 83210
+    },
+    {
+      "epoch": 536.9,
+      "learning_rate": 2.5727598566308246e-05,
+      "loss": 0.2686,
+      "step": 83220
+    },
+    {
+      "epoch": 536.96,
+      "learning_rate": 2.572401433691756e-05,
+      "loss": 0.2597,
+      "step": 83230
+    },
+    {
+      "epoch": 537.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 1.0571321249008179,
+      "eval_runtime": 0.8771,
+      "eval_samples_per_second": 229.157,
+      "eval_steps_per_second": 7.981,
+      "step": 83235
+    },
+    {
+      "epoch": 537.03,
+      "learning_rate": 2.5720430107526883e-05,
+      "loss": 0.2749,
+      "step": 83240
+    },
+    {
+      "epoch": 537.1,
+      "learning_rate": 2.5716845878136204e-05,
+      "loss": 0.2852,
+      "step": 83250
+    },
+    {
+      "epoch": 537.16,
+      "learning_rate": 2.5713261648745523e-05,
+      "loss": 0.2251,
+      "step": 83260
+    },
+    {
+      "epoch": 537.23,
+      "learning_rate": 2.5709677419354837e-05,
+      "loss": 0.2621,
+      "step": 83270
+    },
+    {
+      "epoch": 537.29,
+      "learning_rate": 2.570609318996416e-05,
+      "loss": 0.2371,
+      "step": 83280
+    },
+    {
+      "epoch": 537.35,
+      "learning_rate": 2.570250896057348e-05,
+      "loss": 0.2943,
+      "step": 83290
+    },
+    {
+      "epoch": 537.42,
+      "learning_rate": 2.56989247311828e-05,
+      "loss": 0.2881,
+      "step": 83300
+    },
+    {
+      "epoch": 537.48,
+      "learning_rate": 2.5695340501792114e-05,
+      "loss": 0.2828,
+      "step": 83310
+    },
+    {
+      "epoch": 537.55,
+      "learning_rate": 2.5691756272401435e-05,
+      "loss": 0.2103,
+      "step": 83320
+    },
+    {
+      "epoch": 537.61,
+      "learning_rate": 2.5688172043010757e-05,
+      "loss": 0.2323,
+      "step": 83330
+    },
+    {
+      "epoch": 537.68,
+      "learning_rate": 2.568458781362007e-05,
+      "loss": 0.2595,
+      "step": 83340
+    },
+    {
+      "epoch": 537.74,
+      "learning_rate": 2.568100358422939e-05,
+      "loss": 0.2736,
+      "step": 83350
+    },
+    {
+      "epoch": 537.8,
+      "learning_rate": 2.567741935483871e-05,
+      "loss": 0.1794,
+      "step": 83360
+    },
+    {
+      "epoch": 537.87,
+      "learning_rate": 2.5673835125448033e-05,
+      "loss": 0.2214,
+      "step": 83370
+    },
+    {
+      "epoch": 537.93,
+      "learning_rate": 2.5670250896057348e-05,
+      "loss": 0.2551,
+      "step": 83380
+    },
+    {
+      "epoch": 538.0,
+      "learning_rate": 2.5666666666666666e-05,
+      "loss": 0.2554,
+      "step": 83390
+    },
+    {
+      "epoch": 538.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9905814528465271,
+      "eval_runtime": 0.8939,
+      "eval_samples_per_second": 224.865,
+      "eval_steps_per_second": 7.831,
+      "step": 83390
+    },
+    {
+      "epoch": 538.06,
+      "learning_rate": 2.5663082437275988e-05,
+      "loss": 0.3059,
+      "step": 83400
+    },
+    {
+      "epoch": 538.13,
+      "learning_rate": 2.5659498207885306e-05,
+      "loss": 0.2768,
+      "step": 83410
+    },
+    {
+      "epoch": 538.19,
+      "learning_rate": 2.565591397849462e-05,
+      "loss": 0.2536,
+      "step": 83420
+    },
+    {
+      "epoch": 538.26,
+      "learning_rate": 2.5652329749103942e-05,
+      "loss": 0.2668,
+      "step": 83430
+    },
+    {
+      "epoch": 538.32,
+      "learning_rate": 2.5648745519713264e-05,
+      "loss": 0.2611,
+      "step": 83440
+    },
+    {
+      "epoch": 538.39,
+      "learning_rate": 2.5645161290322582e-05,
+      "loss": 0.271,
+      "step": 83450
+    },
+    {
+      "epoch": 538.45,
+      "learning_rate": 2.5641577060931897e-05,
+      "loss": 0.2106,
+      "step": 83460
+    },
+    {
+      "epoch": 538.51,
+      "learning_rate": 2.563799283154122e-05,
+      "loss": 0.2295,
+      "step": 83470
+    },
+    {
+      "epoch": 538.58,
+      "learning_rate": 2.563440860215054e-05,
+      "loss": 0.2363,
+      "step": 83480
+    },
+    {
+      "epoch": 538.64,
+      "learning_rate": 2.5630824372759858e-05,
+      "loss": 0.2708,
+      "step": 83490
+    },
+    {
+      "epoch": 538.71,
+      "learning_rate": 2.5627240143369173e-05,
+      "loss": 0.2463,
+      "step": 83500
+    },
+    {
+      "epoch": 538.77,
+      "learning_rate": 2.5623655913978495e-05,
+      "loss": 0.2879,
+      "step": 83510
+    },
+    {
+      "epoch": 538.84,
+      "learning_rate": 2.5620071684587816e-05,
+      "loss": 0.1983,
+      "step": 83520
+    },
+    {
+      "epoch": 538.9,
+      "learning_rate": 2.5616487455197134e-05,
+      "loss": 0.2531,
+      "step": 83530
+    },
+    {
+      "epoch": 538.96,
+      "learning_rate": 2.561290322580645e-05,
+      "loss": 0.2752,
+      "step": 83540
+    },
+    {
+      "epoch": 539.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0215286016464233,
+      "eval_runtime": 0.8815,
+      "eval_samples_per_second": 228.031,
+      "eval_steps_per_second": 7.941,
+      "step": 83545
+    },
+    {
+      "epoch": 539.03,
+      "learning_rate": 2.560931899641577e-05,
+      "loss": 0.2482,
+      "step": 83550
+    },
+    {
+      "epoch": 539.1,
+      "learning_rate": 2.5605734767025092e-05,
+      "loss": 0.2213,
+      "step": 83560
+    },
+    {
+      "epoch": 539.16,
+      "learning_rate": 2.560215053763441e-05,
+      "loss": 0.2686,
+      "step": 83570
+    },
+    {
+      "epoch": 539.23,
+      "learning_rate": 2.5598566308243725e-05,
+      "loss": 0.3012,
+      "step": 83580
+    },
+    {
+      "epoch": 539.29,
+      "learning_rate": 2.5594982078853047e-05,
+      "loss": 0.2553,
+      "step": 83590
+    },
+    {
+      "epoch": 539.35,
+      "learning_rate": 2.559139784946237e-05,
+      "loss": 0.2766,
+      "step": 83600
+    },
+    {
+      "epoch": 539.42,
+      "learning_rate": 2.5587813620071687e-05,
+      "loss": 0.2766,
+      "step": 83610
+    },
+    {
+      "epoch": 539.48,
+      "learning_rate": 2.5584229390681e-05,
+      "loss": 0.2227,
+      "step": 83620
+    },
+    {
+      "epoch": 539.55,
+      "learning_rate": 2.5580645161290323e-05,
+      "loss": 0.2982,
+      "step": 83630
+    },
+    {
+      "epoch": 539.61,
+      "learning_rate": 2.5577060931899645e-05,
+      "loss": 0.2166,
+      "step": 83640
+    },
+    {
+      "epoch": 539.68,
+      "learning_rate": 2.5573476702508963e-05,
+      "loss": 0.2502,
+      "step": 83650
+    },
+    {
+      "epoch": 539.74,
+      "learning_rate": 2.5569892473118278e-05,
+      "loss": 0.263,
+      "step": 83660
+    },
+    {
+      "epoch": 539.8,
+      "learning_rate": 2.55663082437276e-05,
+      "loss": 0.2541,
+      "step": 83670
+    },
+    {
+      "epoch": 539.87,
+      "learning_rate": 2.556272401433692e-05,
+      "loss": 0.2378,
+      "step": 83680
+    },
+    {
+      "epoch": 539.93,
+      "learning_rate": 2.555913978494624e-05,
+      "loss": 0.2406,
+      "step": 83690
+    },
+    {
+      "epoch": 540.0,
+      "learning_rate": 2.5555555555555554e-05,
+      "loss": 0.218,
+      "step": 83700
+    },
+    {
+      "epoch": 540.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0872188806533813,
+      "eval_runtime": 0.8969,
+      "eval_samples_per_second": 224.115,
+      "eval_steps_per_second": 7.805,
+      "step": 83700
+    },
+    {
+      "epoch": 540.06,
+      "learning_rate": 2.5551971326164876e-05,
+      "loss": 0.2679,
+      "step": 83710
+    },
+    {
+      "epoch": 540.13,
+      "learning_rate": 2.5548387096774197e-05,
+      "loss": 0.2519,
+      "step": 83720
+    },
+    {
+      "epoch": 540.19,
+      "learning_rate": 2.5544802867383515e-05,
+      "loss": 0.2902,
+      "step": 83730
+    },
+    {
+      "epoch": 540.26,
+      "learning_rate": 2.554121863799283e-05,
+      "loss": 0.2519,
+      "step": 83740
+    },
+    {
+      "epoch": 540.32,
+      "learning_rate": 2.5537634408602152e-05,
+      "loss": 0.242,
+      "step": 83750
+    },
+    {
+      "epoch": 540.39,
+      "learning_rate": 2.553405017921147e-05,
+      "loss": 0.2585,
+      "step": 83760
+    },
+    {
+      "epoch": 540.45,
+      "learning_rate": 2.553046594982079e-05,
+      "loss": 0.3031,
+      "step": 83770
+    },
+    {
+      "epoch": 540.51,
+      "learning_rate": 2.5526881720430106e-05,
+      "loss": 0.2188,
+      "step": 83780
+    },
+    {
+      "epoch": 540.58,
+      "learning_rate": 2.5523297491039428e-05,
+      "loss": 0.2446,
+      "step": 83790
+    },
+    {
+      "epoch": 540.64,
+      "learning_rate": 2.5519713261648746e-05,
+      "loss": 0.2309,
+      "step": 83800
+    },
+    {
+      "epoch": 540.71,
+      "learning_rate": 2.5516129032258068e-05,
+      "loss": 0.2471,
+      "step": 83810
+    },
+    {
+      "epoch": 540.77,
+      "learning_rate": 2.5512544802867383e-05,
+      "loss": 0.2403,
+      "step": 83820
+    },
+    {
+      "epoch": 540.84,
+      "learning_rate": 2.5508960573476704e-05,
+      "loss": 0.249,
+      "step": 83830
+    },
+    {
+      "epoch": 540.9,
+      "learning_rate": 2.5505376344086022e-05,
+      "loss": 0.296,
+      "step": 83840
+    },
+    {
+      "epoch": 540.96,
+      "learning_rate": 2.5501792114695344e-05,
+      "loss": 0.2568,
+      "step": 83850
+    },
+    {
+      "epoch": 541.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.04519784450531,
+      "eval_runtime": 0.9084,
+      "eval_samples_per_second": 221.258,
+      "eval_steps_per_second": 7.705,
+      "step": 83855
+    },
+    {
+      "epoch": 541.03,
+      "learning_rate": 2.549820788530466e-05,
+      "loss": 0.2751,
+      "step": 83860
+    },
+    {
+      "epoch": 541.1,
+      "learning_rate": 2.549462365591398e-05,
+      "loss": 0.2454,
+      "step": 83870
+    },
+    {
+      "epoch": 541.16,
+      "learning_rate": 2.54910394265233e-05,
+      "loss": 0.217,
+      "step": 83880
+    },
+    {
+      "epoch": 541.23,
+      "learning_rate": 2.548745519713262e-05,
+      "loss": 0.285,
+      "step": 83890
+    },
+    {
+      "epoch": 541.29,
+      "learning_rate": 2.5483870967741935e-05,
+      "loss": 0.2639,
+      "step": 83900
+    },
+    {
+      "epoch": 541.35,
+      "learning_rate": 2.5480286738351257e-05,
+      "loss": 0.299,
+      "step": 83910
+    },
+    {
+      "epoch": 541.42,
+      "learning_rate": 2.5476702508960575e-05,
+      "loss": 0.2484,
+      "step": 83920
+    },
+    {
+      "epoch": 541.48,
+      "learning_rate": 2.5473118279569897e-05,
+      "loss": 0.31,
+      "step": 83930
+    },
+    {
+      "epoch": 541.55,
+      "learning_rate": 2.546953405017921e-05,
+      "loss": 0.2104,
+      "step": 83940
+    },
+    {
+      "epoch": 541.61,
+      "learning_rate": 2.5465949820788533e-05,
+      "loss": 0.3159,
+      "step": 83950
+    },
+    {
+      "epoch": 541.68,
+      "learning_rate": 2.546236559139785e-05,
+      "loss": 0.2959,
+      "step": 83960
+    },
+    {
+      "epoch": 541.74,
+      "learning_rate": 2.5458781362007173e-05,
+      "loss": 0.2129,
+      "step": 83970
+    },
+    {
+      "epoch": 541.8,
+      "learning_rate": 2.5455197132616488e-05,
+      "loss": 0.2216,
+      "step": 83980
+    },
+    {
+      "epoch": 541.87,
+      "learning_rate": 2.545161290322581e-05,
+      "loss": 0.2458,
+      "step": 83990
+    },
+    {
+      "epoch": 541.93,
+      "learning_rate": 2.5448028673835127e-05,
+      "loss": 0.2813,
+      "step": 84000
+    },
+    {
+      "epoch": 542.0,
+      "learning_rate": 2.5444444444444442e-05,
+      "loss": 0.2728,
+      "step": 84010
+    },
+    {
+      "epoch": 542.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.995203971862793,
+      "eval_runtime": 0.9201,
+      "eval_samples_per_second": 218.459,
+      "eval_steps_per_second": 7.608,
+      "step": 84010
+    },
+    {
+      "epoch": 542.06,
+      "learning_rate": 2.5440860215053764e-05,
+      "loss": 0.2398,
+      "step": 84020
+    },
+    {
+      "epoch": 542.13,
+      "learning_rate": 2.5437275985663085e-05,
+      "loss": 0.2415,
+      "step": 84030
+    },
+    {
+      "epoch": 542.19,
+      "learning_rate": 2.5433691756272404e-05,
+      "loss": 0.2189,
+      "step": 84040
+    },
+    {
+      "epoch": 542.26,
+      "learning_rate": 2.543010752688172e-05,
+      "loss": 0.2403,
+      "step": 84050
+    },
+    {
+      "epoch": 542.32,
+      "learning_rate": 2.542652329749104e-05,
+      "loss": 0.2212,
+      "step": 84060
+    },
+    {
+      "epoch": 542.39,
+      "learning_rate": 2.542293906810036e-05,
+      "loss": 0.2489,
+      "step": 84070
+    },
+    {
+      "epoch": 542.45,
+      "learning_rate": 2.541935483870968e-05,
+      "loss": 0.2542,
+      "step": 84080
+    },
+    {
+      "epoch": 542.51,
+      "learning_rate": 2.5415770609318995e-05,
+      "loss": 0.2434,
+      "step": 84090
+    },
+    {
+      "epoch": 542.58,
+      "learning_rate": 2.5412186379928316e-05,
+      "loss": 0.2699,
+      "step": 84100
+    },
+    {
+      "epoch": 542.64,
+      "learning_rate": 2.5408602150537634e-05,
+      "loss": 0.2546,
+      "step": 84110
+    },
+    {
+      "epoch": 542.71,
+      "learning_rate": 2.5405017921146956e-05,
+      "loss": 0.2314,
+      "step": 84120
+    },
+    {
+      "epoch": 542.77,
+      "learning_rate": 2.540143369175627e-05,
+      "loss": 0.3343,
+      "step": 84130
+    },
+    {
+      "epoch": 542.84,
+      "learning_rate": 2.5397849462365592e-05,
+      "loss": 0.2052,
+      "step": 84140
+    },
+    {
+      "epoch": 542.9,
+      "learning_rate": 2.539426523297491e-05,
+      "loss": 0.27,
+      "step": 84150
+    },
+    {
+      "epoch": 542.96,
+      "learning_rate": 2.5390681003584232e-05,
+      "loss": 0.2592,
+      "step": 84160
+    },
+    {
+      "epoch": 543.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9522969126701355,
+      "eval_runtime": 0.8729,
+      "eval_samples_per_second": 230.273,
+      "eval_steps_per_second": 8.019,
+      "step": 84165
+    },
+    {
+      "epoch": 543.03,
+      "learning_rate": 2.5387096774193547e-05,
+      "loss": 0.2823,
+      "step": 84170
+    },
+    {
+      "epoch": 543.1,
+      "learning_rate": 2.538351254480287e-05,
+      "loss": 0.2789,
+      "step": 84180
+    },
+    {
+      "epoch": 543.16,
+      "learning_rate": 2.5379928315412187e-05,
+      "loss": 0.2734,
+      "step": 84190
+    },
+    {
+      "epoch": 543.23,
+      "learning_rate": 2.537634408602151e-05,
+      "loss": 0.2775,
+      "step": 84200
+    },
+    {
+      "epoch": 543.29,
+      "learning_rate": 2.5372759856630823e-05,
+      "loss": 0.2341,
+      "step": 84210
+    },
+    {
+      "epoch": 543.35,
+      "learning_rate": 2.5369175627240145e-05,
+      "loss": 0.2499,
+      "step": 84220
+    },
+    {
+      "epoch": 543.42,
+      "learning_rate": 2.5365591397849463e-05,
+      "loss": 0.2491,
+      "step": 84230
+    },
+    {
+      "epoch": 543.48,
+      "learning_rate": 2.5362007168458785e-05,
+      "loss": 0.2506,
+      "step": 84240
+    },
+    {
+      "epoch": 543.55,
+      "learning_rate": 2.53584229390681e-05,
+      "loss": 0.2698,
+      "step": 84250
+    },
+    {
+      "epoch": 543.61,
+      "learning_rate": 2.535483870967742e-05,
+      "loss": 0.2584,
+      "step": 84260
+    },
+    {
+      "epoch": 543.68,
+      "learning_rate": 2.535125448028674e-05,
+      "loss": 0.2937,
+      "step": 84270
+    },
+    {
+      "epoch": 543.74,
+      "learning_rate": 2.534767025089606e-05,
+      "loss": 0.2436,
+      "step": 84280
+    },
+    {
+      "epoch": 543.8,
+      "learning_rate": 2.5344086021505376e-05,
+      "loss": 0.2351,
+      "step": 84290
+    },
+    {
+      "epoch": 543.87,
+      "learning_rate": 2.5340501792114697e-05,
+      "loss": 0.2387,
+      "step": 84300
+    },
+    {
+      "epoch": 543.93,
+      "learning_rate": 2.5336917562724015e-05,
+      "loss": 0.2434,
+      "step": 84310
+    },
+    {
+      "epoch": 544.0,
+      "learning_rate": 2.5333333333333337e-05,
+      "loss": 0.282,
+      "step": 84320
+    },
+    {
+      "epoch": 544.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0198464393615723,
+      "eval_runtime": 0.9058,
+      "eval_samples_per_second": 221.893,
+      "eval_steps_per_second": 7.728,
+      "step": 84320
+    },
+    {
+      "epoch": 544.06,
+      "learning_rate": 2.5329749103942652e-05,
+      "loss": 0.2883,
+      "step": 84330
+    },
+    {
+      "epoch": 544.13,
+      "learning_rate": 2.5326164874551973e-05,
+      "loss": 0.3016,
+      "step": 84340
+    },
+    {
+      "epoch": 544.19,
+      "learning_rate": 2.532258064516129e-05,
+      "loss": 0.2652,
+      "step": 84350
+    },
+    {
+      "epoch": 544.26,
+      "learning_rate": 2.5318996415770613e-05,
+      "loss": 0.2405,
+      "step": 84360
+    },
+    {
+      "epoch": 544.32,
+      "learning_rate": 2.5315412186379928e-05,
+      "loss": 0.2747,
+      "step": 84370
+    },
+    {
+      "epoch": 544.39,
+      "learning_rate": 2.531182795698925e-05,
+      "loss": 0.2932,
+      "step": 84380
+    },
+    {
+      "epoch": 544.45,
+      "learning_rate": 2.5308243727598568e-05,
+      "loss": 0.275,
+      "step": 84390
+    },
+    {
+      "epoch": 544.51,
+      "learning_rate": 2.530465949820789e-05,
+      "loss": 0.2507,
+      "step": 84400
+    },
+    {
+      "epoch": 544.58,
+      "learning_rate": 2.5301075268817204e-05,
+      "loss": 0.2599,
+      "step": 84410
+    },
+    {
+      "epoch": 544.64,
+      "learning_rate": 2.5297491039426522e-05,
+      "loss": 0.2391,
+      "step": 84420
+    },
+    {
+      "epoch": 544.71,
+      "learning_rate": 2.5293906810035844e-05,
+      "loss": 0.2521,
+      "step": 84430
+    },
+    {
+      "epoch": 544.77,
+      "learning_rate": 2.5290322580645166e-05,
+      "loss": 0.3066,
+      "step": 84440
+    },
+    {
+      "epoch": 544.84,
+      "learning_rate": 2.528673835125448e-05,
+      "loss": 0.2178,
+      "step": 84450
+    },
+    {
+      "epoch": 544.9,
+      "learning_rate": 2.52831541218638e-05,
+      "loss": 0.2524,
+      "step": 84460
+    },
+    {
+      "epoch": 544.96,
+      "learning_rate": 2.527956989247312e-05,
+      "loss": 0.2385,
+      "step": 84470
+    },
+    {
+      "epoch": 545.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0097293853759766,
+      "eval_runtime": 0.8827,
+      "eval_samples_per_second": 227.701,
+      "eval_steps_per_second": 7.93,
+      "step": 84475
+    },
+    {
+      "epoch": 545.03,
+      "learning_rate": 2.5275985663082442e-05,
+      "loss": 0.2277,
+      "step": 84480
+    },
+    {
+      "epoch": 545.1,
+      "learning_rate": 2.5272401433691757e-05,
+      "loss": 0.2253,
+      "step": 84490
+    },
+    {
+      "epoch": 545.16,
+      "learning_rate": 2.5268817204301075e-05,
+      "loss": 0.2501,
+      "step": 84500
+    },
+    {
+      "epoch": 545.23,
+      "learning_rate": 2.5265232974910396e-05,
+      "loss": 0.2064,
+      "step": 84510
+    },
+    {
+      "epoch": 545.29,
+      "learning_rate": 2.5261648745519718e-05,
+      "loss": 0.244,
+      "step": 84520
+    },
+    {
+      "epoch": 545.35,
+      "learning_rate": 2.5258064516129033e-05,
+      "loss": 0.204,
+      "step": 84530
+    },
+    {
+      "epoch": 545.42,
+      "learning_rate": 2.525448028673835e-05,
+      "loss": 0.2437,
+      "step": 84540
+    },
+    {
+      "epoch": 545.48,
+      "learning_rate": 2.5250896057347673e-05,
+      "loss": 0.2547,
+      "step": 84550
+    },
+    {
+      "epoch": 545.55,
+      "learning_rate": 2.5247311827956994e-05,
+      "loss": 0.2248,
+      "step": 84560
+    },
+    {
+      "epoch": 545.61,
+      "learning_rate": 2.524372759856631e-05,
+      "loss": 0.2835,
+      "step": 84570
+    },
+    {
+      "epoch": 545.68,
+      "learning_rate": 2.5240143369175627e-05,
+      "loss": 0.1862,
+      "step": 84580
+    },
+    {
+      "epoch": 545.74,
+      "learning_rate": 2.523655913978495e-05,
+      "loss": 0.2881,
+      "step": 84590
+    },
+    {
+      "epoch": 545.8,
+      "learning_rate": 2.523297491039427e-05,
+      "loss": 0.2939,
+      "step": 84600
+    },
+    {
+      "epoch": 545.87,
+      "learning_rate": 2.5229390681003585e-05,
+      "loss": 0.2166,
+      "step": 84610
+    },
+    {
+      "epoch": 545.93,
+      "learning_rate": 2.5225806451612903e-05,
+      "loss": 0.2194,
+      "step": 84620
+    },
+    {
+      "epoch": 546.0,
+      "learning_rate": 2.5222222222222225e-05,
+      "loss": 0.2447,
+      "step": 84630
+    },
+    {
+      "epoch": 546.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.974609375,
+      "eval_runtime": 0.9252,
+      "eval_samples_per_second": 217.252,
+      "eval_steps_per_second": 7.566,
+      "step": 84630
+    },
+    {
+      "epoch": 546.06,
+      "learning_rate": 2.521863799283154e-05,
+      "loss": 0.2585,
+      "step": 84640
+    },
+    {
+      "epoch": 546.13,
+      "learning_rate": 2.521505376344086e-05,
+      "loss": 0.2338,
+      "step": 84650
+    },
+    {
+      "epoch": 546.19,
+      "learning_rate": 2.521146953405018e-05,
+      "loss": 0.3035,
+      "step": 84660
+    },
+    {
+      "epoch": 546.26,
+      "learning_rate": 2.52078853046595e-05,
+      "loss": 0.1988,
+      "step": 84670
+    },
+    {
+      "epoch": 546.32,
+      "learning_rate": 2.5204301075268816e-05,
+      "loss": 0.2648,
+      "step": 84680
+    },
+    {
+      "epoch": 546.39,
+      "learning_rate": 2.5200716845878138e-05,
+      "loss": 0.2016,
+      "step": 84690
+    },
+    {
+      "epoch": 546.45,
+      "learning_rate": 2.5197132616487456e-05,
+      "loss": 0.2138,
+      "step": 84700
+    },
+    {
+      "epoch": 546.51,
+      "learning_rate": 2.5193548387096777e-05,
+      "loss": 0.2352,
+      "step": 84710
+    },
+    {
+      "epoch": 546.58,
+      "learning_rate": 2.5189964157706092e-05,
+      "loss": 0.2971,
+      "step": 84720
+    },
+    {
+      "epoch": 546.64,
+      "learning_rate": 2.5186379928315414e-05,
+      "loss": 0.2638,
+      "step": 84730
+    },
+    {
+      "epoch": 546.71,
+      "learning_rate": 2.5182795698924732e-05,
+      "loss": 0.2102,
+      "step": 84740
+    },
+    {
+      "epoch": 546.77,
+      "learning_rate": 2.5179211469534054e-05,
+      "loss": 0.2628,
+      "step": 84750
+    },
+    {
+      "epoch": 546.84,
+      "learning_rate": 2.517562724014337e-05,
+      "loss": 0.2401,
+      "step": 84760
+    },
+    {
+      "epoch": 546.9,
+      "learning_rate": 2.5172043010752687e-05,
+      "loss": 0.2629,
+      "step": 84770
+    },
+    {
+      "epoch": 546.96,
+      "learning_rate": 2.5168458781362008e-05,
+      "loss": 0.2581,
+      "step": 84780
+    },
+    {
+      "epoch": 547.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9977512359619141,
+      "eval_runtime": 0.8776,
+      "eval_samples_per_second": 229.037,
+      "eval_steps_per_second": 7.976,
+      "step": 84785
+    },
+    {
+      "epoch": 547.03,
+      "learning_rate": 2.516487455197133e-05,
+      "loss": 0.2429,
+      "step": 84790
+    },
+    {
+      "epoch": 547.1,
+      "learning_rate": 2.5161290322580645e-05,
+      "loss": 0.2239,
+      "step": 84800
+    },
+    {
+      "epoch": 547.16,
+      "learning_rate": 2.5157706093189963e-05,
+      "loss": 0.2343,
+      "step": 84810
+    },
+    {
+      "epoch": 547.23,
+      "learning_rate": 2.5154121863799285e-05,
+      "loss": 0.2276,
+      "step": 84820
+    },
+    {
+      "epoch": 547.29,
+      "learning_rate": 2.5150537634408606e-05,
+      "loss": 0.1917,
+      "step": 84830
+    },
+    {
+      "epoch": 547.35,
+      "learning_rate": 2.514695340501792e-05,
+      "loss": 0.2797,
+      "step": 84840
+    },
+    {
+      "epoch": 547.42,
+      "learning_rate": 2.514336917562724e-05,
+      "loss": 0.3011,
+      "step": 84850
+    },
+    {
+      "epoch": 547.48,
+      "learning_rate": 2.513978494623656e-05,
+      "loss": 0.2716,
+      "step": 84860
+    },
+    {
+      "epoch": 547.55,
+      "learning_rate": 2.5136200716845882e-05,
+      "loss": 0.2382,
+      "step": 84870
+    },
+    {
+      "epoch": 547.61,
+      "learning_rate": 2.5132616487455197e-05,
+      "loss": 0.2533,
+      "step": 84880
+    },
+    {
+      "epoch": 547.68,
+      "learning_rate": 2.5129032258064515e-05,
+      "loss": 0.2567,
+      "step": 84890
+    },
+    {
+      "epoch": 547.74,
+      "learning_rate": 2.5125448028673837e-05,
+      "loss": 0.2828,
+      "step": 84900
+    },
+    {
+      "epoch": 547.8,
+      "learning_rate": 2.512186379928316e-05,
+      "loss": 0.2929,
+      "step": 84910
+    },
+    {
+      "epoch": 547.87,
+      "learning_rate": 2.5118279569892473e-05,
+      "loss": 0.3095,
+      "step": 84920
+    },
+    {
+      "epoch": 547.93,
+      "learning_rate": 2.511469534050179e-05,
+      "loss": 0.2044,
+      "step": 84930
+    },
+    {
+      "epoch": 548.0,
+      "learning_rate": 2.5111111111111113e-05,
+      "loss": 0.2402,
+      "step": 84940
+    },
+    {
+      "epoch": 548.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0066843032836914,
+      "eval_runtime": 0.8833,
+      "eval_samples_per_second": 227.543,
+      "eval_steps_per_second": 7.924,
+      "step": 84940
+    },
+    {
+      "epoch": 548.06,
+      "learning_rate": 2.5107526881720435e-05,
+      "loss": 0.3203,
+      "step": 84950
+    },
+    {
+      "epoch": 548.13,
+      "learning_rate": 2.510394265232975e-05,
+      "loss": 0.2395,
+      "step": 84960
+    },
+    {
+      "epoch": 548.19,
+      "learning_rate": 2.5100358422939068e-05,
+      "loss": 0.2557,
+      "step": 84970
+    },
+    {
+      "epoch": 548.26,
+      "learning_rate": 2.509677419354839e-05,
+      "loss": 0.2484,
+      "step": 84980
+    },
+    {
+      "epoch": 548.32,
+      "learning_rate": 2.509318996415771e-05,
+      "loss": 0.2321,
+      "step": 84990
+    },
+    {
+      "epoch": 548.39,
+      "learning_rate": 2.5089605734767026e-05,
+      "loss": 0.2483,
+      "step": 85000
+    },
+    {
+      "epoch": 548.45,
+      "learning_rate": 2.5086021505376344e-05,
+      "loss": 0.2598,
+      "step": 85010
+    },
+    {
+      "epoch": 548.51,
+      "learning_rate": 2.5082437275985666e-05,
+      "loss": 0.2477,
+      "step": 85020
+    },
+    {
+      "epoch": 548.58,
+      "learning_rate": 2.5078853046594987e-05,
+      "loss": 0.3177,
+      "step": 85030
+    },
+    {
+      "epoch": 548.64,
+      "learning_rate": 2.5075268817204302e-05,
+      "loss": 0.304,
+      "step": 85040
+    },
+    {
+      "epoch": 548.71,
+      "learning_rate": 2.507168458781362e-05,
+      "loss": 0.2326,
+      "step": 85050
+    },
+    {
+      "epoch": 548.77,
+      "learning_rate": 2.5068100358422942e-05,
+      "loss": 0.2294,
+      "step": 85060
+    },
+    {
+      "epoch": 548.84,
+      "learning_rate": 2.5064516129032263e-05,
+      "loss": 0.2783,
+      "step": 85070
+    },
+    {
+      "epoch": 548.9,
+      "learning_rate": 2.5060931899641578e-05,
+      "loss": 0.2429,
+      "step": 85080
+    },
+    {
+      "epoch": 548.96,
+      "learning_rate": 2.5057347670250896e-05,
+      "loss": 0.1828,
+      "step": 85090
+    },
+    {
+      "epoch": 549.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9847460985183716,
+      "eval_runtime": 0.8916,
+      "eval_samples_per_second": 225.444,
+      "eval_steps_per_second": 7.851,
+      "step": 85095
+    },
+    {
+      "epoch": 549.03,
+      "learning_rate": 2.5053763440860218e-05,
+      "loss": 0.2452,
+      "step": 85100
+    },
+    {
+      "epoch": 549.1,
+      "learning_rate": 2.5050179211469536e-05,
+      "loss": 0.2411,
+      "step": 85110
+    },
+    {
+      "epoch": 549.16,
+      "learning_rate": 2.504659498207885e-05,
+      "loss": 0.2721,
+      "step": 85120
+    },
+    {
+      "epoch": 549.23,
+      "learning_rate": 2.5043010752688173e-05,
+      "loss": 0.25,
+      "step": 85130
+    },
+    {
+      "epoch": 549.29,
+      "learning_rate": 2.5039426523297494e-05,
+      "loss": 0.2522,
+      "step": 85140
+    },
+    {
+      "epoch": 549.35,
+      "learning_rate": 2.5035842293906812e-05,
+      "loss": 0.2598,
+      "step": 85150
+    },
+    {
+      "epoch": 549.42,
+      "learning_rate": 2.5032258064516127e-05,
+      "loss": 0.2573,
+      "step": 85160
+    },
+    {
+      "epoch": 549.48,
+      "learning_rate": 2.502867383512545e-05,
+      "loss": 0.2577,
+      "step": 85170
+    },
+    {
+      "epoch": 549.55,
+      "learning_rate": 2.502508960573477e-05,
+      "loss": 0.2532,
+      "step": 85180
+    },
+    {
+      "epoch": 549.61,
+      "learning_rate": 2.502150537634409e-05,
+      "loss": 0.2665,
+      "step": 85190
+    },
+    {
+      "epoch": 549.68,
+      "learning_rate": 2.5017921146953403e-05,
+      "loss": 0.2304,
+      "step": 85200
+    },
+    {
+      "epoch": 549.74,
+      "learning_rate": 2.5014336917562725e-05,
+      "loss": 0.2159,
+      "step": 85210
+    },
+    {
+      "epoch": 549.8,
+      "learning_rate": 2.5010752688172047e-05,
+      "loss": 0.2284,
+      "step": 85220
+    },
+    {
+      "epoch": 549.87,
+      "learning_rate": 2.5007168458781365e-05,
+      "loss": 0.259,
+      "step": 85230
+    },
+    {
+      "epoch": 549.93,
+      "learning_rate": 2.500358422939068e-05,
+      "loss": 0.2522,
+      "step": 85240
+    },
+    {
+      "epoch": 550.0,
+      "learning_rate": 2.5e-05,
+      "loss": 0.2511,
+      "step": 85250
+    },
+    {
+      "epoch": 550.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9723673462867737,
+      "eval_runtime": 0.8839,
+      "eval_samples_per_second": 227.403,
+      "eval_steps_per_second": 7.92,
+      "step": 85250
+    },
+    {
+      "epoch": 550.06,
+      "learning_rate": 2.499641577060932e-05,
+      "loss": 0.1947,
+      "step": 85260
+    },
+    {
+      "epoch": 550.13,
+      "learning_rate": 2.4992831541218638e-05,
+      "loss": 0.2666,
+      "step": 85270
+    },
+    {
+      "epoch": 550.19,
+      "learning_rate": 2.4989247311827956e-05,
+      "loss": 0.2236,
+      "step": 85280
+    },
+    {
+      "epoch": 550.26,
+      "learning_rate": 2.4985663082437277e-05,
+      "loss": 0.2548,
+      "step": 85290
+    },
+    {
+      "epoch": 550.32,
+      "learning_rate": 2.4982078853046596e-05,
+      "loss": 0.2635,
+      "step": 85300
+    },
+    {
+      "epoch": 550.39,
+      "learning_rate": 2.4978494623655914e-05,
+      "loss": 0.2413,
+      "step": 85310
+    },
+    {
+      "epoch": 550.45,
+      "learning_rate": 2.4974910394265232e-05,
+      "loss": 0.2654,
+      "step": 85320
+    },
+    {
+      "epoch": 550.51,
+      "learning_rate": 2.4971326164874554e-05,
+      "loss": 0.2419,
+      "step": 85330
+    },
+    {
+      "epoch": 550.58,
+      "learning_rate": 2.4967741935483872e-05,
+      "loss": 0.2705,
+      "step": 85340
+    },
+    {
+      "epoch": 550.64,
+      "learning_rate": 2.496415770609319e-05,
+      "loss": 0.2566,
+      "step": 85350
+    },
+    {
+      "epoch": 550.71,
+      "learning_rate": 2.4960573476702508e-05,
+      "loss": 0.2651,
+      "step": 85360
+    },
+    {
+      "epoch": 550.77,
+      "learning_rate": 2.495698924731183e-05,
+      "loss": 0.2013,
+      "step": 85370
+    },
+    {
+      "epoch": 550.84,
+      "learning_rate": 2.4953405017921148e-05,
+      "loss": 0.2662,
+      "step": 85380
+    },
+    {
+      "epoch": 550.9,
+      "learning_rate": 2.4949820788530466e-05,
+      "loss": 0.2773,
+      "step": 85390
+    },
+    {
+      "epoch": 550.96,
+      "learning_rate": 2.4946236559139784e-05,
+      "loss": 0.2151,
+      "step": 85400
+    },
+    {
+      "epoch": 551.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9613301157951355,
+      "eval_runtime": 0.89,
+      "eval_samples_per_second": 225.838,
+      "eval_steps_per_second": 7.865,
+      "step": 85405
+    },
+    {
+      "epoch": 551.03,
+      "learning_rate": 2.4942652329749106e-05,
+      "loss": 0.2797,
+      "step": 85410
+    },
+    {
+      "epoch": 551.1,
+      "learning_rate": 2.4939068100358424e-05,
+      "loss": 0.2603,
+      "step": 85420
+    },
+    {
+      "epoch": 551.16,
+      "learning_rate": 2.4935483870967742e-05,
+      "loss": 0.2362,
+      "step": 85430
+    },
+    {
+      "epoch": 551.23,
+      "learning_rate": 2.493189964157706e-05,
+      "loss": 0.2408,
+      "step": 85440
+    },
+    {
+      "epoch": 551.29,
+      "learning_rate": 2.4928315412186382e-05,
+      "loss": 0.2655,
+      "step": 85450
+    },
+    {
+      "epoch": 551.35,
+      "learning_rate": 2.49247311827957e-05,
+      "loss": 0.2587,
+      "step": 85460
+    },
+    {
+      "epoch": 551.42,
+      "learning_rate": 2.492114695340502e-05,
+      "loss": 0.2526,
+      "step": 85470
+    },
+    {
+      "epoch": 551.48,
+      "learning_rate": 2.4917562724014337e-05,
+      "loss": 0.1672,
+      "step": 85480
+    },
+    {
+      "epoch": 551.55,
+      "learning_rate": 2.491397849462366e-05,
+      "loss": 0.2284,
+      "step": 85490
+    },
+    {
+      "epoch": 551.61,
+      "learning_rate": 2.4910394265232977e-05,
+      "loss": 0.2488,
+      "step": 85500
+    },
+    {
+      "epoch": 551.68,
+      "learning_rate": 2.4906810035842295e-05,
+      "loss": 0.2237,
+      "step": 85510
+    },
+    {
+      "epoch": 551.74,
+      "learning_rate": 2.4903225806451613e-05,
+      "loss": 0.2261,
+      "step": 85520
+    },
+    {
+      "epoch": 551.8,
+      "learning_rate": 2.4899641577060935e-05,
+      "loss": 0.249,
+      "step": 85530
+    },
+    {
+      "epoch": 551.87,
+      "learning_rate": 2.4896057347670253e-05,
+      "loss": 0.2598,
+      "step": 85540
+    },
+    {
+      "epoch": 551.93,
+      "learning_rate": 2.489247311827957e-05,
+      "loss": 0.2487,
+      "step": 85550
+    },
+    {
+      "epoch": 552.0,
+      "learning_rate": 2.488888888888889e-05,
+      "loss": 0.2349,
+      "step": 85560
+    },
+    {
+      "epoch": 552.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9662097692489624,
+      "eval_runtime": 0.8802,
+      "eval_samples_per_second": 228.347,
+      "eval_steps_per_second": 7.952,
+      "step": 85560
+    },
+    {
+      "epoch": 552.06,
+      "learning_rate": 2.488530465949821e-05,
+      "loss": 0.2407,
+      "step": 85570
+    },
+    {
+      "epoch": 552.13,
+      "learning_rate": 2.488172043010753e-05,
+      "loss": 0.2462,
+      "step": 85580
+    },
+    {
+      "epoch": 552.19,
+      "learning_rate": 2.4878136200716847e-05,
+      "loss": 0.2539,
+      "step": 85590
+    },
+    {
+      "epoch": 552.26,
+      "learning_rate": 2.4874551971326165e-05,
+      "loss": 0.2504,
+      "step": 85600
+    },
+    {
+      "epoch": 552.32,
+      "learning_rate": 2.4870967741935487e-05,
+      "loss": 0.206,
+      "step": 85610
+    },
+    {
+      "epoch": 552.39,
+      "learning_rate": 2.4867383512544802e-05,
+      "loss": 0.2205,
+      "step": 85620
+    },
+    {
+      "epoch": 552.45,
+      "learning_rate": 2.4863799283154123e-05,
+      "loss": 0.2487,
+      "step": 85630
+    },
+    {
+      "epoch": 552.51,
+      "learning_rate": 2.486021505376344e-05,
+      "loss": 0.264,
+      "step": 85640
+    },
+    {
+      "epoch": 552.58,
+      "learning_rate": 2.4856630824372763e-05,
+      "loss": 0.253,
+      "step": 85650
+    },
+    {
+      "epoch": 552.64,
+      "learning_rate": 2.4853046594982078e-05,
+      "loss": 0.2139,
+      "step": 85660
+    },
+    {
+      "epoch": 552.71,
+      "learning_rate": 2.48494623655914e-05,
+      "loss": 0.28,
+      "step": 85670
+    },
+    {
+      "epoch": 552.77,
+      "learning_rate": 2.4845878136200718e-05,
+      "loss": 0.2687,
+      "step": 85680
+    },
+    {
+      "epoch": 552.84,
+      "learning_rate": 2.484229390681004e-05,
+      "loss": 0.2425,
+      "step": 85690
+    },
+    {
+      "epoch": 552.9,
+      "learning_rate": 2.4838709677419354e-05,
+      "loss": 0.212,
+      "step": 85700
+    },
+    {
+      "epoch": 552.96,
+      "learning_rate": 2.4835125448028676e-05,
+      "loss": 0.2219,
+      "step": 85710
+    },
+    {
+      "epoch": 553.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9796000123023987,
+      "eval_runtime": 0.8699,
+      "eval_samples_per_second": 231.051,
+      "eval_steps_per_second": 8.047,
+      "step": 85715
+    },
+    {
+      "epoch": 553.03,
+      "learning_rate": 2.4831541218637994e-05,
+      "loss": 0.2722,
+      "step": 85720
+    },
+    {
+      "epoch": 553.1,
+      "learning_rate": 2.4827956989247316e-05,
+      "loss": 0.2848,
+      "step": 85730
+    },
+    {
+      "epoch": 553.16,
+      "learning_rate": 2.482437275985663e-05,
+      "loss": 0.2152,
+      "step": 85740
+    },
+    {
+      "epoch": 553.23,
+      "learning_rate": 2.4820788530465952e-05,
+      "loss": 0.2378,
+      "step": 85750
+    },
+    {
+      "epoch": 553.29,
+      "learning_rate": 2.481720430107527e-05,
+      "loss": 0.2502,
+      "step": 85760
+    },
+    {
+      "epoch": 553.35,
+      "learning_rate": 2.481362007168459e-05,
+      "loss": 0.2397,
+      "step": 85770
+    },
+    {
+      "epoch": 553.42,
+      "learning_rate": 2.4810035842293907e-05,
+      "loss": 0.2444,
+      "step": 85780
+    },
+    {
+      "epoch": 553.48,
+      "learning_rate": 2.480645161290323e-05,
+      "loss": 0.2786,
+      "step": 85790
+    },
+    {
+      "epoch": 553.55,
+      "learning_rate": 2.4802867383512547e-05,
+      "loss": 0.2099,
+      "step": 85800
+    },
+    {
+      "epoch": 553.61,
+      "learning_rate": 2.4799283154121865e-05,
+      "loss": 0.2175,
+      "step": 85810
+    },
+    {
+      "epoch": 553.68,
+      "learning_rate": 2.4795698924731183e-05,
+      "loss": 0.2041,
+      "step": 85820
+    },
+    {
+      "epoch": 553.74,
+      "learning_rate": 2.4792114695340505e-05,
+      "loss": 0.2842,
+      "step": 85830
+    },
+    {
+      "epoch": 553.8,
+      "learning_rate": 2.4788530465949823e-05,
+      "loss": 0.2523,
+      "step": 85840
+    },
+    {
+      "epoch": 553.87,
+      "learning_rate": 2.478494623655914e-05,
+      "loss": 0.268,
+      "step": 85850
+    },
+    {
+      "epoch": 553.93,
+      "learning_rate": 2.478136200716846e-05,
+      "loss": 0.2557,
+      "step": 85860
+    },
+    {
+      "epoch": 554.0,
+      "learning_rate": 2.477777777777778e-05,
+      "loss": 0.2656,
+      "step": 85870
+    },
+    {
+      "epoch": 554.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9307222366333008,
+      "eval_runtime": 0.8821,
+      "eval_samples_per_second": 227.877,
+      "eval_steps_per_second": 7.936,
+      "step": 85870
+    },
+    {
+      "epoch": 554.06,
+      "learning_rate": 2.47741935483871e-05,
+      "loss": 0.239,
+      "step": 85880
+    },
+    {
+      "epoch": 554.13,
+      "learning_rate": 2.4770609318996417e-05,
+      "loss": 0.2449,
+      "step": 85890
+    },
+    {
+      "epoch": 554.19,
+      "learning_rate": 2.4767025089605735e-05,
+      "loss": 0.2182,
+      "step": 85900
+    },
+    {
+      "epoch": 554.26,
+      "learning_rate": 2.4763440860215054e-05,
+      "loss": 0.2452,
+      "step": 85910
+    },
+    {
+      "epoch": 554.32,
+      "learning_rate": 2.4759856630824375e-05,
+      "loss": 0.2076,
+      "step": 85920
+    },
+    {
+      "epoch": 554.39,
+      "learning_rate": 2.475627240143369e-05,
+      "loss": 0.2263,
+      "step": 85930
+    },
+    {
+      "epoch": 554.45,
+      "learning_rate": 2.475268817204301e-05,
+      "loss": 0.2286,
+      "step": 85940
+    },
+    {
+      "epoch": 554.51,
+      "learning_rate": 2.474910394265233e-05,
+      "loss": 0.2252,
+      "step": 85950
+    },
+    {
+      "epoch": 554.58,
+      "learning_rate": 2.474551971326165e-05,
+      "loss": 0.2597,
+      "step": 85960
+    },
+    {
+      "epoch": 554.64,
+      "learning_rate": 2.4741935483870966e-05,
+      "loss": 0.2716,
+      "step": 85970
+    },
+    {
+      "epoch": 554.71,
+      "learning_rate": 2.4738351254480288e-05,
+      "loss": 0.2812,
+      "step": 85980
+    },
+    {
+      "epoch": 554.77,
+      "learning_rate": 2.4734767025089606e-05,
+      "loss": 0.2947,
+      "step": 85990
+    },
+    {
+      "epoch": 554.84,
+      "learning_rate": 2.4731182795698928e-05,
+      "loss": 0.233,
+      "step": 86000
+    },
+    {
+      "epoch": 554.9,
+      "learning_rate": 2.4727598566308242e-05,
+      "loss": 0.2303,
+      "step": 86010
+    },
+    {
+      "epoch": 554.96,
+      "learning_rate": 2.4724014336917564e-05,
+      "loss": 0.2659,
+      "step": 86020
+    },
+    {
+      "epoch": 555.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9699029922485352,
+      "eval_runtime": 0.8665,
+      "eval_samples_per_second": 231.966,
+      "eval_steps_per_second": 8.078,
+      "step": 86025
+    },
+    {
+      "epoch": 555.03,
+      "learning_rate": 2.4720430107526882e-05,
+      "loss": 0.2554,
+      "step": 86030
+    },
+    {
+      "epoch": 555.1,
+      "learning_rate": 2.4716845878136204e-05,
+      "loss": 0.2305,
+      "step": 86040
+    },
+    {
+      "epoch": 555.16,
+      "learning_rate": 2.471326164874552e-05,
+      "loss": 0.2323,
+      "step": 86050
+    },
+    {
+      "epoch": 555.23,
+      "learning_rate": 2.470967741935484e-05,
+      "loss": 0.219,
+      "step": 86060
+    },
+    {
+      "epoch": 555.29,
+      "learning_rate": 2.470609318996416e-05,
+      "loss": 0.2677,
+      "step": 86070
+    },
+    {
+      "epoch": 555.35,
+      "learning_rate": 2.470250896057348e-05,
+      "loss": 0.2565,
+      "step": 86080
+    },
+    {
+      "epoch": 555.42,
+      "learning_rate": 2.4698924731182795e-05,
+      "loss": 0.274,
+      "step": 86090
+    },
+    {
+      "epoch": 555.48,
+      "learning_rate": 2.4695340501792116e-05,
+      "loss": 0.2166,
+      "step": 86100
+    },
+    {
+      "epoch": 555.55,
+      "learning_rate": 2.4691756272401435e-05,
+      "loss": 0.2876,
+      "step": 86110
+    },
+    {
+      "epoch": 555.61,
+      "learning_rate": 2.4688172043010753e-05,
+      "loss": 0.2768,
+      "step": 86120
+    },
+    {
+      "epoch": 555.68,
+      "learning_rate": 2.468458781362007e-05,
+      "loss": 0.2142,
+      "step": 86130
+    },
+    {
+      "epoch": 555.74,
+      "learning_rate": 2.4681003584229393e-05,
+      "loss": 0.2784,
+      "step": 86140
+    },
+    {
+      "epoch": 555.8,
+      "learning_rate": 2.467741935483871e-05,
+      "loss": 0.2914,
+      "step": 86150
+    },
+    {
+      "epoch": 555.87,
+      "learning_rate": 2.467383512544803e-05,
+      "loss": 0.2349,
+      "step": 86160
+    },
+    {
+      "epoch": 555.93,
+      "learning_rate": 2.4670250896057347e-05,
+      "loss": 0.2299,
+      "step": 86170
+    },
+    {
+      "epoch": 556.0,
+      "learning_rate": 2.466666666666667e-05,
+      "loss": 0.2316,
+      "step": 86180
+    },
+    {
+      "epoch": 556.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9482679963111877,
+      "eval_runtime": 0.902,
+      "eval_samples_per_second": 222.835,
+      "eval_steps_per_second": 7.76,
+      "step": 86180
+    },
+    {
+      "epoch": 556.06,
+      "learning_rate": 2.4663082437275987e-05,
+      "loss": 0.2414,
+      "step": 86190
+    },
+    {
+      "epoch": 556.13,
+      "learning_rate": 2.4659498207885305e-05,
+      "loss": 0.3039,
+      "step": 86200
+    },
+    {
+      "epoch": 556.19,
+      "learning_rate": 2.4655913978494623e-05,
+      "loss": 0.2737,
+      "step": 86210
+    },
+    {
+      "epoch": 556.26,
+      "learning_rate": 2.4652329749103945e-05,
+      "loss": 0.2436,
+      "step": 86220
+    },
+    {
+      "epoch": 556.32,
+      "learning_rate": 2.4648745519713263e-05,
+      "loss": 0.2596,
+      "step": 86230
+    },
+    {
+      "epoch": 556.39,
+      "learning_rate": 2.464516129032258e-05,
+      "loss": 0.2351,
+      "step": 86240
+    },
+    {
+      "epoch": 556.45,
+      "learning_rate": 2.46415770609319e-05,
+      "loss": 0.2661,
+      "step": 86250
+    },
+    {
+      "epoch": 556.51,
+      "learning_rate": 2.463799283154122e-05,
+      "loss": 0.2407,
+      "step": 86260
+    },
+    {
+      "epoch": 556.58,
+      "learning_rate": 2.463440860215054e-05,
+      "loss": 0.2521,
+      "step": 86270
+    },
+    {
+      "epoch": 556.64,
+      "learning_rate": 2.4630824372759858e-05,
+      "loss": 0.2301,
+      "step": 86280
+    },
+    {
+      "epoch": 556.71,
+      "learning_rate": 2.4627240143369176e-05,
+      "loss": 0.2508,
+      "step": 86290
+    },
+    {
+      "epoch": 556.77,
+      "learning_rate": 2.4623655913978497e-05,
+      "loss": 0.1961,
+      "step": 86300
+    },
+    {
+      "epoch": 556.84,
+      "learning_rate": 2.4620071684587816e-05,
+      "loss": 0.2715,
+      "step": 86310
+    },
+    {
+      "epoch": 556.9,
+      "learning_rate": 2.4616487455197134e-05,
+      "loss": 0.265,
+      "step": 86320
+    },
+    {
+      "epoch": 556.96,
+      "learning_rate": 2.4612903225806452e-05,
+      "loss": 0.2463,
+      "step": 86330
+    },
+    {
+      "epoch": 557.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.953129768371582,
+      "eval_runtime": 0.8722,
+      "eval_samples_per_second": 230.456,
+      "eval_steps_per_second": 8.026,
+      "step": 86335
+    },
+    {
+      "epoch": 557.03,
+      "learning_rate": 2.4609318996415774e-05,
+      "loss": 0.2583,
+      "step": 86340
+    },
+    {
+      "epoch": 557.1,
+      "learning_rate": 2.4605734767025092e-05,
+      "loss": 0.2764,
+      "step": 86350
+    },
+    {
+      "epoch": 557.16,
+      "learning_rate": 2.460215053763441e-05,
+      "loss": 0.3013,
+      "step": 86360
+    },
+    {
+      "epoch": 557.23,
+      "learning_rate": 2.4598566308243728e-05,
+      "loss": 0.2218,
+      "step": 86370
+    },
+    {
+      "epoch": 557.29,
+      "learning_rate": 2.459498207885305e-05,
+      "loss": 0.2456,
+      "step": 86380
+    },
+    {
+      "epoch": 557.35,
+      "learning_rate": 2.4591397849462368e-05,
+      "loss": 0.2187,
+      "step": 86390
+    },
+    {
+      "epoch": 557.42,
+      "learning_rate": 2.4587813620071686e-05,
+      "loss": 0.2254,
+      "step": 86400
+    },
+    {
+      "epoch": 557.48,
+      "learning_rate": 2.4584229390681004e-05,
+      "loss": 0.2418,
+      "step": 86410
+    },
+    {
+      "epoch": 557.55,
+      "learning_rate": 2.4580645161290326e-05,
+      "loss": 0.2236,
+      "step": 86420
+    },
+    {
+      "epoch": 557.61,
+      "learning_rate": 2.457706093189964e-05,
+      "loss": 0.2777,
+      "step": 86430
+    },
+    {
+      "epoch": 557.68,
+      "learning_rate": 2.4573476702508962e-05,
+      "loss": 0.295,
+      "step": 86440
+    },
+    {
+      "epoch": 557.74,
+      "learning_rate": 2.456989247311828e-05,
+      "loss": 0.2816,
+      "step": 86450
+    },
+    {
+      "epoch": 557.8,
+      "learning_rate": 2.4566308243727602e-05,
+      "loss": 0.2146,
+      "step": 86460
+    },
+    {
+      "epoch": 557.87,
+      "learning_rate": 2.4562724014336917e-05,
+      "loss": 0.2591,
+      "step": 86470
+    },
+    {
+      "epoch": 557.93,
+      "learning_rate": 2.455913978494624e-05,
+      "loss": 0.3312,
+      "step": 86480
+    },
+    {
+      "epoch": 558.0,
+      "learning_rate": 2.4555555555555557e-05,
+      "loss": 0.2682,
+      "step": 86490
+    },
+    {
+      "epoch": 558.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9768500924110413,
+      "eval_runtime": 0.8472,
+      "eval_samples_per_second": 237.243,
+      "eval_steps_per_second": 8.262,
+      "step": 86490
+    },
+    {
+      "epoch": 558.06,
+      "learning_rate": 2.455197132616488e-05,
+      "loss": 0.2539,
+      "step": 86500
+    },
+    {
+      "epoch": 558.13,
+      "learning_rate": 2.4548387096774193e-05,
+      "loss": 0.256,
+      "step": 86510
+    },
+    {
+      "epoch": 558.19,
+      "learning_rate": 2.4544802867383515e-05,
+      "loss": 0.2339,
+      "step": 86520
+    },
+    {
+      "epoch": 558.26,
+      "learning_rate": 2.4541218637992833e-05,
+      "loss": 0.1943,
+      "step": 86530
+    },
+    {
+      "epoch": 558.32,
+      "learning_rate": 2.453763440860215e-05,
+      "loss": 0.2587,
+      "step": 86540
+    },
+    {
+      "epoch": 558.39,
+      "learning_rate": 2.453405017921147e-05,
+      "loss": 0.2488,
+      "step": 86550
+    },
+    {
+      "epoch": 558.45,
+      "learning_rate": 2.4530465949820788e-05,
+      "loss": 0.2713,
+      "step": 86560
+    },
+    {
+      "epoch": 558.51,
+      "learning_rate": 2.452688172043011e-05,
+      "loss": 0.2167,
+      "step": 86570
+    },
+    {
+      "epoch": 558.58,
+      "learning_rate": 2.4523297491039427e-05,
+      "loss": 0.2279,
+      "step": 86580
+    },
+    {
+      "epoch": 558.64,
+      "learning_rate": 2.4519713261648746e-05,
+      "loss": 0.2665,
+      "step": 86590
+    },
+    {
+      "epoch": 558.71,
+      "learning_rate": 2.4516129032258064e-05,
+      "loss": 0.2456,
+      "step": 86600
+    },
+    {
+      "epoch": 558.77,
+      "learning_rate": 2.4512544802867385e-05,
+      "loss": 0.293,
+      "step": 86610
+    },
+    {
+      "epoch": 558.84,
+      "learning_rate": 2.4508960573476704e-05,
+      "loss": 0.2673,
+      "step": 86620
+    },
+    {
+      "epoch": 558.9,
+      "learning_rate": 2.4505376344086022e-05,
+      "loss": 0.3074,
+      "step": 86630
+    },
+    {
+      "epoch": 558.96,
+      "learning_rate": 2.450179211469534e-05,
+      "loss": 0.2736,
+      "step": 86640
+    },
+    {
+      "epoch": 559.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9772549867630005,
+      "eval_runtime": 0.8654,
+      "eval_samples_per_second": 232.257,
+      "eval_steps_per_second": 8.089,
+      "step": 86645
+    },
+    {
+      "epoch": 559.03,
+      "learning_rate": 2.4498207885304662e-05,
+      "loss": 0.2604,
+      "step": 86650
+    },
+    {
+      "epoch": 559.1,
+      "learning_rate": 2.449462365591398e-05,
+      "loss": 0.2708,
+      "step": 86660
+    },
+    {
+      "epoch": 559.16,
+      "learning_rate": 2.4491039426523298e-05,
+      "loss": 0.2456,
+      "step": 86670
+    },
+    {
+      "epoch": 559.23,
+      "learning_rate": 2.4487455197132616e-05,
+      "loss": 0.231,
+      "step": 86680
+    },
+    {
+      "epoch": 559.29,
+      "learning_rate": 2.4483870967741938e-05,
+      "loss": 0.2595,
+      "step": 86690
+    },
+    {
+      "epoch": 559.35,
+      "learning_rate": 2.4480286738351256e-05,
+      "loss": 0.2528,
+      "step": 86700
+    },
+    {
+      "epoch": 559.42,
+      "learning_rate": 2.4476702508960574e-05,
+      "loss": 0.2463,
+      "step": 86710
+    },
+    {
+      "epoch": 559.48,
+      "learning_rate": 2.4473118279569893e-05,
+      "loss": 0.221,
+      "step": 86720
+    },
+    {
+      "epoch": 559.55,
+      "learning_rate": 2.4469534050179214e-05,
+      "loss": 0.2607,
+      "step": 86730
+    },
+    {
+      "epoch": 559.61,
+      "learning_rate": 2.4465949820788532e-05,
+      "loss": 0.2276,
+      "step": 86740
+    },
+    {
+      "epoch": 559.68,
+      "learning_rate": 2.446236559139785e-05,
+      "loss": 0.2469,
+      "step": 86750
+    },
+    {
+      "epoch": 559.74,
+      "learning_rate": 2.445878136200717e-05,
+      "loss": 0.2386,
+      "step": 86760
+    },
+    {
+      "epoch": 559.8,
+      "learning_rate": 2.445519713261649e-05,
+      "loss": 0.2685,
+      "step": 86770
+    },
+    {
+      "epoch": 559.87,
+      "learning_rate": 2.4451612903225805e-05,
+      "loss": 0.243,
+      "step": 86780
+    },
+    {
+      "epoch": 559.93,
+      "learning_rate": 2.4448028673835127e-05,
+      "loss": 0.2116,
+      "step": 86790
+    },
+    {
+      "epoch": 560.0,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 0.2507,
+      "step": 86800
+    },
+    {
+      "epoch": 560.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9585782289505005,
+      "eval_runtime": 0.8596,
+      "eval_samples_per_second": 233.82,
+      "eval_steps_per_second": 8.143,
+      "step": 86800
+    },
+    {
+      "epoch": 560.06,
+      "learning_rate": 2.4440860215053767e-05,
+      "loss": 0.2351,
+      "step": 86810
+    },
+    {
+      "epoch": 560.13,
+      "learning_rate": 2.443727598566308e-05,
+      "loss": 0.245,
+      "step": 86820
+    },
+    {
+      "epoch": 560.19,
+      "learning_rate": 2.4433691756272403e-05,
+      "loss": 0.265,
+      "step": 86830
+    },
+    {
+      "epoch": 560.26,
+      "learning_rate": 2.443010752688172e-05,
+      "loss": 0.2224,
+      "step": 86840
+    },
+    {
+      "epoch": 560.32,
+      "learning_rate": 2.4426523297491043e-05,
+      "loss": 0.2504,
+      "step": 86850
+    },
+    {
+      "epoch": 560.39,
+      "learning_rate": 2.4422939068100358e-05,
+      "loss": 0.2246,
+      "step": 86860
+    },
+    {
+      "epoch": 560.45,
+      "learning_rate": 2.441935483870968e-05,
+      "loss": 0.2593,
+      "step": 86870
+    },
+    {
+      "epoch": 560.51,
+      "learning_rate": 2.4415770609318997e-05,
+      "loss": 0.2348,
+      "step": 86880
+    },
+    {
+      "epoch": 560.58,
+      "learning_rate": 2.441218637992832e-05,
+      "loss": 0.2333,
+      "step": 86890
+    },
+    {
+      "epoch": 560.64,
+      "learning_rate": 2.4408602150537634e-05,
+      "loss": 0.236,
+      "step": 86900
+    },
+    {
+      "epoch": 560.71,
+      "learning_rate": 2.4405017921146955e-05,
+      "loss": 0.2155,
+      "step": 86910
+    },
+    {
+      "epoch": 560.77,
+      "learning_rate": 2.4401433691756274e-05,
+      "loss": 0.2602,
+      "step": 86920
+    },
+    {
+      "epoch": 560.84,
+      "learning_rate": 2.4397849462365592e-05,
+      "loss": 0.2484,
+      "step": 86930
+    },
+    {
+      "epoch": 560.9,
+      "learning_rate": 2.439426523297491e-05,
+      "loss": 0.233,
+      "step": 86940
+    },
+    {
+      "epoch": 560.96,
+      "learning_rate": 2.439068100358423e-05,
+      "loss": 0.2224,
+      "step": 86950
+    },
+    {
+      "epoch": 561.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9955114722251892,
+      "eval_runtime": 0.8836,
+      "eval_samples_per_second": 227.487,
+      "eval_steps_per_second": 7.922,
+      "step": 86955
+    },
+    {
+      "epoch": 561.03,
+      "learning_rate": 2.438709677419355e-05,
+      "loss": 0.303,
+      "step": 86960
+    },
+    {
+      "epoch": 561.1,
+      "learning_rate": 2.4383512544802868e-05,
+      "loss": 0.2718,
+      "step": 86970
+    },
+    {
+      "epoch": 561.16,
+      "learning_rate": 2.4379928315412186e-05,
+      "loss": 0.2691,
+      "step": 86980
+    },
+    {
+      "epoch": 561.23,
+      "learning_rate": 2.4376344086021508e-05,
+      "loss": 0.2114,
+      "step": 86990
+    },
+    {
+      "epoch": 561.29,
+      "learning_rate": 2.4372759856630826e-05,
+      "loss": 0.2392,
+      "step": 87000
+    },
+    {
+      "epoch": 561.35,
+      "learning_rate": 2.4369175627240144e-05,
+      "loss": 0.248,
+      "step": 87010
+    },
+    {
+      "epoch": 561.42,
+      "learning_rate": 2.4365591397849462e-05,
+      "loss": 0.2909,
+      "step": 87020
+    },
+    {
+      "epoch": 561.48,
+      "learning_rate": 2.4362007168458784e-05,
+      "loss": 0.2586,
+      "step": 87030
+    },
+    {
+      "epoch": 561.55,
+      "learning_rate": 2.4358422939068102e-05,
+      "loss": 0.2594,
+      "step": 87040
+    },
+    {
+      "epoch": 561.61,
+      "learning_rate": 2.435483870967742e-05,
+      "loss": 0.2183,
+      "step": 87050
+    },
+    {
+      "epoch": 561.68,
+      "learning_rate": 2.435125448028674e-05,
+      "loss": 0.2614,
+      "step": 87060
+    },
+    {
+      "epoch": 561.74,
+      "learning_rate": 2.434767025089606e-05,
+      "loss": 0.2347,
+      "step": 87070
+    },
+    {
+      "epoch": 561.8,
+      "learning_rate": 2.434408602150538e-05,
+      "loss": 0.2248,
+      "step": 87080
+    },
+    {
+      "epoch": 561.87,
+      "learning_rate": 2.4340501792114697e-05,
+      "loss": 0.2069,
+      "step": 87090
+    },
+    {
+      "epoch": 561.93,
+      "learning_rate": 2.4336917562724015e-05,
+      "loss": 0.2609,
+      "step": 87100
+    },
+    {
+      "epoch": 562.0,
+      "learning_rate": 2.4333333333333336e-05,
+      "loss": 0.2615,
+      "step": 87110
+    },
+    {
+      "epoch": 562.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9956527352333069,
+      "eval_runtime": 0.8434,
+      "eval_samples_per_second": 238.313,
+      "eval_steps_per_second": 8.299,
+      "step": 87110
+    },
+    {
+      "epoch": 562.06,
+      "learning_rate": 2.4329749103942655e-05,
+      "loss": 0.2692,
+      "step": 87120
+    },
+    {
+      "epoch": 562.13,
+      "learning_rate": 2.4326164874551973e-05,
+      "loss": 0.2413,
+      "step": 87130
+    },
+    {
+      "epoch": 562.19,
+      "learning_rate": 2.432258064516129e-05,
+      "loss": 0.2416,
+      "step": 87140
+    },
+    {
+      "epoch": 562.26,
+      "learning_rate": 2.4318996415770613e-05,
+      "loss": 0.3065,
+      "step": 87150
+    },
+    {
+      "epoch": 562.32,
+      "learning_rate": 2.431541218637993e-05,
+      "loss": 0.2164,
+      "step": 87160
+    },
+    {
+      "epoch": 562.39,
+      "learning_rate": 2.4311827956989246e-05,
+      "loss": 0.2809,
+      "step": 87170
+    },
+    {
+      "epoch": 562.45,
+      "learning_rate": 2.4308243727598567e-05,
+      "loss": 0.2551,
+      "step": 87180
+    },
+    {
+      "epoch": 562.51,
+      "learning_rate": 2.4304659498207885e-05,
+      "loss": 0.2463,
+      "step": 87190
+    },
+    {
+      "epoch": 562.58,
+      "learning_rate": 2.4301075268817207e-05,
+      "loss": 0.227,
+      "step": 87200
+    },
+    {
+      "epoch": 562.64,
+      "learning_rate": 2.4297491039426522e-05,
+      "loss": 0.282,
+      "step": 87210
+    },
+    {
+      "epoch": 562.71,
+      "learning_rate": 2.4293906810035843e-05,
+      "loss": 0.257,
+      "step": 87220
+    },
+    {
+      "epoch": 562.77,
+      "learning_rate": 2.429032258064516e-05,
+      "loss": 0.2641,
+      "step": 87230
+    },
+    {
+      "epoch": 562.84,
+      "learning_rate": 2.4286738351254483e-05,
+      "loss": 0.2623,
+      "step": 87240
+    },
+    {
+      "epoch": 562.9,
+      "learning_rate": 2.4283154121863798e-05,
+      "loss": 0.284,
+      "step": 87250
+    },
+    {
+      "epoch": 562.96,
+      "learning_rate": 2.427956989247312e-05,
+      "loss": 0.274,
+      "step": 87260
+    },
+    {
+      "epoch": 563.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0162380933761597,
+      "eval_runtime": 0.8733,
+      "eval_samples_per_second": 230.166,
+      "eval_steps_per_second": 8.016,
+      "step": 87265
+    },
+    {
+      "epoch": 563.03,
+      "learning_rate": 2.4275985663082438e-05,
+      "loss": 0.2803,
+      "step": 87270
+    },
+    {
+      "epoch": 563.1,
+      "learning_rate": 2.4272401433691756e-05,
+      "loss": 0.2001,
+      "step": 87280
+    },
+    {
+      "epoch": 563.16,
+      "learning_rate": 2.4268817204301074e-05,
+      "loss": 0.2584,
+      "step": 87290
+    },
+    {
+      "epoch": 563.23,
+      "learning_rate": 2.4265232974910396e-05,
+      "loss": 0.2973,
+      "step": 87300
+    },
+    {
+      "epoch": 563.29,
+      "learning_rate": 2.4261648745519714e-05,
+      "loss": 0.2296,
+      "step": 87310
+    },
+    {
+      "epoch": 563.35,
+      "learning_rate": 2.4258064516129032e-05,
+      "loss": 0.243,
+      "step": 87320
+    },
+    {
+      "epoch": 563.42,
+      "learning_rate": 2.425448028673835e-05,
+      "loss": 0.2361,
+      "step": 87330
+    },
+    {
+      "epoch": 563.48,
+      "learning_rate": 2.4250896057347672e-05,
+      "loss": 0.2574,
+      "step": 87340
+    },
+    {
+      "epoch": 563.55,
+      "learning_rate": 2.424731182795699e-05,
+      "loss": 0.2685,
+      "step": 87350
+    },
+    {
+      "epoch": 563.61,
+      "learning_rate": 2.424372759856631e-05,
+      "loss": 0.239,
+      "step": 87360
+    },
+    {
+      "epoch": 563.68,
+      "learning_rate": 2.4240143369175627e-05,
+      "loss": 0.2596,
+      "step": 87370
+    },
+    {
+      "epoch": 563.74,
+      "learning_rate": 2.4236559139784948e-05,
+      "loss": 0.3136,
+      "step": 87380
+    },
+    {
+      "epoch": 563.8,
+      "learning_rate": 2.4232974910394266e-05,
+      "loss": 0.2296,
+      "step": 87390
+    },
+    {
+      "epoch": 563.87,
+      "learning_rate": 2.4229390681003585e-05,
+      "loss": 0.2263,
+      "step": 87400
+    },
+    {
+      "epoch": 563.93,
+      "learning_rate": 2.4225806451612903e-05,
+      "loss": 0.2317,
+      "step": 87410
+    },
+    {
+      "epoch": 564.0,
+      "learning_rate": 2.4222222222222224e-05,
+      "loss": 0.2258,
+      "step": 87420
+    },
+    {
+      "epoch": 564.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.02025306224823,
+      "eval_runtime": 0.8453,
+      "eval_samples_per_second": 237.774,
+      "eval_steps_per_second": 8.281,
+      "step": 87420
+    },
+    {
+      "epoch": 564.06,
+      "learning_rate": 2.4218637992831543e-05,
+      "loss": 0.2684,
+      "step": 87430
+    },
+    {
+      "epoch": 564.13,
+      "learning_rate": 2.421505376344086e-05,
+      "loss": 0.2792,
+      "step": 87440
+    },
+    {
+      "epoch": 564.19,
+      "learning_rate": 2.421146953405018e-05,
+      "loss": 0.2318,
+      "step": 87450
+    },
+    {
+      "epoch": 564.26,
+      "learning_rate": 2.42078853046595e-05,
+      "loss": 0.2919,
+      "step": 87460
+    },
+    {
+      "epoch": 564.32,
+      "learning_rate": 2.420430107526882e-05,
+      "loss": 0.2358,
+      "step": 87470
+    },
+    {
+      "epoch": 564.39,
+      "learning_rate": 2.4200716845878137e-05,
+      "loss": 0.2875,
+      "step": 87480
+    },
+    {
+      "epoch": 564.45,
+      "learning_rate": 2.4197132616487455e-05,
+      "loss": 0.2653,
+      "step": 87490
+    },
+    {
+      "epoch": 564.51,
+      "learning_rate": 2.4193548387096777e-05,
+      "loss": 0.2372,
+      "step": 87500
+    },
+    {
+      "epoch": 564.58,
+      "learning_rate": 2.4189964157706095e-05,
+      "loss": 0.2507,
+      "step": 87510
+    },
+    {
+      "epoch": 564.64,
+      "learning_rate": 2.4186379928315413e-05,
+      "loss": 0.2505,
+      "step": 87520
+    },
+    {
+      "epoch": 564.71,
+      "learning_rate": 2.418279569892473e-05,
+      "loss": 0.2329,
+      "step": 87530
+    },
+    {
+      "epoch": 564.77,
+      "learning_rate": 2.4179211469534053e-05,
+      "loss": 0.2793,
+      "step": 87540
+    },
+    {
+      "epoch": 564.84,
+      "learning_rate": 2.417562724014337e-05,
+      "loss": 0.2808,
+      "step": 87550
+    },
+    {
+      "epoch": 564.9,
+      "learning_rate": 2.417204301075269e-05,
+      "loss": 0.2614,
+      "step": 87560
+    },
+    {
+      "epoch": 564.96,
+      "learning_rate": 2.4168458781362008e-05,
+      "loss": 0.2381,
+      "step": 87570
+    },
+    {
+      "epoch": 565.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0686136484146118,
+      "eval_runtime": 0.8715,
+      "eval_samples_per_second": 230.645,
+      "eval_steps_per_second": 8.032,
+      "step": 87575
+    },
+    {
+      "epoch": 565.03,
+      "learning_rate": 2.416487455197133e-05,
+      "loss": 0.265,
+      "step": 87580
+    },
+    {
+      "epoch": 565.1,
+      "learning_rate": 2.4161290322580648e-05,
+      "loss": 0.236,
+      "step": 87590
+    },
+    {
+      "epoch": 565.16,
+      "learning_rate": 2.4157706093189966e-05,
+      "loss": 0.2633,
+      "step": 87600
+    },
+    {
+      "epoch": 565.23,
+      "learning_rate": 2.4154121863799284e-05,
+      "loss": 0.2723,
+      "step": 87610
+    },
+    {
+      "epoch": 565.29,
+      "learning_rate": 2.4150537634408606e-05,
+      "loss": 0.2458,
+      "step": 87620
+    },
+    {
+      "epoch": 565.35,
+      "learning_rate": 2.414695340501792e-05,
+      "loss": 0.2508,
+      "step": 87630
+    },
+    {
+      "epoch": 565.42,
+      "learning_rate": 2.4143369175627242e-05,
+      "loss": 0.2483,
+      "step": 87640
+    },
+    {
+      "epoch": 565.48,
+      "learning_rate": 2.413978494623656e-05,
+      "loss": 0.2583,
+      "step": 87650
+    },
+    {
+      "epoch": 565.55,
+      "learning_rate": 2.4136200716845882e-05,
+      "loss": 0.2547,
+      "step": 87660
+    },
+    {
+      "epoch": 565.61,
+      "learning_rate": 2.4132616487455197e-05,
+      "loss": 0.2375,
+      "step": 87670
+    },
+    {
+      "epoch": 565.68,
+      "learning_rate": 2.4129032258064518e-05,
+      "loss": 0.2636,
+      "step": 87680
+    },
+    {
+      "epoch": 565.74,
+      "learning_rate": 2.4125448028673836e-05,
+      "loss": 0.2058,
+      "step": 87690
+    },
+    {
+      "epoch": 565.8,
+      "learning_rate": 2.4121863799283158e-05,
+      "loss": 0.2645,
+      "step": 87700
+    },
+    {
+      "epoch": 565.87,
+      "learning_rate": 2.4118279569892473e-05,
+      "loss": 0.3051,
+      "step": 87710
+    },
+    {
+      "epoch": 565.93,
+      "learning_rate": 2.4114695340501794e-05,
+      "loss": 0.2584,
+      "step": 87720
+    },
+    {
+      "epoch": 566.0,
+      "learning_rate": 2.4111111111111113e-05,
+      "loss": 0.2597,
+      "step": 87730
+    },
+    {
+      "epoch": 566.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0300530195236206,
+      "eval_runtime": 0.8683,
+      "eval_samples_per_second": 231.496,
+      "eval_steps_per_second": 8.062,
+      "step": 87730
+    },
+    {
+      "epoch": 566.06,
+      "learning_rate": 2.4107526881720434e-05,
+      "loss": 0.2745,
+      "step": 87740
+    },
+    {
+      "epoch": 566.13,
+      "learning_rate": 2.410394265232975e-05,
+      "loss": 0.2837,
+      "step": 87750
+    },
+    {
+      "epoch": 566.19,
+      "learning_rate": 2.410035842293907e-05,
+      "loss": 0.23,
+      "step": 87760
+    },
+    {
+      "epoch": 566.26,
+      "learning_rate": 2.409677419354839e-05,
+      "loss": 0.2335,
+      "step": 87770
+    },
+    {
+      "epoch": 566.32,
+      "learning_rate": 2.4093189964157707e-05,
+      "loss": 0.2424,
+      "step": 87780
+    },
+    {
+      "epoch": 566.39,
+      "learning_rate": 2.4089605734767025e-05,
+      "loss": 0.2725,
+      "step": 87790
+    },
+    {
+      "epoch": 566.45,
+      "learning_rate": 2.4086021505376347e-05,
+      "loss": 0.27,
+      "step": 87800
+    },
+    {
+      "epoch": 566.51,
+      "learning_rate": 2.4082437275985665e-05,
+      "loss": 0.242,
+      "step": 87810
+    },
+    {
+      "epoch": 566.58,
+      "learning_rate": 2.4078853046594983e-05,
+      "loss": 0.2491,
+      "step": 87820
+    },
+    {
+      "epoch": 566.64,
+      "learning_rate": 2.40752688172043e-05,
+      "loss": 0.2641,
+      "step": 87830
+    },
+    {
+      "epoch": 566.71,
+      "learning_rate": 2.407168458781362e-05,
+      "loss": 0.227,
+      "step": 87840
+    },
+    {
+      "epoch": 566.77,
+      "learning_rate": 2.406810035842294e-05,
+      "loss": 0.2977,
+      "step": 87850
+    },
+    {
+      "epoch": 566.84,
+      "learning_rate": 2.406451612903226e-05,
+      "loss": 0.2684,
+      "step": 87860
+    },
+    {
+      "epoch": 566.9,
+      "learning_rate": 2.4060931899641578e-05,
+      "loss": 0.2502,
+      "step": 87870
+    },
+    {
+      "epoch": 566.96,
+      "learning_rate": 2.4057347670250896e-05,
+      "loss": 0.2702,
+      "step": 87880
+    },
+    {
+      "epoch": 567.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0098071098327637,
+      "eval_runtime": 0.8841,
+      "eval_samples_per_second": 227.35,
+      "eval_steps_per_second": 7.918,
+      "step": 87885
+    },
+    {
+      "epoch": 567.03,
+      "learning_rate": 2.4053763440860217e-05,
+      "loss": 0.2833,
+      "step": 87890
+    },
+    {
+      "epoch": 567.1,
+      "learning_rate": 2.4050179211469536e-05,
+      "loss": 0.2681,
+      "step": 87900
+    },
+    {
+      "epoch": 567.16,
+      "learning_rate": 2.4046594982078854e-05,
+      "loss": 0.2027,
+      "step": 87910
+    },
+    {
+      "epoch": 567.23,
+      "learning_rate": 2.4043010752688172e-05,
+      "loss": 0.2511,
+      "step": 87920
+    },
+    {
+      "epoch": 567.29,
+      "learning_rate": 2.4039426523297494e-05,
+      "loss": 0.2329,
+      "step": 87930
+    },
+    {
+      "epoch": 567.35,
+      "learning_rate": 2.403584229390681e-05,
+      "loss": 0.2414,
+      "step": 87940
+    },
+    {
+      "epoch": 567.42,
+      "learning_rate": 2.403225806451613e-05,
+      "loss": 0.2937,
+      "step": 87950
+    },
+    {
+      "epoch": 567.48,
+      "learning_rate": 2.4028673835125448e-05,
+      "loss": 0.2469,
+      "step": 87960
+    },
+    {
+      "epoch": 567.55,
+      "learning_rate": 2.402508960573477e-05,
+      "loss": 0.2565,
+      "step": 87970
+    },
+    {
+      "epoch": 567.61,
+      "learning_rate": 2.4021505376344085e-05,
+      "loss": 0.2453,
+      "step": 87980
+    },
+    {
+      "epoch": 567.68,
+      "learning_rate": 2.4017921146953406e-05,
+      "loss": 0.2251,
+      "step": 87990
+    },
+    {
+      "epoch": 567.74,
+      "learning_rate": 2.4014336917562724e-05,
+      "loss": 0.2157,
+      "step": 88000
+    },
+    {
+      "epoch": 567.8,
+      "learning_rate": 2.4010752688172046e-05,
+      "loss": 0.2514,
+      "step": 88010
+    },
+    {
+      "epoch": 567.87,
+      "learning_rate": 2.400716845878136e-05,
+      "loss": 0.2138,
+      "step": 88020
+    },
+    {
+      "epoch": 567.93,
+      "learning_rate": 2.4003584229390682e-05,
+      "loss": 0.2433,
+      "step": 88030
+    },
+    {
+      "epoch": 568.0,
+      "learning_rate": 2.4e-05,
+      "loss": 0.2174,
+      "step": 88040
+    },
+    {
+      "epoch": 568.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 1.0045806169509888,
+      "eval_runtime": 0.893,
+      "eval_samples_per_second": 225.072,
+      "eval_steps_per_second": 7.838,
+      "step": 88040
+    },
+    {
+      "epoch": 568.06,
+      "learning_rate": 2.3996415770609322e-05,
+      "loss": 0.2187,
+      "step": 88050
+    },
+    {
+      "epoch": 568.13,
+      "learning_rate": 2.3992831541218637e-05,
+      "loss": 0.2313,
+      "step": 88060
+    },
+    {
+      "epoch": 568.19,
+      "learning_rate": 2.398924731182796e-05,
+      "loss": 0.2533,
+      "step": 88070
+    },
+    {
+      "epoch": 568.26,
+      "learning_rate": 2.3985663082437277e-05,
+      "loss": 0.2648,
+      "step": 88080
+    },
+    {
+      "epoch": 568.32,
+      "learning_rate": 2.39820788530466e-05,
+      "loss": 0.2578,
+      "step": 88090
+    },
+    {
+      "epoch": 568.39,
+      "learning_rate": 2.3978494623655913e-05,
+      "loss": 0.2421,
+      "step": 88100
+    },
+    {
+      "epoch": 568.45,
+      "learning_rate": 2.3974910394265235e-05,
+      "loss": 0.265,
+      "step": 88110
+    },
+    {
+      "epoch": 568.51,
+      "learning_rate": 2.3971326164874553e-05,
+      "loss": 0.2761,
+      "step": 88120
+    },
+    {
+      "epoch": 568.58,
+      "learning_rate": 2.396774193548387e-05,
+      "loss": 0.2163,
+      "step": 88130
+    },
+    {
+      "epoch": 568.64,
+      "learning_rate": 2.396415770609319e-05,
+      "loss": 0.2288,
+      "step": 88140
+    },
+    {
+      "epoch": 568.71,
+      "learning_rate": 2.396057347670251e-05,
+      "loss": 0.2617,
+      "step": 88150
+    },
+    {
+      "epoch": 568.77,
+      "learning_rate": 2.395698924731183e-05,
+      "loss": 0.2685,
+      "step": 88160
+    },
+    {
+      "epoch": 568.84,
+      "learning_rate": 2.3953405017921147e-05,
+      "loss": 0.2068,
+      "step": 88170
+    },
+    {
+      "epoch": 568.9,
+      "learning_rate": 2.3949820788530466e-05,
+      "loss": 0.2648,
+      "step": 88180
+    },
+    {
+      "epoch": 568.96,
+      "learning_rate": 2.3946236559139787e-05,
+      "loss": 0.2683,
+      "step": 88190
+    },
+    {
+      "epoch": 569.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9824224710464478,
+      "eval_runtime": 0.8458,
+      "eval_samples_per_second": 237.632,
+      "eval_steps_per_second": 8.276,
+      "step": 88195
+    },
+    {
+      "epoch": 569.03,
+      "learning_rate": 2.3942652329749105e-05,
+      "loss": 0.291,
+      "step": 88200
+    },
+    {
+      "epoch": 569.1,
+      "learning_rate": 2.3939068100358424e-05,
+      "loss": 0.2689,
+      "step": 88210
+    },
+    {
+      "epoch": 569.16,
+      "learning_rate": 2.3935483870967742e-05,
+      "loss": 0.2212,
+      "step": 88220
+    },
+    {
+      "epoch": 569.23,
+      "learning_rate": 2.3931899641577063e-05,
+      "loss": 0.2589,
+      "step": 88230
+    },
+    {
+      "epoch": 569.29,
+      "learning_rate": 2.392831541218638e-05,
+      "loss": 0.2358,
+      "step": 88240
+    },
+    {
+      "epoch": 569.35,
+      "learning_rate": 2.39247311827957e-05,
+      "loss": 0.2463,
+      "step": 88250
+    },
+    {
+      "epoch": 569.42,
+      "learning_rate": 2.3921146953405018e-05,
+      "loss": 0.2194,
+      "step": 88260
+    },
+    {
+      "epoch": 569.48,
+      "learning_rate": 2.391756272401434e-05,
+      "loss": 0.2309,
+      "step": 88270
+    },
+    {
+      "epoch": 569.55,
+      "learning_rate": 2.3913978494623658e-05,
+      "loss": 0.1892,
+      "step": 88280
+    },
+    {
+      "epoch": 569.61,
+      "learning_rate": 2.3910394265232976e-05,
+      "loss": 0.2497,
+      "step": 88290
+    },
+    {
+      "epoch": 569.68,
+      "learning_rate": 2.3906810035842294e-05,
+      "loss": 0.2426,
+      "step": 88300
+    },
+    {
+      "epoch": 569.74,
+      "learning_rate": 2.3903225806451616e-05,
+      "loss": 0.1998,
+      "step": 88310
+    },
+    {
+      "epoch": 569.8,
+      "learning_rate": 2.3899641577060934e-05,
+      "loss": 0.2523,
+      "step": 88320
+    },
+    {
+      "epoch": 569.87,
+      "learning_rate": 2.3896057347670252e-05,
+      "loss": 0.2801,
+      "step": 88330
+    },
+    {
+      "epoch": 569.93,
+      "learning_rate": 2.389247311827957e-05,
+      "loss": 0.1957,
+      "step": 88340
+    },
+    {
+      "epoch": 570.0,
+      "learning_rate": 2.3888888888888892e-05,
+      "loss": 0.2919,
+      "step": 88350
+    },
+    {
+      "epoch": 570.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9903680682182312,
+      "eval_runtime": 0.8955,
+      "eval_samples_per_second": 224.456,
+      "eval_steps_per_second": 7.817,
+      "step": 88350
+    },
+    {
+      "epoch": 570.06,
+      "learning_rate": 2.388530465949821e-05,
+      "loss": 0.2322,
+      "step": 88360
+    },
+    {
+      "epoch": 570.13,
+      "learning_rate": 2.388172043010753e-05,
+      "loss": 0.2164,
+      "step": 88370
+    },
+    {
+      "epoch": 570.19,
+      "learning_rate": 2.3878136200716847e-05,
+      "loss": 0.2518,
+      "step": 88380
+    },
+    {
+      "epoch": 570.26,
+      "learning_rate": 2.3874551971326168e-05,
+      "loss": 0.239,
+      "step": 88390
+    },
+    {
+      "epoch": 570.32,
+      "learning_rate": 2.3870967741935486e-05,
+      "loss": 0.207,
+      "step": 88400
+    },
+    {
+      "epoch": 570.39,
+      "learning_rate": 2.3867383512544805e-05,
+      "loss": 0.2466,
+      "step": 88410
+    },
+    {
+      "epoch": 570.45,
+      "learning_rate": 2.3863799283154123e-05,
+      "loss": 0.2535,
+      "step": 88420
+    },
+    {
+      "epoch": 570.51,
+      "learning_rate": 2.3860215053763444e-05,
+      "loss": 0.2191,
+      "step": 88430
+    },
+    {
+      "epoch": 570.58,
+      "learning_rate": 2.385663082437276e-05,
+      "loss": 0.2306,
+      "step": 88440
+    },
+    {
+      "epoch": 570.64,
+      "learning_rate": 2.3853046594982078e-05,
+      "loss": 0.2365,
+      "step": 88450
+    },
+    {
+      "epoch": 570.71,
+      "learning_rate": 2.38494623655914e-05,
+      "loss": 0.2432,
+      "step": 88460
+    },
+    {
+      "epoch": 570.77,
+      "learning_rate": 2.3845878136200717e-05,
+      "loss": 0.2255,
+      "step": 88470
+    },
+    {
+      "epoch": 570.84,
+      "learning_rate": 2.3842293906810036e-05,
+      "loss": 0.2226,
+      "step": 88480
+    },
+    {
+      "epoch": 570.9,
+      "learning_rate": 2.3838709677419354e-05,
+      "loss": 0.2506,
+      "step": 88490
+    },
+    {
+      "epoch": 570.96,
+      "learning_rate": 2.3835125448028675e-05,
+      "loss": 0.2829,
+      "step": 88500
+    },
+    {
+      "epoch": 571.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9264712333679199,
+      "eval_runtime": 0.8559,
+      "eval_samples_per_second": 234.828,
+      "eval_steps_per_second": 8.178,
+      "step": 88505
+    },
+    {
+      "epoch": 571.03,
+      "learning_rate": 2.3831541218637994e-05,
+      "loss": 0.2388,
+      "step": 88510
+    },
+    {
+      "epoch": 571.1,
+      "learning_rate": 2.3827956989247312e-05,
+      "loss": 0.2763,
+      "step": 88520
+    },
+    {
+      "epoch": 571.16,
+      "learning_rate": 2.382437275985663e-05,
+      "loss": 0.2162,
+      "step": 88530
+    },
+    {
+      "epoch": 571.23,
+      "learning_rate": 2.382078853046595e-05,
+      "loss": 0.2668,
+      "step": 88540
+    },
+    {
+      "epoch": 571.29,
+      "learning_rate": 2.381720430107527e-05,
+      "loss": 0.214,
+      "step": 88550
+    },
+    {
+      "epoch": 571.35,
+      "learning_rate": 2.3813620071684588e-05,
+      "loss": 0.253,
+      "step": 88560
+    },
+    {
+      "epoch": 571.42,
+      "learning_rate": 2.3810035842293906e-05,
+      "loss": 0.2604,
+      "step": 88570
+    },
+    {
+      "epoch": 571.48,
+      "learning_rate": 2.3806451612903228e-05,
+      "loss": 0.2139,
+      "step": 88580
+    },
+    {
+      "epoch": 571.55,
+      "learning_rate": 2.3802867383512546e-05,
+      "loss": 0.2074,
+      "step": 88590
+    },
+    {
+      "epoch": 571.61,
+      "learning_rate": 2.3799283154121864e-05,
+      "loss": 0.2256,
+      "step": 88600
+    },
+    {
+      "epoch": 571.68,
+      "learning_rate": 2.3795698924731182e-05,
+      "loss": 0.2705,
+      "step": 88610
+    },
+    {
+      "epoch": 571.74,
+      "learning_rate": 2.3792114695340504e-05,
+      "loss": 0.2218,
+      "step": 88620
+    },
+    {
+      "epoch": 571.8,
+      "learning_rate": 2.3788530465949822e-05,
+      "loss": 0.2762,
+      "step": 88630
+    },
+    {
+      "epoch": 571.87,
+      "learning_rate": 2.378494623655914e-05,
+      "loss": 0.2598,
+      "step": 88640
+    },
+    {
+      "epoch": 571.93,
+      "learning_rate": 2.378136200716846e-05,
+      "loss": 0.2138,
+      "step": 88650
+    },
+    {
+      "epoch": 572.0,
+      "learning_rate": 2.377777777777778e-05,
+      "loss": 0.269,
+      "step": 88660
+    },
+    {
+      "epoch": 572.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9498698711395264,
+      "eval_runtime": 0.8853,
+      "eval_samples_per_second": 227.047,
+      "eval_steps_per_second": 7.907,
+      "step": 88660
+    },
+    {
+      "epoch": 572.06,
+      "learning_rate": 2.37741935483871e-05,
+      "loss": 0.2052,
+      "step": 88670
+    },
+    {
+      "epoch": 572.13,
+      "learning_rate": 2.3770609318996417e-05,
+      "loss": 0.2826,
+      "step": 88680
+    },
+    {
+      "epoch": 572.19,
+      "learning_rate": 2.3767025089605735e-05,
+      "loss": 0.2338,
+      "step": 88690
+    },
+    {
+      "epoch": 572.26,
+      "learning_rate": 2.3763440860215056e-05,
+      "loss": 0.2484,
+      "step": 88700
+    },
+    {
+      "epoch": 572.32,
+      "learning_rate": 2.3759856630824375e-05,
+      "loss": 0.2594,
+      "step": 88710
+    },
+    {
+      "epoch": 572.39,
+      "learning_rate": 2.3756272401433693e-05,
+      "loss": 0.2257,
+      "step": 88720
+    },
+    {
+      "epoch": 572.45,
+      "learning_rate": 2.375268817204301e-05,
+      "loss": 0.2142,
+      "step": 88730
+    },
+    {
+      "epoch": 572.51,
+      "learning_rate": 2.3749103942652333e-05,
+      "loss": 0.2396,
+      "step": 88740
+    },
+    {
+      "epoch": 572.58,
+      "learning_rate": 2.374551971326165e-05,
+      "loss": 0.2542,
+      "step": 88750
+    },
+    {
+      "epoch": 572.64,
+      "learning_rate": 2.374193548387097e-05,
+      "loss": 0.2618,
+      "step": 88760
+    },
+    {
+      "epoch": 572.71,
+      "learning_rate": 2.3738351254480287e-05,
+      "loss": 0.218,
+      "step": 88770
+    },
+    {
+      "epoch": 572.77,
+      "learning_rate": 2.373476702508961e-05,
+      "loss": 0.2505,
+      "step": 88780
+    },
+    {
+      "epoch": 572.84,
+      "learning_rate": 2.3731182795698924e-05,
+      "loss": 0.2416,
+      "step": 88790
+    },
+    {
+      "epoch": 572.9,
+      "learning_rate": 2.3727598566308245e-05,
+      "loss": 0.2128,
+      "step": 88800
+    },
+    {
+      "epoch": 572.96,
+      "learning_rate": 2.3724014336917563e-05,
+      "loss": 0.2568,
+      "step": 88810
+    },
+    {
+      "epoch": 573.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9398232698440552,
+      "eval_runtime": 0.8836,
+      "eval_samples_per_second": 227.484,
+      "eval_steps_per_second": 7.922,
+      "step": 88815
+    },
+    {
+      "epoch": 573.03,
+      "learning_rate": 2.3720430107526885e-05,
+      "loss": 0.2407,
+      "step": 88820
+    },
+    {
+      "epoch": 573.1,
+      "learning_rate": 2.37168458781362e-05,
+      "loss": 0.2841,
+      "step": 88830
+    },
+    {
+      "epoch": 573.16,
+      "learning_rate": 2.371326164874552e-05,
+      "loss": 0.2715,
+      "step": 88840
+    },
+    {
+      "epoch": 573.23,
+      "learning_rate": 2.370967741935484e-05,
+      "loss": 0.2367,
+      "step": 88850
+    },
+    {
+      "epoch": 573.29,
+      "learning_rate": 2.370609318996416e-05,
+      "loss": 0.2148,
+      "step": 88860
+    },
+    {
+      "epoch": 573.35,
+      "learning_rate": 2.3702508960573476e-05,
+      "loss": 0.2392,
+      "step": 88870
+    },
+    {
+      "epoch": 573.42,
+      "learning_rate": 2.3698924731182798e-05,
+      "loss": 0.2762,
+      "step": 88880
+    },
+    {
+      "epoch": 573.48,
+      "learning_rate": 2.3695340501792116e-05,
+      "loss": 0.2255,
+      "step": 88890
+    },
+    {
+      "epoch": 573.55,
+      "learning_rate": 2.3691756272401437e-05,
+      "loss": 0.2107,
+      "step": 88900
+    },
+    {
+      "epoch": 573.61,
+      "learning_rate": 2.3688172043010752e-05,
+      "loss": 0.3216,
+      "step": 88910
+    },
+    {
+      "epoch": 573.68,
+      "learning_rate": 2.3684587813620074e-05,
+      "loss": 0.277,
+      "step": 88920
+    },
+    {
+      "epoch": 573.74,
+      "learning_rate": 2.3681003584229392e-05,
+      "loss": 0.2623,
+      "step": 88930
+    },
+    {
+      "epoch": 573.8,
+      "learning_rate": 2.367741935483871e-05,
+      "loss": 0.2519,
+      "step": 88940
+    },
+    {
+      "epoch": 573.87,
+      "learning_rate": 2.367383512544803e-05,
+      "loss": 0.2539,
+      "step": 88950
+    },
+    {
+      "epoch": 573.93,
+      "learning_rate": 2.367025089605735e-05,
+      "loss": 0.2676,
+      "step": 88960
+    },
+    {
+      "epoch": 574.0,
+      "learning_rate": 2.3666666666666668e-05,
+      "loss": 0.2424,
+      "step": 88970
+    },
+    {
+      "epoch": 574.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0076059103012085,
+      "eval_runtime": 0.8716,
+      "eval_samples_per_second": 230.602,
+      "eval_steps_per_second": 8.031,
+      "step": 88970
+    },
+    {
+      "epoch": 574.06,
+      "learning_rate": 2.3663082437275986e-05,
+      "loss": 0.2385,
+      "step": 88980
+    },
+    {
+      "epoch": 574.13,
+      "learning_rate": 2.3659498207885305e-05,
+      "loss": 0.2819,
+      "step": 88990
+    },
+    {
+      "epoch": 574.19,
+      "learning_rate": 2.3655913978494626e-05,
+      "loss": 0.2371,
+      "step": 89000
+    },
+    {
+      "epoch": 574.26,
+      "learning_rate": 2.3652329749103944e-05,
+      "loss": 0.2641,
+      "step": 89010
+    },
+    {
+      "epoch": 574.32,
+      "learning_rate": 2.3648745519713263e-05,
+      "loss": 0.2389,
+      "step": 89020
+    },
+    {
+      "epoch": 574.39,
+      "learning_rate": 2.364516129032258e-05,
+      "loss": 0.2013,
+      "step": 89030
+    },
+    {
+      "epoch": 574.45,
+      "learning_rate": 2.3641577060931902e-05,
+      "loss": 0.2202,
+      "step": 89040
+    },
+    {
+      "epoch": 574.51,
+      "learning_rate": 2.363799283154122e-05,
+      "loss": 0.2333,
+      "step": 89050
+    },
+    {
+      "epoch": 574.58,
+      "learning_rate": 2.363440860215054e-05,
+      "loss": 0.272,
+      "step": 89060
+    },
+    {
+      "epoch": 574.64,
+      "learning_rate": 2.3630824372759857e-05,
+      "loss": 0.2627,
+      "step": 89070
+    },
+    {
+      "epoch": 574.71,
+      "learning_rate": 2.362724014336918e-05,
+      "loss": 0.2924,
+      "step": 89080
+    },
+    {
+      "epoch": 574.77,
+      "learning_rate": 2.3623655913978497e-05,
+      "loss": 0.2202,
+      "step": 89090
+    },
+    {
+      "epoch": 574.84,
+      "learning_rate": 2.3620071684587815e-05,
+      "loss": 0.279,
+      "step": 89100
+    },
+    {
+      "epoch": 574.9,
+      "learning_rate": 2.3616487455197133e-05,
+      "loss": 0.2732,
+      "step": 89110
+    },
+    {
+      "epoch": 574.96,
+      "learning_rate": 2.361290322580645e-05,
+      "loss": 0.2648,
+      "step": 89120
+    },
+    {
+      "epoch": 575.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0090458393096924,
+      "eval_runtime": 0.8719,
+      "eval_samples_per_second": 230.532,
+      "eval_steps_per_second": 8.028,
+      "step": 89125
+    },
+    {
+      "epoch": 575.03,
+      "learning_rate": 2.3609318996415773e-05,
+      "loss": 0.2412,
+      "step": 89130
+    },
+    {
+      "epoch": 575.1,
+      "learning_rate": 2.3605734767025088e-05,
+      "loss": 0.2608,
+      "step": 89140
+    },
+    {
+      "epoch": 575.16,
+      "learning_rate": 2.360215053763441e-05,
+      "loss": 0.2473,
+      "step": 89150
+    },
+    {
+      "epoch": 575.23,
+      "learning_rate": 2.3598566308243728e-05,
+      "loss": 0.2369,
+      "step": 89160
+    },
+    {
+      "epoch": 575.29,
+      "learning_rate": 2.359498207885305e-05,
+      "loss": 0.2912,
+      "step": 89170
+    },
+    {
+      "epoch": 575.35,
+      "learning_rate": 2.3591397849462364e-05,
+      "loss": 0.2304,
+      "step": 89180
+    },
+    {
+      "epoch": 575.42,
+      "learning_rate": 2.3587813620071686e-05,
+      "loss": 0.2542,
+      "step": 89190
+    },
+    {
+      "epoch": 575.48,
+      "learning_rate": 2.3584229390681004e-05,
+      "loss": 0.2341,
+      "step": 89200
+    },
+    {
+      "epoch": 575.55,
+      "learning_rate": 2.3580645161290325e-05,
+      "loss": 0.2542,
+      "step": 89210
+    },
+    {
+      "epoch": 575.61,
+      "learning_rate": 2.357706093189964e-05,
+      "loss": 0.2311,
+      "step": 89220
+    },
+    {
+      "epoch": 575.68,
+      "learning_rate": 2.3573476702508962e-05,
+      "loss": 0.21,
+      "step": 89230
+    },
+    {
+      "epoch": 575.74,
+      "learning_rate": 2.356989247311828e-05,
+      "loss": 0.2263,
+      "step": 89240
+    },
+    {
+      "epoch": 575.8,
+      "learning_rate": 2.35663082437276e-05,
+      "loss": 0.2398,
+      "step": 89250
+    },
+    {
+      "epoch": 575.87,
+      "learning_rate": 2.3562724014336916e-05,
+      "loss": 0.2621,
+      "step": 89260
+    },
+    {
+      "epoch": 575.93,
+      "learning_rate": 2.3559139784946238e-05,
+      "loss": 0.2227,
+      "step": 89270
+    },
+    {
+      "epoch": 576.0,
+      "learning_rate": 2.3555555555555556e-05,
+      "loss": 0.1955,
+      "step": 89280
+    },
+    {
+      "epoch": 576.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0228774547576904,
+      "eval_runtime": 0.8726,
+      "eval_samples_per_second": 230.336,
+      "eval_steps_per_second": 8.022,
+      "step": 89280
+    },
+    {
+      "epoch": 576.06,
+      "learning_rate": 2.3551971326164874e-05,
+      "loss": 0.2279,
+      "step": 89290
+    },
+    {
+      "epoch": 576.13,
+      "learning_rate": 2.3548387096774193e-05,
+      "loss": 0.2449,
+      "step": 89300
+    },
+    {
+      "epoch": 576.19,
+      "learning_rate": 2.3544802867383514e-05,
+      "loss": 0.2527,
+      "step": 89310
+    },
+    {
+      "epoch": 576.26,
+      "learning_rate": 2.3541218637992832e-05,
+      "loss": 0.2382,
+      "step": 89320
+    },
+    {
+      "epoch": 576.32,
+      "learning_rate": 2.353763440860215e-05,
+      "loss": 0.2925,
+      "step": 89330
+    },
+    {
+      "epoch": 576.39,
+      "learning_rate": 2.353405017921147e-05,
+      "loss": 0.2283,
+      "step": 89340
+    },
+    {
+      "epoch": 576.45,
+      "learning_rate": 2.353046594982079e-05,
+      "loss": 0.2442,
+      "step": 89350
+    },
+    {
+      "epoch": 576.51,
+      "learning_rate": 2.352688172043011e-05,
+      "loss": 0.2723,
+      "step": 89360
+    },
+    {
+      "epoch": 576.58,
+      "learning_rate": 2.3523297491039427e-05,
+      "loss": 0.2686,
+      "step": 89370
+    },
+    {
+      "epoch": 576.64,
+      "learning_rate": 2.3519713261648745e-05,
+      "loss": 0.2738,
+      "step": 89380
+    },
+    {
+      "epoch": 576.71,
+      "learning_rate": 2.3516129032258067e-05,
+      "loss": 0.2737,
+      "step": 89390
+    },
+    {
+      "epoch": 576.77,
+      "learning_rate": 2.3512544802867385e-05,
+      "loss": 0.2536,
+      "step": 89400
+    },
+    {
+      "epoch": 576.84,
+      "learning_rate": 2.3508960573476703e-05,
+      "loss": 0.2638,
+      "step": 89410
+    },
+    {
+      "epoch": 576.9,
+      "learning_rate": 2.350537634408602e-05,
+      "loss": 0.1837,
+      "step": 89420
+    },
+    {
+      "epoch": 576.96,
+      "learning_rate": 2.3501792114695343e-05,
+      "loss": 0.2717,
+      "step": 89430
+    },
+    {
+      "epoch": 577.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0373395681381226,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 231.253,
+      "eval_steps_per_second": 8.054,
+      "step": 89435
+    },
+    {
+      "epoch": 577.03,
+      "learning_rate": 2.349820788530466e-05,
+      "loss": 0.2736,
+      "step": 89440
+    },
+    {
+      "epoch": 577.1,
+      "learning_rate": 2.349462365591398e-05,
+      "loss": 0.2169,
+      "step": 89450
+    },
+    {
+      "epoch": 577.16,
+      "learning_rate": 2.3491039426523298e-05,
+      "loss": 0.2472,
+      "step": 89460
+    },
+    {
+      "epoch": 577.23,
+      "learning_rate": 2.348745519713262e-05,
+      "loss": 0.2833,
+      "step": 89470
+    },
+    {
+      "epoch": 577.29,
+      "learning_rate": 2.3483870967741937e-05,
+      "loss": 0.2843,
+      "step": 89480
+    },
+    {
+      "epoch": 577.35,
+      "learning_rate": 2.3480286738351256e-05,
+      "loss": 0.2733,
+      "step": 89490
+    },
+    {
+      "epoch": 577.42,
+      "learning_rate": 2.3476702508960574e-05,
+      "loss": 0.2628,
+      "step": 89500
+    },
+    {
+      "epoch": 577.48,
+      "learning_rate": 2.3473118279569895e-05,
+      "loss": 0.2919,
+      "step": 89510
+    },
+    {
+      "epoch": 577.55,
+      "learning_rate": 2.3469534050179214e-05,
+      "loss": 0.2436,
+      "step": 89520
+    },
+    {
+      "epoch": 577.61,
+      "learning_rate": 2.3465949820788532e-05,
+      "loss": 0.228,
+      "step": 89530
+    },
+    {
+      "epoch": 577.68,
+      "learning_rate": 2.346236559139785e-05,
+      "loss": 0.2755,
+      "step": 89540
+    },
+    {
+      "epoch": 577.74,
+      "learning_rate": 2.345878136200717e-05,
+      "loss": 0.2343,
+      "step": 89550
+    },
+    {
+      "epoch": 577.8,
+      "learning_rate": 2.345519713261649e-05,
+      "loss": 0.3081,
+      "step": 89560
+    },
+    {
+      "epoch": 577.87,
+      "learning_rate": 2.3451612903225808e-05,
+      "loss": 0.2352,
+      "step": 89570
+    },
+    {
+      "epoch": 577.93,
+      "learning_rate": 2.3448028673835126e-05,
+      "loss": 0.2707,
+      "step": 89580
+    },
+    {
+      "epoch": 578.0,
+      "learning_rate": 2.3444444444444448e-05,
+      "loss": 0.2231,
+      "step": 89590
+    },
+    {
+      "epoch": 578.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.036177396774292,
+      "eval_runtime": 0.8966,
+      "eval_samples_per_second": 224.17,
+      "eval_steps_per_second": 7.807,
+      "step": 89590
+    },
+    {
+      "epoch": 578.06,
+      "learning_rate": 2.3440860215053766e-05,
+      "loss": 0.235,
+      "step": 89600
+    },
+    {
+      "epoch": 578.13,
+      "learning_rate": 2.3437275985663084e-05,
+      "loss": 0.2687,
+      "step": 89610
+    },
+    {
+      "epoch": 578.19,
+      "learning_rate": 2.3433691756272402e-05,
+      "loss": 0.2574,
+      "step": 89620
+    },
+    {
+      "epoch": 578.26,
+      "learning_rate": 2.3430107526881724e-05,
+      "loss": 0.2603,
+      "step": 89630
+    },
+    {
+      "epoch": 578.32,
+      "learning_rate": 2.342652329749104e-05,
+      "loss": 0.2206,
+      "step": 89640
+    },
+    {
+      "epoch": 578.39,
+      "learning_rate": 2.342293906810036e-05,
+      "loss": 0.2173,
+      "step": 89650
+    },
+    {
+      "epoch": 578.45,
+      "learning_rate": 2.341935483870968e-05,
+      "loss": 0.2556,
+      "step": 89660
+    },
+    {
+      "epoch": 578.51,
+      "learning_rate": 2.3415770609319e-05,
+      "loss": 0.2328,
+      "step": 89670
+    },
+    {
+      "epoch": 578.58,
+      "learning_rate": 2.3412186379928315e-05,
+      "loss": 0.2798,
+      "step": 89680
+    },
+    {
+      "epoch": 578.64,
+      "learning_rate": 2.3408602150537637e-05,
+      "loss": 0.2688,
+      "step": 89690
+    },
+    {
+      "epoch": 578.71,
+      "learning_rate": 2.3405017921146955e-05,
+      "loss": 0.2046,
+      "step": 89700
+    },
+    {
+      "epoch": 578.77,
+      "learning_rate": 2.3401433691756276e-05,
+      "loss": 0.2387,
+      "step": 89710
+    },
+    {
+      "epoch": 578.84,
+      "learning_rate": 2.339784946236559e-05,
+      "loss": 0.2742,
+      "step": 89720
+    },
+    {
+      "epoch": 578.9,
+      "learning_rate": 2.339426523297491e-05,
+      "loss": 0.2116,
+      "step": 89730
+    },
+    {
+      "epoch": 578.96,
+      "learning_rate": 2.339068100358423e-05,
+      "loss": 0.2334,
+      "step": 89740
+    },
+    {
+      "epoch": 579.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.098311185836792,
+      "eval_runtime": 0.9023,
+      "eval_samples_per_second": 222.771,
+      "eval_steps_per_second": 7.758,
+      "step": 89745
+    },
+    {
+      "epoch": 579.03,
+      "learning_rate": 2.338709677419355e-05,
+      "loss": 0.2553,
+      "step": 89750
+    },
+    {
+      "epoch": 579.1,
+      "learning_rate": 2.3383512544802867e-05,
+      "loss": 0.2951,
+      "step": 89760
+    },
+    {
+      "epoch": 579.16,
+      "learning_rate": 2.3379928315412186e-05,
+      "loss": 0.268,
+      "step": 89770
+    },
+    {
+      "epoch": 579.23,
+      "learning_rate": 2.3376344086021507e-05,
+      "loss": 0.2396,
+      "step": 89780
+    },
+    {
+      "epoch": 579.29,
+      "learning_rate": 2.3372759856630825e-05,
+      "loss": 0.2324,
+      "step": 89790
+    },
+    {
+      "epoch": 579.35,
+      "learning_rate": 2.3369175627240144e-05,
+      "loss": 0.2752,
+      "step": 89800
+    },
+    {
+      "epoch": 579.42,
+      "learning_rate": 2.3365591397849462e-05,
+      "loss": 0.2563,
+      "step": 89810
+    },
+    {
+      "epoch": 579.48,
+      "learning_rate": 2.3362007168458783e-05,
+      "loss": 0.2819,
+      "step": 89820
+    },
+    {
+      "epoch": 579.55,
+      "learning_rate": 2.33584229390681e-05,
+      "loss": 0.2527,
+      "step": 89830
+    },
+    {
+      "epoch": 579.61,
+      "learning_rate": 2.335483870967742e-05,
+      "loss": 0.2379,
+      "step": 89840
+    },
+    {
+      "epoch": 579.68,
+      "learning_rate": 2.3351254480286738e-05,
+      "loss": 0.2217,
+      "step": 89850
+    },
+    {
+      "epoch": 579.74,
+      "learning_rate": 2.334767025089606e-05,
+      "loss": 0.214,
+      "step": 89860
+    },
+    {
+      "epoch": 579.8,
+      "learning_rate": 2.3344086021505378e-05,
+      "loss": 0.183,
+      "step": 89870
+    },
+    {
+      "epoch": 579.87,
+      "learning_rate": 2.3340501792114696e-05,
+      "loss": 0.2216,
+      "step": 89880
+    },
+    {
+      "epoch": 579.93,
+      "learning_rate": 2.3336917562724014e-05,
+      "loss": 0.2343,
+      "step": 89890
+    },
+    {
+      "epoch": 580.0,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 0.2563,
+      "step": 89900
+    },
+    {
+      "epoch": 580.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0996345281600952,
+      "eval_runtime": 0.8509,
+      "eval_samples_per_second": 236.214,
+      "eval_steps_per_second": 8.226,
+      "step": 89900
+    },
+    {
+      "epoch": 580.06,
+      "learning_rate": 2.3329749103942654e-05,
+      "loss": 0.2875,
+      "step": 89910
+    },
+    {
+      "epoch": 580.13,
+      "learning_rate": 2.3326164874551972e-05,
+      "loss": 0.2548,
+      "step": 89920
+    },
+    {
+      "epoch": 580.19,
+      "learning_rate": 2.332258064516129e-05,
+      "loss": 0.2392,
+      "step": 89930
+    },
+    {
+      "epoch": 580.26,
+      "learning_rate": 2.3318996415770612e-05,
+      "loss": 0.2173,
+      "step": 89940
+    },
+    {
+      "epoch": 580.32,
+      "learning_rate": 2.3315412186379927e-05,
+      "loss": 0.2265,
+      "step": 89950
+    },
+    {
+      "epoch": 580.39,
+      "learning_rate": 2.331182795698925e-05,
+      "loss": 0.2407,
+      "step": 89960
+    },
+    {
+      "epoch": 580.45,
+      "learning_rate": 2.3308243727598567e-05,
+      "loss": 0.2535,
+      "step": 89970
+    },
+    {
+      "epoch": 580.51,
+      "learning_rate": 2.3304659498207888e-05,
+      "loss": 0.2837,
+      "step": 89980
+    },
+    {
+      "epoch": 580.58,
+      "learning_rate": 2.3301075268817203e-05,
+      "loss": 0.2695,
+      "step": 89990
+    },
+    {
+      "epoch": 580.64,
+      "learning_rate": 2.3297491039426525e-05,
+      "loss": 0.1967,
+      "step": 90000
+    },
+    {
+      "epoch": 580.71,
+      "learning_rate": 2.3293906810035843e-05,
+      "loss": 0.249,
+      "step": 90010
+    },
+    {
+      "epoch": 580.77,
+      "learning_rate": 2.3290322580645164e-05,
+      "loss": 0.235,
+      "step": 90020
+    },
+    {
+      "epoch": 580.84,
+      "learning_rate": 2.328673835125448e-05,
+      "loss": 0.2358,
+      "step": 90030
+    },
+    {
+      "epoch": 580.9,
+      "learning_rate": 2.32831541218638e-05,
+      "loss": 0.269,
+      "step": 90040
+    },
+    {
+      "epoch": 580.96,
+      "learning_rate": 2.327956989247312e-05,
+      "loss": 0.2658,
+      "step": 90050
+    },
+    {
+      "epoch": 581.0,
+      "eval_accuracy": 0.7313432835820896,
+      "eval_loss": 1.0845876932144165,
+      "eval_runtime": 0.8634,
+      "eval_samples_per_second": 232.804,
+      "eval_steps_per_second": 8.108,
+      "step": 90055
+    },
+    {
+      "epoch": 581.03,
+      "learning_rate": 2.327598566308244e-05,
+      "loss": 0.2724,
+      "step": 90060
+    },
+    {
+      "epoch": 581.1,
+      "learning_rate": 2.3272401433691755e-05,
+      "loss": 0.2367,
+      "step": 90070
+    },
+    {
+      "epoch": 581.16,
+      "learning_rate": 2.3268817204301077e-05,
+      "loss": 0.2266,
+      "step": 90080
+    },
+    {
+      "epoch": 581.23,
+      "learning_rate": 2.3265232974910395e-05,
+      "loss": 0.2539,
+      "step": 90090
+    },
+    {
+      "epoch": 581.29,
+      "learning_rate": 2.3261648745519717e-05,
+      "loss": 0.2017,
+      "step": 90100
+    },
+    {
+      "epoch": 581.35,
+      "learning_rate": 2.325806451612903e-05,
+      "loss": 0.2646,
+      "step": 90110
+    },
+    {
+      "epoch": 581.42,
+      "learning_rate": 2.3254480286738353e-05,
+      "loss": 0.2472,
+      "step": 90120
+    },
+    {
+      "epoch": 581.48,
+      "learning_rate": 2.325089605734767e-05,
+      "loss": 0.2059,
+      "step": 90130
+    },
+    {
+      "epoch": 581.55,
+      "learning_rate": 2.324731182795699e-05,
+      "loss": 0.2573,
+      "step": 90140
+    },
+    {
+      "epoch": 581.61,
+      "learning_rate": 2.3243727598566308e-05,
+      "loss": 0.2392,
+      "step": 90150
+    },
+    {
+      "epoch": 581.68,
+      "learning_rate": 2.324014336917563e-05,
+      "loss": 0.2436,
+      "step": 90160
+    },
+    {
+      "epoch": 581.74,
+      "learning_rate": 2.3236559139784948e-05,
+      "loss": 0.2668,
+      "step": 90170
+    },
+    {
+      "epoch": 581.8,
+      "learning_rate": 2.3232974910394266e-05,
+      "loss": 0.2482,
+      "step": 90180
+    },
+    {
+      "epoch": 581.87,
+      "learning_rate": 2.3229390681003584e-05,
+      "loss": 0.2076,
+      "step": 90190
+    },
+    {
+      "epoch": 581.93,
+      "learning_rate": 2.3225806451612906e-05,
+      "loss": 0.2545,
+      "step": 90200
+    },
+    {
+      "epoch": 582.0,
+      "learning_rate": 2.3222222222222224e-05,
+      "loss": 0.225,
+      "step": 90210
+    },
+    {
+      "epoch": 582.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0747305154800415,
+      "eval_runtime": 0.8857,
+      "eval_samples_per_second": 226.929,
+      "eval_steps_per_second": 7.903,
+      "step": 90210
+    },
+    {
+      "epoch": 582.06,
+      "learning_rate": 2.3218637992831542e-05,
+      "loss": 0.2853,
+      "step": 90220
+    },
+    {
+      "epoch": 582.13,
+      "learning_rate": 2.321505376344086e-05,
+      "loss": 0.2272,
+      "step": 90230
+    },
+    {
+      "epoch": 582.19,
+      "learning_rate": 2.3211469534050182e-05,
+      "loss": 0.2432,
+      "step": 90240
+    },
+    {
+      "epoch": 582.26,
+      "learning_rate": 2.32078853046595e-05,
+      "loss": 0.2808,
+      "step": 90250
+    },
+    {
+      "epoch": 582.32,
+      "learning_rate": 2.3204301075268818e-05,
+      "loss": 0.2223,
+      "step": 90260
+    },
+    {
+      "epoch": 582.39,
+      "learning_rate": 2.3200716845878136e-05,
+      "loss": 0.2184,
+      "step": 90270
+    },
+    {
+      "epoch": 582.45,
+      "learning_rate": 2.3197132616487458e-05,
+      "loss": 0.2471,
+      "step": 90280
+    },
+    {
+      "epoch": 582.51,
+      "learning_rate": 2.3193548387096776e-05,
+      "loss": 0.2745,
+      "step": 90290
+    },
+    {
+      "epoch": 582.58,
+      "learning_rate": 2.3189964157706094e-05,
+      "loss": 0.261,
+      "step": 90300
+    },
+    {
+      "epoch": 582.64,
+      "learning_rate": 2.3186379928315413e-05,
+      "loss": 0.2208,
+      "step": 90310
+    },
+    {
+      "epoch": 582.71,
+      "learning_rate": 2.3182795698924734e-05,
+      "loss": 0.261,
+      "step": 90320
+    },
+    {
+      "epoch": 582.77,
+      "learning_rate": 2.3179211469534053e-05,
+      "loss": 0.2321,
+      "step": 90330
+    },
+    {
+      "epoch": 582.84,
+      "learning_rate": 2.317562724014337e-05,
+      "loss": 0.2368,
+      "step": 90340
+    },
+    {
+      "epoch": 582.9,
+      "learning_rate": 2.317204301075269e-05,
+      "loss": 0.2751,
+      "step": 90350
+    },
+    {
+      "epoch": 582.96,
+      "learning_rate": 2.316845878136201e-05,
+      "loss": 0.2684,
+      "step": 90360
+    },
+    {
+      "epoch": 583.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.1319440603256226,
+      "eval_runtime": 0.8812,
+      "eval_samples_per_second": 228.097,
+      "eval_steps_per_second": 7.944,
+      "step": 90365
+    },
+    {
+      "epoch": 583.03,
+      "learning_rate": 2.316487455197133e-05,
+      "loss": 0.2614,
+      "step": 90370
+    },
+    {
+      "epoch": 583.1,
+      "learning_rate": 2.3161290322580644e-05,
+      "loss": 0.1977,
+      "step": 90380
+    },
+    {
+      "epoch": 583.16,
+      "learning_rate": 2.3157706093189965e-05,
+      "loss": 0.2406,
+      "step": 90390
+    },
+    {
+      "epoch": 583.23,
+      "learning_rate": 2.3154121863799283e-05,
+      "loss": 0.2171,
+      "step": 90400
+    },
+    {
+      "epoch": 583.29,
+      "learning_rate": 2.3150537634408605e-05,
+      "loss": 0.2598,
+      "step": 90410
+    },
+    {
+      "epoch": 583.35,
+      "learning_rate": 2.314695340501792e-05,
+      "loss": 0.2269,
+      "step": 90420
+    },
+    {
+      "epoch": 583.42,
+      "learning_rate": 2.314336917562724e-05,
+      "loss": 0.2719,
+      "step": 90430
+    },
+    {
+      "epoch": 583.48,
+      "learning_rate": 2.313978494623656e-05,
+      "loss": 0.2639,
+      "step": 90440
+    },
+    {
+      "epoch": 583.55,
+      "learning_rate": 2.3136200716845878e-05,
+      "loss": 0.2514,
+      "step": 90450
+    },
+    {
+      "epoch": 583.61,
+      "learning_rate": 2.3132616487455196e-05,
+      "loss": 0.2333,
+      "step": 90460
+    },
+    {
+      "epoch": 583.68,
+      "learning_rate": 2.3129032258064518e-05,
+      "loss": 0.2592,
+      "step": 90470
+    },
+    {
+      "epoch": 583.74,
+      "learning_rate": 2.3125448028673836e-05,
+      "loss": 0.251,
+      "step": 90480
+    },
+    {
+      "epoch": 583.8,
+      "learning_rate": 2.3121863799283154e-05,
+      "loss": 0.2148,
+      "step": 90490
+    },
+    {
+      "epoch": 583.87,
+      "learning_rate": 2.3118279569892472e-05,
+      "loss": 0.2202,
+      "step": 90500
+    },
+    {
+      "epoch": 583.93,
+      "learning_rate": 2.3114695340501794e-05,
+      "loss": 0.2479,
+      "step": 90510
+    },
+    {
+      "epoch": 584.0,
+      "learning_rate": 2.3111111111111112e-05,
+      "loss": 0.2017,
+      "step": 90520
+    },
+    {
+      "epoch": 584.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9989261031150818,
+      "eval_runtime": 0.8559,
+      "eval_samples_per_second": 234.836,
+      "eval_steps_per_second": 8.178,
+      "step": 90520
+    },
+    {
+      "epoch": 584.06,
+      "learning_rate": 2.310752688172043e-05,
+      "loss": 0.2418,
+      "step": 90530
+    },
+    {
+      "epoch": 584.13,
+      "learning_rate": 2.310394265232975e-05,
+      "loss": 0.268,
+      "step": 90540
+    },
+    {
+      "epoch": 584.19,
+      "learning_rate": 2.310035842293907e-05,
+      "loss": 0.2403,
+      "step": 90550
+    },
+    {
+      "epoch": 584.26,
+      "learning_rate": 2.3096774193548388e-05,
+      "loss": 0.2801,
+      "step": 90560
+    },
+    {
+      "epoch": 584.32,
+      "learning_rate": 2.3093189964157706e-05,
+      "loss": 0.2755,
+      "step": 90570
+    },
+    {
+      "epoch": 584.39,
+      "learning_rate": 2.3089605734767025e-05,
+      "loss": 0.2246,
+      "step": 90580
+    },
+    {
+      "epoch": 584.45,
+      "learning_rate": 2.3086021505376346e-05,
+      "loss": 0.2891,
+      "step": 90590
+    },
+    {
+      "epoch": 584.51,
+      "learning_rate": 2.3082437275985664e-05,
+      "loss": 0.2218,
+      "step": 90600
+    },
+    {
+      "epoch": 584.58,
+      "learning_rate": 2.3078853046594983e-05,
+      "loss": 0.2382,
+      "step": 90610
+    },
+    {
+      "epoch": 584.64,
+      "learning_rate": 2.30752688172043e-05,
+      "loss": 0.2631,
+      "step": 90620
+    },
+    {
+      "epoch": 584.71,
+      "learning_rate": 2.3071684587813622e-05,
+      "loss": 0.2515,
+      "step": 90630
+    },
+    {
+      "epoch": 584.77,
+      "learning_rate": 2.306810035842294e-05,
+      "loss": 0.3064,
+      "step": 90640
+    },
+    {
+      "epoch": 584.84,
+      "learning_rate": 2.306451612903226e-05,
+      "loss": 0.2551,
+      "step": 90650
+    },
+    {
+      "epoch": 584.9,
+      "learning_rate": 2.3060931899641577e-05,
+      "loss": 0.2619,
+      "step": 90660
+    },
+    {
+      "epoch": 584.96,
+      "learning_rate": 2.30573476702509e-05,
+      "loss": 0.2799,
+      "step": 90670
+    },
+    {
+      "epoch": 585.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0177987813949585,
+      "eval_runtime": 0.8624,
+      "eval_samples_per_second": 233.068,
+      "eval_steps_per_second": 8.117,
+      "step": 90675
+    },
+    {
+      "epoch": 585.03,
+      "learning_rate": 2.3053763440860217e-05,
+      "loss": 0.2891,
+      "step": 90680
+    },
+    {
+      "epoch": 585.1,
+      "learning_rate": 2.3050179211469535e-05,
+      "loss": 0.2854,
+      "step": 90690
+    },
+    {
+      "epoch": 585.16,
+      "learning_rate": 2.3046594982078853e-05,
+      "loss": 0.2336,
+      "step": 90700
+    },
+    {
+      "epoch": 585.23,
+      "learning_rate": 2.3043010752688175e-05,
+      "loss": 0.3157,
+      "step": 90710
+    },
+    {
+      "epoch": 585.29,
+      "learning_rate": 2.3039426523297493e-05,
+      "loss": 0.2071,
+      "step": 90720
+    },
+    {
+      "epoch": 585.35,
+      "learning_rate": 2.303584229390681e-05,
+      "loss": 0.2233,
+      "step": 90730
+    },
+    {
+      "epoch": 585.42,
+      "learning_rate": 2.303225806451613e-05,
+      "loss": 0.235,
+      "step": 90740
+    },
+    {
+      "epoch": 585.48,
+      "learning_rate": 2.302867383512545e-05,
+      "loss": 0.2781,
+      "step": 90750
+    },
+    {
+      "epoch": 585.55,
+      "learning_rate": 2.302508960573477e-05,
+      "loss": 0.2671,
+      "step": 90760
+    },
+    {
+      "epoch": 585.61,
+      "learning_rate": 2.3021505376344087e-05,
+      "loss": 0.2471,
+      "step": 90770
+    },
+    {
+      "epoch": 585.68,
+      "learning_rate": 2.3017921146953406e-05,
+      "loss": 0.268,
+      "step": 90780
+    },
+    {
+      "epoch": 585.74,
+      "learning_rate": 2.3014336917562727e-05,
+      "loss": 0.2686,
+      "step": 90790
+    },
+    {
+      "epoch": 585.8,
+      "learning_rate": 2.3010752688172042e-05,
+      "loss": 0.2399,
+      "step": 90800
+    },
+    {
+      "epoch": 585.87,
+      "learning_rate": 2.3007168458781364e-05,
+      "loss": 0.2662,
+      "step": 90810
+    },
+    {
+      "epoch": 585.93,
+      "learning_rate": 2.3003584229390682e-05,
+      "loss": 0.2601,
+      "step": 90820
+    },
+    {
+      "epoch": 586.0,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.2344,
+      "step": 90830
+    },
+    {
+      "epoch": 586.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0060404539108276,
+      "eval_runtime": 0.8583,
+      "eval_samples_per_second": 234.174,
+      "eval_steps_per_second": 8.155,
+      "step": 90830
+    },
+    {
+      "epoch": 586.06,
+      "learning_rate": 2.2996415770609318e-05,
+      "loss": 0.2102,
+      "step": 90840
+    },
+    {
+      "epoch": 586.13,
+      "learning_rate": 2.299283154121864e-05,
+      "loss": 0.2418,
+      "step": 90850
+    },
+    {
+      "epoch": 586.19,
+      "learning_rate": 2.2989247311827958e-05,
+      "loss": 0.2484,
+      "step": 90860
+    },
+    {
+      "epoch": 586.26,
+      "learning_rate": 2.298566308243728e-05,
+      "loss": 0.2378,
+      "step": 90870
+    },
+    {
+      "epoch": 586.32,
+      "learning_rate": 2.2982078853046594e-05,
+      "loss": 0.2595,
+      "step": 90880
+    },
+    {
+      "epoch": 586.39,
+      "learning_rate": 2.2978494623655916e-05,
+      "loss": 0.2497,
+      "step": 90890
+    },
+    {
+      "epoch": 586.45,
+      "learning_rate": 2.2974910394265234e-05,
+      "loss": 0.2536,
+      "step": 90900
+    },
+    {
+      "epoch": 586.51,
+      "learning_rate": 2.2971326164874556e-05,
+      "loss": 0.2045,
+      "step": 90910
+    },
+    {
+      "epoch": 586.58,
+      "learning_rate": 2.296774193548387e-05,
+      "loss": 0.2552,
+      "step": 90920
+    },
+    {
+      "epoch": 586.64,
+      "learning_rate": 2.2964157706093192e-05,
+      "loss": 0.2595,
+      "step": 90930
+    },
+    {
+      "epoch": 586.71,
+      "learning_rate": 2.296057347670251e-05,
+      "loss": 0.2315,
+      "step": 90940
+    },
+    {
+      "epoch": 586.77,
+      "learning_rate": 2.295698924731183e-05,
+      "loss": 0.2711,
+      "step": 90950
+    },
+    {
+      "epoch": 586.84,
+      "learning_rate": 2.2953405017921147e-05,
+      "loss": 0.2467,
+      "step": 90960
+    },
+    {
+      "epoch": 586.9,
+      "learning_rate": 2.294982078853047e-05,
+      "loss": 0.2465,
+      "step": 90970
+    },
+    {
+      "epoch": 586.96,
+      "learning_rate": 2.2946236559139787e-05,
+      "loss": 0.2579,
+      "step": 90980
+    },
+    {
+      "epoch": 587.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9840750098228455,
+      "eval_runtime": 0.886,
+      "eval_samples_per_second": 226.851,
+      "eval_steps_per_second": 7.9,
+      "step": 90985
+    },
+    {
+      "epoch": 587.03,
+      "learning_rate": 2.2942652329749105e-05,
+      "loss": 0.2026,
+      "step": 90990
+    },
+    {
+      "epoch": 587.1,
+      "learning_rate": 2.2939068100358423e-05,
+      "loss": 0.2361,
+      "step": 91000
+    },
+    {
+      "epoch": 587.16,
+      "learning_rate": 2.293548387096774e-05,
+      "loss": 0.2166,
+      "step": 91010
+    },
+    {
+      "epoch": 587.23,
+      "learning_rate": 2.2931899641577063e-05,
+      "loss": 0.1654,
+      "step": 91020
+    },
+    {
+      "epoch": 587.29,
+      "learning_rate": 2.292831541218638e-05,
+      "loss": 0.2751,
+      "step": 91030
+    },
+    {
+      "epoch": 587.35,
+      "learning_rate": 2.29247311827957e-05,
+      "loss": 0.2001,
+      "step": 91040
+    },
+    {
+      "epoch": 587.42,
+      "learning_rate": 2.2921146953405017e-05,
+      "loss": 0.2217,
+      "step": 91050
+    },
+    {
+      "epoch": 587.48,
+      "learning_rate": 2.291756272401434e-05,
+      "loss": 0.241,
+      "step": 91060
+    },
+    {
+      "epoch": 587.55,
+      "learning_rate": 2.2913978494623657e-05,
+      "loss": 0.2851,
+      "step": 91070
+    },
+    {
+      "epoch": 587.61,
+      "learning_rate": 2.2910394265232975e-05,
+      "loss": 0.2428,
+      "step": 91080
+    },
+    {
+      "epoch": 587.68,
+      "learning_rate": 2.2906810035842294e-05,
+      "loss": 0.2612,
+      "step": 91090
+    },
+    {
+      "epoch": 587.74,
+      "learning_rate": 2.2903225806451615e-05,
+      "loss": 0.2365,
+      "step": 91100
+    },
+    {
+      "epoch": 587.8,
+      "learning_rate": 2.2899641577060933e-05,
+      "loss": 0.2467,
+      "step": 91110
+    },
+    {
+      "epoch": 587.87,
+      "learning_rate": 2.289605734767025e-05,
+      "loss": 0.2753,
+      "step": 91120
+    },
+    {
+      "epoch": 587.93,
+      "learning_rate": 2.289247311827957e-05,
+      "loss": 0.2454,
+      "step": 91130
+    },
+    {
+      "epoch": 588.0,
+      "learning_rate": 2.288888888888889e-05,
+      "loss": 0.2483,
+      "step": 91140
+    },
+    {
+      "epoch": 588.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0188924074172974,
+      "eval_runtime": 0.876,
+      "eval_samples_per_second": 229.45,
+      "eval_steps_per_second": 7.991,
+      "step": 91140
+    },
+    {
+      "epoch": 588.06,
+      "learning_rate": 2.2885304659498206e-05,
+      "loss": 0.2588,
+      "step": 91150
+    },
+    {
+      "epoch": 588.13,
+      "learning_rate": 2.2881720430107528e-05,
+      "loss": 0.2176,
+      "step": 91160
+    },
+    {
+      "epoch": 588.19,
+      "learning_rate": 2.2878136200716846e-05,
+      "loss": 0.2357,
+      "step": 91170
+    },
+    {
+      "epoch": 588.26,
+      "learning_rate": 2.2874551971326168e-05,
+      "loss": 0.2613,
+      "step": 91180
+    },
+    {
+      "epoch": 588.32,
+      "learning_rate": 2.2870967741935483e-05,
+      "loss": 0.2445,
+      "step": 91190
+    },
+    {
+      "epoch": 588.39,
+      "learning_rate": 2.2867383512544804e-05,
+      "loss": 0.2705,
+      "step": 91200
+    },
+    {
+      "epoch": 588.45,
+      "learning_rate": 2.2863799283154122e-05,
+      "loss": 0.2303,
+      "step": 91210
+    },
+    {
+      "epoch": 588.51,
+      "learning_rate": 2.2860215053763444e-05,
+      "loss": 0.2304,
+      "step": 91220
+    },
+    {
+      "epoch": 588.58,
+      "learning_rate": 2.285663082437276e-05,
+      "loss": 0.2669,
+      "step": 91230
+    },
+    {
+      "epoch": 588.64,
+      "learning_rate": 2.285304659498208e-05,
+      "loss": 0.236,
+      "step": 91240
+    },
+    {
+      "epoch": 588.71,
+      "learning_rate": 2.28494623655914e-05,
+      "loss": 0.2378,
+      "step": 91250
+    },
+    {
+      "epoch": 588.77,
+      "learning_rate": 2.284587813620072e-05,
+      "loss": 0.2919,
+      "step": 91260
+    },
+    {
+      "epoch": 588.84,
+      "learning_rate": 2.2842293906810035e-05,
+      "loss": 0.255,
+      "step": 91270
+    },
+    {
+      "epoch": 588.9,
+      "learning_rate": 2.2838709677419357e-05,
+      "loss": 0.2255,
+      "step": 91280
+    },
+    {
+      "epoch": 588.96,
+      "learning_rate": 2.2835125448028675e-05,
+      "loss": 0.2184,
+      "step": 91290
+    },
+    {
+      "epoch": 589.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.039493203163147,
+      "eval_runtime": 0.8431,
+      "eval_samples_per_second": 238.408,
+      "eval_steps_per_second": 8.303,
+      "step": 91295
+    },
+    {
+      "epoch": 589.03,
+      "learning_rate": 2.2831541218637993e-05,
+      "loss": 0.2728,
+      "step": 91300
+    },
+    {
+      "epoch": 589.1,
+      "learning_rate": 2.282795698924731e-05,
+      "loss": 0.1949,
+      "step": 91310
+    },
+    {
+      "epoch": 589.16,
+      "learning_rate": 2.2824372759856633e-05,
+      "loss": 0.2269,
+      "step": 91320
+    },
+    {
+      "epoch": 589.23,
+      "learning_rate": 2.282078853046595e-05,
+      "loss": 0.2382,
+      "step": 91330
+    },
+    {
+      "epoch": 589.29,
+      "learning_rate": 2.281720430107527e-05,
+      "loss": 0.2289,
+      "step": 91340
+    },
+    {
+      "epoch": 589.35,
+      "learning_rate": 2.2813620071684587e-05,
+      "loss": 0.266,
+      "step": 91350
+    },
+    {
+      "epoch": 589.42,
+      "learning_rate": 2.281003584229391e-05,
+      "loss": 0.2434,
+      "step": 91360
+    },
+    {
+      "epoch": 589.48,
+      "learning_rate": 2.2806451612903227e-05,
+      "loss": 0.2158,
+      "step": 91370
+    },
+    {
+      "epoch": 589.55,
+      "learning_rate": 2.2802867383512545e-05,
+      "loss": 0.2659,
+      "step": 91380
+    },
+    {
+      "epoch": 589.61,
+      "learning_rate": 2.2799283154121864e-05,
+      "loss": 0.2448,
+      "step": 91390
+    },
+    {
+      "epoch": 589.68,
+      "learning_rate": 2.2795698924731185e-05,
+      "loss": 0.2429,
+      "step": 91400
+    },
+    {
+      "epoch": 589.74,
+      "learning_rate": 2.2792114695340503e-05,
+      "loss": 0.2654,
+      "step": 91410
+    },
+    {
+      "epoch": 589.8,
+      "learning_rate": 2.278853046594982e-05,
+      "loss": 0.2777,
+      "step": 91420
+    },
+    {
+      "epoch": 589.87,
+      "learning_rate": 2.278494623655914e-05,
+      "loss": 0.2199,
+      "step": 91430
+    },
+    {
+      "epoch": 589.93,
+      "learning_rate": 2.278136200716846e-05,
+      "loss": 0.2598,
+      "step": 91440
+    },
+    {
+      "epoch": 590.0,
+      "learning_rate": 2.277777777777778e-05,
+      "loss": 0.276,
+      "step": 91450
+    },
+    {
+      "epoch": 590.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.066232442855835,
+      "eval_runtime": 0.877,
+      "eval_samples_per_second": 229.2,
+      "eval_steps_per_second": 7.982,
+      "step": 91450
+    },
+    {
+      "epoch": 590.06,
+      "learning_rate": 2.2774193548387098e-05,
+      "loss": 0.2694,
+      "step": 91460
+    },
+    {
+      "epoch": 590.13,
+      "learning_rate": 2.2770609318996416e-05,
+      "loss": 0.2535,
+      "step": 91470
+    },
+    {
+      "epoch": 590.19,
+      "learning_rate": 2.2767025089605738e-05,
+      "loss": 0.2542,
+      "step": 91480
+    },
+    {
+      "epoch": 590.26,
+      "learning_rate": 2.2763440860215056e-05,
+      "loss": 0.2475,
+      "step": 91490
+    },
+    {
+      "epoch": 590.32,
+      "learning_rate": 2.2759856630824374e-05,
+      "loss": 0.232,
+      "step": 91500
+    },
+    {
+      "epoch": 590.39,
+      "learning_rate": 2.2756272401433692e-05,
+      "loss": 0.1887,
+      "step": 91510
+    },
+    {
+      "epoch": 590.45,
+      "learning_rate": 2.2752688172043014e-05,
+      "loss": 0.2479,
+      "step": 91520
+    },
+    {
+      "epoch": 590.51,
+      "learning_rate": 2.2749103942652332e-05,
+      "loss": 0.2521,
+      "step": 91530
+    },
+    {
+      "epoch": 590.58,
+      "learning_rate": 2.274551971326165e-05,
+      "loss": 0.2473,
+      "step": 91540
+    },
+    {
+      "epoch": 590.64,
+      "learning_rate": 2.274193548387097e-05,
+      "loss": 0.2322,
+      "step": 91550
+    },
+    {
+      "epoch": 590.71,
+      "learning_rate": 2.273835125448029e-05,
+      "loss": 0.247,
+      "step": 91560
+    },
+    {
+      "epoch": 590.77,
+      "learning_rate": 2.2734767025089608e-05,
+      "loss": 0.2735,
+      "step": 91570
+    },
+    {
+      "epoch": 590.84,
+      "learning_rate": 2.2731182795698926e-05,
+      "loss": 0.2383,
+      "step": 91580
+    },
+    {
+      "epoch": 590.9,
+      "learning_rate": 2.2727598566308245e-05,
+      "loss": 0.2715,
+      "step": 91590
+    },
+    {
+      "epoch": 590.96,
+      "learning_rate": 2.2724014336917566e-05,
+      "loss": 0.2274,
+      "step": 91600
+    },
+    {
+      "epoch": 591.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.0669344663619995,
+      "eval_runtime": 0.8838,
+      "eval_samples_per_second": 227.424,
+      "eval_steps_per_second": 7.92,
+      "step": 91605
+    },
+    {
+      "epoch": 591.03,
+      "learning_rate": 2.2720430107526884e-05,
+      "loss": 0.2691,
+      "step": 91610
+    },
+    {
+      "epoch": 591.1,
+      "learning_rate": 2.2716845878136203e-05,
+      "loss": 0.2269,
+      "step": 91620
+    },
+    {
+      "epoch": 591.16,
+      "learning_rate": 2.271326164874552e-05,
+      "loss": 0.2313,
+      "step": 91630
+    },
+    {
+      "epoch": 591.23,
+      "learning_rate": 2.2709677419354842e-05,
+      "loss": 0.1602,
+      "step": 91640
+    },
+    {
+      "epoch": 591.29,
+      "learning_rate": 2.2706093189964157e-05,
+      "loss": 0.2785,
+      "step": 91650
+    },
+    {
+      "epoch": 591.35,
+      "learning_rate": 2.2702508960573475e-05,
+      "loss": 0.2318,
+      "step": 91660
+    },
+    {
+      "epoch": 591.42,
+      "learning_rate": 2.2698924731182797e-05,
+      "loss": 0.2452,
+      "step": 91670
+    },
+    {
+      "epoch": 591.48,
+      "learning_rate": 2.2695340501792115e-05,
+      "loss": 0.2321,
+      "step": 91680
+    },
+    {
+      "epoch": 591.55,
+      "learning_rate": 2.2691756272401433e-05,
+      "loss": 0.2208,
+      "step": 91690
+    },
+    {
+      "epoch": 591.61,
+      "learning_rate": 2.268817204301075e-05,
+      "loss": 0.2375,
+      "step": 91700
+    },
+    {
+      "epoch": 591.68,
+      "learning_rate": 2.2684587813620073e-05,
+      "loss": 0.3033,
+      "step": 91710
+    },
+    {
+      "epoch": 591.74,
+      "learning_rate": 2.268100358422939e-05,
+      "loss": 0.238,
+      "step": 91720
+    },
+    {
+      "epoch": 591.8,
+      "learning_rate": 2.267741935483871e-05,
+      "loss": 0.2637,
+      "step": 91730
+    },
+    {
+      "epoch": 591.87,
+      "learning_rate": 2.2673835125448028e-05,
+      "loss": 0.2418,
+      "step": 91740
+    },
+    {
+      "epoch": 591.93,
+      "learning_rate": 2.267025089605735e-05,
+      "loss": 0.2112,
+      "step": 91750
+    },
+    {
+      "epoch": 592.0,
+      "learning_rate": 2.2666666666666668e-05,
+      "loss": 0.1932,
+      "step": 91760
+    },
+    {
+      "epoch": 592.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0571272373199463,
+      "eval_runtime": 0.8567,
+      "eval_samples_per_second": 234.617,
+      "eval_steps_per_second": 8.171,
+      "step": 91760
+    },
+    {
+      "epoch": 592.06,
+      "learning_rate": 2.2663082437275986e-05,
+      "loss": 0.3023,
+      "step": 91770
+    },
+    {
+      "epoch": 592.13,
+      "learning_rate": 2.2659498207885304e-05,
+      "loss": 0.2329,
+      "step": 91780
+    },
+    {
+      "epoch": 592.19,
+      "learning_rate": 2.2655913978494626e-05,
+      "loss": 0.2482,
+      "step": 91790
+    },
+    {
+      "epoch": 592.26,
+      "learning_rate": 2.2652329749103944e-05,
+      "loss": 0.252,
+      "step": 91800
+    },
+    {
+      "epoch": 592.32,
+      "learning_rate": 2.2648745519713262e-05,
+      "loss": 0.2477,
+      "step": 91810
+    },
+    {
+      "epoch": 592.39,
+      "learning_rate": 2.264516129032258e-05,
+      "loss": 0.2649,
+      "step": 91820
+    },
+    {
+      "epoch": 592.45,
+      "learning_rate": 2.2641577060931902e-05,
+      "loss": 0.2718,
+      "step": 91830
+    },
+    {
+      "epoch": 592.51,
+      "learning_rate": 2.263799283154122e-05,
+      "loss": 0.2248,
+      "step": 91840
+    },
+    {
+      "epoch": 592.58,
+      "learning_rate": 2.2634408602150538e-05,
+      "loss": 0.2173,
+      "step": 91850
+    },
+    {
+      "epoch": 592.64,
+      "learning_rate": 2.2630824372759856e-05,
+      "loss": 0.2552,
+      "step": 91860
+    },
+    {
+      "epoch": 592.71,
+      "learning_rate": 2.2627240143369178e-05,
+      "loss": 0.1814,
+      "step": 91870
+    },
+    {
+      "epoch": 592.77,
+      "learning_rate": 2.2623655913978496e-05,
+      "loss": 0.2349,
+      "step": 91880
+    },
+    {
+      "epoch": 592.84,
+      "learning_rate": 2.2620071684587814e-05,
+      "loss": 0.1849,
+      "step": 91890
+    },
+    {
+      "epoch": 592.9,
+      "learning_rate": 2.2616487455197133e-05,
+      "loss": 0.2583,
+      "step": 91900
+    },
+    {
+      "epoch": 592.96,
+      "learning_rate": 2.2612903225806454e-05,
+      "loss": 0.2318,
+      "step": 91910
+    },
+    {
+      "epoch": 593.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0403748750686646,
+      "eval_runtime": 0.8777,
+      "eval_samples_per_second": 229.003,
+      "eval_steps_per_second": 7.975,
+      "step": 91915
+    },
+    {
+      "epoch": 593.03,
+      "learning_rate": 2.2609318996415772e-05,
+      "loss": 0.2875,
+      "step": 91920
+    },
+    {
+      "epoch": 593.1,
+      "learning_rate": 2.260573476702509e-05,
+      "loss": 0.2532,
+      "step": 91930
+    },
+    {
+      "epoch": 593.16,
+      "learning_rate": 2.260215053763441e-05,
+      "loss": 0.2173,
+      "step": 91940
+    },
+    {
+      "epoch": 593.23,
+      "learning_rate": 2.259856630824373e-05,
+      "loss": 0.2647,
+      "step": 91950
+    },
+    {
+      "epoch": 593.29,
+      "learning_rate": 2.259498207885305e-05,
+      "loss": 0.2292,
+      "step": 91960
+    },
+    {
+      "epoch": 593.35,
+      "learning_rate": 2.2591397849462367e-05,
+      "loss": 0.2192,
+      "step": 91970
+    },
+    {
+      "epoch": 593.42,
+      "learning_rate": 2.2587813620071685e-05,
+      "loss": 0.2666,
+      "step": 91980
+    },
+    {
+      "epoch": 593.48,
+      "learning_rate": 2.2584229390681007e-05,
+      "loss": 0.2153,
+      "step": 91990
+    },
+    {
+      "epoch": 593.55,
+      "learning_rate": 2.258064516129032e-05,
+      "loss": 0.2256,
+      "step": 92000
+    },
+    {
+      "epoch": 593.61,
+      "learning_rate": 2.2577060931899643e-05,
+      "loss": 0.274,
+      "step": 92010
+    },
+    {
+      "epoch": 593.68,
+      "learning_rate": 2.257347670250896e-05,
+      "loss": 0.2326,
+      "step": 92020
+    },
+    {
+      "epoch": 593.74,
+      "learning_rate": 2.2569892473118283e-05,
+      "loss": 0.26,
+      "step": 92030
+    },
+    {
+      "epoch": 593.8,
+      "learning_rate": 2.2566308243727598e-05,
+      "loss": 0.286,
+      "step": 92040
+    },
+    {
+      "epoch": 593.87,
+      "learning_rate": 2.256272401433692e-05,
+      "loss": 0.2457,
+      "step": 92050
+    },
+    {
+      "epoch": 593.93,
+      "learning_rate": 2.2559139784946237e-05,
+      "loss": 0.2612,
+      "step": 92060
+    },
+    {
+      "epoch": 594.0,
+      "learning_rate": 2.255555555555556e-05,
+      "loss": 0.2234,
+      "step": 92070
+    },
+    {
+      "epoch": 594.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0175690650939941,
+      "eval_runtime": 0.8903,
+      "eval_samples_per_second": 225.76,
+      "eval_steps_per_second": 7.862,
+      "step": 92070
+    },
+    {
+      "epoch": 594.06,
+      "learning_rate": 2.2551971326164874e-05,
+      "loss": 0.229,
+      "step": 92080
+    },
+    {
+      "epoch": 594.13,
+      "learning_rate": 2.2548387096774195e-05,
+      "loss": 0.2209,
+      "step": 92090
+    },
+    {
+      "epoch": 594.19,
+      "learning_rate": 2.2544802867383514e-05,
+      "loss": 0.2386,
+      "step": 92100
+    },
+    {
+      "epoch": 594.26,
+      "learning_rate": 2.2541218637992835e-05,
+      "loss": 0.2856,
+      "step": 92110
+    },
+    {
+      "epoch": 594.32,
+      "learning_rate": 2.253763440860215e-05,
+      "loss": 0.2569,
+      "step": 92120
+    },
+    {
+      "epoch": 594.39,
+      "learning_rate": 2.253405017921147e-05,
+      "loss": 0.2517,
+      "step": 92130
+    },
+    {
+      "epoch": 594.45,
+      "learning_rate": 2.253046594982079e-05,
+      "loss": 0.2342,
+      "step": 92140
+    },
+    {
+      "epoch": 594.51,
+      "learning_rate": 2.2526881720430108e-05,
+      "loss": 0.1855,
+      "step": 92150
+    },
+    {
+      "epoch": 594.58,
+      "learning_rate": 2.2523297491039426e-05,
+      "loss": 0.2627,
+      "step": 92160
+    },
+    {
+      "epoch": 594.64,
+      "learning_rate": 2.2519713261648748e-05,
+      "loss": 0.255,
+      "step": 92170
+    },
+    {
+      "epoch": 594.71,
+      "learning_rate": 2.2516129032258066e-05,
+      "loss": 0.2432,
+      "step": 92180
+    },
+    {
+      "epoch": 594.77,
+      "learning_rate": 2.2512544802867384e-05,
+      "loss": 0.2426,
+      "step": 92190
+    },
+    {
+      "epoch": 594.84,
+      "learning_rate": 2.2508960573476703e-05,
+      "loss": 0.2515,
+      "step": 92200
+    },
+    {
+      "epoch": 594.9,
+      "learning_rate": 2.2505376344086024e-05,
+      "loss": 0.2511,
+      "step": 92210
+    },
+    {
+      "epoch": 594.96,
+      "learning_rate": 2.2501792114695342e-05,
+      "loss": 0.2852,
+      "step": 92220
+    },
+    {
+      "epoch": 595.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.045558214187622,
+      "eval_runtime": 0.858,
+      "eval_samples_per_second": 234.27,
+      "eval_steps_per_second": 8.159,
+      "step": 92225
+    },
+    {
+      "epoch": 595.03,
+      "learning_rate": 2.249820788530466e-05,
+      "loss": 0.3099,
+      "step": 92230
+    },
+    {
+      "epoch": 595.1,
+      "learning_rate": 2.249462365591398e-05,
+      "loss": 0.213,
+      "step": 92240
+    },
+    {
+      "epoch": 595.16,
+      "learning_rate": 2.24910394265233e-05,
+      "loss": 0.2706,
+      "step": 92250
+    },
+    {
+      "epoch": 595.23,
+      "learning_rate": 2.248745519713262e-05,
+      "loss": 0.3121,
+      "step": 92260
+    },
+    {
+      "epoch": 595.29,
+      "learning_rate": 2.2483870967741937e-05,
+      "loss": 0.2403,
+      "step": 92270
+    },
+    {
+      "epoch": 595.35,
+      "learning_rate": 2.2480286738351255e-05,
+      "loss": 0.262,
+      "step": 92280
+    },
+    {
+      "epoch": 595.42,
+      "learning_rate": 2.2476702508960573e-05,
+      "loss": 0.2025,
+      "step": 92290
+    },
+    {
+      "epoch": 595.48,
+      "learning_rate": 2.2473118279569895e-05,
+      "loss": 0.2289,
+      "step": 92300
+    },
+    {
+      "epoch": 595.55,
+      "learning_rate": 2.246953405017921e-05,
+      "loss": 0.2266,
+      "step": 92310
+    },
+    {
+      "epoch": 595.61,
+      "learning_rate": 2.246594982078853e-05,
+      "loss": 0.2242,
+      "step": 92320
+    },
+    {
+      "epoch": 595.68,
+      "learning_rate": 2.246236559139785e-05,
+      "loss": 0.2591,
+      "step": 92330
+    },
+    {
+      "epoch": 595.74,
+      "learning_rate": 2.245878136200717e-05,
+      "loss": 0.2367,
+      "step": 92340
+    },
+    {
+      "epoch": 595.8,
+      "learning_rate": 2.2455197132616486e-05,
+      "loss": 0.2039,
+      "step": 92350
+    },
+    {
+      "epoch": 595.87,
+      "learning_rate": 2.2451612903225807e-05,
+      "loss": 0.2374,
+      "step": 92360
+    },
+    {
+      "epoch": 595.93,
+      "learning_rate": 2.2448028673835126e-05,
+      "loss": 0.2309,
+      "step": 92370
+    },
+    {
+      "epoch": 596.0,
+      "learning_rate": 2.2444444444444447e-05,
+      "loss": 0.2734,
+      "step": 92380
+    },
+    {
+      "epoch": 596.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0127131938934326,
+      "eval_runtime": 0.8513,
+      "eval_samples_per_second": 236.109,
+      "eval_steps_per_second": 8.223,
+      "step": 92380
+    },
+    {
+      "epoch": 596.06,
+      "learning_rate": 2.2440860215053762e-05,
+      "loss": 0.2972,
+      "step": 92390
+    },
+    {
+      "epoch": 596.13,
+      "learning_rate": 2.2437275985663084e-05,
+      "loss": 0.2245,
+      "step": 92400
+    },
+    {
+      "epoch": 596.19,
+      "learning_rate": 2.2433691756272402e-05,
+      "loss": 0.2378,
+      "step": 92410
+    },
+    {
+      "epoch": 596.26,
+      "learning_rate": 2.2430107526881723e-05,
+      "loss": 0.2213,
+      "step": 92420
+    },
+    {
+      "epoch": 596.32,
+      "learning_rate": 2.2426523297491038e-05,
+      "loss": 0.2218,
+      "step": 92430
+    },
+    {
+      "epoch": 596.39,
+      "learning_rate": 2.242293906810036e-05,
+      "loss": 0.3143,
+      "step": 92440
+    },
+    {
+      "epoch": 596.45,
+      "learning_rate": 2.2419354838709678e-05,
+      "loss": 0.2457,
+      "step": 92450
+    },
+    {
+      "epoch": 596.51,
+      "learning_rate": 2.2415770609318996e-05,
+      "loss": 0.2221,
+      "step": 92460
+    },
+    {
+      "epoch": 596.58,
+      "learning_rate": 2.2412186379928314e-05,
+      "loss": 0.2435,
+      "step": 92470
+    },
+    {
+      "epoch": 596.64,
+      "learning_rate": 2.2408602150537636e-05,
+      "loss": 0.2539,
+      "step": 92480
+    },
+    {
+      "epoch": 596.71,
+      "learning_rate": 2.2405017921146954e-05,
+      "loss": 0.2258,
+      "step": 92490
+    },
+    {
+      "epoch": 596.77,
+      "learning_rate": 2.2401433691756272e-05,
+      "loss": 0.2169,
+      "step": 92500
+    },
+    {
+      "epoch": 596.84,
+      "learning_rate": 2.239784946236559e-05,
+      "loss": 0.2727,
+      "step": 92510
+    },
+    {
+      "epoch": 596.9,
+      "learning_rate": 2.2394265232974912e-05,
+      "loss": 0.2927,
+      "step": 92520
+    },
+    {
+      "epoch": 596.96,
+      "learning_rate": 2.239068100358423e-05,
+      "loss": 0.2336,
+      "step": 92530
+    },
+    {
+      "epoch": 597.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0206955671310425,
+      "eval_runtime": 0.8586,
+      "eval_samples_per_second": 234.098,
+      "eval_steps_per_second": 8.153,
+      "step": 92535
+    },
+    {
+      "epoch": 597.03,
+      "learning_rate": 2.238709677419355e-05,
+      "loss": 0.2679,
+      "step": 92540
+    },
+    {
+      "epoch": 597.1,
+      "learning_rate": 2.2383512544802867e-05,
+      "loss": 0.2029,
+      "step": 92550
+    },
+    {
+      "epoch": 597.16,
+      "learning_rate": 2.237992831541219e-05,
+      "loss": 0.2961,
+      "step": 92560
+    },
+    {
+      "epoch": 597.23,
+      "learning_rate": 2.2376344086021507e-05,
+      "loss": 0.2849,
+      "step": 92570
+    },
+    {
+      "epoch": 597.29,
+      "learning_rate": 2.2372759856630825e-05,
+      "loss": 0.2249,
+      "step": 92580
+    },
+    {
+      "epoch": 597.35,
+      "learning_rate": 2.2369175627240143e-05,
+      "loss": 0.2324,
+      "step": 92590
+    },
+    {
+      "epoch": 597.42,
+      "learning_rate": 2.2365591397849465e-05,
+      "loss": 0.266,
+      "step": 92600
+    },
+    {
+      "epoch": 597.48,
+      "learning_rate": 2.2362007168458783e-05,
+      "loss": 0.2314,
+      "step": 92610
+    },
+    {
+      "epoch": 597.55,
+      "learning_rate": 2.23584229390681e-05,
+      "loss": 0.286,
+      "step": 92620
+    },
+    {
+      "epoch": 597.61,
+      "learning_rate": 2.235483870967742e-05,
+      "loss": 0.2145,
+      "step": 92630
+    },
+    {
+      "epoch": 597.68,
+      "learning_rate": 2.235125448028674e-05,
+      "loss": 0.2155,
+      "step": 92640
+    },
+    {
+      "epoch": 597.74,
+      "learning_rate": 2.234767025089606e-05,
+      "loss": 0.2841,
+      "step": 92650
+    },
+    {
+      "epoch": 597.8,
+      "learning_rate": 2.2344086021505377e-05,
+      "loss": 0.224,
+      "step": 92660
+    },
+    {
+      "epoch": 597.87,
+      "learning_rate": 2.2340501792114695e-05,
+      "loss": 0.24,
+      "step": 92670
+    },
+    {
+      "epoch": 597.93,
+      "learning_rate": 2.2336917562724017e-05,
+      "loss": 0.2884,
+      "step": 92680
+    },
+    {
+      "epoch": 598.0,
+      "learning_rate": 2.2333333333333335e-05,
+      "loss": 0.2967,
+      "step": 92690
+    },
+    {
+      "epoch": 598.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.0480026006698608,
+      "eval_runtime": 0.894,
+      "eval_samples_per_second": 224.837,
+      "eval_steps_per_second": 7.83,
+      "step": 92690
+    },
+    {
+      "epoch": 598.06,
+      "learning_rate": 2.2329749103942653e-05,
+      "loss": 0.2434,
+      "step": 92700
+    },
+    {
+      "epoch": 598.13,
+      "learning_rate": 2.232616487455197e-05,
+      "loss": 0.2787,
+      "step": 92710
+    },
+    {
+      "epoch": 598.19,
+      "learning_rate": 2.2322580645161293e-05,
+      "loss": 0.2914,
+      "step": 92720
+    },
+    {
+      "epoch": 598.26,
+      "learning_rate": 2.231899641577061e-05,
+      "loss": 0.2286,
+      "step": 92730
+    },
+    {
+      "epoch": 598.32,
+      "learning_rate": 2.231541218637993e-05,
+      "loss": 0.2692,
+      "step": 92740
+    },
+    {
+      "epoch": 598.39,
+      "learning_rate": 2.2311827956989248e-05,
+      "loss": 0.3183,
+      "step": 92750
+    },
+    {
+      "epoch": 598.45,
+      "learning_rate": 2.230824372759857e-05,
+      "loss": 0.239,
+      "step": 92760
+    },
+    {
+      "epoch": 598.51,
+      "learning_rate": 2.2304659498207888e-05,
+      "loss": 0.2322,
+      "step": 92770
+    },
+    {
+      "epoch": 598.58,
+      "learning_rate": 2.2301075268817206e-05,
+      "loss": 0.2251,
+      "step": 92780
+    },
+    {
+      "epoch": 598.64,
+      "learning_rate": 2.2297491039426524e-05,
+      "loss": 0.2237,
+      "step": 92790
+    },
+    {
+      "epoch": 598.71,
+      "learning_rate": 2.2293906810035846e-05,
+      "loss": 0.2468,
+      "step": 92800
+    },
+    {
+      "epoch": 598.77,
+      "learning_rate": 2.229032258064516e-05,
+      "loss": 0.228,
+      "step": 92810
+    },
+    {
+      "epoch": 598.84,
+      "learning_rate": 2.2286738351254482e-05,
+      "loss": 0.2487,
+      "step": 92820
+    },
+    {
+      "epoch": 598.9,
+      "learning_rate": 2.22831541218638e-05,
+      "loss": 0.2494,
+      "step": 92830
+    },
+    {
+      "epoch": 598.96,
+      "learning_rate": 2.2279569892473122e-05,
+      "loss": 0.2279,
+      "step": 92840
+    },
+    {
+      "epoch": 599.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.070110559463501,
+      "eval_runtime": 0.8926,
+      "eval_samples_per_second": 225.196,
+      "eval_steps_per_second": 7.843,
+      "step": 92845
+    },
+    {
+      "epoch": 599.03,
+      "learning_rate": 2.2275985663082437e-05,
+      "loss": 0.219,
+      "step": 92850
+    },
+    {
+      "epoch": 599.1,
+      "learning_rate": 2.2272401433691758e-05,
+      "loss": 0.2374,
+      "step": 92860
+    },
+    {
+      "epoch": 599.16,
+      "learning_rate": 2.2268817204301076e-05,
+      "loss": 0.2361,
+      "step": 92870
+    },
+    {
+      "epoch": 599.23,
+      "learning_rate": 2.2265232974910398e-05,
+      "loss": 0.2435,
+      "step": 92880
+    },
+    {
+      "epoch": 599.29,
+      "learning_rate": 2.2261648745519713e-05,
+      "loss": 0.3004,
+      "step": 92890
+    },
+    {
+      "epoch": 599.35,
+      "learning_rate": 2.2258064516129034e-05,
+      "loss": 0.2391,
+      "step": 92900
+    },
+    {
+      "epoch": 599.42,
+      "learning_rate": 2.2254480286738353e-05,
+      "loss": 0.225,
+      "step": 92910
+    },
+    {
+      "epoch": 599.48,
+      "learning_rate": 2.2250896057347674e-05,
+      "loss": 0.2004,
+      "step": 92920
+    },
+    {
+      "epoch": 599.55,
+      "learning_rate": 2.224731182795699e-05,
+      "loss": 0.2358,
+      "step": 92930
+    },
+    {
+      "epoch": 599.61,
+      "learning_rate": 2.2243727598566307e-05,
+      "loss": 0.2469,
+      "step": 92940
+    },
+    {
+      "epoch": 599.68,
+      "learning_rate": 2.224014336917563e-05,
+      "loss": 0.2201,
+      "step": 92950
+    },
+    {
+      "epoch": 599.74,
+      "learning_rate": 2.2236559139784947e-05,
+      "loss": 0.2249,
+      "step": 92960
+    },
+    {
+      "epoch": 599.8,
+      "learning_rate": 2.2232974910394265e-05,
+      "loss": 0.2794,
+      "step": 92970
+    },
+    {
+      "epoch": 599.87,
+      "learning_rate": 2.2229390681003583e-05,
+      "loss": 0.2012,
+      "step": 92980
+    },
+    {
+      "epoch": 599.93,
+      "learning_rate": 2.2225806451612905e-05,
+      "loss": 0.2115,
+      "step": 92990
+    },
+    {
+      "epoch": 600.0,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.2054,
+      "step": 93000
+    },
+    {
+      "epoch": 600.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.078826904296875,
+      "eval_runtime": 0.8625,
+      "eval_samples_per_second": 233.036,
+      "eval_steps_per_second": 8.116,
+      "step": 93000
+    },
+    {
+      "epoch": 600.06,
+      "learning_rate": 2.221863799283154e-05,
+      "loss": 0.2747,
+      "step": 93010
+    },
+    {
+      "epoch": 600.13,
+      "learning_rate": 2.221505376344086e-05,
+      "loss": 0.2653,
+      "step": 93020
+    },
+    {
+      "epoch": 600.19,
+      "learning_rate": 2.221146953405018e-05,
+      "loss": 0.2022,
+      "step": 93030
+    },
+    {
+      "epoch": 600.26,
+      "learning_rate": 2.22078853046595e-05,
+      "loss": 0.2185,
+      "step": 93040
+    },
+    {
+      "epoch": 600.32,
+      "learning_rate": 2.2204301075268818e-05,
+      "loss": 0.2422,
+      "step": 93050
+    },
+    {
+      "epoch": 600.39,
+      "learning_rate": 2.2200716845878136e-05,
+      "loss": 0.2495,
+      "step": 93060
+    },
+    {
+      "epoch": 600.45,
+      "learning_rate": 2.2197132616487457e-05,
+      "loss": 0.2113,
+      "step": 93070
+    },
+    {
+      "epoch": 600.51,
+      "learning_rate": 2.2193548387096776e-05,
+      "loss": 0.1918,
+      "step": 93080
+    },
+    {
+      "epoch": 600.58,
+      "learning_rate": 2.2189964157706094e-05,
+      "loss": 0.2646,
+      "step": 93090
+    },
+    {
+      "epoch": 600.64,
+      "learning_rate": 2.2186379928315412e-05,
+      "loss": 0.2557,
+      "step": 93100
+    },
+    {
+      "epoch": 600.71,
+      "learning_rate": 2.2182795698924734e-05,
+      "loss": 0.2485,
+      "step": 93110
+    },
+    {
+      "epoch": 600.77,
+      "learning_rate": 2.2179211469534052e-05,
+      "loss": 0.2432,
+      "step": 93120
+    },
+    {
+      "epoch": 600.84,
+      "learning_rate": 2.217562724014337e-05,
+      "loss": 0.2422,
+      "step": 93130
+    },
+    {
+      "epoch": 600.9,
+      "learning_rate": 2.217204301075269e-05,
+      "loss": 0.2258,
+      "step": 93140
+    },
+    {
+      "epoch": 600.96,
+      "learning_rate": 2.216845878136201e-05,
+      "loss": 0.2407,
+      "step": 93150
+    },
+    {
+      "epoch": 601.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0734915733337402,
+      "eval_runtime": 0.8729,
+      "eval_samples_per_second": 230.259,
+      "eval_steps_per_second": 8.019,
+      "step": 93155
+    },
+    {
+      "epoch": 601.03,
+      "learning_rate": 2.2164874551971325e-05,
+      "loss": 0.2655,
+      "step": 93160
+    },
+    {
+      "epoch": 601.1,
+      "learning_rate": 2.2161290322580646e-05,
+      "loss": 0.2486,
+      "step": 93170
+    },
+    {
+      "epoch": 601.16,
+      "learning_rate": 2.2157706093189965e-05,
+      "loss": 0.2296,
+      "step": 93180
+    },
+    {
+      "epoch": 601.23,
+      "learning_rate": 2.2154121863799286e-05,
+      "loss": 0.2231,
+      "step": 93190
+    },
+    {
+      "epoch": 601.29,
+      "learning_rate": 2.21505376344086e-05,
+      "loss": 0.2381,
+      "step": 93200
+    },
+    {
+      "epoch": 601.35,
+      "learning_rate": 2.2146953405017923e-05,
+      "loss": 0.2521,
+      "step": 93210
+    },
+    {
+      "epoch": 601.42,
+      "learning_rate": 2.214336917562724e-05,
+      "loss": 0.2287,
+      "step": 93220
+    },
+    {
+      "epoch": 601.48,
+      "learning_rate": 2.2139784946236562e-05,
+      "loss": 0.245,
+      "step": 93230
+    },
+    {
+      "epoch": 601.55,
+      "learning_rate": 2.2136200716845877e-05,
+      "loss": 0.2392,
+      "step": 93240
+    },
+    {
+      "epoch": 601.61,
+      "learning_rate": 2.21326164874552e-05,
+      "loss": 0.2263,
+      "step": 93250
+    },
+    {
+      "epoch": 601.68,
+      "learning_rate": 2.2129032258064517e-05,
+      "loss": 0.2382,
+      "step": 93260
+    },
+    {
+      "epoch": 601.74,
+      "learning_rate": 2.212544802867384e-05,
+      "loss": 0.2269,
+      "step": 93270
+    },
+    {
+      "epoch": 601.8,
+      "learning_rate": 2.2121863799283153e-05,
+      "loss": 0.2393,
+      "step": 93280
+    },
+    {
+      "epoch": 601.87,
+      "learning_rate": 2.2118279569892475e-05,
+      "loss": 0.2883,
+      "step": 93290
+    },
+    {
+      "epoch": 601.93,
+      "learning_rate": 2.2114695340501793e-05,
+      "loss": 0.2407,
+      "step": 93300
+    },
+    {
+      "epoch": 602.0,
+      "learning_rate": 2.211111111111111e-05,
+      "loss": 0.2738,
+      "step": 93310
+    },
+    {
+      "epoch": 602.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 1.1207984685897827,
+      "eval_runtime": 0.8562,
+      "eval_samples_per_second": 234.759,
+      "eval_steps_per_second": 8.176,
+      "step": 93310
+    },
+    {
+      "epoch": 602.06,
+      "learning_rate": 2.210752688172043e-05,
+      "loss": 0.2625,
+      "step": 93320
+    },
+    {
+      "epoch": 602.13,
+      "learning_rate": 2.210394265232975e-05,
+      "loss": 0.2635,
+      "step": 93330
+    },
+    {
+      "epoch": 602.19,
+      "learning_rate": 2.210035842293907e-05,
+      "loss": 0.2457,
+      "step": 93340
+    },
+    {
+      "epoch": 602.26,
+      "learning_rate": 2.2096774193548388e-05,
+      "loss": 0.2312,
+      "step": 93350
+    },
+    {
+      "epoch": 602.32,
+      "learning_rate": 2.2093189964157706e-05,
+      "loss": 0.2692,
+      "step": 93360
+    },
+    {
+      "epoch": 602.39,
+      "learning_rate": 2.2089605734767027e-05,
+      "loss": 0.2195,
+      "step": 93370
+    },
+    {
+      "epoch": 602.45,
+      "learning_rate": 2.2086021505376346e-05,
+      "loss": 0.2237,
+      "step": 93380
+    },
+    {
+      "epoch": 602.51,
+      "learning_rate": 2.2082437275985664e-05,
+      "loss": 0.2233,
+      "step": 93390
+    },
+    {
+      "epoch": 602.58,
+      "learning_rate": 2.2078853046594982e-05,
+      "loss": 0.2566,
+      "step": 93400
+    },
+    {
+      "epoch": 602.64,
+      "learning_rate": 2.2075268817204304e-05,
+      "loss": 0.2842,
+      "step": 93410
+    },
+    {
+      "epoch": 602.71,
+      "learning_rate": 2.2071684587813622e-05,
+      "loss": 0.2197,
+      "step": 93420
+    },
+    {
+      "epoch": 602.77,
+      "learning_rate": 2.206810035842294e-05,
+      "loss": 0.2368,
+      "step": 93430
+    },
+    {
+      "epoch": 602.84,
+      "learning_rate": 2.2064516129032258e-05,
+      "loss": 0.2847,
+      "step": 93440
+    },
+    {
+      "epoch": 602.9,
+      "learning_rate": 2.206093189964158e-05,
+      "loss": 0.2482,
+      "step": 93450
+    },
+    {
+      "epoch": 602.96,
+      "learning_rate": 2.2057347670250898e-05,
+      "loss": 0.2413,
+      "step": 93460
+    },
+    {
+      "epoch": 603.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0706510543823242,
+      "eval_runtime": 0.8656,
+      "eval_samples_per_second": 232.217,
+      "eval_steps_per_second": 8.087,
+      "step": 93465
+    },
+    {
+      "epoch": 603.03,
+      "learning_rate": 2.2053763440860216e-05,
+      "loss": 0.2667,
+      "step": 93470
+    },
+    {
+      "epoch": 603.1,
+      "learning_rate": 2.2050179211469534e-05,
+      "loss": 0.2522,
+      "step": 93480
+    },
+    {
+      "epoch": 603.16,
+      "learning_rate": 2.2046594982078856e-05,
+      "loss": 0.2491,
+      "step": 93490
+    },
+    {
+      "epoch": 603.23,
+      "learning_rate": 2.2043010752688174e-05,
+      "loss": 0.2144,
+      "step": 93500
+    },
+    {
+      "epoch": 603.29,
+      "learning_rate": 2.2039426523297492e-05,
+      "loss": 0.239,
+      "step": 93510
+    },
+    {
+      "epoch": 603.35,
+      "learning_rate": 2.203584229390681e-05,
+      "loss": 0.235,
+      "step": 93520
+    },
+    {
+      "epoch": 603.42,
+      "learning_rate": 2.2032258064516132e-05,
+      "loss": 0.2318,
+      "step": 93530
+    },
+    {
+      "epoch": 603.48,
+      "learning_rate": 2.202867383512545e-05,
+      "loss": 0.2364,
+      "step": 93540
+    },
+    {
+      "epoch": 603.55,
+      "learning_rate": 2.202508960573477e-05,
+      "loss": 0.269,
+      "step": 93550
+    },
+    {
+      "epoch": 603.61,
+      "learning_rate": 2.2021505376344087e-05,
+      "loss": 0.2662,
+      "step": 93560
+    },
+    {
+      "epoch": 603.68,
+      "learning_rate": 2.2017921146953405e-05,
+      "loss": 0.261,
+      "step": 93570
+    },
+    {
+      "epoch": 603.74,
+      "learning_rate": 2.2014336917562727e-05,
+      "loss": 0.2718,
+      "step": 93580
+    },
+    {
+      "epoch": 603.8,
+      "learning_rate": 2.201075268817204e-05,
+      "loss": 0.2333,
+      "step": 93590
+    },
+    {
+      "epoch": 603.87,
+      "learning_rate": 2.2007168458781363e-05,
+      "loss": 0.2581,
+      "step": 93600
+    },
+    {
+      "epoch": 603.93,
+      "learning_rate": 2.200358422939068e-05,
+      "loss": 0.2095,
+      "step": 93610
+    },
+    {
+      "epoch": 604.0,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.2885,
+      "step": 93620
+    },
+    {
+      "epoch": 604.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 1.043394684791565,
+      "eval_runtime": 0.8774,
+      "eval_samples_per_second": 229.097,
+      "eval_steps_per_second": 7.979,
+      "step": 93620
+    },
+    {
+      "epoch": 604.06,
+      "learning_rate": 2.1996415770609318e-05,
+      "loss": 0.2073,
+      "step": 93630
+    },
+    {
+      "epoch": 604.13,
+      "learning_rate": 2.199283154121864e-05,
+      "loss": 0.2656,
+      "step": 93640
+    },
+    {
+      "epoch": 604.19,
+      "learning_rate": 2.1989247311827957e-05,
+      "loss": 0.2446,
+      "step": 93650
+    },
+    {
+      "epoch": 604.26,
+      "learning_rate": 2.1985663082437276e-05,
+      "loss": 0.2564,
+      "step": 93660
+    },
+    {
+      "epoch": 604.32,
+      "learning_rate": 2.1982078853046594e-05,
+      "loss": 0.1883,
+      "step": 93670
+    },
+    {
+      "epoch": 604.39,
+      "learning_rate": 2.1978494623655915e-05,
+      "loss": 0.2606,
+      "step": 93680
+    },
+    {
+      "epoch": 604.45,
+      "learning_rate": 2.1974910394265234e-05,
+      "loss": 0.2463,
+      "step": 93690
+    },
+    {
+      "epoch": 604.51,
+      "learning_rate": 2.1971326164874552e-05,
+      "loss": 0.2529,
+      "step": 93700
+    },
+    {
+      "epoch": 604.58,
+      "learning_rate": 2.196774193548387e-05,
+      "loss": 0.2676,
+      "step": 93710
+    },
+    {
+      "epoch": 604.64,
+      "learning_rate": 2.196415770609319e-05,
+      "loss": 0.2079,
+      "step": 93720
+    },
+    {
+      "epoch": 604.71,
+      "learning_rate": 2.196057347670251e-05,
+      "loss": 0.1858,
+      "step": 93730
+    },
+    {
+      "epoch": 604.77,
+      "learning_rate": 2.1956989247311828e-05,
+      "loss": 0.256,
+      "step": 93740
+    },
+    {
+      "epoch": 604.84,
+      "learning_rate": 2.1953405017921146e-05,
+      "loss": 0.2923,
+      "step": 93750
+    },
+    {
+      "epoch": 604.9,
+      "learning_rate": 2.1949820788530468e-05,
+      "loss": 0.2314,
+      "step": 93760
+    },
+    {
+      "epoch": 604.96,
+      "learning_rate": 2.1946236559139786e-05,
+      "loss": 0.2753,
+      "step": 93770
+    },
+    {
+      "epoch": 605.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0223796367645264,
+      "eval_runtime": 0.8768,
+      "eval_samples_per_second": 229.239,
+      "eval_steps_per_second": 7.983,
+      "step": 93775
+    },
+    {
+      "epoch": 605.03,
+      "learning_rate": 2.1942652329749104e-05,
+      "loss": 0.2003,
+      "step": 93780
+    },
+    {
+      "epoch": 605.1,
+      "learning_rate": 2.1939068100358422e-05,
+      "loss": 0.2667,
+      "step": 93790
+    },
+    {
+      "epoch": 605.16,
+      "learning_rate": 2.1935483870967744e-05,
+      "loss": 0.3062,
+      "step": 93800
+    },
+    {
+      "epoch": 605.23,
+      "learning_rate": 2.1931899641577062e-05,
+      "loss": 0.2729,
+      "step": 93810
+    },
+    {
+      "epoch": 605.29,
+      "learning_rate": 2.192831541218638e-05,
+      "loss": 0.2651,
+      "step": 93820
+    },
+    {
+      "epoch": 605.35,
+      "learning_rate": 2.19247311827957e-05,
+      "loss": 0.225,
+      "step": 93830
+    },
+    {
+      "epoch": 605.42,
+      "learning_rate": 2.192114695340502e-05,
+      "loss": 0.2837,
+      "step": 93840
+    },
+    {
+      "epoch": 605.48,
+      "learning_rate": 2.191756272401434e-05,
+      "loss": 0.2778,
+      "step": 93850
+    },
+    {
+      "epoch": 605.55,
+      "learning_rate": 2.1913978494623657e-05,
+      "loss": 0.2741,
+      "step": 93860
+    },
+    {
+      "epoch": 605.61,
+      "learning_rate": 2.1910394265232975e-05,
+      "loss": 0.2708,
+      "step": 93870
+    },
+    {
+      "epoch": 605.68,
+      "learning_rate": 2.1906810035842296e-05,
+      "loss": 0.2829,
+      "step": 93880
+    },
+    {
+      "epoch": 605.74,
+      "learning_rate": 2.1903225806451615e-05,
+      "loss": 0.264,
+      "step": 93890
+    },
+    {
+      "epoch": 605.8,
+      "learning_rate": 2.1899641577060933e-05,
+      "loss": 0.2319,
+      "step": 93900
+    },
+    {
+      "epoch": 605.87,
+      "learning_rate": 2.189605734767025e-05,
+      "loss": 0.2349,
+      "step": 93910
+    },
+    {
+      "epoch": 605.93,
+      "learning_rate": 2.1892473118279573e-05,
+      "loss": 0.2231,
+      "step": 93920
+    },
+    {
+      "epoch": 606.0,
+      "learning_rate": 2.188888888888889e-05,
+      "loss": 0.2308,
+      "step": 93930
+    },
+    {
+      "epoch": 606.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.013787865638733,
+      "eval_runtime": 0.8561,
+      "eval_samples_per_second": 234.781,
+      "eval_steps_per_second": 8.176,
+      "step": 93930
+    },
+    {
+      "epoch": 606.06,
+      "learning_rate": 2.188530465949821e-05,
+      "loss": 0.2977,
+      "step": 93940
+    },
+    {
+      "epoch": 606.13,
+      "learning_rate": 2.1881720430107527e-05,
+      "loss": 0.2651,
+      "step": 93950
+    },
+    {
+      "epoch": 606.19,
+      "learning_rate": 2.187813620071685e-05,
+      "loss": 0.2213,
+      "step": 93960
+    },
+    {
+      "epoch": 606.26,
+      "learning_rate": 2.1874551971326167e-05,
+      "loss": 0.2058,
+      "step": 93970
+    },
+    {
+      "epoch": 606.32,
+      "learning_rate": 2.1870967741935485e-05,
+      "loss": 0.2289,
+      "step": 93980
+    },
+    {
+      "epoch": 606.39,
+      "learning_rate": 2.1867383512544804e-05,
+      "loss": 0.2184,
+      "step": 93990
+    },
+    {
+      "epoch": 606.45,
+      "learning_rate": 2.1863799283154125e-05,
+      "loss": 0.2363,
+      "step": 94000
+    },
+    {
+      "epoch": 606.51,
+      "learning_rate": 2.186021505376344e-05,
+      "loss": 0.2187,
+      "step": 94010
+    },
+    {
+      "epoch": 606.58,
+      "learning_rate": 2.185663082437276e-05,
+      "loss": 0.2333,
+      "step": 94020
+    },
+    {
+      "epoch": 606.64,
+      "learning_rate": 2.185304659498208e-05,
+      "loss": 0.2349,
+      "step": 94030
+    },
+    {
+      "epoch": 606.71,
+      "learning_rate": 2.18494623655914e-05,
+      "loss": 0.2491,
+      "step": 94040
+    },
+    {
+      "epoch": 606.77,
+      "learning_rate": 2.1845878136200716e-05,
+      "loss": 0.2269,
+      "step": 94050
+    },
+    {
+      "epoch": 606.84,
+      "learning_rate": 2.1842293906810038e-05,
+      "loss": 0.2539,
+      "step": 94060
+    },
+    {
+      "epoch": 606.9,
+      "learning_rate": 2.1838709677419356e-05,
+      "loss": 0.2701,
+      "step": 94070
+    },
+    {
+      "epoch": 606.96,
+      "learning_rate": 2.1835125448028678e-05,
+      "loss": 0.2397,
+      "step": 94080
+    },
+    {
+      "epoch": 607.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9919421672821045,
+      "eval_runtime": 0.8693,
+      "eval_samples_per_second": 231.221,
+      "eval_steps_per_second": 8.052,
+      "step": 94085
+    },
+    {
+      "epoch": 607.03,
+      "learning_rate": 2.1831541218637992e-05,
+      "loss": 0.2726,
+      "step": 94090
+    },
+    {
+      "epoch": 607.1,
+      "learning_rate": 2.1827956989247314e-05,
+      "loss": 0.2226,
+      "step": 94100
+    },
+    {
+      "epoch": 607.16,
+      "learning_rate": 2.1824372759856632e-05,
+      "loss": 0.2368,
+      "step": 94110
+    },
+    {
+      "epoch": 607.23,
+      "learning_rate": 2.1820788530465954e-05,
+      "loss": 0.2461,
+      "step": 94120
+    },
+    {
+      "epoch": 607.29,
+      "learning_rate": 2.181720430107527e-05,
+      "loss": 0.3022,
+      "step": 94130
+    },
+    {
+      "epoch": 607.35,
+      "learning_rate": 2.181362007168459e-05,
+      "loss": 0.2749,
+      "step": 94140
+    },
+    {
+      "epoch": 607.42,
+      "learning_rate": 2.181003584229391e-05,
+      "loss": 0.2228,
+      "step": 94150
+    },
+    {
+      "epoch": 607.48,
+      "learning_rate": 2.1806451612903227e-05,
+      "loss": 0.2707,
+      "step": 94160
+    },
+    {
+      "epoch": 607.55,
+      "learning_rate": 2.1802867383512545e-05,
+      "loss": 0.2407,
+      "step": 94170
+    },
+    {
+      "epoch": 607.61,
+      "learning_rate": 2.1799283154121866e-05,
+      "loss": 0.2417,
+      "step": 94180
+    },
+    {
+      "epoch": 607.68,
+      "learning_rate": 2.1795698924731185e-05,
+      "loss": 0.2628,
+      "step": 94190
+    },
+    {
+      "epoch": 607.74,
+      "learning_rate": 2.1792114695340503e-05,
+      "loss": 0.2541,
+      "step": 94200
+    },
+    {
+      "epoch": 607.8,
+      "learning_rate": 2.178853046594982e-05,
+      "loss": 0.216,
+      "step": 94210
+    },
+    {
+      "epoch": 607.87,
+      "learning_rate": 2.178494623655914e-05,
+      "loss": 0.1942,
+      "step": 94220
+    },
+    {
+      "epoch": 607.93,
+      "learning_rate": 2.178136200716846e-05,
+      "loss": 0.2388,
+      "step": 94230
+    },
+    {
+      "epoch": 608.0,
+      "learning_rate": 2.177777777777778e-05,
+      "loss": 0.2793,
+      "step": 94240
+    },
+    {
+      "epoch": 608.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0313565731048584,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.48,
+      "eval_steps_per_second": 8.027,
+      "step": 94240
+    },
+    {
+      "epoch": 608.06,
+      "learning_rate": 2.1774193548387097e-05,
+      "loss": 0.2911,
+      "step": 94250
+    },
+    {
+      "epoch": 608.13,
+      "learning_rate": 2.1770609318996415e-05,
+      "loss": 0.3042,
+      "step": 94260
+    },
+    {
+      "epoch": 608.19,
+      "learning_rate": 2.1767025089605737e-05,
+      "loss": 0.2272,
+      "step": 94270
+    },
+    {
+      "epoch": 608.26,
+      "learning_rate": 2.1763440860215055e-05,
+      "loss": 0.2548,
+      "step": 94280
+    },
+    {
+      "epoch": 608.32,
+      "learning_rate": 2.1759856630824373e-05,
+      "loss": 0.2467,
+      "step": 94290
+    },
+    {
+      "epoch": 608.39,
+      "learning_rate": 2.175627240143369e-05,
+      "loss": 0.2567,
+      "step": 94300
+    },
+    {
+      "epoch": 608.45,
+      "learning_rate": 2.1752688172043013e-05,
+      "loss": 0.232,
+      "step": 94310
+    },
+    {
+      "epoch": 608.51,
+      "learning_rate": 2.1749103942652328e-05,
+      "loss": 0.3088,
+      "step": 94320
+    },
+    {
+      "epoch": 608.58,
+      "learning_rate": 2.174551971326165e-05,
+      "loss": 0.2567,
+      "step": 94330
+    },
+    {
+      "epoch": 608.64,
+      "learning_rate": 2.1741935483870968e-05,
+      "loss": 0.2201,
+      "step": 94340
+    },
+    {
+      "epoch": 608.71,
+      "learning_rate": 2.173835125448029e-05,
+      "loss": 0.299,
+      "step": 94350
+    },
+    {
+      "epoch": 608.77,
+      "learning_rate": 2.1734767025089604e-05,
+      "loss": 0.2569,
+      "step": 94360
+    },
+    {
+      "epoch": 608.84,
+      "learning_rate": 2.1731182795698926e-05,
+      "loss": 0.2524,
+      "step": 94370
+    },
+    {
+      "epoch": 608.9,
+      "learning_rate": 2.1727598566308244e-05,
+      "loss": 0.2115,
+      "step": 94380
+    },
+    {
+      "epoch": 608.96,
+      "learning_rate": 2.1724014336917566e-05,
+      "loss": 0.2233,
+      "step": 94390
+    },
+    {
+      "epoch": 609.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 1.0028773546218872,
+      "eval_runtime": 0.8792,
+      "eval_samples_per_second": 228.615,
+      "eval_steps_per_second": 7.962,
+      "step": 94395
+    },
+    {
+      "epoch": 609.03,
+      "learning_rate": 2.172043010752688e-05,
+      "loss": 0.283,
+      "step": 94400
+    },
+    {
+      "epoch": 609.1,
+      "learning_rate": 2.1716845878136202e-05,
+      "loss": 0.2064,
+      "step": 94410
+    },
+    {
+      "epoch": 609.16,
+      "learning_rate": 2.171326164874552e-05,
+      "loss": 0.2602,
+      "step": 94420
+    },
+    {
+      "epoch": 609.23,
+      "learning_rate": 2.1709677419354842e-05,
+      "loss": 0.2557,
+      "step": 94430
+    },
+    {
+      "epoch": 609.29,
+      "learning_rate": 2.1706093189964157e-05,
+      "loss": 0.2443,
+      "step": 94440
+    },
+    {
+      "epoch": 609.35,
+      "learning_rate": 2.1702508960573478e-05,
+      "loss": 0.2717,
+      "step": 94450
+    },
+    {
+      "epoch": 609.42,
+      "learning_rate": 2.1698924731182796e-05,
+      "loss": 0.2358,
+      "step": 94460
+    },
+    {
+      "epoch": 609.48,
+      "learning_rate": 2.1695340501792115e-05,
+      "loss": 0.214,
+      "step": 94470
+    },
+    {
+      "epoch": 609.55,
+      "learning_rate": 2.1691756272401433e-05,
+      "loss": 0.2688,
+      "step": 94480
+    },
+    {
+      "epoch": 609.61,
+      "learning_rate": 2.1688172043010754e-05,
+      "loss": 0.2403,
+      "step": 94490
+    },
+    {
+      "epoch": 609.68,
+      "learning_rate": 2.1684587813620073e-05,
+      "loss": 0.2775,
+      "step": 94500
+    },
+    {
+      "epoch": 609.74,
+      "learning_rate": 2.168100358422939e-05,
+      "loss": 0.1873,
+      "step": 94510
+    },
+    {
+      "epoch": 609.8,
+      "learning_rate": 2.167741935483871e-05,
+      "loss": 0.2742,
+      "step": 94520
+    },
+    {
+      "epoch": 609.87,
+      "learning_rate": 2.167383512544803e-05,
+      "loss": 0.2256,
+      "step": 94530
+    },
+    {
+      "epoch": 609.93,
+      "learning_rate": 2.167025089605735e-05,
+      "loss": 0.2472,
+      "step": 94540
+    },
+    {
+      "epoch": 610.0,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 0.2328,
+      "step": 94550
+    },
+    {
+      "epoch": 610.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9752150774002075,
+      "eval_runtime": 0.8601,
+      "eval_samples_per_second": 233.689,
+      "eval_steps_per_second": 8.138,
+      "step": 94550
+    },
+    {
+      "epoch": 610.06,
+      "learning_rate": 2.1663082437275985e-05,
+      "loss": 0.2578,
+      "step": 94560
+    },
+    {
+      "epoch": 610.13,
+      "learning_rate": 2.1659498207885307e-05,
+      "loss": 0.258,
+      "step": 94570
+    },
+    {
+      "epoch": 610.19,
+      "learning_rate": 2.1655913978494625e-05,
+      "loss": 0.2718,
+      "step": 94580
+    },
+    {
+      "epoch": 610.26,
+      "learning_rate": 2.1652329749103943e-05,
+      "loss": 0.2584,
+      "step": 94590
+    },
+    {
+      "epoch": 610.32,
+      "learning_rate": 2.164874551971326e-05,
+      "loss": 0.2263,
+      "step": 94600
+    },
+    {
+      "epoch": 610.39,
+      "learning_rate": 2.1645161290322583e-05,
+      "loss": 0.25,
+      "step": 94610
+    },
+    {
+      "epoch": 610.45,
+      "learning_rate": 2.16415770609319e-05,
+      "loss": 0.2367,
+      "step": 94620
+    },
+    {
+      "epoch": 610.51,
+      "learning_rate": 2.163799283154122e-05,
+      "loss": 0.3311,
+      "step": 94630
+    },
+    {
+      "epoch": 610.58,
+      "learning_rate": 2.1634408602150538e-05,
+      "loss": 0.2282,
+      "step": 94640
+    },
+    {
+      "epoch": 610.64,
+      "learning_rate": 2.163082437275986e-05,
+      "loss": 0.2623,
+      "step": 94650
+    },
+    {
+      "epoch": 610.71,
+      "learning_rate": 2.1627240143369177e-05,
+      "loss": 0.1993,
+      "step": 94660
+    },
+    {
+      "epoch": 610.77,
+      "learning_rate": 2.1623655913978496e-05,
+      "loss": 0.2506,
+      "step": 94670
+    },
+    {
+      "epoch": 610.84,
+      "learning_rate": 2.1620071684587814e-05,
+      "loss": 0.251,
+      "step": 94680
+    },
+    {
+      "epoch": 610.9,
+      "learning_rate": 2.1616487455197135e-05,
+      "loss": 0.2832,
+      "step": 94690
+    },
+    {
+      "epoch": 610.96,
+      "learning_rate": 2.1612903225806454e-05,
+      "loss": 0.2364,
+      "step": 94700
+    },
+    {
+      "epoch": 611.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9505053758621216,
+      "eval_runtime": 0.8851,
+      "eval_samples_per_second": 227.08,
+      "eval_steps_per_second": 7.908,
+      "step": 94705
+    },
+    {
+      "epoch": 611.03,
+      "learning_rate": 2.1609318996415772e-05,
+      "loss": 0.2673,
+      "step": 94710
+    },
+    {
+      "epoch": 611.1,
+      "learning_rate": 2.160573476702509e-05,
+      "loss": 0.2231,
+      "step": 94720
+    },
+    {
+      "epoch": 611.16,
+      "learning_rate": 2.160215053763441e-05,
+      "loss": 0.2224,
+      "step": 94730
+    },
+    {
+      "epoch": 611.23,
+      "learning_rate": 2.159856630824373e-05,
+      "loss": 0.2344,
+      "step": 94740
+    },
+    {
+      "epoch": 611.29,
+      "learning_rate": 2.1594982078853048e-05,
+      "loss": 0.2128,
+      "step": 94750
+    },
+    {
+      "epoch": 611.35,
+      "learning_rate": 2.1591397849462366e-05,
+      "loss": 0.2012,
+      "step": 94760
+    },
+    {
+      "epoch": 611.42,
+      "learning_rate": 2.1587813620071688e-05,
+      "loss": 0.2669,
+      "step": 94770
+    },
+    {
+      "epoch": 611.48,
+      "learning_rate": 2.1584229390681006e-05,
+      "loss": 0.2226,
+      "step": 94780
+    },
+    {
+      "epoch": 611.55,
+      "learning_rate": 2.1580645161290324e-05,
+      "loss": 0.2862,
+      "step": 94790
+    },
+    {
+      "epoch": 611.61,
+      "learning_rate": 2.1577060931899642e-05,
+      "loss": 0.2531,
+      "step": 94800
+    },
+    {
+      "epoch": 611.68,
+      "learning_rate": 2.1573476702508964e-05,
+      "loss": 0.2086,
+      "step": 94810
+    },
+    {
+      "epoch": 611.74,
+      "learning_rate": 2.156989247311828e-05,
+      "loss": 0.2621,
+      "step": 94820
+    },
+    {
+      "epoch": 611.8,
+      "learning_rate": 2.15663082437276e-05,
+      "loss": 0.2286,
+      "step": 94830
+    },
+    {
+      "epoch": 611.87,
+      "learning_rate": 2.156272401433692e-05,
+      "loss": 0.1891,
+      "step": 94840
+    },
+    {
+      "epoch": 611.93,
+      "learning_rate": 2.1559139784946237e-05,
+      "loss": 0.2371,
+      "step": 94850
+    },
+    {
+      "epoch": 612.0,
+      "learning_rate": 2.1555555555555555e-05,
+      "loss": 0.3001,
+      "step": 94860
+    },
+    {
+      "epoch": 612.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.907511830329895,
+      "eval_runtime": 0.9035,
+      "eval_samples_per_second": 222.458,
+      "eval_steps_per_second": 7.747,
+      "step": 94860
+    },
+    {
+      "epoch": 612.06,
+      "learning_rate": 2.1551971326164873e-05,
+      "loss": 0.2501,
+      "step": 94870
+    },
+    {
+      "epoch": 612.13,
+      "learning_rate": 2.1548387096774195e-05,
+      "loss": 0.2315,
+      "step": 94880
+    },
+    {
+      "epoch": 612.19,
+      "learning_rate": 2.1544802867383513e-05,
+      "loss": 0.2723,
+      "step": 94890
+    },
+    {
+      "epoch": 612.26,
+      "learning_rate": 2.154121863799283e-05,
+      "loss": 0.242,
+      "step": 94900
+    },
+    {
+      "epoch": 612.32,
+      "learning_rate": 2.153763440860215e-05,
+      "loss": 0.231,
+      "step": 94910
+    },
+    {
+      "epoch": 612.39,
+      "learning_rate": 2.153405017921147e-05,
+      "loss": 0.2471,
+      "step": 94920
+    },
+    {
+      "epoch": 612.45,
+      "learning_rate": 2.153046594982079e-05,
+      "loss": 0.2498,
+      "step": 94930
+    },
+    {
+      "epoch": 612.51,
+      "learning_rate": 2.1526881720430108e-05,
+      "loss": 0.2673,
+      "step": 94940
+    },
+    {
+      "epoch": 612.58,
+      "learning_rate": 2.1523297491039426e-05,
+      "loss": 0.257,
+      "step": 94950
+    },
+    {
+      "epoch": 612.64,
+      "learning_rate": 2.1519713261648747e-05,
+      "loss": 0.2253,
+      "step": 94960
+    },
+    {
+      "epoch": 612.71,
+      "learning_rate": 2.1516129032258066e-05,
+      "loss": 0.2334,
+      "step": 94970
+    },
+    {
+      "epoch": 612.77,
+      "learning_rate": 2.1512544802867384e-05,
+      "loss": 0.2454,
+      "step": 94980
+    },
+    {
+      "epoch": 612.84,
+      "learning_rate": 2.1508960573476702e-05,
+      "loss": 0.2722,
+      "step": 94990
+    },
+    {
+      "epoch": 612.9,
+      "learning_rate": 2.1505376344086024e-05,
+      "loss": 0.2144,
+      "step": 95000
+    },
+    {
+      "epoch": 612.96,
+      "learning_rate": 2.1501792114695342e-05,
+      "loss": 0.2396,
+      "step": 95010
+    },
+    {
+      "epoch": 613.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9856950044631958,
+      "eval_runtime": 0.886,
+      "eval_samples_per_second": 226.85,
+      "eval_steps_per_second": 7.9,
+      "step": 95015
+    },
+    {
+      "epoch": 613.03,
+      "learning_rate": 2.149820788530466e-05,
+      "loss": 0.2791,
+      "step": 95020
+    },
+    {
+      "epoch": 613.1,
+      "learning_rate": 2.1494623655913978e-05,
+      "loss": 0.1984,
+      "step": 95030
+    },
+    {
+      "epoch": 613.16,
+      "learning_rate": 2.14910394265233e-05,
+      "loss": 0.1837,
+      "step": 95040
+    },
+    {
+      "epoch": 613.23,
+      "learning_rate": 2.1487455197132618e-05,
+      "loss": 0.2415,
+      "step": 95050
+    },
+    {
+      "epoch": 613.29,
+      "learning_rate": 2.1483870967741936e-05,
+      "loss": 0.2408,
+      "step": 95060
+    },
+    {
+      "epoch": 613.35,
+      "learning_rate": 2.1480286738351254e-05,
+      "loss": 0.2268,
+      "step": 95070
+    },
+    {
+      "epoch": 613.42,
+      "learning_rate": 2.1476702508960576e-05,
+      "loss": 0.2,
+      "step": 95080
+    },
+    {
+      "epoch": 613.48,
+      "learning_rate": 2.1473118279569894e-05,
+      "loss": 0.1764,
+      "step": 95090
+    },
+    {
+      "epoch": 613.55,
+      "learning_rate": 2.1469534050179212e-05,
+      "loss": 0.2939,
+      "step": 95100
+    },
+    {
+      "epoch": 613.61,
+      "learning_rate": 2.146594982078853e-05,
+      "loss": 0.2421,
+      "step": 95110
+    },
+    {
+      "epoch": 613.68,
+      "learning_rate": 2.1462365591397852e-05,
+      "loss": 0.2533,
+      "step": 95120
+    },
+    {
+      "epoch": 613.74,
+      "learning_rate": 2.145878136200717e-05,
+      "loss": 0.2262,
+      "step": 95130
+    },
+    {
+      "epoch": 613.8,
+      "learning_rate": 2.145519713261649e-05,
+      "loss": 0.2463,
+      "step": 95140
+    },
+    {
+      "epoch": 613.87,
+      "learning_rate": 2.1451612903225807e-05,
+      "loss": 0.2192,
+      "step": 95150
+    },
+    {
+      "epoch": 613.93,
+      "learning_rate": 2.144802867383513e-05,
+      "loss": 0.1796,
+      "step": 95160
+    },
+    {
+      "epoch": 614.0,
+      "learning_rate": 2.1444444444444443e-05,
+      "loss": 0.2564,
+      "step": 95170
+    },
+    {
+      "epoch": 614.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.916637659072876,
+      "eval_runtime": 0.8873,
+      "eval_samples_per_second": 226.542,
+      "eval_steps_per_second": 7.89,
+      "step": 95170
+    },
+    {
+      "epoch": 614.06,
+      "learning_rate": 2.1440860215053765e-05,
+      "loss": 0.2947,
+      "step": 95180
+    },
+    {
+      "epoch": 614.13,
+      "learning_rate": 2.1437275985663083e-05,
+      "loss": 0.2201,
+      "step": 95190
+    },
+    {
+      "epoch": 614.19,
+      "learning_rate": 2.1433691756272405e-05,
+      "loss": 0.2189,
+      "step": 95200
+    },
+    {
+      "epoch": 614.26,
+      "learning_rate": 2.143010752688172e-05,
+      "loss": 0.2935,
+      "step": 95210
+    },
+    {
+      "epoch": 614.32,
+      "learning_rate": 2.142652329749104e-05,
+      "loss": 0.2662,
+      "step": 95220
+    },
+    {
+      "epoch": 614.39,
+      "learning_rate": 2.142293906810036e-05,
+      "loss": 0.3023,
+      "step": 95230
+    },
+    {
+      "epoch": 614.45,
+      "learning_rate": 2.141935483870968e-05,
+      "loss": 0.269,
+      "step": 95240
+    },
+    {
+      "epoch": 614.51,
+      "learning_rate": 2.1415770609318996e-05,
+      "loss": 0.2294,
+      "step": 95250
+    },
+    {
+      "epoch": 614.58,
+      "learning_rate": 2.1412186379928317e-05,
+      "loss": 0.263,
+      "step": 95260
+    },
+    {
+      "epoch": 614.64,
+      "learning_rate": 2.1408602150537635e-05,
+      "loss": 0.316,
+      "step": 95270
+    },
+    {
+      "epoch": 614.71,
+      "learning_rate": 2.1405017921146957e-05,
+      "loss": 0.2273,
+      "step": 95280
+    },
+    {
+      "epoch": 614.77,
+      "learning_rate": 2.1401433691756272e-05,
+      "loss": 0.2693,
+      "step": 95290
+    },
+    {
+      "epoch": 614.84,
+      "learning_rate": 2.1397849462365593e-05,
+      "loss": 0.2361,
+      "step": 95300
+    },
+    {
+      "epoch": 614.9,
+      "learning_rate": 2.139426523297491e-05,
+      "loss": 0.1981,
+      "step": 95310
+    },
+    {
+      "epoch": 614.96,
+      "learning_rate": 2.139068100358423e-05,
+      "loss": 0.2594,
+      "step": 95320
+    },
+    {
+      "epoch": 615.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8926971554756165,
+      "eval_runtime": 0.8756,
+      "eval_samples_per_second": 229.546,
+      "eval_steps_per_second": 7.994,
+      "step": 95325
+    },
+    {
+      "epoch": 615.03,
+      "learning_rate": 2.1387096774193548e-05,
+      "loss": 0.2392,
+      "step": 95330
+    },
+    {
+      "epoch": 615.1,
+      "learning_rate": 2.138351254480287e-05,
+      "loss": 0.2203,
+      "step": 95340
+    },
+    {
+      "epoch": 615.16,
+      "learning_rate": 2.1379928315412188e-05,
+      "loss": 0.2405,
+      "step": 95350
+    },
+    {
+      "epoch": 615.23,
+      "learning_rate": 2.1376344086021506e-05,
+      "loss": 0.245,
+      "step": 95360
+    },
+    {
+      "epoch": 615.29,
+      "learning_rate": 2.1372759856630824e-05,
+      "loss": 0.2803,
+      "step": 95370
+    },
+    {
+      "epoch": 615.35,
+      "learning_rate": 2.1369175627240146e-05,
+      "loss": 0.2482,
+      "step": 95380
+    },
+    {
+      "epoch": 615.42,
+      "learning_rate": 2.1365591397849464e-05,
+      "loss": 0.2414,
+      "step": 95390
+    },
+    {
+      "epoch": 615.48,
+      "learning_rate": 2.1362007168458782e-05,
+      "loss": 0.2924,
+      "step": 95400
+    },
+    {
+      "epoch": 615.55,
+      "learning_rate": 2.13584229390681e-05,
+      "loss": 0.2665,
+      "step": 95410
+    },
+    {
+      "epoch": 615.61,
+      "learning_rate": 2.1354838709677422e-05,
+      "loss": 0.2264,
+      "step": 95420
+    },
+    {
+      "epoch": 615.68,
+      "learning_rate": 2.135125448028674e-05,
+      "loss": 0.2256,
+      "step": 95430
+    },
+    {
+      "epoch": 615.74,
+      "learning_rate": 2.134767025089606e-05,
+      "loss": 0.246,
+      "step": 95440
+    },
+    {
+      "epoch": 615.8,
+      "learning_rate": 2.1344086021505377e-05,
+      "loss": 0.2467,
+      "step": 95450
+    },
+    {
+      "epoch": 615.87,
+      "learning_rate": 2.1340501792114698e-05,
+      "loss": 0.2686,
+      "step": 95460
+    },
+    {
+      "epoch": 615.93,
+      "learning_rate": 2.1336917562724016e-05,
+      "loss": 0.2425,
+      "step": 95470
+    },
+    {
+      "epoch": 616.0,
+      "learning_rate": 2.1333333333333335e-05,
+      "loss": 0.2285,
+      "step": 95480
+    },
+    {
+      "epoch": 616.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8763977289199829,
+      "eval_runtime": 0.8709,
+      "eval_samples_per_second": 230.794,
+      "eval_steps_per_second": 8.038,
+      "step": 95480
+    },
+    {
+      "epoch": 616.06,
+      "learning_rate": 2.1329749103942653e-05,
+      "loss": 0.2756,
+      "step": 95490
+    },
+    {
+      "epoch": 616.13,
+      "learning_rate": 2.132616487455197e-05,
+      "loss": 0.2665,
+      "step": 95500
+    },
+    {
+      "epoch": 616.19,
+      "learning_rate": 2.1322580645161293e-05,
+      "loss": 0.2733,
+      "step": 95510
+    },
+    {
+      "epoch": 616.26,
+      "learning_rate": 2.1318996415770607e-05,
+      "loss": 0.2662,
+      "step": 95520
+    },
+    {
+      "epoch": 616.32,
+      "learning_rate": 2.131541218637993e-05,
+      "loss": 0.2246,
+      "step": 95530
+    },
+    {
+      "epoch": 616.39,
+      "learning_rate": 2.1311827956989247e-05,
+      "loss": 0.2573,
+      "step": 95540
+    },
+    {
+      "epoch": 616.45,
+      "learning_rate": 2.130824372759857e-05,
+      "loss": 0.29,
+      "step": 95550
+    },
+    {
+      "epoch": 616.51,
+      "learning_rate": 2.1304659498207884e-05,
+      "loss": 0.2315,
+      "step": 95560
+    },
+    {
+      "epoch": 616.58,
+      "learning_rate": 2.1301075268817205e-05,
+      "loss": 0.2453,
+      "step": 95570
+    },
+    {
+      "epoch": 616.64,
+      "learning_rate": 2.1297491039426523e-05,
+      "loss": 0.2137,
+      "step": 95580
+    },
+    {
+      "epoch": 616.71,
+      "learning_rate": 2.1293906810035845e-05,
+      "loss": 0.2213,
+      "step": 95590
+    },
+    {
+      "epoch": 616.77,
+      "learning_rate": 2.129032258064516e-05,
+      "loss": 0.2276,
+      "step": 95600
+    },
+    {
+      "epoch": 616.84,
+      "learning_rate": 2.128673835125448e-05,
+      "loss": 0.245,
+      "step": 95610
+    },
+    {
+      "epoch": 616.9,
+      "learning_rate": 2.12831541218638e-05,
+      "loss": 0.2195,
+      "step": 95620
+    },
+    {
+      "epoch": 616.96,
+      "learning_rate": 2.127956989247312e-05,
+      "loss": 0.2996,
+      "step": 95630
+    },
+    {
+      "epoch": 617.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9115207195281982,
+      "eval_runtime": 0.88,
+      "eval_samples_per_second": 228.399,
+      "eval_steps_per_second": 7.954,
+      "step": 95635
+    },
+    {
+      "epoch": 617.03,
+      "learning_rate": 2.1275985663082436e-05,
+      "loss": 0.2257,
+      "step": 95640
+    },
+    {
+      "epoch": 617.1,
+      "learning_rate": 2.1272401433691758e-05,
+      "loss": 0.2155,
+      "step": 95650
+    },
+    {
+      "epoch": 617.16,
+      "learning_rate": 2.1268817204301076e-05,
+      "loss": 0.2453,
+      "step": 95660
+    },
+    {
+      "epoch": 617.23,
+      "learning_rate": 2.1265232974910394e-05,
+      "loss": 0.2215,
+      "step": 95670
+    },
+    {
+      "epoch": 617.29,
+      "learning_rate": 2.1261648745519712e-05,
+      "loss": 0.2401,
+      "step": 95680
+    },
+    {
+      "epoch": 617.35,
+      "learning_rate": 2.1258064516129034e-05,
+      "loss": 0.2523,
+      "step": 95690
+    },
+    {
+      "epoch": 617.42,
+      "learning_rate": 2.1254480286738352e-05,
+      "loss": 0.2171,
+      "step": 95700
+    },
+    {
+      "epoch": 617.48,
+      "learning_rate": 2.125089605734767e-05,
+      "loss": 0.2683,
+      "step": 95710
+    },
+    {
+      "epoch": 617.55,
+      "learning_rate": 2.124731182795699e-05,
+      "loss": 0.2331,
+      "step": 95720
+    },
+    {
+      "epoch": 617.61,
+      "learning_rate": 2.124372759856631e-05,
+      "loss": 0.269,
+      "step": 95730
+    },
+    {
+      "epoch": 617.68,
+      "learning_rate": 2.1240143369175628e-05,
+      "loss": 0.258,
+      "step": 95740
+    },
+    {
+      "epoch": 617.74,
+      "learning_rate": 2.1236559139784946e-05,
+      "loss": 0.2176,
+      "step": 95750
+    },
+    {
+      "epoch": 617.8,
+      "learning_rate": 2.1232974910394265e-05,
+      "loss": 0.2085,
+      "step": 95760
+    },
+    {
+      "epoch": 617.87,
+      "learning_rate": 2.1229390681003586e-05,
+      "loss": 0.2522,
+      "step": 95770
+    },
+    {
+      "epoch": 617.93,
+      "learning_rate": 2.1225806451612904e-05,
+      "loss": 0.2071,
+      "step": 95780
+    },
+    {
+      "epoch": 618.0,
+      "learning_rate": 2.1222222222222223e-05,
+      "loss": 0.2388,
+      "step": 95790
+    },
+    {
+      "epoch": 618.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9528090357780457,
+      "eval_runtime": 0.8592,
+      "eval_samples_per_second": 233.951,
+      "eval_steps_per_second": 8.148,
+      "step": 95790
+    },
+    {
+      "epoch": 618.06,
+      "learning_rate": 2.121863799283154e-05,
+      "loss": 0.1985,
+      "step": 95800
+    },
+    {
+      "epoch": 618.13,
+      "learning_rate": 2.1215053763440862e-05,
+      "loss": 0.237,
+      "step": 95810
+    },
+    {
+      "epoch": 618.19,
+      "learning_rate": 2.121146953405018e-05,
+      "loss": 0.1936,
+      "step": 95820
+    },
+    {
+      "epoch": 618.26,
+      "learning_rate": 2.12078853046595e-05,
+      "loss": 0.2426,
+      "step": 95830
+    },
+    {
+      "epoch": 618.32,
+      "learning_rate": 2.1204301075268817e-05,
+      "loss": 0.245,
+      "step": 95840
+    },
+    {
+      "epoch": 618.39,
+      "learning_rate": 2.120071684587814e-05,
+      "loss": 0.2526,
+      "step": 95850
+    },
+    {
+      "epoch": 618.45,
+      "learning_rate": 2.1197132616487457e-05,
+      "loss": 0.2933,
+      "step": 95860
+    },
+    {
+      "epoch": 618.51,
+      "learning_rate": 2.1193548387096775e-05,
+      "loss": 0.2523,
+      "step": 95870
+    },
+    {
+      "epoch": 618.58,
+      "learning_rate": 2.1189964157706093e-05,
+      "loss": 0.1984,
+      "step": 95880
+    },
+    {
+      "epoch": 618.64,
+      "learning_rate": 2.1186379928315415e-05,
+      "loss": 0.2234,
+      "step": 95890
+    },
+    {
+      "epoch": 618.71,
+      "learning_rate": 2.1182795698924733e-05,
+      "loss": 0.2632,
+      "step": 95900
+    },
+    {
+      "epoch": 618.77,
+      "learning_rate": 2.117921146953405e-05,
+      "loss": 0.228,
+      "step": 95910
+    },
+    {
+      "epoch": 618.84,
+      "learning_rate": 2.117562724014337e-05,
+      "loss": 0.1995,
+      "step": 95920
+    },
+    {
+      "epoch": 618.9,
+      "learning_rate": 2.117204301075269e-05,
+      "loss": 0.2466,
+      "step": 95930
+    },
+    {
+      "epoch": 618.96,
+      "learning_rate": 2.116845878136201e-05,
+      "loss": 0.2528,
+      "step": 95940
+    },
+    {
+      "epoch": 619.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8952215909957886,
+      "eval_runtime": 0.8915,
+      "eval_samples_per_second": 225.474,
+      "eval_steps_per_second": 7.852,
+      "step": 95945
+    },
+    {
+      "epoch": 619.03,
+      "learning_rate": 2.1164874551971328e-05,
+      "loss": 0.2453,
+      "step": 95950
+    },
+    {
+      "epoch": 619.1,
+      "learning_rate": 2.1161290322580646e-05,
+      "loss": 0.2264,
+      "step": 95960
+    },
+    {
+      "epoch": 619.16,
+      "learning_rate": 2.1157706093189967e-05,
+      "loss": 0.2738,
+      "step": 95970
+    },
+    {
+      "epoch": 619.23,
+      "learning_rate": 2.1154121863799286e-05,
+      "loss": 0.2452,
+      "step": 95980
+    },
+    {
+      "epoch": 619.29,
+      "learning_rate": 2.1150537634408604e-05,
+      "loss": 0.2361,
+      "step": 95990
+    },
+    {
+      "epoch": 619.35,
+      "learning_rate": 2.1146953405017922e-05,
+      "loss": 0.2258,
+      "step": 96000
+    },
+    {
+      "epoch": 619.42,
+      "learning_rate": 2.1143369175627244e-05,
+      "loss": 0.2463,
+      "step": 96010
+    },
+    {
+      "epoch": 619.48,
+      "learning_rate": 2.113978494623656e-05,
+      "loss": 0.2156,
+      "step": 96020
+    },
+    {
+      "epoch": 619.55,
+      "learning_rate": 2.113620071684588e-05,
+      "loss": 0.2178,
+      "step": 96030
+    },
+    {
+      "epoch": 619.61,
+      "learning_rate": 2.1132616487455198e-05,
+      "loss": 0.2119,
+      "step": 96040
+    },
+    {
+      "epoch": 619.68,
+      "learning_rate": 2.112903225806452e-05,
+      "loss": 0.2106,
+      "step": 96050
+    },
+    {
+      "epoch": 619.74,
+      "learning_rate": 2.1125448028673835e-05,
+      "loss": 0.3019,
+      "step": 96060
+    },
+    {
+      "epoch": 619.8,
+      "learning_rate": 2.1121863799283156e-05,
+      "loss": 0.2344,
+      "step": 96070
+    },
+    {
+      "epoch": 619.87,
+      "learning_rate": 2.1118279569892474e-05,
+      "loss": 0.3237,
+      "step": 96080
+    },
+    {
+      "epoch": 619.93,
+      "learning_rate": 2.1114695340501796e-05,
+      "loss": 0.2108,
+      "step": 96090
+    },
+    {
+      "epoch": 620.0,
+      "learning_rate": 2.111111111111111e-05,
+      "loss": 0.2728,
+      "step": 96100
+    },
+    {
+      "epoch": 620.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9580175280570984,
+      "eval_runtime": 0.8552,
+      "eval_samples_per_second": 235.031,
+      "eval_steps_per_second": 8.185,
+      "step": 96100
+    },
+    {
+      "epoch": 620.06,
+      "learning_rate": 2.1107526881720432e-05,
+      "loss": 0.2479,
+      "step": 96110
+    },
+    {
+      "epoch": 620.13,
+      "learning_rate": 2.110394265232975e-05,
+      "loss": 0.2156,
+      "step": 96120
+    },
+    {
+      "epoch": 620.19,
+      "learning_rate": 2.110035842293907e-05,
+      "loss": 0.2211,
+      "step": 96130
+    },
+    {
+      "epoch": 620.26,
+      "learning_rate": 2.1096774193548387e-05,
+      "loss": 0.2449,
+      "step": 96140
+    },
+    {
+      "epoch": 620.32,
+      "learning_rate": 2.1093189964157705e-05,
+      "loss": 0.2106,
+      "step": 96150
+    },
+    {
+      "epoch": 620.39,
+      "learning_rate": 2.1089605734767027e-05,
+      "loss": 0.2242,
+      "step": 96160
+    },
+    {
+      "epoch": 620.45,
+      "learning_rate": 2.1086021505376345e-05,
+      "loss": 0.2573,
+      "step": 96170
+    },
+    {
+      "epoch": 620.51,
+      "learning_rate": 2.1082437275985663e-05,
+      "loss": 0.204,
+      "step": 96180
+    },
+    {
+      "epoch": 620.58,
+      "learning_rate": 2.107885304659498e-05,
+      "loss": 0.2285,
+      "step": 96190
+    },
+    {
+      "epoch": 620.64,
+      "learning_rate": 2.1075268817204303e-05,
+      "loss": 0.2317,
+      "step": 96200
+    },
+    {
+      "epoch": 620.71,
+      "learning_rate": 2.107168458781362e-05,
+      "loss": 0.2662,
+      "step": 96210
+    },
+    {
+      "epoch": 620.77,
+      "learning_rate": 2.106810035842294e-05,
+      "loss": 0.2716,
+      "step": 96220
+    },
+    {
+      "epoch": 620.84,
+      "learning_rate": 2.1064516129032258e-05,
+      "loss": 0.2512,
+      "step": 96230
+    },
+    {
+      "epoch": 620.9,
+      "learning_rate": 2.106093189964158e-05,
+      "loss": 0.2395,
+      "step": 96240
+    },
+    {
+      "epoch": 620.96,
+      "learning_rate": 2.1057347670250897e-05,
+      "loss": 0.2326,
+      "step": 96250
+    },
+    {
+      "epoch": 621.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9124412536621094,
+      "eval_runtime": 0.8816,
+      "eval_samples_per_second": 228.007,
+      "eval_steps_per_second": 7.941,
+      "step": 96255
+    },
+    {
+      "epoch": 621.03,
+      "learning_rate": 2.1053763440860216e-05,
+      "loss": 0.2505,
+      "step": 96260
+    },
+    {
+      "epoch": 621.1,
+      "learning_rate": 2.1050179211469534e-05,
+      "loss": 0.2562,
+      "step": 96270
+    },
+    {
+      "epoch": 621.16,
+      "learning_rate": 2.1046594982078855e-05,
+      "loss": 0.2687,
+      "step": 96280
+    },
+    {
+      "epoch": 621.23,
+      "learning_rate": 2.1043010752688174e-05,
+      "loss": 0.2281,
+      "step": 96290
+    },
+    {
+      "epoch": 621.29,
+      "learning_rate": 2.1039426523297492e-05,
+      "loss": 0.2778,
+      "step": 96300
+    },
+    {
+      "epoch": 621.35,
+      "learning_rate": 2.103584229390681e-05,
+      "loss": 0.281,
+      "step": 96310
+    },
+    {
+      "epoch": 621.42,
+      "learning_rate": 2.103225806451613e-05,
+      "loss": 0.2538,
+      "step": 96320
+    },
+    {
+      "epoch": 621.48,
+      "learning_rate": 2.1028673835125446e-05,
+      "loss": 0.2358,
+      "step": 96330
+    },
+    {
+      "epoch": 621.55,
+      "learning_rate": 2.1025089605734768e-05,
+      "loss": 0.2452,
+      "step": 96340
+    },
+    {
+      "epoch": 621.61,
+      "learning_rate": 2.1021505376344086e-05,
+      "loss": 0.202,
+      "step": 96350
+    },
+    {
+      "epoch": 621.68,
+      "learning_rate": 2.1017921146953408e-05,
+      "loss": 0.2442,
+      "step": 96360
+    },
+    {
+      "epoch": 621.74,
+      "learning_rate": 2.1014336917562723e-05,
+      "loss": 0.2488,
+      "step": 96370
+    },
+    {
+      "epoch": 621.8,
+      "learning_rate": 2.1010752688172044e-05,
+      "loss": 0.2774,
+      "step": 96380
+    },
+    {
+      "epoch": 621.87,
+      "learning_rate": 2.1007168458781362e-05,
+      "loss": 0.1966,
+      "step": 96390
+    },
+    {
+      "epoch": 621.93,
+      "learning_rate": 2.1003584229390684e-05,
+      "loss": 0.2314,
+      "step": 96400
+    },
+    {
+      "epoch": 622.0,
+      "learning_rate": 2.1e-05,
+      "loss": 0.2516,
+      "step": 96410
+    },
+    {
+      "epoch": 622.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9475574493408203,
+      "eval_runtime": 0.8559,
+      "eval_samples_per_second": 234.847,
+      "eval_steps_per_second": 8.179,
+      "step": 96410
+    },
+    {
+      "epoch": 622.06,
+      "learning_rate": 2.099641577060932e-05,
+      "loss": 0.2701,
+      "step": 96420
+    },
+    {
+      "epoch": 622.13,
+      "learning_rate": 2.099283154121864e-05,
+      "loss": 0.2471,
+      "step": 96430
+    },
+    {
+      "epoch": 622.19,
+      "learning_rate": 2.098924731182796e-05,
+      "loss": 0.2232,
+      "step": 96440
+    },
+    {
+      "epoch": 622.26,
+      "learning_rate": 2.0985663082437275e-05,
+      "loss": 0.2464,
+      "step": 96450
+    },
+    {
+      "epoch": 622.32,
+      "learning_rate": 2.0982078853046597e-05,
+      "loss": 0.2143,
+      "step": 96460
+    },
+    {
+      "epoch": 622.39,
+      "learning_rate": 2.0978494623655915e-05,
+      "loss": 0.2513,
+      "step": 96470
+    },
+    {
+      "epoch": 622.45,
+      "learning_rate": 2.0974910394265233e-05,
+      "loss": 0.2208,
+      "step": 96480
+    },
+    {
+      "epoch": 622.51,
+      "learning_rate": 2.097132616487455e-05,
+      "loss": 0.2961,
+      "step": 96490
+    },
+    {
+      "epoch": 622.58,
+      "learning_rate": 2.0967741935483873e-05,
+      "loss": 0.2024,
+      "step": 96500
+    },
+    {
+      "epoch": 622.64,
+      "learning_rate": 2.096415770609319e-05,
+      "loss": 0.2599,
+      "step": 96510
+    },
+    {
+      "epoch": 622.71,
+      "learning_rate": 2.096057347670251e-05,
+      "loss": 0.2168,
+      "step": 96520
+    },
+    {
+      "epoch": 622.77,
+      "learning_rate": 2.0956989247311827e-05,
+      "loss": 0.2335,
+      "step": 96530
+    },
+    {
+      "epoch": 622.84,
+      "learning_rate": 2.095340501792115e-05,
+      "loss": 0.2611,
+      "step": 96540
+    },
+    {
+      "epoch": 622.9,
+      "learning_rate": 2.0949820788530467e-05,
+      "loss": 0.2245,
+      "step": 96550
+    },
+    {
+      "epoch": 622.96,
+      "learning_rate": 2.0946236559139785e-05,
+      "loss": 0.2256,
+      "step": 96560
+    },
+    {
+      "epoch": 623.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9561653137207031,
+      "eval_runtime": 0.8631,
+      "eval_samples_per_second": 232.873,
+      "eval_steps_per_second": 8.11,
+      "step": 96565
+    },
+    {
+      "epoch": 623.03,
+      "learning_rate": 2.0942652329749104e-05,
+      "loss": 0.2713,
+      "step": 96570
+    },
+    {
+      "epoch": 623.1,
+      "learning_rate": 2.0939068100358425e-05,
+      "loss": 0.2312,
+      "step": 96580
+    },
+    {
+      "epoch": 623.16,
+      "learning_rate": 2.0935483870967743e-05,
+      "loss": 0.2579,
+      "step": 96590
+    },
+    {
+      "epoch": 623.23,
+      "learning_rate": 2.093189964157706e-05,
+      "loss": 0.26,
+      "step": 96600
+    },
+    {
+      "epoch": 623.29,
+      "learning_rate": 2.092831541218638e-05,
+      "loss": 0.2099,
+      "step": 96610
+    },
+    {
+      "epoch": 623.35,
+      "learning_rate": 2.09247311827957e-05,
+      "loss": 0.2621,
+      "step": 96620
+    },
+    {
+      "epoch": 623.42,
+      "learning_rate": 2.092114695340502e-05,
+      "loss": 0.207,
+      "step": 96630
+    },
+    {
+      "epoch": 623.48,
+      "learning_rate": 2.0917562724014338e-05,
+      "loss": 0.2352,
+      "step": 96640
+    },
+    {
+      "epoch": 623.55,
+      "learning_rate": 2.0913978494623656e-05,
+      "loss": 0.2406,
+      "step": 96650
+    },
+    {
+      "epoch": 623.61,
+      "learning_rate": 2.0910394265232978e-05,
+      "loss": 0.2467,
+      "step": 96660
+    },
+    {
+      "epoch": 623.68,
+      "learning_rate": 2.0906810035842296e-05,
+      "loss": 0.2563,
+      "step": 96670
+    },
+    {
+      "epoch": 623.74,
+      "learning_rate": 2.0903225806451614e-05,
+      "loss": 0.2765,
+      "step": 96680
+    },
+    {
+      "epoch": 623.8,
+      "learning_rate": 2.0899641577060932e-05,
+      "loss": 0.192,
+      "step": 96690
+    },
+    {
+      "epoch": 623.87,
+      "learning_rate": 2.0896057347670254e-05,
+      "loss": 0.2533,
+      "step": 96700
+    },
+    {
+      "epoch": 623.93,
+      "learning_rate": 2.0892473118279572e-05,
+      "loss": 0.2164,
+      "step": 96710
+    },
+    {
+      "epoch": 624.0,
+      "learning_rate": 2.088888888888889e-05,
+      "loss": 0.2386,
+      "step": 96720
+    },
+    {
+      "epoch": 624.0,
+      "eval_accuracy": 0.8009950248756219,
+      "eval_loss": 0.9499607086181641,
+      "eval_runtime": 0.8573,
+      "eval_samples_per_second": 234.445,
+      "eval_steps_per_second": 8.165,
+      "step": 96720
+    },
+    {
+      "epoch": 624.06,
+      "learning_rate": 2.088530465949821e-05,
+      "loss": 0.2777,
+      "step": 96730
+    },
+    {
+      "epoch": 624.13,
+      "learning_rate": 2.088172043010753e-05,
+      "loss": 0.2287,
+      "step": 96740
+    },
+    {
+      "epoch": 624.19,
+      "learning_rate": 2.0878136200716848e-05,
+      "loss": 0.2798,
+      "step": 96750
+    },
+    {
+      "epoch": 624.26,
+      "learning_rate": 2.0874551971326167e-05,
+      "loss": 0.2114,
+      "step": 96760
+    },
+    {
+      "epoch": 624.32,
+      "learning_rate": 2.0870967741935485e-05,
+      "loss": 0.2235,
+      "step": 96770
+    },
+    {
+      "epoch": 624.39,
+      "learning_rate": 2.0867383512544803e-05,
+      "loss": 0.2318,
+      "step": 96780
+    },
+    {
+      "epoch": 624.45,
+      "learning_rate": 2.0863799283154125e-05,
+      "loss": 0.259,
+      "step": 96790
+    },
+    {
+      "epoch": 624.51,
+      "learning_rate": 2.086021505376344e-05,
+      "loss": 0.2595,
+      "step": 96800
+    },
+    {
+      "epoch": 624.58,
+      "learning_rate": 2.085663082437276e-05,
+      "loss": 0.2213,
+      "step": 96810
+    },
+    {
+      "epoch": 624.64,
+      "learning_rate": 2.085304659498208e-05,
+      "loss": 0.2112,
+      "step": 96820
+    },
+    {
+      "epoch": 624.71,
+      "learning_rate": 2.0849462365591397e-05,
+      "loss": 0.2399,
+      "step": 96830
+    },
+    {
+      "epoch": 624.77,
+      "learning_rate": 2.0845878136200716e-05,
+      "loss": 0.2018,
+      "step": 96840
+    },
+    {
+      "epoch": 624.84,
+      "learning_rate": 2.0842293906810037e-05,
+      "loss": 0.2572,
+      "step": 96850
+    },
+    {
+      "epoch": 624.9,
+      "learning_rate": 2.0838709677419355e-05,
+      "loss": 0.2414,
+      "step": 96860
+    },
+    {
+      "epoch": 624.96,
+      "learning_rate": 2.0835125448028674e-05,
+      "loss": 0.2267,
+      "step": 96870
+    },
+    {
+      "epoch": 625.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.956499457359314,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 231.259,
+      "eval_steps_per_second": 8.054,
+      "step": 96875
+    },
+    {
+      "epoch": 625.03,
+      "learning_rate": 2.0831541218637992e-05,
+      "loss": 0.2426,
+      "step": 96880
+    },
+    {
+      "epoch": 625.1,
+      "learning_rate": 2.0827956989247313e-05,
+      "loss": 0.2461,
+      "step": 96890
+    },
+    {
+      "epoch": 625.16,
+      "learning_rate": 2.082437275985663e-05,
+      "loss": 0.2778,
+      "step": 96900
+    },
+    {
+      "epoch": 625.23,
+      "learning_rate": 2.082078853046595e-05,
+      "loss": 0.2658,
+      "step": 96910
+    },
+    {
+      "epoch": 625.29,
+      "learning_rate": 2.0817204301075268e-05,
+      "loss": 0.2286,
+      "step": 96920
+    },
+    {
+      "epoch": 625.35,
+      "learning_rate": 2.081362007168459e-05,
+      "loss": 0.273,
+      "step": 96930
+    },
+    {
+      "epoch": 625.42,
+      "learning_rate": 2.0810035842293908e-05,
+      "loss": 0.2515,
+      "step": 96940
+    },
+    {
+      "epoch": 625.48,
+      "learning_rate": 2.0806451612903226e-05,
+      "loss": 0.2446,
+      "step": 96950
+    },
+    {
+      "epoch": 625.55,
+      "learning_rate": 2.0802867383512544e-05,
+      "loss": 0.2342,
+      "step": 96960
+    },
+    {
+      "epoch": 625.61,
+      "learning_rate": 2.0799283154121866e-05,
+      "loss": 0.2697,
+      "step": 96970
+    },
+    {
+      "epoch": 625.68,
+      "learning_rate": 2.0795698924731184e-05,
+      "loss": 0.2334,
+      "step": 96980
+    },
+    {
+      "epoch": 625.74,
+      "learning_rate": 2.0792114695340502e-05,
+      "loss": 0.2516,
+      "step": 96990
+    },
+    {
+      "epoch": 625.8,
+      "learning_rate": 2.078853046594982e-05,
+      "loss": 0.2166,
+      "step": 97000
+    },
+    {
+      "epoch": 625.87,
+      "learning_rate": 2.0784946236559142e-05,
+      "loss": 0.2372,
+      "step": 97010
+    },
+    {
+      "epoch": 625.93,
+      "learning_rate": 2.078136200716846e-05,
+      "loss": 0.2363,
+      "step": 97020
+    },
+    {
+      "epoch": 626.0,
+      "learning_rate": 2.077777777777778e-05,
+      "loss": 0.2277,
+      "step": 97030
+    },
+    {
+      "epoch": 626.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9514741897583008,
+      "eval_runtime": 0.8675,
+      "eval_samples_per_second": 231.695,
+      "eval_steps_per_second": 8.069,
+      "step": 97030
+    },
+    {
+      "epoch": 626.06,
+      "learning_rate": 2.0774193548387097e-05,
+      "loss": 0.2681,
+      "step": 97040
+    },
+    {
+      "epoch": 626.13,
+      "learning_rate": 2.0770609318996418e-05,
+      "loss": 0.2586,
+      "step": 97050
+    },
+    {
+      "epoch": 626.19,
+      "learning_rate": 2.0767025089605736e-05,
+      "loss": 0.2539,
+      "step": 97060
+    },
+    {
+      "epoch": 626.26,
+      "learning_rate": 2.0763440860215055e-05,
+      "loss": 0.2632,
+      "step": 97070
+    },
+    {
+      "epoch": 626.32,
+      "learning_rate": 2.0759856630824373e-05,
+      "loss": 0.2612,
+      "step": 97080
+    },
+    {
+      "epoch": 626.39,
+      "learning_rate": 2.0756272401433694e-05,
+      "loss": 0.2329,
+      "step": 97090
+    },
+    {
+      "epoch": 626.45,
+      "learning_rate": 2.0752688172043013e-05,
+      "loss": 0.2573,
+      "step": 97100
+    },
+    {
+      "epoch": 626.51,
+      "learning_rate": 2.074910394265233e-05,
+      "loss": 0.2546,
+      "step": 97110
+    },
+    {
+      "epoch": 626.58,
+      "learning_rate": 2.074551971326165e-05,
+      "loss": 0.2045,
+      "step": 97120
+    },
+    {
+      "epoch": 626.64,
+      "learning_rate": 2.074193548387097e-05,
+      "loss": 0.2113,
+      "step": 97130
+    },
+    {
+      "epoch": 626.71,
+      "learning_rate": 2.073835125448029e-05,
+      "loss": 0.2996,
+      "step": 97140
+    },
+    {
+      "epoch": 626.77,
+      "learning_rate": 2.0734767025089607e-05,
+      "loss": 0.2316,
+      "step": 97150
+    },
+    {
+      "epoch": 626.84,
+      "learning_rate": 2.0731182795698925e-05,
+      "loss": 0.2422,
+      "step": 97160
+    },
+    {
+      "epoch": 626.9,
+      "learning_rate": 2.0727598566308247e-05,
+      "loss": 0.217,
+      "step": 97170
+    },
+    {
+      "epoch": 626.96,
+      "learning_rate": 2.072401433691756e-05,
+      "loss": 0.246,
+      "step": 97180
+    },
+    {
+      "epoch": 627.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9422449469566345,
+      "eval_runtime": 0.8731,
+      "eval_samples_per_second": 230.214,
+      "eval_steps_per_second": 8.017,
+      "step": 97185
+    },
+    {
+      "epoch": 627.03,
+      "learning_rate": 2.0720430107526883e-05,
+      "loss": 0.2313,
+      "step": 97190
+    },
+    {
+      "epoch": 627.1,
+      "learning_rate": 2.07168458781362e-05,
+      "loss": 0.2545,
+      "step": 97200
+    },
+    {
+      "epoch": 627.16,
+      "learning_rate": 2.0713261648745523e-05,
+      "loss": 0.2764,
+      "step": 97210
+    },
+    {
+      "epoch": 627.23,
+      "learning_rate": 2.0709677419354838e-05,
+      "loss": 0.2737,
+      "step": 97220
+    },
+    {
+      "epoch": 627.29,
+      "learning_rate": 2.070609318996416e-05,
+      "loss": 0.2412,
+      "step": 97230
+    },
+    {
+      "epoch": 627.35,
+      "learning_rate": 2.0702508960573478e-05,
+      "loss": 0.2402,
+      "step": 97240
+    },
+    {
+      "epoch": 627.42,
+      "learning_rate": 2.06989247311828e-05,
+      "loss": 0.1983,
+      "step": 97250
+    },
+    {
+      "epoch": 627.48,
+      "learning_rate": 2.0695340501792114e-05,
+      "loss": 0.2757,
+      "step": 97260
+    },
+    {
+      "epoch": 627.55,
+      "learning_rate": 2.0691756272401436e-05,
+      "loss": 0.2164,
+      "step": 97270
+    },
+    {
+      "epoch": 627.61,
+      "learning_rate": 2.0688172043010754e-05,
+      "loss": 0.2624,
+      "step": 97280
+    },
+    {
+      "epoch": 627.68,
+      "learning_rate": 2.0684587813620075e-05,
+      "loss": 0.2473,
+      "step": 97290
+    },
+    {
+      "epoch": 627.74,
+      "learning_rate": 2.068100358422939e-05,
+      "loss": 0.2575,
+      "step": 97300
+    },
+    {
+      "epoch": 627.8,
+      "learning_rate": 2.0677419354838712e-05,
+      "loss": 0.2713,
+      "step": 97310
+    },
+    {
+      "epoch": 627.87,
+      "learning_rate": 2.067383512544803e-05,
+      "loss": 0.2673,
+      "step": 97320
+    },
+    {
+      "epoch": 627.93,
+      "learning_rate": 2.0670250896057348e-05,
+      "loss": 0.2511,
+      "step": 97330
+    },
+    {
+      "epoch": 628.0,
+      "learning_rate": 2.0666666666666666e-05,
+      "loss": 0.2166,
+      "step": 97340
+    },
+    {
+      "epoch": 628.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9369572401046753,
+      "eval_runtime": 0.9132,
+      "eval_samples_per_second": 220.106,
+      "eval_steps_per_second": 7.665,
+      "step": 97340
+    },
+    {
+      "epoch": 628.06,
+      "learning_rate": 2.0663082437275988e-05,
+      "loss": 0.2494,
+      "step": 97350
+    },
+    {
+      "epoch": 628.13,
+      "learning_rate": 2.0659498207885306e-05,
+      "loss": 0.2233,
+      "step": 97360
+    },
+    {
+      "epoch": 628.19,
+      "learning_rate": 2.0655913978494624e-05,
+      "loss": 0.2507,
+      "step": 97370
+    },
+    {
+      "epoch": 628.26,
+      "learning_rate": 2.0652329749103943e-05,
+      "loss": 0.2406,
+      "step": 97380
+    },
+    {
+      "epoch": 628.32,
+      "learning_rate": 2.0648745519713264e-05,
+      "loss": 0.2655,
+      "step": 97390
+    },
+    {
+      "epoch": 628.39,
+      "learning_rate": 2.0645161290322582e-05,
+      "loss": 0.2516,
+      "step": 97400
+    },
+    {
+      "epoch": 628.45,
+      "learning_rate": 2.06415770609319e-05,
+      "loss": 0.2319,
+      "step": 97410
+    },
+    {
+      "epoch": 628.51,
+      "learning_rate": 2.063799283154122e-05,
+      "loss": 0.2186,
+      "step": 97420
+    },
+    {
+      "epoch": 628.58,
+      "learning_rate": 2.0634408602150537e-05,
+      "loss": 0.2171,
+      "step": 97430
+    },
+    {
+      "epoch": 628.64,
+      "learning_rate": 2.063082437275986e-05,
+      "loss": 0.2635,
+      "step": 97440
+    },
+    {
+      "epoch": 628.71,
+      "learning_rate": 2.0627240143369177e-05,
+      "loss": 0.2569,
+      "step": 97450
+    },
+    {
+      "epoch": 628.77,
+      "learning_rate": 2.0623655913978495e-05,
+      "loss": 0.2395,
+      "step": 97460
+    },
+    {
+      "epoch": 628.84,
+      "learning_rate": 2.0620071684587813e-05,
+      "loss": 0.2203,
+      "step": 97470
+    },
+    {
+      "epoch": 628.9,
+      "learning_rate": 2.0616487455197135e-05,
+      "loss": 0.2885,
+      "step": 97480
+    },
+    {
+      "epoch": 628.96,
+      "learning_rate": 2.0612903225806453e-05,
+      "loss": 0.2297,
+      "step": 97490
+    },
+    {
+      "epoch": 629.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.927139937877655,
+      "eval_runtime": 0.8662,
+      "eval_samples_per_second": 232.044,
+      "eval_steps_per_second": 8.081,
+      "step": 97495
+    },
+    {
+      "epoch": 629.03,
+      "learning_rate": 2.060931899641577e-05,
+      "loss": 0.2669,
+      "step": 97500
+    },
+    {
+      "epoch": 629.1,
+      "learning_rate": 2.060573476702509e-05,
+      "loss": 0.2769,
+      "step": 97510
+    },
+    {
+      "epoch": 629.16,
+      "learning_rate": 2.060215053763441e-05,
+      "loss": 0.2455,
+      "step": 97520
+    },
+    {
+      "epoch": 629.23,
+      "learning_rate": 2.0598566308243726e-05,
+      "loss": 0.239,
+      "step": 97530
+    },
+    {
+      "epoch": 629.29,
+      "learning_rate": 2.0594982078853047e-05,
+      "loss": 0.2118,
+      "step": 97540
+    },
+    {
+      "epoch": 629.35,
+      "learning_rate": 2.0591397849462366e-05,
+      "loss": 0.1819,
+      "step": 97550
+    },
+    {
+      "epoch": 629.42,
+      "learning_rate": 2.0587813620071687e-05,
+      "loss": 0.2257,
+      "step": 97560
+    },
+    {
+      "epoch": 629.48,
+      "learning_rate": 2.0584229390681002e-05,
+      "loss": 0.2603,
+      "step": 97570
+    },
+    {
+      "epoch": 629.55,
+      "learning_rate": 2.0580645161290324e-05,
+      "loss": 0.2519,
+      "step": 97580
+    },
+    {
+      "epoch": 629.61,
+      "learning_rate": 2.0577060931899642e-05,
+      "loss": 0.2616,
+      "step": 97590
+    },
+    {
+      "epoch": 629.68,
+      "learning_rate": 2.0573476702508963e-05,
+      "loss": 0.281,
+      "step": 97600
+    },
+    {
+      "epoch": 629.74,
+      "learning_rate": 2.0569892473118278e-05,
+      "loss": 0.2071,
+      "step": 97610
+    },
+    {
+      "epoch": 629.8,
+      "learning_rate": 2.05663082437276e-05,
+      "loss": 0.2317,
+      "step": 97620
+    },
+    {
+      "epoch": 629.87,
+      "learning_rate": 2.0562724014336918e-05,
+      "loss": 0.1945,
+      "step": 97630
+    },
+    {
+      "epoch": 629.93,
+      "learning_rate": 2.055913978494624e-05,
+      "loss": 0.2751,
+      "step": 97640
+    },
+    {
+      "epoch": 630.0,
+      "learning_rate": 2.0555555555555555e-05,
+      "loss": 0.3154,
+      "step": 97650
+    },
+    {
+      "epoch": 630.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9398110508918762,
+      "eval_runtime": 0.8706,
+      "eval_samples_per_second": 230.869,
+      "eval_steps_per_second": 8.04,
+      "step": 97650
+    },
+    {
+      "epoch": 630.06,
+      "learning_rate": 2.0551971326164876e-05,
+      "loss": 0.2432,
+      "step": 97660
+    },
+    {
+      "epoch": 630.13,
+      "learning_rate": 2.0548387096774194e-05,
+      "loss": 0.2239,
+      "step": 97670
+    },
+    {
+      "epoch": 630.19,
+      "learning_rate": 2.0544802867383513e-05,
+      "loss": 0.2075,
+      "step": 97680
+    },
+    {
+      "epoch": 630.26,
+      "learning_rate": 2.054121863799283e-05,
+      "loss": 0.2285,
+      "step": 97690
+    },
+    {
+      "epoch": 630.32,
+      "learning_rate": 2.0537634408602152e-05,
+      "loss": 0.2494,
+      "step": 97700
+    },
+    {
+      "epoch": 630.39,
+      "learning_rate": 2.053405017921147e-05,
+      "loss": 0.2358,
+      "step": 97710
+    },
+    {
+      "epoch": 630.45,
+      "learning_rate": 2.053046594982079e-05,
+      "loss": 0.2547,
+      "step": 97720
+    },
+    {
+      "epoch": 630.51,
+      "learning_rate": 2.0526881720430107e-05,
+      "loss": 0.2058,
+      "step": 97730
+    },
+    {
+      "epoch": 630.58,
+      "learning_rate": 2.052329749103943e-05,
+      "loss": 0.2291,
+      "step": 97740
+    },
+    {
+      "epoch": 630.64,
+      "learning_rate": 2.0519713261648747e-05,
+      "loss": 0.2439,
+      "step": 97750
+    },
+    {
+      "epoch": 630.71,
+      "learning_rate": 2.0516129032258065e-05,
+      "loss": 0.2294,
+      "step": 97760
+    },
+    {
+      "epoch": 630.77,
+      "learning_rate": 2.0512544802867383e-05,
+      "loss": 0.2503,
+      "step": 97770
+    },
+    {
+      "epoch": 630.84,
+      "learning_rate": 2.0508960573476705e-05,
+      "loss": 0.2725,
+      "step": 97780
+    },
+    {
+      "epoch": 630.9,
+      "learning_rate": 2.0505376344086023e-05,
+      "loss": 0.2136,
+      "step": 97790
+    },
+    {
+      "epoch": 630.96,
+      "learning_rate": 2.050179211469534e-05,
+      "loss": 0.2179,
+      "step": 97800
+    },
+    {
+      "epoch": 631.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8988078832626343,
+      "eval_runtime": 0.873,
+      "eval_samples_per_second": 230.252,
+      "eval_steps_per_second": 8.019,
+      "step": 97805
+    },
+    {
+      "epoch": 631.03,
+      "learning_rate": 2.049820788530466e-05,
+      "loss": 0.238,
+      "step": 97810
+    },
+    {
+      "epoch": 631.1,
+      "learning_rate": 2.049462365591398e-05,
+      "loss": 0.2757,
+      "step": 97820
+    },
+    {
+      "epoch": 631.16,
+      "learning_rate": 2.04910394265233e-05,
+      "loss": 0.2097,
+      "step": 97830
+    },
+    {
+      "epoch": 631.23,
+      "learning_rate": 2.0487455197132617e-05,
+      "loss": 0.2765,
+      "step": 97840
+    },
+    {
+      "epoch": 631.29,
+      "learning_rate": 2.0483870967741936e-05,
+      "loss": 0.2496,
+      "step": 97850
+    },
+    {
+      "epoch": 631.35,
+      "learning_rate": 2.0480286738351257e-05,
+      "loss": 0.2357,
+      "step": 97860
+    },
+    {
+      "epoch": 631.42,
+      "learning_rate": 2.0476702508960575e-05,
+      "loss": 0.2324,
+      "step": 97870
+    },
+    {
+      "epoch": 631.48,
+      "learning_rate": 2.0473118279569894e-05,
+      "loss": 0.2296,
+      "step": 97880
+    },
+    {
+      "epoch": 631.55,
+      "learning_rate": 2.0469534050179212e-05,
+      "loss": 0.2169,
+      "step": 97890
+    },
+    {
+      "epoch": 631.61,
+      "learning_rate": 2.0465949820788533e-05,
+      "loss": 0.2538,
+      "step": 97900
+    },
+    {
+      "epoch": 631.68,
+      "learning_rate": 2.046236559139785e-05,
+      "loss": 0.274,
+      "step": 97910
+    },
+    {
+      "epoch": 631.74,
+      "learning_rate": 2.045878136200717e-05,
+      "loss": 0.2348,
+      "step": 97920
+    },
+    {
+      "epoch": 631.8,
+      "learning_rate": 2.0455197132616488e-05,
+      "loss": 0.2754,
+      "step": 97930
+    },
+    {
+      "epoch": 631.87,
+      "learning_rate": 2.045161290322581e-05,
+      "loss": 0.2319,
+      "step": 97940
+    },
+    {
+      "epoch": 631.93,
+      "learning_rate": 2.0448028673835128e-05,
+      "loss": 0.2641,
+      "step": 97950
+    },
+    {
+      "epoch": 632.0,
+      "learning_rate": 2.0444444444444446e-05,
+      "loss": 0.2315,
+      "step": 97960
+    },
+    {
+      "epoch": 632.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9267097115516663,
+      "eval_runtime": 0.9034,
+      "eval_samples_per_second": 222.499,
+      "eval_steps_per_second": 7.749,
+      "step": 97960
+    },
+    {
+      "epoch": 632.06,
+      "learning_rate": 2.0440860215053764e-05,
+      "loss": 0.333,
+      "step": 97970
+    },
+    {
+      "epoch": 632.13,
+      "learning_rate": 2.0437275985663086e-05,
+      "loss": 0.2792,
+      "step": 97980
+    },
+    {
+      "epoch": 632.19,
+      "learning_rate": 2.0433691756272404e-05,
+      "loss": 0.2145,
+      "step": 97990
+    },
+    {
+      "epoch": 632.26,
+      "learning_rate": 2.0430107526881722e-05,
+      "loss": 0.1924,
+      "step": 98000
+    },
+    {
+      "epoch": 632.32,
+      "learning_rate": 2.042652329749104e-05,
+      "loss": 0.2661,
+      "step": 98010
+    },
+    {
+      "epoch": 632.39,
+      "learning_rate": 2.0422939068100362e-05,
+      "loss": 0.2122,
+      "step": 98020
+    },
+    {
+      "epoch": 632.45,
+      "learning_rate": 2.0419354838709677e-05,
+      "loss": 0.2917,
+      "step": 98030
+    },
+    {
+      "epoch": 632.51,
+      "learning_rate": 2.0415770609319e-05,
+      "loss": 0.2652,
+      "step": 98040
+    },
+    {
+      "epoch": 632.58,
+      "learning_rate": 2.0412186379928317e-05,
+      "loss": 0.2625,
+      "step": 98050
+    },
+    {
+      "epoch": 632.64,
+      "learning_rate": 2.0408602150537635e-05,
+      "loss": 0.2481,
+      "step": 98060
+    },
+    {
+      "epoch": 632.71,
+      "learning_rate": 2.0405017921146953e-05,
+      "loss": 0.233,
+      "step": 98070
+    },
+    {
+      "epoch": 632.77,
+      "learning_rate": 2.040143369175627e-05,
+      "loss": 0.2519,
+      "step": 98080
+    },
+    {
+      "epoch": 632.84,
+      "learning_rate": 2.0397849462365593e-05,
+      "loss": 0.248,
+      "step": 98090
+    },
+    {
+      "epoch": 632.9,
+      "learning_rate": 2.039426523297491e-05,
+      "loss": 0.2133,
+      "step": 98100
+    },
+    {
+      "epoch": 632.96,
+      "learning_rate": 2.039068100358423e-05,
+      "loss": 0.1975,
+      "step": 98110
+    },
+    {
+      "epoch": 633.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9631497859954834,
+      "eval_runtime": 0.8846,
+      "eval_samples_per_second": 227.228,
+      "eval_steps_per_second": 7.913,
+      "step": 98115
+    },
+    {
+      "epoch": 633.03,
+      "learning_rate": 2.0387096774193547e-05,
+      "loss": 0.2557,
+      "step": 98120
+    },
+    {
+      "epoch": 633.1,
+      "learning_rate": 2.038351254480287e-05,
+      "loss": 0.2239,
+      "step": 98130
+    },
+    {
+      "epoch": 633.16,
+      "learning_rate": 2.0379928315412187e-05,
+      "loss": 0.2507,
+      "step": 98140
+    },
+    {
+      "epoch": 633.23,
+      "learning_rate": 2.0376344086021505e-05,
+      "loss": 0.2571,
+      "step": 98150
+    },
+    {
+      "epoch": 633.29,
+      "learning_rate": 2.0372759856630824e-05,
+      "loss": 0.2409,
+      "step": 98160
+    },
+    {
+      "epoch": 633.35,
+      "learning_rate": 2.0369175627240145e-05,
+      "loss": 0.2356,
+      "step": 98170
+    },
+    {
+      "epoch": 633.42,
+      "learning_rate": 2.0365591397849463e-05,
+      "loss": 0.234,
+      "step": 98180
+    },
+    {
+      "epoch": 633.48,
+      "learning_rate": 2.036200716845878e-05,
+      "loss": 0.2726,
+      "step": 98190
+    },
+    {
+      "epoch": 633.55,
+      "learning_rate": 2.03584229390681e-05,
+      "loss": 0.2572,
+      "step": 98200
+    },
+    {
+      "epoch": 633.61,
+      "learning_rate": 2.035483870967742e-05,
+      "loss": 0.2518,
+      "step": 98210
+    },
+    {
+      "epoch": 633.68,
+      "learning_rate": 2.035125448028674e-05,
+      "loss": 0.2604,
+      "step": 98220
+    },
+    {
+      "epoch": 633.74,
+      "learning_rate": 2.0347670250896058e-05,
+      "loss": 0.1968,
+      "step": 98230
+    },
+    {
+      "epoch": 633.8,
+      "learning_rate": 2.0344086021505376e-05,
+      "loss": 0.2374,
+      "step": 98240
+    },
+    {
+      "epoch": 633.87,
+      "learning_rate": 2.0340501792114698e-05,
+      "loss": 0.2104,
+      "step": 98250
+    },
+    {
+      "epoch": 633.93,
+      "learning_rate": 2.0336917562724016e-05,
+      "loss": 0.2472,
+      "step": 98260
+    },
+    {
+      "epoch": 634.0,
+      "learning_rate": 2.0333333333333334e-05,
+      "loss": 0.2101,
+      "step": 98270
+    },
+    {
+      "epoch": 634.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.981904149055481,
+      "eval_runtime": 0.8595,
+      "eval_samples_per_second": 233.87,
+      "eval_steps_per_second": 8.145,
+      "step": 98270
+    },
+    {
+      "epoch": 634.06,
+      "learning_rate": 2.0329749103942652e-05,
+      "loss": 0.2984,
+      "step": 98280
+    },
+    {
+      "epoch": 634.13,
+      "learning_rate": 2.0326164874551974e-05,
+      "loss": 0.2252,
+      "step": 98290
+    },
+    {
+      "epoch": 634.19,
+      "learning_rate": 2.0322580645161292e-05,
+      "loss": 0.2303,
+      "step": 98300
+    },
+    {
+      "epoch": 634.26,
+      "learning_rate": 2.031899641577061e-05,
+      "loss": 0.2422,
+      "step": 98310
+    },
+    {
+      "epoch": 634.32,
+      "learning_rate": 2.031541218637993e-05,
+      "loss": 0.2169,
+      "step": 98320
+    },
+    {
+      "epoch": 634.39,
+      "learning_rate": 2.031182795698925e-05,
+      "loss": 0.2428,
+      "step": 98330
+    },
+    {
+      "epoch": 634.45,
+      "learning_rate": 2.0308243727598565e-05,
+      "loss": 0.2477,
+      "step": 98340
+    },
+    {
+      "epoch": 634.51,
+      "learning_rate": 2.0304659498207886e-05,
+      "loss": 0.1879,
+      "step": 98350
+    },
+    {
+      "epoch": 634.58,
+      "learning_rate": 2.0301075268817205e-05,
+      "loss": 0.2411,
+      "step": 98360
+    },
+    {
+      "epoch": 634.64,
+      "learning_rate": 2.0297491039426526e-05,
+      "loss": 0.2331,
+      "step": 98370
+    },
+    {
+      "epoch": 634.71,
+      "learning_rate": 2.029390681003584e-05,
+      "loss": 0.2512,
+      "step": 98380
+    },
+    {
+      "epoch": 634.77,
+      "learning_rate": 2.0290322580645163e-05,
+      "loss": 0.2335,
+      "step": 98390
+    },
+    {
+      "epoch": 634.84,
+      "learning_rate": 2.028673835125448e-05,
+      "loss": 0.2134,
+      "step": 98400
+    },
+    {
+      "epoch": 634.9,
+      "learning_rate": 2.0283154121863802e-05,
+      "loss": 0.2081,
+      "step": 98410
+    },
+    {
+      "epoch": 634.96,
+      "learning_rate": 2.0279569892473117e-05,
+      "loss": 0.2755,
+      "step": 98420
+    },
+    {
+      "epoch": 635.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9725988507270813,
+      "eval_runtime": 0.8667,
+      "eval_samples_per_second": 231.925,
+      "eval_steps_per_second": 8.077,
+      "step": 98425
+    },
+    {
+      "epoch": 635.03,
+      "learning_rate": 2.027598566308244e-05,
+      "loss": 0.2895,
+      "step": 98430
+    },
+    {
+      "epoch": 635.1,
+      "learning_rate": 2.0272401433691757e-05,
+      "loss": 0.2684,
+      "step": 98440
+    },
+    {
+      "epoch": 635.16,
+      "learning_rate": 2.026881720430108e-05,
+      "loss": 0.2053,
+      "step": 98450
+    },
+    {
+      "epoch": 635.23,
+      "learning_rate": 2.0265232974910393e-05,
+      "loss": 0.2293,
+      "step": 98460
+    },
+    {
+      "epoch": 635.29,
+      "learning_rate": 2.0261648745519715e-05,
+      "loss": 0.2919,
+      "step": 98470
+    },
+    {
+      "epoch": 635.35,
+      "learning_rate": 2.0258064516129033e-05,
+      "loss": 0.2469,
+      "step": 98480
+    },
+    {
+      "epoch": 635.42,
+      "learning_rate": 2.0254480286738355e-05,
+      "loss": 0.2203,
+      "step": 98490
+    },
+    {
+      "epoch": 635.48,
+      "learning_rate": 2.025089605734767e-05,
+      "loss": 0.2816,
+      "step": 98500
+    },
+    {
+      "epoch": 635.55,
+      "learning_rate": 2.024731182795699e-05,
+      "loss": 0.2362,
+      "step": 98510
+    },
+    {
+      "epoch": 635.61,
+      "learning_rate": 2.024372759856631e-05,
+      "loss": 0.2633,
+      "step": 98520
+    },
+    {
+      "epoch": 635.68,
+      "learning_rate": 2.0240143369175628e-05,
+      "loss": 0.2826,
+      "step": 98530
+    },
+    {
+      "epoch": 635.74,
+      "learning_rate": 2.0236559139784946e-05,
+      "loss": 0.243,
+      "step": 98540
+    },
+    {
+      "epoch": 635.8,
+      "learning_rate": 2.0232974910394267e-05,
+      "loss": 0.2293,
+      "step": 98550
+    },
+    {
+      "epoch": 635.87,
+      "learning_rate": 2.0229390681003586e-05,
+      "loss": 0.2237,
+      "step": 98560
+    },
+    {
+      "epoch": 635.93,
+      "learning_rate": 2.0225806451612904e-05,
+      "loss": 0.2644,
+      "step": 98570
+    },
+    {
+      "epoch": 636.0,
+      "learning_rate": 2.0222222222222222e-05,
+      "loss": 0.2789,
+      "step": 98580
+    },
+    {
+      "epoch": 636.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0003150701522827,
+      "eval_runtime": 0.8762,
+      "eval_samples_per_second": 229.39,
+      "eval_steps_per_second": 7.989,
+      "step": 98580
+    },
+    {
+      "epoch": 636.06,
+      "learning_rate": 2.0218637992831544e-05,
+      "loss": 0.2839,
+      "step": 98590
+    },
+    {
+      "epoch": 636.13,
+      "learning_rate": 2.0215053763440862e-05,
+      "loss": 0.2342,
+      "step": 98600
+    },
+    {
+      "epoch": 636.19,
+      "learning_rate": 2.021146953405018e-05,
+      "loss": 0.2773,
+      "step": 98610
+    },
+    {
+      "epoch": 636.26,
+      "learning_rate": 2.02078853046595e-05,
+      "loss": 0.2426,
+      "step": 98620
+    },
+    {
+      "epoch": 636.32,
+      "learning_rate": 2.020430107526882e-05,
+      "loss": 0.2374,
+      "step": 98630
+    },
+    {
+      "epoch": 636.39,
+      "learning_rate": 2.0200716845878138e-05,
+      "loss": 0.2146,
+      "step": 98640
+    },
+    {
+      "epoch": 636.45,
+      "learning_rate": 2.0197132616487456e-05,
+      "loss": 0.2405,
+      "step": 98650
+    },
+    {
+      "epoch": 636.51,
+      "learning_rate": 2.0193548387096775e-05,
+      "loss": 0.267,
+      "step": 98660
+    },
+    {
+      "epoch": 636.58,
+      "learning_rate": 2.0189964157706096e-05,
+      "loss": 0.2382,
+      "step": 98670
+    },
+    {
+      "epoch": 636.64,
+      "learning_rate": 2.0186379928315414e-05,
+      "loss": 0.2548,
+      "step": 98680
+    },
+    {
+      "epoch": 636.71,
+      "learning_rate": 2.018279569892473e-05,
+      "loss": 0.2165,
+      "step": 98690
+    },
+    {
+      "epoch": 636.77,
+      "learning_rate": 2.017921146953405e-05,
+      "loss": 0.201,
+      "step": 98700
+    },
+    {
+      "epoch": 636.84,
+      "learning_rate": 2.017562724014337e-05,
+      "loss": 0.2335,
+      "step": 98710
+    },
+    {
+      "epoch": 636.9,
+      "learning_rate": 2.017204301075269e-05,
+      "loss": 0.2721,
+      "step": 98720
+    },
+    {
+      "epoch": 636.96,
+      "learning_rate": 2.0168458781362005e-05,
+      "loss": 0.2414,
+      "step": 98730
+    },
+    {
+      "epoch": 637.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0073676109313965,
+      "eval_runtime": 0.8713,
+      "eval_samples_per_second": 230.701,
+      "eval_steps_per_second": 8.034,
+      "step": 98735
+    },
+    {
+      "epoch": 637.03,
+      "learning_rate": 2.0164874551971327e-05,
+      "loss": 0.2741,
+      "step": 98740
+    },
+    {
+      "epoch": 637.1,
+      "learning_rate": 2.0161290322580645e-05,
+      "loss": 0.2172,
+      "step": 98750
+    },
+    {
+      "epoch": 637.16,
+      "learning_rate": 2.0157706093189967e-05,
+      "loss": 0.2316,
+      "step": 98760
+    },
+    {
+      "epoch": 637.23,
+      "learning_rate": 2.015412186379928e-05,
+      "loss": 0.2073,
+      "step": 98770
+    },
+    {
+      "epoch": 637.29,
+      "learning_rate": 2.0150537634408603e-05,
+      "loss": 0.2833,
+      "step": 98780
+    },
+    {
+      "epoch": 637.35,
+      "learning_rate": 2.014695340501792e-05,
+      "loss": 0.2114,
+      "step": 98790
+    },
+    {
+      "epoch": 637.42,
+      "learning_rate": 2.0143369175627243e-05,
+      "loss": 0.2435,
+      "step": 98800
+    },
+    {
+      "epoch": 637.48,
+      "learning_rate": 2.0139784946236558e-05,
+      "loss": 0.2499,
+      "step": 98810
+    },
+    {
+      "epoch": 637.55,
+      "learning_rate": 2.013620071684588e-05,
+      "loss": 0.2393,
+      "step": 98820
+    },
+    {
+      "epoch": 637.61,
+      "learning_rate": 2.0132616487455198e-05,
+      "loss": 0.2474,
+      "step": 98830
+    },
+    {
+      "epoch": 637.68,
+      "learning_rate": 2.0129032258064516e-05,
+      "loss": 0.228,
+      "step": 98840
+    },
+    {
+      "epoch": 637.74,
+      "learning_rate": 2.0125448028673834e-05,
+      "loss": 0.3352,
+      "step": 98850
+    },
+    {
+      "epoch": 637.8,
+      "learning_rate": 2.0121863799283156e-05,
+      "loss": 0.2231,
+      "step": 98860
+    },
+    {
+      "epoch": 637.87,
+      "learning_rate": 2.0118279569892474e-05,
+      "loss": 0.2542,
+      "step": 98870
+    },
+    {
+      "epoch": 637.93,
+      "learning_rate": 2.0114695340501792e-05,
+      "loss": 0.2338,
+      "step": 98880
+    },
+    {
+      "epoch": 638.0,
+      "learning_rate": 2.011111111111111e-05,
+      "loss": 0.2195,
+      "step": 98890
+    },
+    {
+      "epoch": 638.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9718272089958191,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.485,
+      "eval_steps_per_second": 8.027,
+      "step": 98890
+    },
+    {
+      "epoch": 638.06,
+      "learning_rate": 2.0107526881720432e-05,
+      "loss": 0.2398,
+      "step": 98900
+    },
+    {
+      "epoch": 638.13,
+      "learning_rate": 2.010394265232975e-05,
+      "loss": 0.2416,
+      "step": 98910
+    },
+    {
+      "epoch": 638.19,
+      "learning_rate": 2.0100358422939068e-05,
+      "loss": 0.229,
+      "step": 98920
+    },
+    {
+      "epoch": 638.26,
+      "learning_rate": 2.0096774193548386e-05,
+      "loss": 0.2683,
+      "step": 98930
+    },
+    {
+      "epoch": 638.32,
+      "learning_rate": 2.0093189964157708e-05,
+      "loss": 0.2271,
+      "step": 98940
+    },
+    {
+      "epoch": 638.39,
+      "learning_rate": 2.0089605734767026e-05,
+      "loss": 0.2662,
+      "step": 98950
+    },
+    {
+      "epoch": 638.45,
+      "learning_rate": 2.0086021505376344e-05,
+      "loss": 0.2724,
+      "step": 98960
+    },
+    {
+      "epoch": 638.51,
+      "learning_rate": 2.0082437275985663e-05,
+      "loss": 0.2041,
+      "step": 98970
+    },
+    {
+      "epoch": 638.58,
+      "learning_rate": 2.0078853046594984e-05,
+      "loss": 0.2493,
+      "step": 98980
+    },
+    {
+      "epoch": 638.64,
+      "learning_rate": 2.0075268817204302e-05,
+      "loss": 0.2075,
+      "step": 98990
+    },
+    {
+      "epoch": 638.71,
+      "learning_rate": 2.007168458781362e-05,
+      "loss": 0.2432,
+      "step": 99000
+    },
+    {
+      "epoch": 638.77,
+      "learning_rate": 2.006810035842294e-05,
+      "loss": 0.2079,
+      "step": 99010
+    },
+    {
+      "epoch": 638.84,
+      "learning_rate": 2.006451612903226e-05,
+      "loss": 0.2131,
+      "step": 99020
+    },
+    {
+      "epoch": 638.9,
+      "learning_rate": 2.006093189964158e-05,
+      "loss": 0.222,
+      "step": 99030
+    },
+    {
+      "epoch": 638.96,
+      "learning_rate": 2.0057347670250897e-05,
+      "loss": 0.234,
+      "step": 99040
+    },
+    {
+      "epoch": 639.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0192784070968628,
+      "eval_runtime": 0.8772,
+      "eval_samples_per_second": 229.133,
+      "eval_steps_per_second": 7.98,
+      "step": 99045
+    },
+    {
+      "epoch": 639.03,
+      "learning_rate": 2.0053763440860215e-05,
+      "loss": 0.2969,
+      "step": 99050
+    },
+    {
+      "epoch": 639.1,
+      "learning_rate": 2.0050179211469537e-05,
+      "loss": 0.2463,
+      "step": 99060
+    },
+    {
+      "epoch": 639.16,
+      "learning_rate": 2.0046594982078855e-05,
+      "loss": 0.2428,
+      "step": 99070
+    },
+    {
+      "epoch": 639.23,
+      "learning_rate": 2.0043010752688173e-05,
+      "loss": 0.2218,
+      "step": 99080
+    },
+    {
+      "epoch": 639.29,
+      "learning_rate": 2.003942652329749e-05,
+      "loss": 0.2252,
+      "step": 99090
+    },
+    {
+      "epoch": 639.35,
+      "learning_rate": 2.0035842293906813e-05,
+      "loss": 0.2589,
+      "step": 99100
+    },
+    {
+      "epoch": 639.42,
+      "learning_rate": 2.003225806451613e-05,
+      "loss": 0.2351,
+      "step": 99110
+    },
+    {
+      "epoch": 639.48,
+      "learning_rate": 2.002867383512545e-05,
+      "loss": 0.2275,
+      "step": 99120
+    },
+    {
+      "epoch": 639.55,
+      "learning_rate": 2.0025089605734767e-05,
+      "loss": 0.2381,
+      "step": 99130
+    },
+    {
+      "epoch": 639.61,
+      "learning_rate": 2.002150537634409e-05,
+      "loss": 0.2159,
+      "step": 99140
+    },
+    {
+      "epoch": 639.68,
+      "learning_rate": 2.0017921146953407e-05,
+      "loss": 0.2558,
+      "step": 99150
+    },
+    {
+      "epoch": 639.74,
+      "learning_rate": 2.0014336917562725e-05,
+      "loss": 0.2486,
+      "step": 99160
+    },
+    {
+      "epoch": 639.8,
+      "learning_rate": 2.0010752688172044e-05,
+      "loss": 0.1883,
+      "step": 99170
+    },
+    {
+      "epoch": 639.87,
+      "learning_rate": 2.0007168458781365e-05,
+      "loss": 0.238,
+      "step": 99180
+    },
+    {
+      "epoch": 639.93,
+      "learning_rate": 2.000358422939068e-05,
+      "loss": 0.2232,
+      "step": 99190
+    },
+    {
+      "epoch": 640.0,
+      "learning_rate": 2e-05,
+      "loss": 0.2291,
+      "step": 99200
+    },
+    {
+      "epoch": 640.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0392063856124878,
+      "eval_runtime": 0.8711,
+      "eval_samples_per_second": 230.732,
+      "eval_steps_per_second": 8.035,
+      "step": 99200
+    },
+    {
+      "epoch": 640.06,
+      "learning_rate": 1.999641577060932e-05,
+      "loss": 0.2399,
+      "step": 99210
+    },
+    {
+      "epoch": 640.13,
+      "learning_rate": 1.999283154121864e-05,
+      "loss": 0.2324,
+      "step": 99220
+    },
+    {
+      "epoch": 640.19,
+      "learning_rate": 1.9989247311827956e-05,
+      "loss": 0.2599,
+      "step": 99230
+    },
+    {
+      "epoch": 640.26,
+      "learning_rate": 1.9985663082437278e-05,
+      "loss": 0.2129,
+      "step": 99240
+    },
+    {
+      "epoch": 640.32,
+      "learning_rate": 1.9982078853046596e-05,
+      "loss": 0.2307,
+      "step": 99250
+    },
+    {
+      "epoch": 640.39,
+      "learning_rate": 1.9978494623655918e-05,
+      "loss": 0.2407,
+      "step": 99260
+    },
+    {
+      "epoch": 640.45,
+      "learning_rate": 1.9974910394265232e-05,
+      "loss": 0.196,
+      "step": 99270
+    },
+    {
+      "epoch": 640.51,
+      "learning_rate": 1.9971326164874554e-05,
+      "loss": 0.2718,
+      "step": 99280
+    },
+    {
+      "epoch": 640.58,
+      "learning_rate": 1.9967741935483872e-05,
+      "loss": 0.2594,
+      "step": 99290
+    },
+    {
+      "epoch": 640.64,
+      "learning_rate": 1.9964157706093194e-05,
+      "loss": 0.1936,
+      "step": 99300
+    },
+    {
+      "epoch": 640.71,
+      "learning_rate": 1.996057347670251e-05,
+      "loss": 0.2491,
+      "step": 99310
+    },
+    {
+      "epoch": 640.77,
+      "learning_rate": 1.995698924731183e-05,
+      "loss": 0.2578,
+      "step": 99320
+    },
+    {
+      "epoch": 640.84,
+      "learning_rate": 1.995340501792115e-05,
+      "loss": 0.2386,
+      "step": 99330
+    },
+    {
+      "epoch": 640.9,
+      "learning_rate": 1.9949820788530467e-05,
+      "loss": 0.2518,
+      "step": 99340
+    },
+    {
+      "epoch": 640.96,
+      "learning_rate": 1.9946236559139785e-05,
+      "loss": 0.2277,
+      "step": 99350
+    },
+    {
+      "epoch": 641.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9483751058578491,
+      "eval_runtime": 0.8946,
+      "eval_samples_per_second": 224.676,
+      "eval_steps_per_second": 7.825,
+      "step": 99355
+    },
+    {
+      "epoch": 641.03,
+      "learning_rate": 1.9942652329749103e-05,
+      "loss": 0.2648,
+      "step": 99360
+    },
+    {
+      "epoch": 641.1,
+      "learning_rate": 1.9939068100358425e-05,
+      "loss": 0.2297,
+      "step": 99370
+    },
+    {
+      "epoch": 641.16,
+      "learning_rate": 1.9935483870967743e-05,
+      "loss": 0.2544,
+      "step": 99380
+    },
+    {
+      "epoch": 641.23,
+      "learning_rate": 1.993189964157706e-05,
+      "loss": 0.2425,
+      "step": 99390
+    },
+    {
+      "epoch": 641.29,
+      "learning_rate": 1.992831541218638e-05,
+      "loss": 0.247,
+      "step": 99400
+    },
+    {
+      "epoch": 641.35,
+      "learning_rate": 1.99247311827957e-05,
+      "loss": 0.2003,
+      "step": 99410
+    },
+    {
+      "epoch": 641.42,
+      "learning_rate": 1.992114695340502e-05,
+      "loss": 0.2161,
+      "step": 99420
+    },
+    {
+      "epoch": 641.48,
+      "learning_rate": 1.9917562724014337e-05,
+      "loss": 0.2257,
+      "step": 99430
+    },
+    {
+      "epoch": 641.55,
+      "learning_rate": 1.9913978494623655e-05,
+      "loss": 0.2217,
+      "step": 99440
+    },
+    {
+      "epoch": 641.61,
+      "learning_rate": 1.9910394265232977e-05,
+      "loss": 0.2752,
+      "step": 99450
+    },
+    {
+      "epoch": 641.68,
+      "learning_rate": 1.9906810035842295e-05,
+      "loss": 0.2331,
+      "step": 99460
+    },
+    {
+      "epoch": 641.74,
+      "learning_rate": 1.9903225806451613e-05,
+      "loss": 0.2054,
+      "step": 99470
+    },
+    {
+      "epoch": 641.8,
+      "learning_rate": 1.9899641577060932e-05,
+      "loss": 0.2398,
+      "step": 99480
+    },
+    {
+      "epoch": 641.87,
+      "learning_rate": 1.9896057347670253e-05,
+      "loss": 0.2351,
+      "step": 99490
+    },
+    {
+      "epoch": 641.93,
+      "learning_rate": 1.989247311827957e-05,
+      "loss": 0.242,
+      "step": 99500
+    },
+    {
+      "epoch": 642.0,
+      "learning_rate": 1.988888888888889e-05,
+      "loss": 0.2233,
+      "step": 99510
+    },
+    {
+      "epoch": 642.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9749283790588379,
+      "eval_runtime": 0.8965,
+      "eval_samples_per_second": 224.209,
+      "eval_steps_per_second": 7.808,
+      "step": 99510
+    },
+    {
+      "epoch": 642.06,
+      "learning_rate": 1.9885304659498208e-05,
+      "loss": 0.264,
+      "step": 99520
+    },
+    {
+      "epoch": 642.13,
+      "learning_rate": 1.988172043010753e-05,
+      "loss": 0.2598,
+      "step": 99530
+    },
+    {
+      "epoch": 642.19,
+      "learning_rate": 1.9878136200716844e-05,
+      "loss": 0.2347,
+      "step": 99540
+    },
+    {
+      "epoch": 642.26,
+      "learning_rate": 1.9874551971326166e-05,
+      "loss": 0.2913,
+      "step": 99550
+    },
+    {
+      "epoch": 642.32,
+      "learning_rate": 1.9870967741935484e-05,
+      "loss": 0.2006,
+      "step": 99560
+    },
+    {
+      "epoch": 642.39,
+      "learning_rate": 1.9867383512544806e-05,
+      "loss": 0.2527,
+      "step": 99570
+    },
+    {
+      "epoch": 642.45,
+      "learning_rate": 1.986379928315412e-05,
+      "loss": 0.2241,
+      "step": 99580
+    },
+    {
+      "epoch": 642.51,
+      "learning_rate": 1.9860215053763442e-05,
+      "loss": 0.2333,
+      "step": 99590
+    },
+    {
+      "epoch": 642.58,
+      "learning_rate": 1.985663082437276e-05,
+      "loss": 0.2395,
+      "step": 99600
+    },
+    {
+      "epoch": 642.64,
+      "learning_rate": 1.9853046594982082e-05,
+      "loss": 0.2393,
+      "step": 99610
+    },
+    {
+      "epoch": 642.71,
+      "learning_rate": 1.9849462365591397e-05,
+      "loss": 0.2236,
+      "step": 99620
+    },
+    {
+      "epoch": 642.77,
+      "learning_rate": 1.984587813620072e-05,
+      "loss": 0.2631,
+      "step": 99630
+    },
+    {
+      "epoch": 642.84,
+      "learning_rate": 1.9842293906810037e-05,
+      "loss": 0.2136,
+      "step": 99640
+    },
+    {
+      "epoch": 642.9,
+      "learning_rate": 1.9838709677419358e-05,
+      "loss": 0.2437,
+      "step": 99650
+    },
+    {
+      "epoch": 642.96,
+      "learning_rate": 1.9835125448028673e-05,
+      "loss": 0.2945,
+      "step": 99660
+    },
+    {
+      "epoch": 643.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0419642925262451,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.119,
+      "eval_steps_per_second": 8.153,
+      "step": 99665
+    },
+    {
+      "epoch": 643.03,
+      "learning_rate": 1.9831541218637995e-05,
+      "loss": 0.2309,
+      "step": 99670
+    },
+    {
+      "epoch": 643.1,
+      "learning_rate": 1.9827956989247313e-05,
+      "loss": 0.2717,
+      "step": 99680
+    },
+    {
+      "epoch": 643.16,
+      "learning_rate": 1.982437275985663e-05,
+      "loss": 0.213,
+      "step": 99690
+    },
+    {
+      "epoch": 643.23,
+      "learning_rate": 1.982078853046595e-05,
+      "loss": 0.2467,
+      "step": 99700
+    },
+    {
+      "epoch": 643.29,
+      "learning_rate": 1.981720430107527e-05,
+      "loss": 0.2628,
+      "step": 99710
+    },
+    {
+      "epoch": 643.35,
+      "learning_rate": 1.981362007168459e-05,
+      "loss": 0.2596,
+      "step": 99720
+    },
+    {
+      "epoch": 643.42,
+      "learning_rate": 1.9810035842293907e-05,
+      "loss": 0.2316,
+      "step": 99730
+    },
+    {
+      "epoch": 643.48,
+      "learning_rate": 1.9806451612903225e-05,
+      "loss": 0.2407,
+      "step": 99740
+    },
+    {
+      "epoch": 643.55,
+      "learning_rate": 1.9802867383512547e-05,
+      "loss": 0.2263,
+      "step": 99750
+    },
+    {
+      "epoch": 643.61,
+      "learning_rate": 1.9799283154121865e-05,
+      "loss": 0.2677,
+      "step": 99760
+    },
+    {
+      "epoch": 643.68,
+      "learning_rate": 1.9795698924731183e-05,
+      "loss": 0.2243,
+      "step": 99770
+    },
+    {
+      "epoch": 643.74,
+      "learning_rate": 1.97921146953405e-05,
+      "loss": 0.2458,
+      "step": 99780
+    },
+    {
+      "epoch": 643.8,
+      "learning_rate": 1.9788530465949823e-05,
+      "loss": 0.2314,
+      "step": 99790
+    },
+    {
+      "epoch": 643.87,
+      "learning_rate": 1.978494623655914e-05,
+      "loss": 0.2599,
+      "step": 99800
+    },
+    {
+      "epoch": 643.93,
+      "learning_rate": 1.978136200716846e-05,
+      "loss": 0.2534,
+      "step": 99810
+    },
+    {
+      "epoch": 644.0,
+      "learning_rate": 1.9777777777777778e-05,
+      "loss": 0.1878,
+      "step": 99820
+    },
+    {
+      "epoch": 644.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.986842155456543,
+      "eval_runtime": 0.8652,
+      "eval_samples_per_second": 232.329,
+      "eval_steps_per_second": 8.091,
+      "step": 99820
+    },
+    {
+      "epoch": 644.06,
+      "learning_rate": 1.97741935483871e-05,
+      "loss": 0.2213,
+      "step": 99830
+    },
+    {
+      "epoch": 644.13,
+      "learning_rate": 1.9770609318996418e-05,
+      "loss": 0.2328,
+      "step": 99840
+    },
+    {
+      "epoch": 644.19,
+      "learning_rate": 1.9767025089605736e-05,
+      "loss": 0.2222,
+      "step": 99850
+    },
+    {
+      "epoch": 644.26,
+      "learning_rate": 1.9763440860215054e-05,
+      "loss": 0.2676,
+      "step": 99860
+    },
+    {
+      "epoch": 644.32,
+      "learning_rate": 1.9759856630824376e-05,
+      "loss": 0.2178,
+      "step": 99870
+    },
+    {
+      "epoch": 644.39,
+      "learning_rate": 1.9756272401433694e-05,
+      "loss": 0.2423,
+      "step": 99880
+    },
+    {
+      "epoch": 644.45,
+      "learning_rate": 1.9752688172043012e-05,
+      "loss": 0.2829,
+      "step": 99890
+    },
+    {
+      "epoch": 644.51,
+      "learning_rate": 1.974910394265233e-05,
+      "loss": 0.2355,
+      "step": 99900
+    },
+    {
+      "epoch": 644.58,
+      "learning_rate": 1.9745519713261652e-05,
+      "loss": 0.2324,
+      "step": 99910
+    },
+    {
+      "epoch": 644.64,
+      "learning_rate": 1.974193548387097e-05,
+      "loss": 0.2155,
+      "step": 99920
+    },
+    {
+      "epoch": 644.71,
+      "learning_rate": 1.9738351254480288e-05,
+      "loss": 0.2261,
+      "step": 99930
+    },
+    {
+      "epoch": 644.77,
+      "learning_rate": 1.9734767025089606e-05,
+      "loss": 0.2505,
+      "step": 99940
+    },
+    {
+      "epoch": 644.84,
+      "learning_rate": 1.9731182795698928e-05,
+      "loss": 0.2678,
+      "step": 99950
+    },
+    {
+      "epoch": 644.9,
+      "learning_rate": 1.9727598566308246e-05,
+      "loss": 0.1928,
+      "step": 99960
+    },
+    {
+      "epoch": 644.96,
+      "learning_rate": 1.972401433691756e-05,
+      "loss": 0.2441,
+      "step": 99970
+    },
+    {
+      "epoch": 645.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9590407013893127,
+      "eval_runtime": 0.8742,
+      "eval_samples_per_second": 229.921,
+      "eval_steps_per_second": 8.007,
+      "step": 99975
+    },
+    {
+      "epoch": 645.03,
+      "learning_rate": 1.9720430107526883e-05,
+      "loss": 0.2849,
+      "step": 99980
+    },
+    {
+      "epoch": 645.1,
+      "learning_rate": 1.97168458781362e-05,
+      "loss": 0.2883,
+      "step": 99990
+    },
+    {
+      "epoch": 645.16,
+      "learning_rate": 1.9713261648745522e-05,
+      "loss": 0.2267,
+      "step": 100000
+    },
+    {
+      "epoch": 645.23,
+      "learning_rate": 1.9709677419354837e-05,
+      "loss": 0.2349,
+      "step": 100010
+    },
+    {
+      "epoch": 645.29,
+      "learning_rate": 1.970609318996416e-05,
+      "loss": 0.231,
+      "step": 100020
+    },
+    {
+      "epoch": 645.35,
+      "learning_rate": 1.9702508960573477e-05,
+      "loss": 0.2293,
+      "step": 100030
+    },
+    {
+      "epoch": 645.42,
+      "learning_rate": 1.9698924731182795e-05,
+      "loss": 0.2349,
+      "step": 100040
+    },
+    {
+      "epoch": 645.48,
+      "learning_rate": 1.9695340501792113e-05,
+      "loss": 0.2568,
+      "step": 100050
+    },
+    {
+      "epoch": 645.55,
+      "learning_rate": 1.9691756272401435e-05,
+      "loss": 0.201,
+      "step": 100060
+    },
+    {
+      "epoch": 645.61,
+      "learning_rate": 1.9688172043010753e-05,
+      "loss": 0.2514,
+      "step": 100070
+    },
+    {
+      "epoch": 645.68,
+      "learning_rate": 1.968458781362007e-05,
+      "loss": 0.2258,
+      "step": 100080
+    },
+    {
+      "epoch": 645.74,
+      "learning_rate": 1.968100358422939e-05,
+      "loss": 0.2972,
+      "step": 100090
+    },
+    {
+      "epoch": 645.8,
+      "learning_rate": 1.967741935483871e-05,
+      "loss": 0.2376,
+      "step": 100100
+    },
+    {
+      "epoch": 645.87,
+      "learning_rate": 1.967383512544803e-05,
+      "loss": 0.2398,
+      "step": 100110
+    },
+    {
+      "epoch": 645.93,
+      "learning_rate": 1.9670250896057348e-05,
+      "loss": 0.2574,
+      "step": 100120
+    },
+    {
+      "epoch": 646.0,
+      "learning_rate": 1.9666666666666666e-05,
+      "loss": 0.1881,
+      "step": 100130
+    },
+    {
+      "epoch": 646.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.962207555770874,
+      "eval_runtime": 0.8728,
+      "eval_samples_per_second": 230.286,
+      "eval_steps_per_second": 8.02,
+      "step": 100130
+    },
+    {
+      "epoch": 646.06,
+      "learning_rate": 1.9663082437275987e-05,
+      "loss": 0.1958,
+      "step": 100140
+    },
+    {
+      "epoch": 646.13,
+      "learning_rate": 1.9659498207885306e-05,
+      "loss": 0.2571,
+      "step": 100150
+    },
+    {
+      "epoch": 646.19,
+      "learning_rate": 1.9655913978494624e-05,
+      "loss": 0.2616,
+      "step": 100160
+    },
+    {
+      "epoch": 646.26,
+      "learning_rate": 1.9652329749103942e-05,
+      "loss": 0.2771,
+      "step": 100170
+    },
+    {
+      "epoch": 646.32,
+      "learning_rate": 1.9648745519713264e-05,
+      "loss": 0.2068,
+      "step": 100180
+    },
+    {
+      "epoch": 646.39,
+      "learning_rate": 1.9645161290322582e-05,
+      "loss": 0.2528,
+      "step": 100190
+    },
+    {
+      "epoch": 646.45,
+      "learning_rate": 1.96415770609319e-05,
+      "loss": 0.2284,
+      "step": 100200
+    },
+    {
+      "epoch": 646.51,
+      "learning_rate": 1.9637992831541218e-05,
+      "loss": 0.2563,
+      "step": 100210
+    },
+    {
+      "epoch": 646.58,
+      "learning_rate": 1.963440860215054e-05,
+      "loss": 0.2423,
+      "step": 100220
+    },
+    {
+      "epoch": 646.64,
+      "learning_rate": 1.9630824372759858e-05,
+      "loss": 0.2236,
+      "step": 100230
+    },
+    {
+      "epoch": 646.71,
+      "learning_rate": 1.9627240143369176e-05,
+      "loss": 0.2662,
+      "step": 100240
+    },
+    {
+      "epoch": 646.77,
+      "learning_rate": 1.9623655913978494e-05,
+      "loss": 0.2107,
+      "step": 100250
+    },
+    {
+      "epoch": 646.84,
+      "learning_rate": 1.9620071684587816e-05,
+      "loss": 0.2169,
+      "step": 100260
+    },
+    {
+      "epoch": 646.9,
+      "learning_rate": 1.9616487455197134e-05,
+      "loss": 0.2412,
+      "step": 100270
+    },
+    {
+      "epoch": 646.96,
+      "learning_rate": 1.9612903225806452e-05,
+      "loss": 0.2301,
+      "step": 100280
+    },
+    {
+      "epoch": 647.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9129219651222229,
+      "eval_runtime": 0.8514,
+      "eval_samples_per_second": 236.09,
+      "eval_steps_per_second": 8.222,
+      "step": 100285
+    },
+    {
+      "epoch": 647.03,
+      "learning_rate": 1.960931899641577e-05,
+      "loss": 0.309,
+      "step": 100290
+    },
+    {
+      "epoch": 647.1,
+      "learning_rate": 1.9605734767025092e-05,
+      "loss": 0.224,
+      "step": 100300
+    },
+    {
+      "epoch": 647.16,
+      "learning_rate": 1.960215053763441e-05,
+      "loss": 0.2344,
+      "step": 100310
+    },
+    {
+      "epoch": 647.23,
+      "learning_rate": 1.959856630824373e-05,
+      "loss": 0.2363,
+      "step": 100320
+    },
+    {
+      "epoch": 647.29,
+      "learning_rate": 1.9594982078853047e-05,
+      "loss": 0.216,
+      "step": 100330
+    },
+    {
+      "epoch": 647.35,
+      "learning_rate": 1.959139784946237e-05,
+      "loss": 0.2549,
+      "step": 100340
+    },
+    {
+      "epoch": 647.42,
+      "learning_rate": 1.9587813620071683e-05,
+      "loss": 0.2481,
+      "step": 100350
+    },
+    {
+      "epoch": 647.48,
+      "learning_rate": 1.9584229390681005e-05,
+      "loss": 0.2328,
+      "step": 100360
+    },
+    {
+      "epoch": 647.55,
+      "learning_rate": 1.9580645161290323e-05,
+      "loss": 0.216,
+      "step": 100370
+    },
+    {
+      "epoch": 647.61,
+      "learning_rate": 1.9577060931899645e-05,
+      "loss": 0.2087,
+      "step": 100380
+    },
+    {
+      "epoch": 647.68,
+      "learning_rate": 1.957347670250896e-05,
+      "loss": 0.2033,
+      "step": 100390
+    },
+    {
+      "epoch": 647.74,
+      "learning_rate": 1.956989247311828e-05,
+      "loss": 0.162,
+      "step": 100400
+    },
+    {
+      "epoch": 647.8,
+      "learning_rate": 1.95663082437276e-05,
+      "loss": 0.2256,
+      "step": 100410
+    },
+    {
+      "epoch": 647.87,
+      "learning_rate": 1.956272401433692e-05,
+      "loss": 0.2273,
+      "step": 100420
+    },
+    {
+      "epoch": 647.93,
+      "learning_rate": 1.9559139784946236e-05,
+      "loss": 0.285,
+      "step": 100430
+    },
+    {
+      "epoch": 648.0,
+      "learning_rate": 1.9555555555555557e-05,
+      "loss": 0.2336,
+      "step": 100440
+    },
+    {
+      "epoch": 648.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9354284405708313,
+      "eval_runtime": 0.8711,
+      "eval_samples_per_second": 230.751,
+      "eval_steps_per_second": 8.036,
+      "step": 100440
+    },
+    {
+      "epoch": 648.06,
+      "learning_rate": 1.9551971326164876e-05,
+      "loss": 0.3357,
+      "step": 100450
+    },
+    {
+      "epoch": 648.13,
+      "learning_rate": 1.9548387096774197e-05,
+      "loss": 0.2353,
+      "step": 100460
+    },
+    {
+      "epoch": 648.19,
+      "learning_rate": 1.9544802867383512e-05,
+      "loss": 0.2321,
+      "step": 100470
+    },
+    {
+      "epoch": 648.26,
+      "learning_rate": 1.9541218637992834e-05,
+      "loss": 0.2322,
+      "step": 100480
+    },
+    {
+      "epoch": 648.32,
+      "learning_rate": 1.9537634408602152e-05,
+      "loss": 0.2151,
+      "step": 100490
+    },
+    {
+      "epoch": 648.39,
+      "learning_rate": 1.9534050179211473e-05,
+      "loss": 0.2129,
+      "step": 100500
+    },
+    {
+      "epoch": 648.45,
+      "learning_rate": 1.9530465949820788e-05,
+      "loss": 0.2434,
+      "step": 100510
+    },
+    {
+      "epoch": 648.51,
+      "learning_rate": 1.952688172043011e-05,
+      "loss": 0.2245,
+      "step": 100520
+    },
+    {
+      "epoch": 648.58,
+      "learning_rate": 1.9523297491039428e-05,
+      "loss": 0.2568,
+      "step": 100530
+    },
+    {
+      "epoch": 648.64,
+      "learning_rate": 1.9519713261648746e-05,
+      "loss": 0.2251,
+      "step": 100540
+    },
+    {
+      "epoch": 648.71,
+      "learning_rate": 1.9516129032258064e-05,
+      "loss": 0.1914,
+      "step": 100550
+    },
+    {
+      "epoch": 648.77,
+      "learning_rate": 1.9512544802867386e-05,
+      "loss": 0.1886,
+      "step": 100560
+    },
+    {
+      "epoch": 648.84,
+      "learning_rate": 1.9508960573476704e-05,
+      "loss": 0.2561,
+      "step": 100570
+    },
+    {
+      "epoch": 648.9,
+      "learning_rate": 1.9505376344086022e-05,
+      "loss": 0.2704,
+      "step": 100580
+    },
+    {
+      "epoch": 648.96,
+      "learning_rate": 1.950179211469534e-05,
+      "loss": 0.2027,
+      "step": 100590
+    },
+    {
+      "epoch": 649.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9889823794364929,
+      "eval_runtime": 0.8704,
+      "eval_samples_per_second": 230.919,
+      "eval_steps_per_second": 8.042,
+      "step": 100595
+    },
+    {
+      "epoch": 649.03,
+      "learning_rate": 1.9498207885304662e-05,
+      "loss": 0.2238,
+      "step": 100600
+    },
+    {
+      "epoch": 649.1,
+      "learning_rate": 1.949462365591398e-05,
+      "loss": 0.2632,
+      "step": 100610
+    },
+    {
+      "epoch": 649.16,
+      "learning_rate": 1.94910394265233e-05,
+      "loss": 0.2591,
+      "step": 100620
+    },
+    {
+      "epoch": 649.23,
+      "learning_rate": 1.9487455197132617e-05,
+      "loss": 0.2334,
+      "step": 100630
+    },
+    {
+      "epoch": 649.29,
+      "learning_rate": 1.9483870967741935e-05,
+      "loss": 0.2317,
+      "step": 100640
+    },
+    {
+      "epoch": 649.35,
+      "learning_rate": 1.9480286738351257e-05,
+      "loss": 0.2563,
+      "step": 100650
+    },
+    {
+      "epoch": 649.42,
+      "learning_rate": 1.9476702508960575e-05,
+      "loss": 0.2476,
+      "step": 100660
+    },
+    {
+      "epoch": 649.48,
+      "learning_rate": 1.9473118279569893e-05,
+      "loss": 0.242,
+      "step": 100670
+    },
+    {
+      "epoch": 649.55,
+      "learning_rate": 1.946953405017921e-05,
+      "loss": 0.243,
+      "step": 100680
+    },
+    {
+      "epoch": 649.61,
+      "learning_rate": 1.9465949820788533e-05,
+      "loss": 0.2762,
+      "step": 100690
+    },
+    {
+      "epoch": 649.68,
+      "learning_rate": 1.9462365591397848e-05,
+      "loss": 0.23,
+      "step": 100700
+    },
+    {
+      "epoch": 649.74,
+      "learning_rate": 1.945878136200717e-05,
+      "loss": 0.222,
+      "step": 100710
+    },
+    {
+      "epoch": 649.8,
+      "learning_rate": 1.9455197132616487e-05,
+      "loss": 0.2669,
+      "step": 100720
+    },
+    {
+      "epoch": 649.87,
+      "learning_rate": 1.945161290322581e-05,
+      "loss": 0.2437,
+      "step": 100730
+    },
+    {
+      "epoch": 649.93,
+      "learning_rate": 1.9448028673835124e-05,
+      "loss": 0.2752,
+      "step": 100740
+    },
+    {
+      "epoch": 650.0,
+      "learning_rate": 1.9444444444444445e-05,
+      "loss": 0.2837,
+      "step": 100750
+    },
+    {
+      "epoch": 650.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9972453713417053,
+      "eval_runtime": 0.8906,
+      "eval_samples_per_second": 225.682,
+      "eval_steps_per_second": 7.86,
+      "step": 100750
+    },
+    {
+      "epoch": 650.06,
+      "learning_rate": 1.9440860215053764e-05,
+      "loss": 0.2139,
+      "step": 100760
+    },
+    {
+      "epoch": 650.13,
+      "learning_rate": 1.9437275985663085e-05,
+      "loss": 0.2207,
+      "step": 100770
+    },
+    {
+      "epoch": 650.19,
+      "learning_rate": 1.94336917562724e-05,
+      "loss": 0.2261,
+      "step": 100780
+    },
+    {
+      "epoch": 650.26,
+      "learning_rate": 1.943010752688172e-05,
+      "loss": 0.234,
+      "step": 100790
+    },
+    {
+      "epoch": 650.32,
+      "learning_rate": 1.942652329749104e-05,
+      "loss": 0.2341,
+      "step": 100800
+    },
+    {
+      "epoch": 650.39,
+      "learning_rate": 1.942293906810036e-05,
+      "loss": 0.2241,
+      "step": 100810
+    },
+    {
+      "epoch": 650.45,
+      "learning_rate": 1.9419354838709676e-05,
+      "loss": 0.2247,
+      "step": 100820
+    },
+    {
+      "epoch": 650.51,
+      "learning_rate": 1.9415770609318998e-05,
+      "loss": 0.2694,
+      "step": 100830
+    },
+    {
+      "epoch": 650.58,
+      "learning_rate": 1.9412186379928316e-05,
+      "loss": 0.2505,
+      "step": 100840
+    },
+    {
+      "epoch": 650.64,
+      "learning_rate": 1.9408602150537634e-05,
+      "loss": 0.1975,
+      "step": 100850
+    },
+    {
+      "epoch": 650.71,
+      "learning_rate": 1.9405017921146952e-05,
+      "loss": 0.2542,
+      "step": 100860
+    },
+    {
+      "epoch": 650.77,
+      "learning_rate": 1.9401433691756274e-05,
+      "loss": 0.2183,
+      "step": 100870
+    },
+    {
+      "epoch": 650.84,
+      "learning_rate": 1.9397849462365592e-05,
+      "loss": 0.2372,
+      "step": 100880
+    },
+    {
+      "epoch": 650.9,
+      "learning_rate": 1.939426523297491e-05,
+      "loss": 0.2591,
+      "step": 100890
+    },
+    {
+      "epoch": 650.96,
+      "learning_rate": 1.939068100358423e-05,
+      "loss": 0.2293,
+      "step": 100900
+    },
+    {
+      "epoch": 651.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9634412527084351,
+      "eval_runtime": 0.8551,
+      "eval_samples_per_second": 235.054,
+      "eval_steps_per_second": 8.186,
+      "step": 100905
+    },
+    {
+      "epoch": 651.03,
+      "learning_rate": 1.938709677419355e-05,
+      "loss": 0.2588,
+      "step": 100910
+    },
+    {
+      "epoch": 651.1,
+      "learning_rate": 1.938351254480287e-05,
+      "loss": 0.2167,
+      "step": 100920
+    },
+    {
+      "epoch": 651.16,
+      "learning_rate": 1.9379928315412187e-05,
+      "loss": 0.1884,
+      "step": 100930
+    },
+    {
+      "epoch": 651.23,
+      "learning_rate": 1.9376344086021505e-05,
+      "loss": 0.232,
+      "step": 100940
+    },
+    {
+      "epoch": 651.29,
+      "learning_rate": 1.9372759856630826e-05,
+      "loss": 0.2638,
+      "step": 100950
+    },
+    {
+      "epoch": 651.35,
+      "learning_rate": 1.9369175627240145e-05,
+      "loss": 0.2324,
+      "step": 100960
+    },
+    {
+      "epoch": 651.42,
+      "learning_rate": 1.9365591397849463e-05,
+      "loss": 0.2328,
+      "step": 100970
+    },
+    {
+      "epoch": 651.48,
+      "learning_rate": 1.936200716845878e-05,
+      "loss": 0.2362,
+      "step": 100980
+    },
+    {
+      "epoch": 651.55,
+      "learning_rate": 1.9358422939068103e-05,
+      "loss": 0.2009,
+      "step": 100990
+    },
+    {
+      "epoch": 651.61,
+      "learning_rate": 1.935483870967742e-05,
+      "loss": 0.2401,
+      "step": 101000
+    },
+    {
+      "epoch": 651.68,
+      "learning_rate": 1.935125448028674e-05,
+      "loss": 0.2569,
+      "step": 101010
+    },
+    {
+      "epoch": 651.74,
+      "learning_rate": 1.9347670250896057e-05,
+      "loss": 0.2326,
+      "step": 101020
+    },
+    {
+      "epoch": 651.8,
+      "learning_rate": 1.934408602150538e-05,
+      "loss": 0.2604,
+      "step": 101030
+    },
+    {
+      "epoch": 651.87,
+      "learning_rate": 1.9340501792114697e-05,
+      "loss": 0.2485,
+      "step": 101040
+    },
+    {
+      "epoch": 651.93,
+      "learning_rate": 1.9336917562724015e-05,
+      "loss": 0.1733,
+      "step": 101050
+    },
+    {
+      "epoch": 652.0,
+      "learning_rate": 1.9333333333333333e-05,
+      "loss": 0.257,
+      "step": 101060
+    },
+    {
+      "epoch": 652.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9752476811408997,
+      "eval_runtime": 0.8605,
+      "eval_samples_per_second": 233.592,
+      "eval_steps_per_second": 8.135,
+      "step": 101060
+    },
+    {
+      "epoch": 652.06,
+      "learning_rate": 1.9329749103942655e-05,
+      "loss": 0.2909,
+      "step": 101070
+    },
+    {
+      "epoch": 652.13,
+      "learning_rate": 1.9326164874551973e-05,
+      "loss": 0.2559,
+      "step": 101080
+    },
+    {
+      "epoch": 652.19,
+      "learning_rate": 1.932258064516129e-05,
+      "loss": 0.2431,
+      "step": 101090
+    },
+    {
+      "epoch": 652.26,
+      "learning_rate": 1.931899641577061e-05,
+      "loss": 0.23,
+      "step": 101100
+    },
+    {
+      "epoch": 652.32,
+      "learning_rate": 1.931541218637993e-05,
+      "loss": 0.1668,
+      "step": 101110
+    },
+    {
+      "epoch": 652.39,
+      "learning_rate": 1.931182795698925e-05,
+      "loss": 0.2689,
+      "step": 101120
+    },
+    {
+      "epoch": 652.45,
+      "learning_rate": 1.9308243727598568e-05,
+      "loss": 0.2017,
+      "step": 101130
+    },
+    {
+      "epoch": 652.51,
+      "learning_rate": 1.9304659498207886e-05,
+      "loss": 0.2508,
+      "step": 101140
+    },
+    {
+      "epoch": 652.58,
+      "learning_rate": 1.9301075268817207e-05,
+      "loss": 0.1876,
+      "step": 101150
+    },
+    {
+      "epoch": 652.64,
+      "learning_rate": 1.9297491039426526e-05,
+      "loss": 0.2224,
+      "step": 101160
+    },
+    {
+      "epoch": 652.71,
+      "learning_rate": 1.9293906810035844e-05,
+      "loss": 0.235,
+      "step": 101170
+    },
+    {
+      "epoch": 652.77,
+      "learning_rate": 1.9290322580645162e-05,
+      "loss": 0.2236,
+      "step": 101180
+    },
+    {
+      "epoch": 652.84,
+      "learning_rate": 1.9286738351254484e-05,
+      "loss": 0.2748,
+      "step": 101190
+    },
+    {
+      "epoch": 652.9,
+      "learning_rate": 1.92831541218638e-05,
+      "loss": 0.228,
+      "step": 101200
+    },
+    {
+      "epoch": 652.96,
+      "learning_rate": 1.927956989247312e-05,
+      "loss": 0.2436,
+      "step": 101210
+    },
+    {
+      "epoch": 653.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9379125833511353,
+      "eval_runtime": 0.8793,
+      "eval_samples_per_second": 228.598,
+      "eval_steps_per_second": 7.961,
+      "step": 101215
+    },
+    {
+      "epoch": 653.03,
+      "learning_rate": 1.9275985663082438e-05,
+      "loss": 0.2185,
+      "step": 101220
+    },
+    {
+      "epoch": 653.1,
+      "learning_rate": 1.927240143369176e-05,
+      "loss": 0.2168,
+      "step": 101230
+    },
+    {
+      "epoch": 653.16,
+      "learning_rate": 1.9268817204301075e-05,
+      "loss": 0.2549,
+      "step": 101240
+    },
+    {
+      "epoch": 653.23,
+      "learning_rate": 1.9265232974910393e-05,
+      "loss": 0.239,
+      "step": 101250
+    },
+    {
+      "epoch": 653.29,
+      "learning_rate": 1.9261648745519714e-05,
+      "loss": 0.2125,
+      "step": 101260
+    },
+    {
+      "epoch": 653.35,
+      "learning_rate": 1.9258064516129033e-05,
+      "loss": 0.2376,
+      "step": 101270
+    },
+    {
+      "epoch": 653.42,
+      "learning_rate": 1.925448028673835e-05,
+      "loss": 0.2337,
+      "step": 101280
+    },
+    {
+      "epoch": 653.48,
+      "learning_rate": 1.925089605734767e-05,
+      "loss": 0.2434,
+      "step": 101290
+    },
+    {
+      "epoch": 653.55,
+      "learning_rate": 1.924731182795699e-05,
+      "loss": 0.2685,
+      "step": 101300
+    },
+    {
+      "epoch": 653.61,
+      "learning_rate": 1.924372759856631e-05,
+      "loss": 0.2437,
+      "step": 101310
+    },
+    {
+      "epoch": 653.68,
+      "learning_rate": 1.9240143369175627e-05,
+      "loss": 0.2164,
+      "step": 101320
+    },
+    {
+      "epoch": 653.74,
+      "learning_rate": 1.9236559139784945e-05,
+      "loss": 0.2784,
+      "step": 101330
+    },
+    {
+      "epoch": 653.8,
+      "learning_rate": 1.9232974910394267e-05,
+      "loss": 0.2238,
+      "step": 101340
+    },
+    {
+      "epoch": 653.87,
+      "learning_rate": 1.9229390681003585e-05,
+      "loss": 0.2961,
+      "step": 101350
+    },
+    {
+      "epoch": 653.93,
+      "learning_rate": 1.9225806451612903e-05,
+      "loss": 0.1886,
+      "step": 101360
+    },
+    {
+      "epoch": 654.0,
+      "learning_rate": 1.922222222222222e-05,
+      "loss": 0.2145,
+      "step": 101370
+    },
+    {
+      "epoch": 654.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9256795048713684,
+      "eval_runtime": 0.8572,
+      "eval_samples_per_second": 234.485,
+      "eval_steps_per_second": 8.166,
+      "step": 101370
+    },
+    {
+      "epoch": 654.06,
+      "learning_rate": 1.9218637992831543e-05,
+      "loss": 0.2336,
+      "step": 101380
+    },
+    {
+      "epoch": 654.13,
+      "learning_rate": 1.921505376344086e-05,
+      "loss": 0.2331,
+      "step": 101390
+    },
+    {
+      "epoch": 654.19,
+      "learning_rate": 1.921146953405018e-05,
+      "loss": 0.1928,
+      "step": 101400
+    },
+    {
+      "epoch": 654.26,
+      "learning_rate": 1.9207885304659498e-05,
+      "loss": 0.2281,
+      "step": 101410
+    },
+    {
+      "epoch": 654.32,
+      "learning_rate": 1.920430107526882e-05,
+      "loss": 0.2641,
+      "step": 101420
+    },
+    {
+      "epoch": 654.39,
+      "learning_rate": 1.9200716845878138e-05,
+      "loss": 0.216,
+      "step": 101430
+    },
+    {
+      "epoch": 654.45,
+      "learning_rate": 1.9197132616487456e-05,
+      "loss": 0.2316,
+      "step": 101440
+    },
+    {
+      "epoch": 654.51,
+      "learning_rate": 1.9193548387096774e-05,
+      "loss": 0.2535,
+      "step": 101450
+    },
+    {
+      "epoch": 654.58,
+      "learning_rate": 1.9189964157706096e-05,
+      "loss": 0.2832,
+      "step": 101460
+    },
+    {
+      "epoch": 654.64,
+      "learning_rate": 1.9186379928315414e-05,
+      "loss": 0.2188,
+      "step": 101470
+    },
+    {
+      "epoch": 654.71,
+      "learning_rate": 1.9182795698924732e-05,
+      "loss": 0.2657,
+      "step": 101480
+    },
+    {
+      "epoch": 654.77,
+      "learning_rate": 1.917921146953405e-05,
+      "loss": 0.3218,
+      "step": 101490
+    },
+    {
+      "epoch": 654.84,
+      "learning_rate": 1.9175627240143372e-05,
+      "loss": 0.2372,
+      "step": 101500
+    },
+    {
+      "epoch": 654.9,
+      "learning_rate": 1.917204301075269e-05,
+      "loss": 0.2026,
+      "step": 101510
+    },
+    {
+      "epoch": 654.96,
+      "learning_rate": 1.9168458781362008e-05,
+      "loss": 0.2625,
+      "step": 101520
+    },
+    {
+      "epoch": 655.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9090712666511536,
+      "eval_runtime": 0.8739,
+      "eval_samples_per_second": 230.015,
+      "eval_steps_per_second": 8.01,
+      "step": 101525
+    },
+    {
+      "epoch": 655.03,
+      "learning_rate": 1.9164874551971326e-05,
+      "loss": 0.2183,
+      "step": 101530
+    },
+    {
+      "epoch": 655.1,
+      "learning_rate": 1.9161290322580648e-05,
+      "loss": 0.1849,
+      "step": 101540
+    },
+    {
+      "epoch": 655.16,
+      "learning_rate": 1.9157706093189963e-05,
+      "loss": 0.303,
+      "step": 101550
+    },
+    {
+      "epoch": 655.23,
+      "learning_rate": 1.9154121863799284e-05,
+      "loss": 0.2543,
+      "step": 101560
+    },
+    {
+      "epoch": 655.29,
+      "learning_rate": 1.9150537634408603e-05,
+      "loss": 0.1867,
+      "step": 101570
+    },
+    {
+      "epoch": 655.35,
+      "learning_rate": 1.9146953405017924e-05,
+      "loss": 0.2079,
+      "step": 101580
+    },
+    {
+      "epoch": 655.42,
+      "learning_rate": 1.914336917562724e-05,
+      "loss": 0.2633,
+      "step": 101590
+    },
+    {
+      "epoch": 655.48,
+      "learning_rate": 1.913978494623656e-05,
+      "loss": 0.2833,
+      "step": 101600
+    },
+    {
+      "epoch": 655.55,
+      "learning_rate": 1.913620071684588e-05,
+      "loss": 0.1766,
+      "step": 101610
+    },
+    {
+      "epoch": 655.61,
+      "learning_rate": 1.91326164874552e-05,
+      "loss": 0.2292,
+      "step": 101620
+    },
+    {
+      "epoch": 655.68,
+      "learning_rate": 1.9129032258064515e-05,
+      "loss": 0.254,
+      "step": 101630
+    },
+    {
+      "epoch": 655.74,
+      "learning_rate": 1.9125448028673837e-05,
+      "loss": 0.1959,
+      "step": 101640
+    },
+    {
+      "epoch": 655.8,
+      "learning_rate": 1.9121863799283155e-05,
+      "loss": 0.2472,
+      "step": 101650
+    },
+    {
+      "epoch": 655.87,
+      "learning_rate": 1.9118279569892477e-05,
+      "loss": 0.2573,
+      "step": 101660
+    },
+    {
+      "epoch": 655.93,
+      "learning_rate": 1.911469534050179e-05,
+      "loss": 0.2199,
+      "step": 101670
+    },
+    {
+      "epoch": 656.0,
+      "learning_rate": 1.9111111111111113e-05,
+      "loss": 0.2518,
+      "step": 101680
+    },
+    {
+      "epoch": 656.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9539374113082886,
+      "eval_runtime": 0.8688,
+      "eval_samples_per_second": 231.358,
+      "eval_steps_per_second": 8.057,
+      "step": 101680
+    },
+    {
+      "epoch": 656.06,
+      "learning_rate": 1.910752688172043e-05,
+      "loss": 0.2491,
+      "step": 101690
+    },
+    {
+      "epoch": 656.13,
+      "learning_rate": 1.910394265232975e-05,
+      "loss": 0.2257,
+      "step": 101700
+    },
+    {
+      "epoch": 656.19,
+      "learning_rate": 1.9100358422939068e-05,
+      "loss": 0.2668,
+      "step": 101710
+    },
+    {
+      "epoch": 656.26,
+      "learning_rate": 1.909677419354839e-05,
+      "loss": 0.2405,
+      "step": 101720
+    },
+    {
+      "epoch": 656.32,
+      "learning_rate": 1.9093189964157707e-05,
+      "loss": 0.2329,
+      "step": 101730
+    },
+    {
+      "epoch": 656.39,
+      "learning_rate": 1.9089605734767026e-05,
+      "loss": 0.213,
+      "step": 101740
+    },
+    {
+      "epoch": 656.45,
+      "learning_rate": 1.9086021505376344e-05,
+      "loss": 0.1997,
+      "step": 101750
+    },
+    {
+      "epoch": 656.51,
+      "learning_rate": 1.9082437275985665e-05,
+      "loss": 0.2169,
+      "step": 101760
+    },
+    {
+      "epoch": 656.58,
+      "learning_rate": 1.9078853046594984e-05,
+      "loss": 0.2263,
+      "step": 101770
+    },
+    {
+      "epoch": 656.64,
+      "learning_rate": 1.9075268817204302e-05,
+      "loss": 0.2579,
+      "step": 101780
+    },
+    {
+      "epoch": 656.71,
+      "learning_rate": 1.907168458781362e-05,
+      "loss": 0.2071,
+      "step": 101790
+    },
+    {
+      "epoch": 656.77,
+      "learning_rate": 1.906810035842294e-05,
+      "loss": 0.2178,
+      "step": 101800
+    },
+    {
+      "epoch": 656.84,
+      "learning_rate": 1.906451612903226e-05,
+      "loss": 0.2473,
+      "step": 101810
+    },
+    {
+      "epoch": 656.9,
+      "learning_rate": 1.9060931899641578e-05,
+      "loss": 0.2435,
+      "step": 101820
+    },
+    {
+      "epoch": 656.96,
+      "learning_rate": 1.9057347670250896e-05,
+      "loss": 0.2416,
+      "step": 101830
+    },
+    {
+      "epoch": 657.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9461241960525513,
+      "eval_runtime": 0.861,
+      "eval_samples_per_second": 233.457,
+      "eval_steps_per_second": 8.13,
+      "step": 101835
+    },
+    {
+      "epoch": 657.03,
+      "learning_rate": 1.9053763440860218e-05,
+      "loss": 0.2015,
+      "step": 101840
+    },
+    {
+      "epoch": 657.1,
+      "learning_rate": 1.9050179211469536e-05,
+      "loss": 0.2259,
+      "step": 101850
+    },
+    {
+      "epoch": 657.16,
+      "learning_rate": 1.9046594982078854e-05,
+      "loss": 0.2591,
+      "step": 101860
+    },
+    {
+      "epoch": 657.23,
+      "learning_rate": 1.9043010752688172e-05,
+      "loss": 0.2807,
+      "step": 101870
+    },
+    {
+      "epoch": 657.29,
+      "learning_rate": 1.9039426523297494e-05,
+      "loss": 0.2611,
+      "step": 101880
+    },
+    {
+      "epoch": 657.35,
+      "learning_rate": 1.9035842293906812e-05,
+      "loss": 0.2209,
+      "step": 101890
+    },
+    {
+      "epoch": 657.42,
+      "learning_rate": 1.9032258064516127e-05,
+      "loss": 0.261,
+      "step": 101900
+    },
+    {
+      "epoch": 657.48,
+      "learning_rate": 1.902867383512545e-05,
+      "loss": 0.1805,
+      "step": 101910
+    },
+    {
+      "epoch": 657.55,
+      "learning_rate": 1.9025089605734767e-05,
+      "loss": 0.1833,
+      "step": 101920
+    },
+    {
+      "epoch": 657.61,
+      "learning_rate": 1.902150537634409e-05,
+      "loss": 0.208,
+      "step": 101930
+    },
+    {
+      "epoch": 657.68,
+      "learning_rate": 1.9017921146953403e-05,
+      "loss": 0.2187,
+      "step": 101940
+    },
+    {
+      "epoch": 657.74,
+      "learning_rate": 1.9014336917562725e-05,
+      "loss": 0.2184,
+      "step": 101950
+    },
+    {
+      "epoch": 657.8,
+      "learning_rate": 1.9010752688172043e-05,
+      "loss": 0.2272,
+      "step": 101960
+    },
+    {
+      "epoch": 657.87,
+      "learning_rate": 1.9007168458781365e-05,
+      "loss": 0.2174,
+      "step": 101970
+    },
+    {
+      "epoch": 657.93,
+      "learning_rate": 1.900358422939068e-05,
+      "loss": 0.2394,
+      "step": 101980
+    },
+    {
+      "epoch": 658.0,
+      "learning_rate": 1.9e-05,
+      "loss": 0.2745,
+      "step": 101990
+    },
+    {
+      "epoch": 658.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9528095126152039,
+      "eval_runtime": 0.8605,
+      "eval_samples_per_second": 233.59,
+      "eval_steps_per_second": 8.135,
+      "step": 101990
+    },
+    {
+      "epoch": 658.06,
+      "learning_rate": 1.899641577060932e-05,
+      "loss": 0.2875,
+      "step": 102000
+    },
+    {
+      "epoch": 658.13,
+      "learning_rate": 1.899283154121864e-05,
+      "loss": 0.273,
+      "step": 102010
+    },
+    {
+      "epoch": 658.19,
+      "learning_rate": 1.8989247311827956e-05,
+      "loss": 0.2376,
+      "step": 102020
+    },
+    {
+      "epoch": 658.26,
+      "learning_rate": 1.8985663082437277e-05,
+      "loss": 0.2461,
+      "step": 102030
+    },
+    {
+      "epoch": 658.32,
+      "learning_rate": 1.8982078853046595e-05,
+      "loss": 0.2572,
+      "step": 102040
+    },
+    {
+      "epoch": 658.39,
+      "learning_rate": 1.8978494623655914e-05,
+      "loss": 0.2465,
+      "step": 102050
+    },
+    {
+      "epoch": 658.45,
+      "learning_rate": 1.8974910394265232e-05,
+      "loss": 0.2455,
+      "step": 102060
+    },
+    {
+      "epoch": 658.51,
+      "learning_rate": 1.8971326164874553e-05,
+      "loss": 0.2351,
+      "step": 102070
+    },
+    {
+      "epoch": 658.58,
+      "learning_rate": 1.896774193548387e-05,
+      "loss": 0.219,
+      "step": 102080
+    },
+    {
+      "epoch": 658.64,
+      "learning_rate": 1.896415770609319e-05,
+      "loss": 0.2229,
+      "step": 102090
+    },
+    {
+      "epoch": 658.71,
+      "learning_rate": 1.8960573476702508e-05,
+      "loss": 0.2387,
+      "step": 102100
+    },
+    {
+      "epoch": 658.77,
+      "learning_rate": 1.895698924731183e-05,
+      "loss": 0.2832,
+      "step": 102110
+    },
+    {
+      "epoch": 658.84,
+      "learning_rate": 1.8953405017921148e-05,
+      "loss": 0.2418,
+      "step": 102120
+    },
+    {
+      "epoch": 658.9,
+      "learning_rate": 1.8949820788530466e-05,
+      "loss": 0.2205,
+      "step": 102130
+    },
+    {
+      "epoch": 658.96,
+      "learning_rate": 1.8946236559139784e-05,
+      "loss": 0.2577,
+      "step": 102140
+    },
+    {
+      "epoch": 659.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9552705883979797,
+      "eval_runtime": 0.8651,
+      "eval_samples_per_second": 232.333,
+      "eval_steps_per_second": 8.091,
+      "step": 102145
+    },
+    {
+      "epoch": 659.03,
+      "learning_rate": 1.8942652329749106e-05,
+      "loss": 0.2044,
+      "step": 102150
+    },
+    {
+      "epoch": 659.1,
+      "learning_rate": 1.8939068100358424e-05,
+      "loss": 0.2214,
+      "step": 102160
+    },
+    {
+      "epoch": 659.16,
+      "learning_rate": 1.8935483870967742e-05,
+      "loss": 0.1861,
+      "step": 102170
+    },
+    {
+      "epoch": 659.23,
+      "learning_rate": 1.893189964157706e-05,
+      "loss": 0.2177,
+      "step": 102180
+    },
+    {
+      "epoch": 659.29,
+      "learning_rate": 1.8928315412186382e-05,
+      "loss": 0.2348,
+      "step": 102190
+    },
+    {
+      "epoch": 659.35,
+      "learning_rate": 1.89247311827957e-05,
+      "loss": 0.1932,
+      "step": 102200
+    },
+    {
+      "epoch": 659.42,
+      "learning_rate": 1.892114695340502e-05,
+      "loss": 0.219,
+      "step": 102210
+    },
+    {
+      "epoch": 659.48,
+      "learning_rate": 1.8917562724014337e-05,
+      "loss": 0.2243,
+      "step": 102220
+    },
+    {
+      "epoch": 659.55,
+      "learning_rate": 1.8913978494623658e-05,
+      "loss": 0.2211,
+      "step": 102230
+    },
+    {
+      "epoch": 659.61,
+      "learning_rate": 1.8910394265232976e-05,
+      "loss": 0.2096,
+      "step": 102240
+    },
+    {
+      "epoch": 659.68,
+      "learning_rate": 1.8906810035842295e-05,
+      "loss": 0.2375,
+      "step": 102250
+    },
+    {
+      "epoch": 659.74,
+      "learning_rate": 1.8903225806451613e-05,
+      "loss": 0.245,
+      "step": 102260
+    },
+    {
+      "epoch": 659.8,
+      "learning_rate": 1.8899641577060935e-05,
+      "loss": 0.2306,
+      "step": 102270
+    },
+    {
+      "epoch": 659.87,
+      "learning_rate": 1.8896057347670253e-05,
+      "loss": 0.2632,
+      "step": 102280
+    },
+    {
+      "epoch": 659.93,
+      "learning_rate": 1.889247311827957e-05,
+      "loss": 0.1935,
+      "step": 102290
+    },
+    {
+      "epoch": 660.0,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 0.2065,
+      "step": 102300
+    },
+    {
+      "epoch": 660.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9131225347518921,
+      "eval_runtime": 0.8779,
+      "eval_samples_per_second": 228.945,
+      "eval_steps_per_second": 7.973,
+      "step": 102300
+    },
+    {
+      "epoch": 660.06,
+      "learning_rate": 1.888530465949821e-05,
+      "loss": 0.2383,
+      "step": 102310
+    },
+    {
+      "epoch": 660.13,
+      "learning_rate": 1.888172043010753e-05,
+      "loss": 0.2162,
+      "step": 102320
+    },
+    {
+      "epoch": 660.19,
+      "learning_rate": 1.8878136200716847e-05,
+      "loss": 0.193,
+      "step": 102330
+    },
+    {
+      "epoch": 660.26,
+      "learning_rate": 1.8874551971326165e-05,
+      "loss": 0.2688,
+      "step": 102340
+    },
+    {
+      "epoch": 660.32,
+      "learning_rate": 1.8870967741935487e-05,
+      "loss": 0.2405,
+      "step": 102350
+    },
+    {
+      "epoch": 660.39,
+      "learning_rate": 1.8867383512544802e-05,
+      "loss": 0.2342,
+      "step": 102360
+    },
+    {
+      "epoch": 660.45,
+      "learning_rate": 1.8863799283154123e-05,
+      "loss": 0.2461,
+      "step": 102370
+    },
+    {
+      "epoch": 660.51,
+      "learning_rate": 1.886021505376344e-05,
+      "loss": 0.2386,
+      "step": 102380
+    },
+    {
+      "epoch": 660.58,
+      "learning_rate": 1.8856630824372763e-05,
+      "loss": 0.2402,
+      "step": 102390
+    },
+    {
+      "epoch": 660.64,
+      "learning_rate": 1.8853046594982078e-05,
+      "loss": 0.1863,
+      "step": 102400
+    },
+    {
+      "epoch": 660.71,
+      "learning_rate": 1.88494623655914e-05,
+      "loss": 0.258,
+      "step": 102410
+    },
+    {
+      "epoch": 660.77,
+      "learning_rate": 1.8845878136200718e-05,
+      "loss": 0.2164,
+      "step": 102420
+    },
+    {
+      "epoch": 660.84,
+      "learning_rate": 1.884229390681004e-05,
+      "loss": 0.2613,
+      "step": 102430
+    },
+    {
+      "epoch": 660.9,
+      "learning_rate": 1.8838709677419354e-05,
+      "loss": 0.2284,
+      "step": 102440
+    },
+    {
+      "epoch": 660.96,
+      "learning_rate": 1.8835125448028676e-05,
+      "loss": 0.2063,
+      "step": 102450
+    },
+    {
+      "epoch": 661.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8982571363449097,
+      "eval_runtime": 0.8892,
+      "eval_samples_per_second": 226.038,
+      "eval_steps_per_second": 7.872,
+      "step": 102455
+    },
+    {
+      "epoch": 661.03,
+      "learning_rate": 1.8831541218637994e-05,
+      "loss": 0.2491,
+      "step": 102460
+    },
+    {
+      "epoch": 661.1,
+      "learning_rate": 1.8827956989247316e-05,
+      "loss": 0.2046,
+      "step": 102470
+    },
+    {
+      "epoch": 661.16,
+      "learning_rate": 1.882437275985663e-05,
+      "loss": 0.2126,
+      "step": 102480
+    },
+    {
+      "epoch": 661.23,
+      "learning_rate": 1.8820788530465952e-05,
+      "loss": 0.2141,
+      "step": 102490
+    },
+    {
+      "epoch": 661.29,
+      "learning_rate": 1.881720430107527e-05,
+      "loss": 0.2607,
+      "step": 102500
+    },
+    {
+      "epoch": 661.35,
+      "learning_rate": 1.8813620071684592e-05,
+      "loss": 0.2452,
+      "step": 102510
+    },
+    {
+      "epoch": 661.42,
+      "learning_rate": 1.8810035842293907e-05,
+      "loss": 0.2043,
+      "step": 102520
+    },
+    {
+      "epoch": 661.48,
+      "learning_rate": 1.8806451612903225e-05,
+      "loss": 0.2497,
+      "step": 102530
+    },
+    {
+      "epoch": 661.55,
+      "learning_rate": 1.8802867383512546e-05,
+      "loss": 0.2472,
+      "step": 102540
+    },
+    {
+      "epoch": 661.61,
+      "learning_rate": 1.8799283154121865e-05,
+      "loss": 0.25,
+      "step": 102550
+    },
+    {
+      "epoch": 661.68,
+      "learning_rate": 1.8795698924731183e-05,
+      "loss": 0.2443,
+      "step": 102560
+    },
+    {
+      "epoch": 661.74,
+      "learning_rate": 1.87921146953405e-05,
+      "loss": 0.2307,
+      "step": 102570
+    },
+    {
+      "epoch": 661.8,
+      "learning_rate": 1.8788530465949823e-05,
+      "loss": 0.2361,
+      "step": 102580
+    },
+    {
+      "epoch": 661.87,
+      "learning_rate": 1.878494623655914e-05,
+      "loss": 0.2192,
+      "step": 102590
+    },
+    {
+      "epoch": 661.93,
+      "learning_rate": 1.878136200716846e-05,
+      "loss": 0.2422,
+      "step": 102600
+    },
+    {
+      "epoch": 662.0,
+      "learning_rate": 1.8777777777777777e-05,
+      "loss": 0.2763,
+      "step": 102610
+    },
+    {
+      "epoch": 662.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9211325645446777,
+      "eval_runtime": 0.8887,
+      "eval_samples_per_second": 226.179,
+      "eval_steps_per_second": 7.877,
+      "step": 102610
+    },
+    {
+      "epoch": 662.06,
+      "learning_rate": 1.87741935483871e-05,
+      "loss": 0.2315,
+      "step": 102620
+    },
+    {
+      "epoch": 662.13,
+      "learning_rate": 1.8770609318996417e-05,
+      "loss": 0.2564,
+      "step": 102630
+    },
+    {
+      "epoch": 662.19,
+      "learning_rate": 1.8767025089605735e-05,
+      "loss": 0.2174,
+      "step": 102640
+    },
+    {
+      "epoch": 662.26,
+      "learning_rate": 1.8763440860215053e-05,
+      "loss": 0.2457,
+      "step": 102650
+    },
+    {
+      "epoch": 662.32,
+      "learning_rate": 1.8759856630824375e-05,
+      "loss": 0.2657,
+      "step": 102660
+    },
+    {
+      "epoch": 662.39,
+      "learning_rate": 1.8756272401433693e-05,
+      "loss": 0.258,
+      "step": 102670
+    },
+    {
+      "epoch": 662.45,
+      "learning_rate": 1.875268817204301e-05,
+      "loss": 0.2259,
+      "step": 102680
+    },
+    {
+      "epoch": 662.51,
+      "learning_rate": 1.874910394265233e-05,
+      "loss": 0.2559,
+      "step": 102690
+    },
+    {
+      "epoch": 662.58,
+      "learning_rate": 1.874551971326165e-05,
+      "loss": 0.2524,
+      "step": 102700
+    },
+    {
+      "epoch": 662.64,
+      "learning_rate": 1.8741935483870966e-05,
+      "loss": 0.2801,
+      "step": 102710
+    },
+    {
+      "epoch": 662.71,
+      "learning_rate": 1.8738351254480288e-05,
+      "loss": 0.1913,
+      "step": 102720
+    },
+    {
+      "epoch": 662.77,
+      "learning_rate": 1.8734767025089606e-05,
+      "loss": 0.1976,
+      "step": 102730
+    },
+    {
+      "epoch": 662.84,
+      "learning_rate": 1.8731182795698927e-05,
+      "loss": 0.2824,
+      "step": 102740
+    },
+    {
+      "epoch": 662.9,
+      "learning_rate": 1.8727598566308242e-05,
+      "loss": 0.2632,
+      "step": 102750
+    },
+    {
+      "epoch": 662.96,
+      "learning_rate": 1.8724014336917564e-05,
+      "loss": 0.2653,
+      "step": 102760
+    },
+    {
+      "epoch": 663.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9743476510047913,
+      "eval_runtime": 0.9659,
+      "eval_samples_per_second": 208.106,
+      "eval_steps_per_second": 7.247,
+      "step": 102765
+    },
+    {
+      "epoch": 663.03,
+      "learning_rate": 1.8720430107526882e-05,
+      "loss": 0.258,
+      "step": 102770
+    },
+    {
+      "epoch": 663.1,
+      "learning_rate": 1.8716845878136204e-05,
+      "loss": 0.2497,
+      "step": 102780
+    },
+    {
+      "epoch": 663.16,
+      "learning_rate": 1.871326164874552e-05,
+      "loss": 0.2468,
+      "step": 102790
+    },
+    {
+      "epoch": 663.23,
+      "learning_rate": 1.870967741935484e-05,
+      "loss": 0.2145,
+      "step": 102800
+    },
+    {
+      "epoch": 663.29,
+      "learning_rate": 1.8706093189964158e-05,
+      "loss": 0.2282,
+      "step": 102810
+    },
+    {
+      "epoch": 663.35,
+      "learning_rate": 1.870250896057348e-05,
+      "loss": 0.1858,
+      "step": 102820
+    },
+    {
+      "epoch": 663.42,
+      "learning_rate": 1.8698924731182795e-05,
+      "loss": 0.2639,
+      "step": 102830
+    },
+    {
+      "epoch": 663.48,
+      "learning_rate": 1.8695340501792116e-05,
+      "loss": 0.2285,
+      "step": 102840
+    },
+    {
+      "epoch": 663.55,
+      "learning_rate": 1.8691756272401434e-05,
+      "loss": 0.2621,
+      "step": 102850
+    },
+    {
+      "epoch": 663.61,
+      "learning_rate": 1.8688172043010753e-05,
+      "loss": 0.2251,
+      "step": 102860
+    },
+    {
+      "epoch": 663.68,
+      "learning_rate": 1.868458781362007e-05,
+      "loss": 0.2337,
+      "step": 102870
+    },
+    {
+      "epoch": 663.74,
+      "learning_rate": 1.8681003584229392e-05,
+      "loss": 0.2176,
+      "step": 102880
+    },
+    {
+      "epoch": 663.8,
+      "learning_rate": 1.867741935483871e-05,
+      "loss": 0.2266,
+      "step": 102890
+    },
+    {
+      "epoch": 663.87,
+      "learning_rate": 1.867383512544803e-05,
+      "loss": 0.2304,
+      "step": 102900
+    },
+    {
+      "epoch": 663.93,
+      "learning_rate": 1.8670250896057347e-05,
+      "loss": 0.2385,
+      "step": 102910
+    },
+    {
+      "epoch": 664.0,
+      "learning_rate": 1.866666666666667e-05,
+      "loss": 0.2328,
+      "step": 102920
+    },
+    {
+      "epoch": 664.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9069110751152039,
+      "eval_runtime": 0.8781,
+      "eval_samples_per_second": 228.894,
+      "eval_steps_per_second": 7.971,
+      "step": 102920
+    },
+    {
+      "epoch": 664.06,
+      "learning_rate": 1.8663082437275987e-05,
+      "loss": 0.2577,
+      "step": 102930
+    },
+    {
+      "epoch": 664.13,
+      "learning_rate": 1.8659498207885305e-05,
+      "loss": 0.2192,
+      "step": 102940
+    },
+    {
+      "epoch": 664.19,
+      "learning_rate": 1.8655913978494623e-05,
+      "loss": 0.2282,
+      "step": 102950
+    },
+    {
+      "epoch": 664.26,
+      "learning_rate": 1.8652329749103945e-05,
+      "loss": 0.2189,
+      "step": 102960
+    },
+    {
+      "epoch": 664.32,
+      "learning_rate": 1.8648745519713263e-05,
+      "loss": 0.2539,
+      "step": 102970
+    },
+    {
+      "epoch": 664.39,
+      "learning_rate": 1.864516129032258e-05,
+      "loss": 0.2372,
+      "step": 102980
+    },
+    {
+      "epoch": 664.45,
+      "learning_rate": 1.86415770609319e-05,
+      "loss": 0.209,
+      "step": 102990
+    },
+    {
+      "epoch": 664.51,
+      "learning_rate": 1.863799283154122e-05,
+      "loss": 0.2469,
+      "step": 103000
+    },
+    {
+      "epoch": 664.58,
+      "learning_rate": 1.863440860215054e-05,
+      "loss": 0.2327,
+      "step": 103010
+    },
+    {
+      "epoch": 664.64,
+      "learning_rate": 1.8630824372759857e-05,
+      "loss": 0.2003,
+      "step": 103020
+    },
+    {
+      "epoch": 664.71,
+      "learning_rate": 1.8627240143369176e-05,
+      "loss": 0.2588,
+      "step": 103030
+    },
+    {
+      "epoch": 664.77,
+      "learning_rate": 1.8623655913978497e-05,
+      "loss": 0.207,
+      "step": 103040
+    },
+    {
+      "epoch": 664.84,
+      "learning_rate": 1.8620071684587815e-05,
+      "loss": 0.2266,
+      "step": 103050
+    },
+    {
+      "epoch": 664.9,
+      "learning_rate": 1.8616487455197134e-05,
+      "loss": 0.2835,
+      "step": 103060
+    },
+    {
+      "epoch": 664.96,
+      "learning_rate": 1.8612903225806452e-05,
+      "loss": 0.2224,
+      "step": 103070
+    },
+    {
+      "epoch": 665.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9446420073509216,
+      "eval_runtime": 0.8658,
+      "eval_samples_per_second": 232.145,
+      "eval_steps_per_second": 8.085,
+      "step": 103075
+    },
+    {
+      "epoch": 665.03,
+      "learning_rate": 1.8609318996415773e-05,
+      "loss": 0.2052,
+      "step": 103080
+    },
+    {
+      "epoch": 665.1,
+      "learning_rate": 1.860573476702509e-05,
+      "loss": 0.2381,
+      "step": 103090
+    },
+    {
+      "epoch": 665.16,
+      "learning_rate": 1.860215053763441e-05,
+      "loss": 0.2316,
+      "step": 103100
+    },
+    {
+      "epoch": 665.23,
+      "learning_rate": 1.8598566308243728e-05,
+      "loss": 0.2489,
+      "step": 103110
+    },
+    {
+      "epoch": 665.29,
+      "learning_rate": 1.859498207885305e-05,
+      "loss": 0.2603,
+      "step": 103120
+    },
+    {
+      "epoch": 665.35,
+      "learning_rate": 1.8591397849462368e-05,
+      "loss": 0.216,
+      "step": 103130
+    },
+    {
+      "epoch": 665.42,
+      "learning_rate": 1.8587813620071686e-05,
+      "loss": 0.2125,
+      "step": 103140
+    },
+    {
+      "epoch": 665.48,
+      "learning_rate": 1.8584229390681004e-05,
+      "loss": 0.1811,
+      "step": 103150
+    },
+    {
+      "epoch": 665.55,
+      "learning_rate": 1.8580645161290326e-05,
+      "loss": 0.238,
+      "step": 103160
+    },
+    {
+      "epoch": 665.61,
+      "learning_rate": 1.8577060931899644e-05,
+      "loss": 0.2411,
+      "step": 103170
+    },
+    {
+      "epoch": 665.68,
+      "learning_rate": 1.857347670250896e-05,
+      "loss": 0.23,
+      "step": 103180
+    },
+    {
+      "epoch": 665.74,
+      "learning_rate": 1.856989247311828e-05,
+      "loss": 0.2396,
+      "step": 103190
+    },
+    {
+      "epoch": 665.8,
+      "learning_rate": 1.85663082437276e-05,
+      "loss": 0.2425,
+      "step": 103200
+    },
+    {
+      "epoch": 665.87,
+      "learning_rate": 1.8562724014336917e-05,
+      "loss": 0.2229,
+      "step": 103210
+    },
+    {
+      "epoch": 665.93,
+      "learning_rate": 1.8559139784946235e-05,
+      "loss": 0.2506,
+      "step": 103220
+    },
+    {
+      "epoch": 666.0,
+      "learning_rate": 1.8555555555555557e-05,
+      "loss": 0.2267,
+      "step": 103230
+    },
+    {
+      "epoch": 666.0,
+      "eval_accuracy": 0.8009950248756219,
+      "eval_loss": 0.9552135467529297,
+      "eval_runtime": 0.8681,
+      "eval_samples_per_second": 231.536,
+      "eval_steps_per_second": 8.063,
+      "step": 103230
+    },
+    {
+      "epoch": 666.06,
+      "learning_rate": 1.8551971326164875e-05,
+      "loss": 0.2301,
+      "step": 103240
+    },
+    {
+      "epoch": 666.13,
+      "learning_rate": 1.8548387096774193e-05,
+      "loss": 0.2538,
+      "step": 103250
+    },
+    {
+      "epoch": 666.19,
+      "learning_rate": 1.854480286738351e-05,
+      "loss": 0.2016,
+      "step": 103260
+    },
+    {
+      "epoch": 666.26,
+      "learning_rate": 1.8541218637992833e-05,
+      "loss": 0.2308,
+      "step": 103270
+    },
+    {
+      "epoch": 666.32,
+      "learning_rate": 1.853763440860215e-05,
+      "loss": 0.2117,
+      "step": 103280
+    },
+    {
+      "epoch": 666.39,
+      "learning_rate": 1.853405017921147e-05,
+      "loss": 0.2407,
+      "step": 103290
+    },
+    {
+      "epoch": 666.45,
+      "learning_rate": 1.8530465949820788e-05,
+      "loss": 0.2411,
+      "step": 103300
+    },
+    {
+      "epoch": 666.51,
+      "learning_rate": 1.852688172043011e-05,
+      "loss": 0.2312,
+      "step": 103310
+    },
+    {
+      "epoch": 666.58,
+      "learning_rate": 1.8523297491039427e-05,
+      "loss": 0.2229,
+      "step": 103320
+    },
+    {
+      "epoch": 666.64,
+      "learning_rate": 1.8519713261648746e-05,
+      "loss": 0.1785,
+      "step": 103330
+    },
+    {
+      "epoch": 666.71,
+      "learning_rate": 1.8516129032258064e-05,
+      "loss": 0.2608,
+      "step": 103340
+    },
+    {
+      "epoch": 666.77,
+      "learning_rate": 1.8512544802867385e-05,
+      "loss": 0.2279,
+      "step": 103350
+    },
+    {
+      "epoch": 666.84,
+      "learning_rate": 1.8508960573476704e-05,
+      "loss": 0.2122,
+      "step": 103360
+    },
+    {
+      "epoch": 666.9,
+      "learning_rate": 1.8505376344086022e-05,
+      "loss": 0.1906,
+      "step": 103370
+    },
+    {
+      "epoch": 666.96,
+      "learning_rate": 1.850179211469534e-05,
+      "loss": 0.2225,
+      "step": 103380
+    },
+    {
+      "epoch": 667.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9777212738990784,
+      "eval_runtime": 0.8602,
+      "eval_samples_per_second": 233.676,
+      "eval_steps_per_second": 8.138,
+      "step": 103385
+    },
+    {
+      "epoch": 667.03,
+      "learning_rate": 1.849820788530466e-05,
+      "loss": 0.2628,
+      "step": 103390
+    },
+    {
+      "epoch": 667.1,
+      "learning_rate": 1.849462365591398e-05,
+      "loss": 0.2475,
+      "step": 103400
+    },
+    {
+      "epoch": 667.16,
+      "learning_rate": 1.8491039426523298e-05,
+      "loss": 0.2405,
+      "step": 103410
+    },
+    {
+      "epoch": 667.23,
+      "learning_rate": 1.8487455197132616e-05,
+      "loss": 0.1825,
+      "step": 103420
+    },
+    {
+      "epoch": 667.29,
+      "learning_rate": 1.8483870967741938e-05,
+      "loss": 0.2508,
+      "step": 103430
+    },
+    {
+      "epoch": 667.35,
+      "learning_rate": 1.8480286738351256e-05,
+      "loss": 0.2523,
+      "step": 103440
+    },
+    {
+      "epoch": 667.42,
+      "learning_rate": 1.8476702508960574e-05,
+      "loss": 0.2486,
+      "step": 103450
+    },
+    {
+      "epoch": 667.48,
+      "learning_rate": 1.8473118279569892e-05,
+      "loss": 0.2645,
+      "step": 103460
+    },
+    {
+      "epoch": 667.55,
+      "learning_rate": 1.8469534050179214e-05,
+      "loss": 0.269,
+      "step": 103470
+    },
+    {
+      "epoch": 667.61,
+      "learning_rate": 1.8465949820788532e-05,
+      "loss": 0.2281,
+      "step": 103480
+    },
+    {
+      "epoch": 667.68,
+      "learning_rate": 1.846236559139785e-05,
+      "loss": 0.191,
+      "step": 103490
+    },
+    {
+      "epoch": 667.74,
+      "learning_rate": 1.845878136200717e-05,
+      "loss": 0.2342,
+      "step": 103500
+    },
+    {
+      "epoch": 667.8,
+      "learning_rate": 1.845519713261649e-05,
+      "loss": 0.2374,
+      "step": 103510
+    },
+    {
+      "epoch": 667.87,
+      "learning_rate": 1.845161290322581e-05,
+      "loss": 0.2351,
+      "step": 103520
+    },
+    {
+      "epoch": 667.93,
+      "learning_rate": 1.8448028673835127e-05,
+      "loss": 0.2473,
+      "step": 103530
+    },
+    {
+      "epoch": 668.0,
+      "learning_rate": 1.8444444444444445e-05,
+      "loss": 0.2189,
+      "step": 103540
+    },
+    {
+      "epoch": 668.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9595470428466797,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.12,
+      "eval_steps_per_second": 8.153,
+      "step": 103540
+    },
+    {
+      "epoch": 668.06,
+      "learning_rate": 1.8440860215053766e-05,
+      "loss": 0.2341,
+      "step": 103550
+    },
+    {
+      "epoch": 668.13,
+      "learning_rate": 1.843727598566308e-05,
+      "loss": 0.2795,
+      "step": 103560
+    },
+    {
+      "epoch": 668.19,
+      "learning_rate": 1.8433691756272403e-05,
+      "loss": 0.2016,
+      "step": 103570
+    },
+    {
+      "epoch": 668.26,
+      "learning_rate": 1.843010752688172e-05,
+      "loss": 0.2053,
+      "step": 103580
+    },
+    {
+      "epoch": 668.32,
+      "learning_rate": 1.8426523297491043e-05,
+      "loss": 0.2536,
+      "step": 103590
+    },
+    {
+      "epoch": 668.39,
+      "learning_rate": 1.8422939068100357e-05,
+      "loss": 0.2701,
+      "step": 103600
+    },
+    {
+      "epoch": 668.45,
+      "learning_rate": 1.841935483870968e-05,
+      "loss": 0.2045,
+      "step": 103610
+    },
+    {
+      "epoch": 668.51,
+      "learning_rate": 1.8415770609318997e-05,
+      "loss": 0.2321,
+      "step": 103620
+    },
+    {
+      "epoch": 668.58,
+      "learning_rate": 1.841218637992832e-05,
+      "loss": 0.2228,
+      "step": 103630
+    },
+    {
+      "epoch": 668.64,
+      "learning_rate": 1.8408602150537634e-05,
+      "loss": 0.2021,
+      "step": 103640
+    },
+    {
+      "epoch": 668.71,
+      "learning_rate": 1.8405017921146955e-05,
+      "loss": 0.2279,
+      "step": 103650
+    },
+    {
+      "epoch": 668.77,
+      "learning_rate": 1.8401433691756273e-05,
+      "loss": 0.2133,
+      "step": 103660
+    },
+    {
+      "epoch": 668.84,
+      "learning_rate": 1.8397849462365595e-05,
+      "loss": 0.2133,
+      "step": 103670
+    },
+    {
+      "epoch": 668.9,
+      "learning_rate": 1.839426523297491e-05,
+      "loss": 0.2854,
+      "step": 103680
+    },
+    {
+      "epoch": 668.96,
+      "learning_rate": 1.839068100358423e-05,
+      "loss": 0.2476,
+      "step": 103690
+    },
+    {
+      "epoch": 669.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9305344223976135,
+      "eval_runtime": 0.8942,
+      "eval_samples_per_second": 224.784,
+      "eval_steps_per_second": 7.828,
+      "step": 103695
+    },
+    {
+      "epoch": 669.03,
+      "learning_rate": 1.838709677419355e-05,
+      "loss": 0.211,
+      "step": 103700
+    },
+    {
+      "epoch": 669.1,
+      "learning_rate": 1.8383512544802868e-05,
+      "loss": 0.2584,
+      "step": 103710
+    },
+    {
+      "epoch": 669.16,
+      "learning_rate": 1.8379928315412186e-05,
+      "loss": 0.2377,
+      "step": 103720
+    },
+    {
+      "epoch": 669.23,
+      "learning_rate": 1.8376344086021508e-05,
+      "loss": 0.2283,
+      "step": 103730
+    },
+    {
+      "epoch": 669.29,
+      "learning_rate": 1.8372759856630826e-05,
+      "loss": 0.2336,
+      "step": 103740
+    },
+    {
+      "epoch": 669.35,
+      "learning_rate": 1.8369175627240144e-05,
+      "loss": 0.2212,
+      "step": 103750
+    },
+    {
+      "epoch": 669.42,
+      "learning_rate": 1.8365591397849462e-05,
+      "loss": 0.2364,
+      "step": 103760
+    },
+    {
+      "epoch": 669.48,
+      "learning_rate": 1.8362007168458784e-05,
+      "loss": 0.2388,
+      "step": 103770
+    },
+    {
+      "epoch": 669.55,
+      "learning_rate": 1.8358422939068102e-05,
+      "loss": 0.2229,
+      "step": 103780
+    },
+    {
+      "epoch": 669.61,
+      "learning_rate": 1.835483870967742e-05,
+      "loss": 0.2369,
+      "step": 103790
+    },
+    {
+      "epoch": 669.68,
+      "learning_rate": 1.835125448028674e-05,
+      "loss": 0.2019,
+      "step": 103800
+    },
+    {
+      "epoch": 669.74,
+      "learning_rate": 1.8347670250896057e-05,
+      "loss": 0.2058,
+      "step": 103810
+    },
+    {
+      "epoch": 669.8,
+      "learning_rate": 1.8344086021505378e-05,
+      "loss": 0.2369,
+      "step": 103820
+    },
+    {
+      "epoch": 669.87,
+      "learning_rate": 1.8340501792114696e-05,
+      "loss": 0.2436,
+      "step": 103830
+    },
+    {
+      "epoch": 669.93,
+      "learning_rate": 1.8336917562724015e-05,
+      "loss": 0.2582,
+      "step": 103840
+    },
+    {
+      "epoch": 670.0,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 0.2226,
+      "step": 103850
+    },
+    {
+      "epoch": 670.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.007832407951355,
+      "eval_runtime": 0.8705,
+      "eval_samples_per_second": 230.909,
+      "eval_steps_per_second": 8.042,
+      "step": 103850
+    },
+    {
+      "epoch": 670.06,
+      "learning_rate": 1.8329749103942654e-05,
+      "loss": 0.2907,
+      "step": 103860
+    },
+    {
+      "epoch": 670.13,
+      "learning_rate": 1.832616487455197e-05,
+      "loss": 0.223,
+      "step": 103870
+    },
+    {
+      "epoch": 670.19,
+      "learning_rate": 1.832258064516129e-05,
+      "loss": 0.2383,
+      "step": 103880
+    },
+    {
+      "epoch": 670.26,
+      "learning_rate": 1.831899641577061e-05,
+      "loss": 0.2142,
+      "step": 103890
+    },
+    {
+      "epoch": 670.32,
+      "learning_rate": 1.831541218637993e-05,
+      "loss": 0.233,
+      "step": 103900
+    },
+    {
+      "epoch": 670.39,
+      "learning_rate": 1.8311827956989245e-05,
+      "loss": 0.2409,
+      "step": 103910
+    },
+    {
+      "epoch": 670.45,
+      "learning_rate": 1.8308243727598567e-05,
+      "loss": 0.1811,
+      "step": 103920
+    },
+    {
+      "epoch": 670.51,
+      "learning_rate": 1.8304659498207885e-05,
+      "loss": 0.2247,
+      "step": 103930
+    },
+    {
+      "epoch": 670.58,
+      "learning_rate": 1.8301075268817207e-05,
+      "loss": 0.265,
+      "step": 103940
+    },
+    {
+      "epoch": 670.64,
+      "learning_rate": 1.829749103942652e-05,
+      "loss": 0.2071,
+      "step": 103950
+    },
+    {
+      "epoch": 670.71,
+      "learning_rate": 1.8293906810035843e-05,
+      "loss": 0.2869,
+      "step": 103960
+    },
+    {
+      "epoch": 670.77,
+      "learning_rate": 1.829032258064516e-05,
+      "loss": 0.203,
+      "step": 103970
+    },
+    {
+      "epoch": 670.84,
+      "learning_rate": 1.8286738351254483e-05,
+      "loss": 0.24,
+      "step": 103980
+    },
+    {
+      "epoch": 670.9,
+      "learning_rate": 1.8283154121863798e-05,
+      "loss": 0.2603,
+      "step": 103990
+    },
+    {
+      "epoch": 670.96,
+      "learning_rate": 1.827956989247312e-05,
+      "loss": 0.2599,
+      "step": 104000
+    },
+    {
+      "epoch": 671.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9557844996452332,
+      "eval_runtime": 0.8686,
+      "eval_samples_per_second": 231.413,
+      "eval_steps_per_second": 8.059,
+      "step": 104005
+    },
+    {
+      "epoch": 671.03,
+      "learning_rate": 1.8275985663082438e-05,
+      "loss": 0.2564,
+      "step": 104010
+    },
+    {
+      "epoch": 671.1,
+      "learning_rate": 1.827240143369176e-05,
+      "loss": 0.2495,
+      "step": 104020
+    },
+    {
+      "epoch": 671.16,
+      "learning_rate": 1.8268817204301074e-05,
+      "loss": 0.2373,
+      "step": 104030
+    },
+    {
+      "epoch": 671.23,
+      "learning_rate": 1.8265232974910396e-05,
+      "loss": 0.1745,
+      "step": 104040
+    },
+    {
+      "epoch": 671.29,
+      "learning_rate": 1.8261648745519714e-05,
+      "loss": 0.2386,
+      "step": 104050
+    },
+    {
+      "epoch": 671.35,
+      "learning_rate": 1.8258064516129032e-05,
+      "loss": 0.1862,
+      "step": 104060
+    },
+    {
+      "epoch": 671.42,
+      "learning_rate": 1.825448028673835e-05,
+      "loss": 0.2558,
+      "step": 104070
+    },
+    {
+      "epoch": 671.48,
+      "learning_rate": 1.8250896057347672e-05,
+      "loss": 0.2108,
+      "step": 104080
+    },
+    {
+      "epoch": 671.55,
+      "learning_rate": 1.824731182795699e-05,
+      "loss": 0.1931,
+      "step": 104090
+    },
+    {
+      "epoch": 671.61,
+      "learning_rate": 1.824372759856631e-05,
+      "loss": 0.2027,
+      "step": 104100
+    },
+    {
+      "epoch": 671.68,
+      "learning_rate": 1.8240143369175627e-05,
+      "loss": 0.1927,
+      "step": 104110
+    },
+    {
+      "epoch": 671.74,
+      "learning_rate": 1.8236559139784948e-05,
+      "loss": 0.2472,
+      "step": 104120
+    },
+    {
+      "epoch": 671.8,
+      "learning_rate": 1.8232974910394266e-05,
+      "loss": 0.2615,
+      "step": 104130
+    },
+    {
+      "epoch": 671.87,
+      "learning_rate": 1.8229390681003585e-05,
+      "loss": 0.2143,
+      "step": 104140
+    },
+    {
+      "epoch": 671.93,
+      "learning_rate": 1.8225806451612903e-05,
+      "loss": 0.2551,
+      "step": 104150
+    },
+    {
+      "epoch": 672.0,
+      "learning_rate": 1.8222222222222224e-05,
+      "loss": 0.1975,
+      "step": 104160
+    },
+    {
+      "epoch": 672.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9723358154296875,
+      "eval_runtime": 0.8669,
+      "eval_samples_per_second": 231.866,
+      "eval_steps_per_second": 8.075,
+      "step": 104160
+    },
+    {
+      "epoch": 672.06,
+      "learning_rate": 1.8218637992831543e-05,
+      "loss": 0.3328,
+      "step": 104170
+    },
+    {
+      "epoch": 672.13,
+      "learning_rate": 1.821505376344086e-05,
+      "loss": 0.2491,
+      "step": 104180
+    },
+    {
+      "epoch": 672.19,
+      "learning_rate": 1.821146953405018e-05,
+      "loss": 0.2203,
+      "step": 104190
+    },
+    {
+      "epoch": 672.26,
+      "learning_rate": 1.82078853046595e-05,
+      "loss": 0.2446,
+      "step": 104200
+    },
+    {
+      "epoch": 672.32,
+      "learning_rate": 1.820430107526882e-05,
+      "loss": 0.2409,
+      "step": 104210
+    },
+    {
+      "epoch": 672.39,
+      "learning_rate": 1.8200716845878137e-05,
+      "loss": 0.2092,
+      "step": 104220
+    },
+    {
+      "epoch": 672.45,
+      "learning_rate": 1.8197132616487455e-05,
+      "loss": 0.2075,
+      "step": 104230
+    },
+    {
+      "epoch": 672.51,
+      "learning_rate": 1.8193548387096777e-05,
+      "loss": 0.2207,
+      "step": 104240
+    },
+    {
+      "epoch": 672.58,
+      "learning_rate": 1.8189964157706095e-05,
+      "loss": 0.2447,
+      "step": 104250
+    },
+    {
+      "epoch": 672.64,
+      "learning_rate": 1.8186379928315413e-05,
+      "loss": 0.2477,
+      "step": 104260
+    },
+    {
+      "epoch": 672.71,
+      "learning_rate": 1.818279569892473e-05,
+      "loss": 0.2711,
+      "step": 104270
+    },
+    {
+      "epoch": 672.77,
+      "learning_rate": 1.8179211469534053e-05,
+      "loss": 0.2135,
+      "step": 104280
+    },
+    {
+      "epoch": 672.84,
+      "learning_rate": 1.817562724014337e-05,
+      "loss": 0.2311,
+      "step": 104290
+    },
+    {
+      "epoch": 672.9,
+      "learning_rate": 1.817204301075269e-05,
+      "loss": 0.2288,
+      "step": 104300
+    },
+    {
+      "epoch": 672.96,
+      "learning_rate": 1.8168458781362008e-05,
+      "loss": 0.2691,
+      "step": 104310
+    },
+    {
+      "epoch": 673.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9633730053901672,
+      "eval_runtime": 0.8691,
+      "eval_samples_per_second": 231.269,
+      "eval_steps_per_second": 8.054,
+      "step": 104315
+    },
+    {
+      "epoch": 673.03,
+      "learning_rate": 1.816487455197133e-05,
+      "loss": 0.2656,
+      "step": 104320
+    },
+    {
+      "epoch": 673.1,
+      "learning_rate": 1.8161290322580647e-05,
+      "loss": 0.2823,
+      "step": 104330
+    },
+    {
+      "epoch": 673.16,
+      "learning_rate": 1.8157706093189966e-05,
+      "loss": 0.2523,
+      "step": 104340
+    },
+    {
+      "epoch": 673.23,
+      "learning_rate": 1.8154121863799284e-05,
+      "loss": 0.277,
+      "step": 104350
+    },
+    {
+      "epoch": 673.29,
+      "learning_rate": 1.8150537634408605e-05,
+      "loss": 0.1969,
+      "step": 104360
+    },
+    {
+      "epoch": 673.35,
+      "learning_rate": 1.814695340501792e-05,
+      "loss": 0.2463,
+      "step": 104370
+    },
+    {
+      "epoch": 673.42,
+      "learning_rate": 1.8143369175627242e-05,
+      "loss": 0.2451,
+      "step": 104380
+    },
+    {
+      "epoch": 673.48,
+      "learning_rate": 1.813978494623656e-05,
+      "loss": 0.2255,
+      "step": 104390
+    },
+    {
+      "epoch": 673.55,
+      "learning_rate": 1.813620071684588e-05,
+      "loss": 0.2411,
+      "step": 104400
+    },
+    {
+      "epoch": 673.61,
+      "learning_rate": 1.8132616487455196e-05,
+      "loss": 0.2535,
+      "step": 104410
+    },
+    {
+      "epoch": 673.68,
+      "learning_rate": 1.8129032258064518e-05,
+      "loss": 0.2372,
+      "step": 104420
+    },
+    {
+      "epoch": 673.74,
+      "learning_rate": 1.8125448028673836e-05,
+      "loss": 0.2463,
+      "step": 104430
+    },
+    {
+      "epoch": 673.8,
+      "learning_rate": 1.8121863799283158e-05,
+      "loss": 0.2055,
+      "step": 104440
+    },
+    {
+      "epoch": 673.87,
+      "learning_rate": 1.8118279569892473e-05,
+      "loss": 0.1835,
+      "step": 104450
+    },
+    {
+      "epoch": 673.93,
+      "learning_rate": 1.811469534050179e-05,
+      "loss": 0.217,
+      "step": 104460
+    },
+    {
+      "epoch": 674.0,
+      "learning_rate": 1.8111111111111112e-05,
+      "loss": 0.2136,
+      "step": 104470
+    },
+    {
+      "epoch": 674.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9864404797554016,
+      "eval_runtime": 0.9168,
+      "eval_samples_per_second": 219.25,
+      "eval_steps_per_second": 7.636,
+      "step": 104470
+    },
+    {
+      "epoch": 674.06,
+      "learning_rate": 1.810752688172043e-05,
+      "loss": 0.2419,
+      "step": 104480
+    },
+    {
+      "epoch": 674.13,
+      "learning_rate": 1.810394265232975e-05,
+      "loss": 0.1937,
+      "step": 104490
+    },
+    {
+      "epoch": 674.19,
+      "learning_rate": 1.8100358422939067e-05,
+      "loss": 0.1957,
+      "step": 104500
+    },
+    {
+      "epoch": 674.26,
+      "learning_rate": 1.809677419354839e-05,
+      "loss": 0.2492,
+      "step": 104510
+    },
+    {
+      "epoch": 674.32,
+      "learning_rate": 1.8093189964157707e-05,
+      "loss": 0.2184,
+      "step": 104520
+    },
+    {
+      "epoch": 674.39,
+      "learning_rate": 1.8089605734767025e-05,
+      "loss": 0.2934,
+      "step": 104530
+    },
+    {
+      "epoch": 674.45,
+      "learning_rate": 1.8086021505376343e-05,
+      "loss": 0.2505,
+      "step": 104540
+    },
+    {
+      "epoch": 674.51,
+      "learning_rate": 1.8082437275985665e-05,
+      "loss": 0.2435,
+      "step": 104550
+    },
+    {
+      "epoch": 674.58,
+      "learning_rate": 1.8078853046594983e-05,
+      "loss": 0.2365,
+      "step": 104560
+    },
+    {
+      "epoch": 674.64,
+      "learning_rate": 1.80752688172043e-05,
+      "loss": 0.2192,
+      "step": 104570
+    },
+    {
+      "epoch": 674.71,
+      "learning_rate": 1.807168458781362e-05,
+      "loss": 0.2357,
+      "step": 104580
+    },
+    {
+      "epoch": 674.77,
+      "learning_rate": 1.806810035842294e-05,
+      "loss": 0.2396,
+      "step": 104590
+    },
+    {
+      "epoch": 674.84,
+      "learning_rate": 1.806451612903226e-05,
+      "loss": 0.2371,
+      "step": 104600
+    },
+    {
+      "epoch": 674.9,
+      "learning_rate": 1.8060931899641577e-05,
+      "loss": 0.2511,
+      "step": 104610
+    },
+    {
+      "epoch": 674.96,
+      "learning_rate": 1.8057347670250896e-05,
+      "loss": 0.1907,
+      "step": 104620
+    },
+    {
+      "epoch": 675.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9293041229248047,
+      "eval_runtime": 0.8723,
+      "eval_samples_per_second": 230.426,
+      "eval_steps_per_second": 8.025,
+      "step": 104625
+    },
+    {
+      "epoch": 675.03,
+      "learning_rate": 1.8053763440860217e-05,
+      "loss": 0.2684,
+      "step": 104630
+    },
+    {
+      "epoch": 675.1,
+      "learning_rate": 1.8050179211469535e-05,
+      "loss": 0.1917,
+      "step": 104640
+    },
+    {
+      "epoch": 675.16,
+      "learning_rate": 1.8046594982078854e-05,
+      "loss": 0.256,
+      "step": 104650
+    },
+    {
+      "epoch": 675.23,
+      "learning_rate": 1.8043010752688172e-05,
+      "loss": 0.2455,
+      "step": 104660
+    },
+    {
+      "epoch": 675.29,
+      "learning_rate": 1.8039426523297493e-05,
+      "loss": 0.2198,
+      "step": 104670
+    },
+    {
+      "epoch": 675.35,
+      "learning_rate": 1.803584229390681e-05,
+      "loss": 0.247,
+      "step": 104680
+    },
+    {
+      "epoch": 675.42,
+      "learning_rate": 1.803225806451613e-05,
+      "loss": 0.2421,
+      "step": 104690
+    },
+    {
+      "epoch": 675.48,
+      "learning_rate": 1.8028673835125448e-05,
+      "loss": 0.2291,
+      "step": 104700
+    },
+    {
+      "epoch": 675.55,
+      "learning_rate": 1.802508960573477e-05,
+      "loss": 0.2623,
+      "step": 104710
+    },
+    {
+      "epoch": 675.61,
+      "learning_rate": 1.8021505376344084e-05,
+      "loss": 0.2046,
+      "step": 104720
+    },
+    {
+      "epoch": 675.68,
+      "learning_rate": 1.8017921146953406e-05,
+      "loss": 0.2434,
+      "step": 104730
+    },
+    {
+      "epoch": 675.74,
+      "learning_rate": 1.8014336917562724e-05,
+      "loss": 0.2061,
+      "step": 104740
+    },
+    {
+      "epoch": 675.8,
+      "learning_rate": 1.8010752688172046e-05,
+      "loss": 0.2347,
+      "step": 104750
+    },
+    {
+      "epoch": 675.87,
+      "learning_rate": 1.800716845878136e-05,
+      "loss": 0.2424,
+      "step": 104760
+    },
+    {
+      "epoch": 675.93,
+      "learning_rate": 1.8003584229390682e-05,
+      "loss": 0.2559,
+      "step": 104770
+    },
+    {
+      "epoch": 676.0,
+      "learning_rate": 1.8e-05,
+      "loss": 0.2373,
+      "step": 104780
+    },
+    {
+      "epoch": 676.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9303658604621887,
+      "eval_runtime": 0.8914,
+      "eval_samples_per_second": 225.486,
+      "eval_steps_per_second": 7.853,
+      "step": 104780
+    },
+    {
+      "epoch": 676.06,
+      "learning_rate": 1.7996415770609322e-05,
+      "loss": 0.2533,
+      "step": 104790
+    },
+    {
+      "epoch": 676.13,
+      "learning_rate": 1.7992831541218637e-05,
+      "loss": 0.2459,
+      "step": 104800
+    },
+    {
+      "epoch": 676.19,
+      "learning_rate": 1.798924731182796e-05,
+      "loss": 0.2932,
+      "step": 104810
+    },
+    {
+      "epoch": 676.26,
+      "learning_rate": 1.7985663082437277e-05,
+      "loss": 0.208,
+      "step": 104820
+    },
+    {
+      "epoch": 676.32,
+      "learning_rate": 1.7982078853046598e-05,
+      "loss": 0.1894,
+      "step": 104830
+    },
+    {
+      "epoch": 676.39,
+      "learning_rate": 1.7978494623655913e-05,
+      "loss": 0.2141,
+      "step": 104840
+    },
+    {
+      "epoch": 676.45,
+      "learning_rate": 1.7974910394265235e-05,
+      "loss": 0.2573,
+      "step": 104850
+    },
+    {
+      "epoch": 676.51,
+      "learning_rate": 1.7971326164874553e-05,
+      "loss": 0.2573,
+      "step": 104860
+    },
+    {
+      "epoch": 676.58,
+      "learning_rate": 1.796774193548387e-05,
+      "loss": 0.1978,
+      "step": 104870
+    },
+    {
+      "epoch": 676.64,
+      "learning_rate": 1.796415770609319e-05,
+      "loss": 0.208,
+      "step": 104880
+    },
+    {
+      "epoch": 676.71,
+      "learning_rate": 1.796057347670251e-05,
+      "loss": 0.2232,
+      "step": 104890
+    },
+    {
+      "epoch": 676.77,
+      "learning_rate": 1.795698924731183e-05,
+      "loss": 0.2346,
+      "step": 104900
+    },
+    {
+      "epoch": 676.84,
+      "learning_rate": 1.7953405017921147e-05,
+      "loss": 0.237,
+      "step": 104910
+    },
+    {
+      "epoch": 676.9,
+      "learning_rate": 1.7949820788530465e-05,
+      "loss": 0.2499,
+      "step": 104920
+    },
+    {
+      "epoch": 676.96,
+      "learning_rate": 1.7946236559139787e-05,
+      "loss": 0.211,
+      "step": 104930
+    },
+    {
+      "epoch": 677.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9860677123069763,
+      "eval_runtime": 0.8835,
+      "eval_samples_per_second": 227.498,
+      "eval_steps_per_second": 7.923,
+      "step": 104935
+    },
+    {
+      "epoch": 677.03,
+      "learning_rate": 1.7942652329749105e-05,
+      "loss": 0.245,
+      "step": 104940
+    },
+    {
+      "epoch": 677.1,
+      "learning_rate": 1.7939068100358423e-05,
+      "loss": 0.2435,
+      "step": 104950
+    },
+    {
+      "epoch": 677.16,
+      "learning_rate": 1.7935483870967742e-05,
+      "loss": 0.2319,
+      "step": 104960
+    },
+    {
+      "epoch": 677.23,
+      "learning_rate": 1.7931899641577063e-05,
+      "loss": 0.2033,
+      "step": 104970
+    },
+    {
+      "epoch": 677.29,
+      "learning_rate": 1.792831541218638e-05,
+      "loss": 0.2307,
+      "step": 104980
+    },
+    {
+      "epoch": 677.35,
+      "learning_rate": 1.79247311827957e-05,
+      "loss": 0.2264,
+      "step": 104990
+    },
+    {
+      "epoch": 677.42,
+      "learning_rate": 1.7921146953405018e-05,
+      "loss": 0.2214,
+      "step": 105000
+    },
+    {
+      "epoch": 677.48,
+      "learning_rate": 1.791756272401434e-05,
+      "loss": 0.2615,
+      "step": 105010
+    },
+    {
+      "epoch": 677.55,
+      "learning_rate": 1.7913978494623658e-05,
+      "loss": 0.2737,
+      "step": 105020
+    },
+    {
+      "epoch": 677.61,
+      "learning_rate": 1.7910394265232976e-05,
+      "loss": 0.2354,
+      "step": 105030
+    },
+    {
+      "epoch": 677.68,
+      "learning_rate": 1.7906810035842294e-05,
+      "loss": 0.2261,
+      "step": 105040
+    },
+    {
+      "epoch": 677.74,
+      "learning_rate": 1.7903225806451616e-05,
+      "loss": 0.2282,
+      "step": 105050
+    },
+    {
+      "epoch": 677.8,
+      "learning_rate": 1.7899641577060934e-05,
+      "loss": 0.2406,
+      "step": 105060
+    },
+    {
+      "epoch": 677.87,
+      "learning_rate": 1.7896057347670252e-05,
+      "loss": 0.2613,
+      "step": 105070
+    },
+    {
+      "epoch": 677.93,
+      "learning_rate": 1.789247311827957e-05,
+      "loss": 0.243,
+      "step": 105080
+    },
+    {
+      "epoch": 678.0,
+      "learning_rate": 1.788888888888889e-05,
+      "loss": 0.2264,
+      "step": 105090
+    },
+    {
+      "epoch": 678.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9436492323875427,
+      "eval_runtime": 0.9091,
+      "eval_samples_per_second": 221.1,
+      "eval_steps_per_second": 7.7,
+      "step": 105090
+    },
+    {
+      "epoch": 678.06,
+      "learning_rate": 1.788530465949821e-05,
+      "loss": 0.2166,
+      "step": 105100
+    },
+    {
+      "epoch": 678.13,
+      "learning_rate": 1.7881720430107525e-05,
+      "loss": 0.2175,
+      "step": 105110
+    },
+    {
+      "epoch": 678.19,
+      "learning_rate": 1.7878136200716847e-05,
+      "loss": 0.192,
+      "step": 105120
+    },
+    {
+      "epoch": 678.26,
+      "learning_rate": 1.7874551971326165e-05,
+      "loss": 0.2519,
+      "step": 105130
+    },
+    {
+      "epoch": 678.32,
+      "learning_rate": 1.7870967741935486e-05,
+      "loss": 0.2696,
+      "step": 105140
+    },
+    {
+      "epoch": 678.39,
+      "learning_rate": 1.78673835125448e-05,
+      "loss": 0.212,
+      "step": 105150
+    },
+    {
+      "epoch": 678.45,
+      "learning_rate": 1.7863799283154123e-05,
+      "loss": 0.2283,
+      "step": 105160
+    },
+    {
+      "epoch": 678.51,
+      "learning_rate": 1.786021505376344e-05,
+      "loss": 0.2279,
+      "step": 105170
+    },
+    {
+      "epoch": 678.58,
+      "learning_rate": 1.7856630824372763e-05,
+      "loss": 0.2092,
+      "step": 105180
+    },
+    {
+      "epoch": 678.64,
+      "learning_rate": 1.7853046594982077e-05,
+      "loss": 0.22,
+      "step": 105190
+    },
+    {
+      "epoch": 678.71,
+      "learning_rate": 1.78494623655914e-05,
+      "loss": 0.2233,
+      "step": 105200
+    },
+    {
+      "epoch": 678.77,
+      "learning_rate": 1.7845878136200717e-05,
+      "loss": 0.1997,
+      "step": 105210
+    },
+    {
+      "epoch": 678.84,
+      "learning_rate": 1.7842293906810035e-05,
+      "loss": 0.1881,
+      "step": 105220
+    },
+    {
+      "epoch": 678.9,
+      "learning_rate": 1.7838709677419354e-05,
+      "loss": 0.2376,
+      "step": 105230
+    },
+    {
+      "epoch": 678.96,
+      "learning_rate": 1.7835125448028675e-05,
+      "loss": 0.2032,
+      "step": 105240
+    },
+    {
+      "epoch": 679.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9397395253181458,
+      "eval_runtime": 0.8852,
+      "eval_samples_per_second": 227.065,
+      "eval_steps_per_second": 7.908,
+      "step": 105245
+    },
+    {
+      "epoch": 679.03,
+      "learning_rate": 1.7831541218637993e-05,
+      "loss": 0.2417,
+      "step": 105250
+    },
+    {
+      "epoch": 679.1,
+      "learning_rate": 1.782795698924731e-05,
+      "loss": 0.1813,
+      "step": 105260
+    },
+    {
+      "epoch": 679.16,
+      "learning_rate": 1.782437275985663e-05,
+      "loss": 0.2184,
+      "step": 105270
+    },
+    {
+      "epoch": 679.23,
+      "learning_rate": 1.782078853046595e-05,
+      "loss": 0.2145,
+      "step": 105280
+    },
+    {
+      "epoch": 679.29,
+      "learning_rate": 1.781720430107527e-05,
+      "loss": 0.2063,
+      "step": 105290
+    },
+    {
+      "epoch": 679.35,
+      "learning_rate": 1.7813620071684588e-05,
+      "loss": 0.2275,
+      "step": 105300
+    },
+    {
+      "epoch": 679.42,
+      "learning_rate": 1.7810035842293906e-05,
+      "loss": 0.2118,
+      "step": 105310
+    },
+    {
+      "epoch": 679.48,
+      "learning_rate": 1.7806451612903228e-05,
+      "loss": 0.2687,
+      "step": 105320
+    },
+    {
+      "epoch": 679.55,
+      "learning_rate": 1.7802867383512546e-05,
+      "loss": 0.2749,
+      "step": 105330
+    },
+    {
+      "epoch": 679.61,
+      "learning_rate": 1.7799283154121864e-05,
+      "loss": 0.2301,
+      "step": 105340
+    },
+    {
+      "epoch": 679.68,
+      "learning_rate": 1.7795698924731182e-05,
+      "loss": 0.2529,
+      "step": 105350
+    },
+    {
+      "epoch": 679.74,
+      "learning_rate": 1.7792114695340504e-05,
+      "loss": 0.241,
+      "step": 105360
+    },
+    {
+      "epoch": 679.8,
+      "learning_rate": 1.7788530465949822e-05,
+      "loss": 0.2586,
+      "step": 105370
+    },
+    {
+      "epoch": 679.87,
+      "learning_rate": 1.778494623655914e-05,
+      "loss": 0.2594,
+      "step": 105380
+    },
+    {
+      "epoch": 679.93,
+      "learning_rate": 1.778136200716846e-05,
+      "loss": 0.2437,
+      "step": 105390
+    },
+    {
+      "epoch": 680.0,
+      "learning_rate": 1.777777777777778e-05,
+      "loss": 0.2289,
+      "step": 105400
+    },
+    {
+      "epoch": 680.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9875171184539795,
+      "eval_runtime": 0.8804,
+      "eval_samples_per_second": 228.304,
+      "eval_steps_per_second": 7.951,
+      "step": 105400
+    },
+    {
+      "epoch": 680.06,
+      "learning_rate": 1.7774193548387098e-05,
+      "loss": 0.2801,
+      "step": 105410
+    },
+    {
+      "epoch": 680.13,
+      "learning_rate": 1.7770609318996416e-05,
+      "loss": 0.206,
+      "step": 105420
+    },
+    {
+      "epoch": 680.19,
+      "learning_rate": 1.7767025089605735e-05,
+      "loss": 0.235,
+      "step": 105430
+    },
+    {
+      "epoch": 680.26,
+      "learning_rate": 1.7763440860215056e-05,
+      "loss": 0.2384,
+      "step": 105440
+    },
+    {
+      "epoch": 680.32,
+      "learning_rate": 1.7759856630824374e-05,
+      "loss": 0.2245,
+      "step": 105450
+    },
+    {
+      "epoch": 680.39,
+      "learning_rate": 1.7756272401433693e-05,
+      "loss": 0.2628,
+      "step": 105460
+    },
+    {
+      "epoch": 680.45,
+      "learning_rate": 1.775268817204301e-05,
+      "loss": 0.2255,
+      "step": 105470
+    },
+    {
+      "epoch": 680.51,
+      "learning_rate": 1.7749103942652332e-05,
+      "loss": 0.2192,
+      "step": 105480
+    },
+    {
+      "epoch": 680.58,
+      "learning_rate": 1.774551971326165e-05,
+      "loss": 0.2721,
+      "step": 105490
+    },
+    {
+      "epoch": 680.64,
+      "learning_rate": 1.774193548387097e-05,
+      "loss": 0.2363,
+      "step": 105500
+    },
+    {
+      "epoch": 680.71,
+      "learning_rate": 1.7738351254480287e-05,
+      "loss": 0.2149,
+      "step": 105510
+    },
+    {
+      "epoch": 680.77,
+      "learning_rate": 1.773476702508961e-05,
+      "loss": 0.2549,
+      "step": 105520
+    },
+    {
+      "epoch": 680.84,
+      "learning_rate": 1.7731182795698927e-05,
+      "loss": 0.2341,
+      "step": 105530
+    },
+    {
+      "epoch": 680.9,
+      "learning_rate": 1.7727598566308245e-05,
+      "loss": 0.2195,
+      "step": 105540
+    },
+    {
+      "epoch": 680.96,
+      "learning_rate": 1.7724014336917563e-05,
+      "loss": 0.2035,
+      "step": 105550
+    },
+    {
+      "epoch": 681.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9824814200401306,
+      "eval_runtime": 0.8925,
+      "eval_samples_per_second": 225.221,
+      "eval_steps_per_second": 7.844,
+      "step": 105555
+    },
+    {
+      "epoch": 681.03,
+      "learning_rate": 1.7720430107526885e-05,
+      "loss": 0.2258,
+      "step": 105560
+    },
+    {
+      "epoch": 681.1,
+      "learning_rate": 1.77168458781362e-05,
+      "loss": 0.2122,
+      "step": 105570
+    },
+    {
+      "epoch": 681.16,
+      "learning_rate": 1.771326164874552e-05,
+      "loss": 0.2457,
+      "step": 105580
+    },
+    {
+      "epoch": 681.23,
+      "learning_rate": 1.770967741935484e-05,
+      "loss": 0.2014,
+      "step": 105590
+    },
+    {
+      "epoch": 681.29,
+      "learning_rate": 1.770609318996416e-05,
+      "loss": 0.2137,
+      "step": 105600
+    },
+    {
+      "epoch": 681.35,
+      "learning_rate": 1.7702508960573476e-05,
+      "loss": 0.2569,
+      "step": 105610
+    },
+    {
+      "epoch": 681.42,
+      "learning_rate": 1.7698924731182797e-05,
+      "loss": 0.2341,
+      "step": 105620
+    },
+    {
+      "epoch": 681.48,
+      "learning_rate": 1.7695340501792116e-05,
+      "loss": 0.2353,
+      "step": 105630
+    },
+    {
+      "epoch": 681.55,
+      "learning_rate": 1.7691756272401437e-05,
+      "loss": 0.2534,
+      "step": 105640
+    },
+    {
+      "epoch": 681.61,
+      "learning_rate": 1.7688172043010752e-05,
+      "loss": 0.2264,
+      "step": 105650
+    },
+    {
+      "epoch": 681.68,
+      "learning_rate": 1.7684587813620074e-05,
+      "loss": 0.1985,
+      "step": 105660
+    },
+    {
+      "epoch": 681.74,
+      "learning_rate": 1.7681003584229392e-05,
+      "loss": 0.2122,
+      "step": 105670
+    },
+    {
+      "epoch": 681.8,
+      "learning_rate": 1.7677419354838713e-05,
+      "loss": 0.2341,
+      "step": 105680
+    },
+    {
+      "epoch": 681.87,
+      "learning_rate": 1.7673835125448028e-05,
+      "loss": 0.1923,
+      "step": 105690
+    },
+    {
+      "epoch": 681.93,
+      "learning_rate": 1.767025089605735e-05,
+      "loss": 0.2451,
+      "step": 105700
+    },
+    {
+      "epoch": 682.0,
+      "learning_rate": 1.7666666666666668e-05,
+      "loss": 0.1997,
+      "step": 105710
+    },
+    {
+      "epoch": 682.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9347655773162842,
+      "eval_runtime": 0.8732,
+      "eval_samples_per_second": 230.187,
+      "eval_steps_per_second": 8.016,
+      "step": 105710
+    },
+    {
+      "epoch": 682.06,
+      "learning_rate": 1.7663082437275986e-05,
+      "loss": 0.2217,
+      "step": 105720
+    },
+    {
+      "epoch": 682.13,
+      "learning_rate": 1.7659498207885304e-05,
+      "loss": 0.2318,
+      "step": 105730
+    },
+    {
+      "epoch": 682.19,
+      "learning_rate": 1.7655913978494623e-05,
+      "loss": 0.2518,
+      "step": 105740
+    },
+    {
+      "epoch": 682.26,
+      "learning_rate": 1.7652329749103944e-05,
+      "loss": 0.2658,
+      "step": 105750
+    },
+    {
+      "epoch": 682.32,
+      "learning_rate": 1.7648745519713262e-05,
+      "loss": 0.2734,
+      "step": 105760
+    },
+    {
+      "epoch": 682.39,
+      "learning_rate": 1.764516129032258e-05,
+      "loss": 0.218,
+      "step": 105770
+    },
+    {
+      "epoch": 682.45,
+      "learning_rate": 1.76415770609319e-05,
+      "loss": 0.2094,
+      "step": 105780
+    },
+    {
+      "epoch": 682.51,
+      "learning_rate": 1.763799283154122e-05,
+      "loss": 0.2777,
+      "step": 105790
+    },
+    {
+      "epoch": 682.58,
+      "learning_rate": 1.763440860215054e-05,
+      "loss": 0.2865,
+      "step": 105800
+    },
+    {
+      "epoch": 682.64,
+      "learning_rate": 1.7630824372759857e-05,
+      "loss": 0.2292,
+      "step": 105810
+    },
+    {
+      "epoch": 682.71,
+      "learning_rate": 1.7627240143369175e-05,
+      "loss": 0.2091,
+      "step": 105820
+    },
+    {
+      "epoch": 682.77,
+      "learning_rate": 1.7623655913978497e-05,
+      "loss": 0.2182,
+      "step": 105830
+    },
+    {
+      "epoch": 682.84,
+      "learning_rate": 1.7620071684587815e-05,
+      "loss": 0.2031,
+      "step": 105840
+    },
+    {
+      "epoch": 682.9,
+      "learning_rate": 1.7616487455197133e-05,
+      "loss": 0.2275,
+      "step": 105850
+    },
+    {
+      "epoch": 682.96,
+      "learning_rate": 1.761290322580645e-05,
+      "loss": 0.2181,
+      "step": 105860
+    },
+    {
+      "epoch": 683.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.97990483045578,
+      "eval_runtime": 0.8778,
+      "eval_samples_per_second": 228.98,
+      "eval_steps_per_second": 7.974,
+      "step": 105865
+    },
+    {
+      "epoch": 683.03,
+      "learning_rate": 1.7609318996415773e-05,
+      "loss": 0.2789,
+      "step": 105870
+    },
+    {
+      "epoch": 683.1,
+      "learning_rate": 1.7605734767025088e-05,
+      "loss": 0.2605,
+      "step": 105880
+    },
+    {
+      "epoch": 683.16,
+      "learning_rate": 1.760215053763441e-05,
+      "loss": 0.2098,
+      "step": 105890
+    },
+    {
+      "epoch": 683.23,
+      "learning_rate": 1.7598566308243728e-05,
+      "loss": 0.1857,
+      "step": 105900
+    },
+    {
+      "epoch": 683.29,
+      "learning_rate": 1.759498207885305e-05,
+      "loss": 0.2152,
+      "step": 105910
+    },
+    {
+      "epoch": 683.35,
+      "learning_rate": 1.7591397849462364e-05,
+      "loss": 0.2093,
+      "step": 105920
+    },
+    {
+      "epoch": 683.42,
+      "learning_rate": 1.7587813620071686e-05,
+      "loss": 0.1693,
+      "step": 105930
+    },
+    {
+      "epoch": 683.48,
+      "learning_rate": 1.7584229390681004e-05,
+      "loss": 0.2409,
+      "step": 105940
+    },
+    {
+      "epoch": 683.55,
+      "learning_rate": 1.7580645161290325e-05,
+      "loss": 0.1906,
+      "step": 105950
+    },
+    {
+      "epoch": 683.61,
+      "learning_rate": 1.757706093189964e-05,
+      "loss": 0.2391,
+      "step": 105960
+    },
+    {
+      "epoch": 683.68,
+      "learning_rate": 1.7573476702508962e-05,
+      "loss": 0.192,
+      "step": 105970
+    },
+    {
+      "epoch": 683.74,
+      "learning_rate": 1.756989247311828e-05,
+      "loss": 0.2489,
+      "step": 105980
+    },
+    {
+      "epoch": 683.8,
+      "learning_rate": 1.75663082437276e-05,
+      "loss": 0.2097,
+      "step": 105990
+    },
+    {
+      "epoch": 683.87,
+      "learning_rate": 1.7562724014336916e-05,
+      "loss": 0.2049,
+      "step": 106000
+    },
+    {
+      "epoch": 683.93,
+      "learning_rate": 1.7559139784946238e-05,
+      "loss": 0.2195,
+      "step": 106010
+    },
+    {
+      "epoch": 684.0,
+      "learning_rate": 1.7555555555555556e-05,
+      "loss": 0.2637,
+      "step": 106020
+    },
+    {
+      "epoch": 684.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.985089898109436,
+      "eval_runtime": 0.8764,
+      "eval_samples_per_second": 229.342,
+      "eval_steps_per_second": 7.987,
+      "step": 106020
+    },
+    {
+      "epoch": 684.06,
+      "learning_rate": 1.7551971326164878e-05,
+      "loss": 0.2466,
+      "step": 106030
+    },
+    {
+      "epoch": 684.13,
+      "learning_rate": 1.7548387096774193e-05,
+      "loss": 0.2515,
+      "step": 106040
+    },
+    {
+      "epoch": 684.19,
+      "learning_rate": 1.7544802867383514e-05,
+      "loss": 0.2733,
+      "step": 106050
+    },
+    {
+      "epoch": 684.26,
+      "learning_rate": 1.7541218637992832e-05,
+      "loss": 0.2356,
+      "step": 106060
+    },
+    {
+      "epoch": 684.32,
+      "learning_rate": 1.753763440860215e-05,
+      "loss": 0.207,
+      "step": 106070
+    },
+    {
+      "epoch": 684.39,
+      "learning_rate": 1.753405017921147e-05,
+      "loss": 0.2274,
+      "step": 106080
+    },
+    {
+      "epoch": 684.45,
+      "learning_rate": 1.753046594982079e-05,
+      "loss": 0.1975,
+      "step": 106090
+    },
+    {
+      "epoch": 684.51,
+      "learning_rate": 1.752688172043011e-05,
+      "loss": 0.2349,
+      "step": 106100
+    },
+    {
+      "epoch": 684.58,
+      "learning_rate": 1.7523297491039427e-05,
+      "loss": 0.1931,
+      "step": 106110
+    },
+    {
+      "epoch": 684.64,
+      "learning_rate": 1.7519713261648745e-05,
+      "loss": 0.2414,
+      "step": 106120
+    },
+    {
+      "epoch": 684.71,
+      "learning_rate": 1.7516129032258067e-05,
+      "loss": 0.2397,
+      "step": 106130
+    },
+    {
+      "epoch": 684.77,
+      "learning_rate": 1.7512544802867385e-05,
+      "loss": 0.2074,
+      "step": 106140
+    },
+    {
+      "epoch": 684.84,
+      "learning_rate": 1.7508960573476703e-05,
+      "loss": 0.2693,
+      "step": 106150
+    },
+    {
+      "epoch": 684.9,
+      "learning_rate": 1.750537634408602e-05,
+      "loss": 0.2277,
+      "step": 106160
+    },
+    {
+      "epoch": 684.96,
+      "learning_rate": 1.7501792114695343e-05,
+      "loss": 0.2288,
+      "step": 106170
+    },
+    {
+      "epoch": 685.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9742719531059265,
+      "eval_runtime": 0.8908,
+      "eval_samples_per_second": 225.648,
+      "eval_steps_per_second": 7.858,
+      "step": 106175
+    },
+    {
+      "epoch": 685.03,
+      "learning_rate": 1.749820788530466e-05,
+      "loss": 0.2085,
+      "step": 106180
+    },
+    {
+      "epoch": 685.1,
+      "learning_rate": 1.749462365591398e-05,
+      "loss": 0.2263,
+      "step": 106190
+    },
+    {
+      "epoch": 685.16,
+      "learning_rate": 1.7491039426523297e-05,
+      "loss": 0.1965,
+      "step": 106200
+    },
+    {
+      "epoch": 685.23,
+      "learning_rate": 1.748745519713262e-05,
+      "loss": 0.2227,
+      "step": 106210
+    },
+    {
+      "epoch": 685.29,
+      "learning_rate": 1.7483870967741937e-05,
+      "loss": 0.2696,
+      "step": 106220
+    },
+    {
+      "epoch": 685.35,
+      "learning_rate": 1.7480286738351255e-05,
+      "loss": 0.1814,
+      "step": 106230
+    },
+    {
+      "epoch": 685.42,
+      "learning_rate": 1.7476702508960574e-05,
+      "loss": 0.281,
+      "step": 106240
+    },
+    {
+      "epoch": 685.48,
+      "learning_rate": 1.7473118279569895e-05,
+      "loss": 0.2133,
+      "step": 106250
+    },
+    {
+      "epoch": 685.55,
+      "learning_rate": 1.7469534050179213e-05,
+      "loss": 0.2577,
+      "step": 106260
+    },
+    {
+      "epoch": 685.61,
+      "learning_rate": 1.746594982078853e-05,
+      "loss": 0.2469,
+      "step": 106270
+    },
+    {
+      "epoch": 685.68,
+      "learning_rate": 1.746236559139785e-05,
+      "loss": 0.2331,
+      "step": 106280
+    },
+    {
+      "epoch": 685.74,
+      "learning_rate": 1.745878136200717e-05,
+      "loss": 0.2478,
+      "step": 106290
+    },
+    {
+      "epoch": 685.8,
+      "learning_rate": 1.745519713261649e-05,
+      "loss": 0.2262,
+      "step": 106300
+    },
+    {
+      "epoch": 685.87,
+      "learning_rate": 1.7451612903225808e-05,
+      "loss": 0.2112,
+      "step": 106310
+    },
+    {
+      "epoch": 685.93,
+      "learning_rate": 1.7448028673835126e-05,
+      "loss": 0.2417,
+      "step": 106320
+    },
+    {
+      "epoch": 686.0,
+      "learning_rate": 1.7444444444444448e-05,
+      "loss": 0.2251,
+      "step": 106330
+    },
+    {
+      "epoch": 686.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9770976901054382,
+      "eval_runtime": 0.8726,
+      "eval_samples_per_second": 230.353,
+      "eval_steps_per_second": 8.022,
+      "step": 106330
+    },
+    {
+      "epoch": 686.06,
+      "learning_rate": 1.7440860215053766e-05,
+      "loss": 0.2229,
+      "step": 106340
+    },
+    {
+      "epoch": 686.13,
+      "learning_rate": 1.7437275985663084e-05,
+      "loss": 0.2416,
+      "step": 106350
+    },
+    {
+      "epoch": 686.19,
+      "learning_rate": 1.7433691756272402e-05,
+      "loss": 0.241,
+      "step": 106360
+    },
+    {
+      "epoch": 686.26,
+      "learning_rate": 1.743010752688172e-05,
+      "loss": 0.2505,
+      "step": 106370
+    },
+    {
+      "epoch": 686.32,
+      "learning_rate": 1.742652329749104e-05,
+      "loss": 0.2421,
+      "step": 106380
+    },
+    {
+      "epoch": 686.39,
+      "learning_rate": 1.7422939068100357e-05,
+      "loss": 0.2729,
+      "step": 106390
+    },
+    {
+      "epoch": 686.45,
+      "learning_rate": 1.741935483870968e-05,
+      "loss": 0.2171,
+      "step": 106400
+    },
+    {
+      "epoch": 686.51,
+      "learning_rate": 1.7415770609318997e-05,
+      "loss": 0.2731,
+      "step": 106410
+    },
+    {
+      "epoch": 686.58,
+      "learning_rate": 1.7412186379928315e-05,
+      "loss": 0.2323,
+      "step": 106420
+    },
+    {
+      "epoch": 686.64,
+      "learning_rate": 1.7408602150537633e-05,
+      "loss": 0.2451,
+      "step": 106430
+    },
+    {
+      "epoch": 686.71,
+      "learning_rate": 1.7405017921146955e-05,
+      "loss": 0.2238,
+      "step": 106440
+    },
+    {
+      "epoch": 686.77,
+      "learning_rate": 1.7401433691756273e-05,
+      "loss": 0.1941,
+      "step": 106450
+    },
+    {
+      "epoch": 686.84,
+      "learning_rate": 1.739784946236559e-05,
+      "loss": 0.2453,
+      "step": 106460
+    },
+    {
+      "epoch": 686.9,
+      "learning_rate": 1.739426523297491e-05,
+      "loss": 0.2477,
+      "step": 106470
+    },
+    {
+      "epoch": 686.96,
+      "learning_rate": 1.739068100358423e-05,
+      "loss": 0.2467,
+      "step": 106480
+    },
+    {
+      "epoch": 687.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.975865364074707,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 230.71,
+      "eval_steps_per_second": 8.035,
+      "step": 106485
+    },
+    {
+      "epoch": 687.03,
+      "learning_rate": 1.738709677419355e-05,
+      "loss": 0.2379,
+      "step": 106490
+    },
+    {
+      "epoch": 687.1,
+      "learning_rate": 1.7383512544802867e-05,
+      "loss": 0.2234,
+      "step": 106500
+    },
+    {
+      "epoch": 687.16,
+      "learning_rate": 1.7379928315412185e-05,
+      "loss": 0.2262,
+      "step": 106510
+    },
+    {
+      "epoch": 687.23,
+      "learning_rate": 1.7376344086021507e-05,
+      "loss": 0.2479,
+      "step": 106520
+    },
+    {
+      "epoch": 687.29,
+      "learning_rate": 1.7372759856630825e-05,
+      "loss": 0.2035,
+      "step": 106530
+    },
+    {
+      "epoch": 687.35,
+      "learning_rate": 1.7369175627240143e-05,
+      "loss": 0.2462,
+      "step": 106540
+    },
+    {
+      "epoch": 687.42,
+      "learning_rate": 1.736559139784946e-05,
+      "loss": 0.24,
+      "step": 106550
+    },
+    {
+      "epoch": 687.48,
+      "learning_rate": 1.7362007168458783e-05,
+      "loss": 0.1705,
+      "step": 106560
+    },
+    {
+      "epoch": 687.55,
+      "learning_rate": 1.73584229390681e-05,
+      "loss": 0.1826,
+      "step": 106570
+    },
+    {
+      "epoch": 687.61,
+      "learning_rate": 1.735483870967742e-05,
+      "loss": 0.2616,
+      "step": 106580
+    },
+    {
+      "epoch": 687.68,
+      "learning_rate": 1.7351254480286738e-05,
+      "loss": 0.2699,
+      "step": 106590
+    },
+    {
+      "epoch": 687.74,
+      "learning_rate": 1.734767025089606e-05,
+      "loss": 0.2086,
+      "step": 106600
+    },
+    {
+      "epoch": 687.8,
+      "learning_rate": 1.7344086021505378e-05,
+      "loss": 0.2223,
+      "step": 106610
+    },
+    {
+      "epoch": 687.87,
+      "learning_rate": 1.7340501792114696e-05,
+      "loss": 0.2721,
+      "step": 106620
+    },
+    {
+      "epoch": 687.93,
+      "learning_rate": 1.7336917562724014e-05,
+      "loss": 0.2196,
+      "step": 106630
+    },
+    {
+      "epoch": 688.0,
+      "learning_rate": 1.7333333333333336e-05,
+      "loss": 0.1985,
+      "step": 106640
+    },
+    {
+      "epoch": 688.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9627480506896973,
+      "eval_runtime": 0.8744,
+      "eval_samples_per_second": 229.866,
+      "eval_steps_per_second": 8.005,
+      "step": 106640
+    },
+    {
+      "epoch": 688.06,
+      "learning_rate": 1.7329749103942654e-05,
+      "loss": 0.2419,
+      "step": 106650
+    },
+    {
+      "epoch": 688.13,
+      "learning_rate": 1.7326164874551972e-05,
+      "loss": 0.139,
+      "step": 106660
+    },
+    {
+      "epoch": 688.19,
+      "learning_rate": 1.732258064516129e-05,
+      "loss": 0.2018,
+      "step": 106670
+    },
+    {
+      "epoch": 688.26,
+      "learning_rate": 1.7318996415770612e-05,
+      "loss": 0.2121,
+      "step": 106680
+    },
+    {
+      "epoch": 688.32,
+      "learning_rate": 1.731541218637993e-05,
+      "loss": 0.1928,
+      "step": 106690
+    },
+    {
+      "epoch": 688.39,
+      "learning_rate": 1.7311827956989248e-05,
+      "loss": 0.2482,
+      "step": 106700
+    },
+    {
+      "epoch": 688.45,
+      "learning_rate": 1.7308243727598566e-05,
+      "loss": 0.2022,
+      "step": 106710
+    },
+    {
+      "epoch": 688.51,
+      "learning_rate": 1.7304659498207888e-05,
+      "loss": 0.2662,
+      "step": 106720
+    },
+    {
+      "epoch": 688.58,
+      "learning_rate": 1.7301075268817203e-05,
+      "loss": 0.2146,
+      "step": 106730
+    },
+    {
+      "epoch": 688.64,
+      "learning_rate": 1.7297491039426524e-05,
+      "loss": 0.2589,
+      "step": 106740
+    },
+    {
+      "epoch": 688.71,
+      "learning_rate": 1.7293906810035843e-05,
+      "loss": 0.2396,
+      "step": 106750
+    },
+    {
+      "epoch": 688.77,
+      "learning_rate": 1.7290322580645164e-05,
+      "loss": 0.2031,
+      "step": 106760
+    },
+    {
+      "epoch": 688.84,
+      "learning_rate": 1.728673835125448e-05,
+      "loss": 0.2664,
+      "step": 106770
+    },
+    {
+      "epoch": 688.9,
+      "learning_rate": 1.72831541218638e-05,
+      "loss": 0.1862,
+      "step": 106780
+    },
+    {
+      "epoch": 688.96,
+      "learning_rate": 1.727956989247312e-05,
+      "loss": 0.2419,
+      "step": 106790
+    },
+    {
+      "epoch": 689.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9423263072967529,
+      "eval_runtime": 0.931,
+      "eval_samples_per_second": 215.894,
+      "eval_steps_per_second": 7.519,
+      "step": 106795
+    },
+    {
+      "epoch": 689.03,
+      "learning_rate": 1.727598566308244e-05,
+      "loss": 0.2764,
+      "step": 106800
+    },
+    {
+      "epoch": 689.1,
+      "learning_rate": 1.7272401433691755e-05,
+      "loss": 0.2347,
+      "step": 106810
+    },
+    {
+      "epoch": 689.16,
+      "learning_rate": 1.7268817204301077e-05,
+      "loss": 0.1997,
+      "step": 106820
+    },
+    {
+      "epoch": 689.23,
+      "learning_rate": 1.7265232974910395e-05,
+      "loss": 0.2211,
+      "step": 106830
+    },
+    {
+      "epoch": 689.29,
+      "learning_rate": 1.7261648745519717e-05,
+      "loss": 0.2268,
+      "step": 106840
+    },
+    {
+      "epoch": 689.35,
+      "learning_rate": 1.725806451612903e-05,
+      "loss": 0.2271,
+      "step": 106850
+    },
+    {
+      "epoch": 689.42,
+      "learning_rate": 1.7254480286738353e-05,
+      "loss": 0.2322,
+      "step": 106860
+    },
+    {
+      "epoch": 689.48,
+      "learning_rate": 1.725089605734767e-05,
+      "loss": 0.2162,
+      "step": 106870
+    },
+    {
+      "epoch": 689.55,
+      "learning_rate": 1.724731182795699e-05,
+      "loss": 0.3017,
+      "step": 106880
+    },
+    {
+      "epoch": 689.61,
+      "learning_rate": 1.7243727598566308e-05,
+      "loss": 0.2555,
+      "step": 106890
+    },
+    {
+      "epoch": 689.68,
+      "learning_rate": 1.724014336917563e-05,
+      "loss": 0.2421,
+      "step": 106900
+    },
+    {
+      "epoch": 689.74,
+      "learning_rate": 1.7236559139784948e-05,
+      "loss": 0.198,
+      "step": 106910
+    },
+    {
+      "epoch": 689.8,
+      "learning_rate": 1.7232974910394266e-05,
+      "loss": 0.2633,
+      "step": 106920
+    },
+    {
+      "epoch": 689.87,
+      "learning_rate": 1.7229390681003584e-05,
+      "loss": 0.263,
+      "step": 106930
+    },
+    {
+      "epoch": 689.93,
+      "learning_rate": 1.7225806451612906e-05,
+      "loss": 0.2329,
+      "step": 106940
+    },
+    {
+      "epoch": 690.0,
+      "learning_rate": 1.7222222222222224e-05,
+      "loss": 0.2343,
+      "step": 106950
+    },
+    {
+      "epoch": 690.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9535319209098816,
+      "eval_runtime": 0.8879,
+      "eval_samples_per_second": 226.366,
+      "eval_steps_per_second": 7.883,
+      "step": 106950
+    },
+    {
+      "epoch": 690.06,
+      "learning_rate": 1.7218637992831542e-05,
+      "loss": 0.2292,
+      "step": 106960
+    },
+    {
+      "epoch": 690.13,
+      "learning_rate": 1.721505376344086e-05,
+      "loss": 0.194,
+      "step": 106970
+    },
+    {
+      "epoch": 690.19,
+      "learning_rate": 1.7211469534050182e-05,
+      "loss": 0.2114,
+      "step": 106980
+    },
+    {
+      "epoch": 690.26,
+      "learning_rate": 1.72078853046595e-05,
+      "loss": 0.1918,
+      "step": 106990
+    },
+    {
+      "epoch": 690.32,
+      "learning_rate": 1.7204301075268818e-05,
+      "loss": 0.2326,
+      "step": 107000
+    },
+    {
+      "epoch": 690.39,
+      "learning_rate": 1.7200716845878136e-05,
+      "loss": 0.2121,
+      "step": 107010
+    },
+    {
+      "epoch": 690.45,
+      "learning_rate": 1.7197132616487455e-05,
+      "loss": 0.2266,
+      "step": 107020
+    },
+    {
+      "epoch": 690.51,
+      "learning_rate": 1.7193548387096776e-05,
+      "loss": 0.2539,
+      "step": 107030
+    },
+    {
+      "epoch": 690.58,
+      "learning_rate": 1.7189964157706094e-05,
+      "loss": 0.2312,
+      "step": 107040
+    },
+    {
+      "epoch": 690.64,
+      "learning_rate": 1.7186379928315413e-05,
+      "loss": 0.2311,
+      "step": 107050
+    },
+    {
+      "epoch": 690.71,
+      "learning_rate": 1.718279569892473e-05,
+      "loss": 0.2076,
+      "step": 107060
+    },
+    {
+      "epoch": 690.77,
+      "learning_rate": 1.7179211469534052e-05,
+      "loss": 0.2423,
+      "step": 107070
+    },
+    {
+      "epoch": 690.84,
+      "learning_rate": 1.7175627240143367e-05,
+      "loss": 0.1829,
+      "step": 107080
+    },
+    {
+      "epoch": 690.9,
+      "learning_rate": 1.717204301075269e-05,
+      "loss": 0.2255,
+      "step": 107090
+    },
+    {
+      "epoch": 690.96,
+      "learning_rate": 1.7168458781362007e-05,
+      "loss": 0.2673,
+      "step": 107100
+    },
+    {
+      "epoch": 691.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9410730004310608,
+      "eval_runtime": 0.9218,
+      "eval_samples_per_second": 218.042,
+      "eval_steps_per_second": 7.594,
+      "step": 107105
+    },
+    {
+      "epoch": 691.03,
+      "learning_rate": 1.716487455197133e-05,
+      "loss": 0.2465,
+      "step": 107110
+    },
+    {
+      "epoch": 691.1,
+      "learning_rate": 1.7161290322580643e-05,
+      "loss": 0.2066,
+      "step": 107120
+    },
+    {
+      "epoch": 691.16,
+      "learning_rate": 1.7157706093189965e-05,
+      "loss": 0.1976,
+      "step": 107130
+    },
+    {
+      "epoch": 691.23,
+      "learning_rate": 1.7154121863799283e-05,
+      "loss": 0.205,
+      "step": 107140
+    },
+    {
+      "epoch": 691.29,
+      "learning_rate": 1.7150537634408605e-05,
+      "loss": 0.2246,
+      "step": 107150
+    },
+    {
+      "epoch": 691.35,
+      "learning_rate": 1.714695340501792e-05,
+      "loss": 0.2739,
+      "step": 107160
+    },
+    {
+      "epoch": 691.42,
+      "learning_rate": 1.714336917562724e-05,
+      "loss": 0.2387,
+      "step": 107170
+    },
+    {
+      "epoch": 691.48,
+      "learning_rate": 1.713978494623656e-05,
+      "loss": 0.2131,
+      "step": 107180
+    },
+    {
+      "epoch": 691.55,
+      "learning_rate": 1.713620071684588e-05,
+      "loss": 0.235,
+      "step": 107190
+    },
+    {
+      "epoch": 691.61,
+      "learning_rate": 1.7132616487455196e-05,
+      "loss": 0.1701,
+      "step": 107200
+    },
+    {
+      "epoch": 691.68,
+      "learning_rate": 1.7129032258064517e-05,
+      "loss": 0.2139,
+      "step": 107210
+    },
+    {
+      "epoch": 691.74,
+      "learning_rate": 1.7125448028673836e-05,
+      "loss": 0.2433,
+      "step": 107220
+    },
+    {
+      "epoch": 691.8,
+      "learning_rate": 1.7121863799283154e-05,
+      "loss": 0.2344,
+      "step": 107230
+    },
+    {
+      "epoch": 691.87,
+      "learning_rate": 1.7118279569892472e-05,
+      "loss": 0.2174,
+      "step": 107240
+    },
+    {
+      "epoch": 691.93,
+      "learning_rate": 1.7114695340501794e-05,
+      "loss": 0.2347,
+      "step": 107250
+    },
+    {
+      "epoch": 692.0,
+      "learning_rate": 1.7111111111111112e-05,
+      "loss": 0.2378,
+      "step": 107260
+    },
+    {
+      "epoch": 692.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9001235365867615,
+      "eval_runtime": 0.8689,
+      "eval_samples_per_second": 231.327,
+      "eval_steps_per_second": 8.056,
+      "step": 107260
+    },
+    {
+      "epoch": 692.06,
+      "learning_rate": 1.710752688172043e-05,
+      "loss": 0.2607,
+      "step": 107270
+    },
+    {
+      "epoch": 692.13,
+      "learning_rate": 1.7103942652329748e-05,
+      "loss": 0.1904,
+      "step": 107280
+    },
+    {
+      "epoch": 692.19,
+      "learning_rate": 1.710035842293907e-05,
+      "loss": 0.1951,
+      "step": 107290
+    },
+    {
+      "epoch": 692.26,
+      "learning_rate": 1.7096774193548388e-05,
+      "loss": 0.2621,
+      "step": 107300
+    },
+    {
+      "epoch": 692.32,
+      "learning_rate": 1.7093189964157706e-05,
+      "loss": 0.247,
+      "step": 107310
+    },
+    {
+      "epoch": 692.39,
+      "learning_rate": 1.7089605734767024e-05,
+      "loss": 0.2223,
+      "step": 107320
+    },
+    {
+      "epoch": 692.45,
+      "learning_rate": 1.7086021505376346e-05,
+      "loss": 0.1844,
+      "step": 107330
+    },
+    {
+      "epoch": 692.51,
+      "learning_rate": 1.7082437275985664e-05,
+      "loss": 0.2661,
+      "step": 107340
+    },
+    {
+      "epoch": 692.58,
+      "learning_rate": 1.7078853046594982e-05,
+      "loss": 0.2285,
+      "step": 107350
+    },
+    {
+      "epoch": 692.64,
+      "learning_rate": 1.70752688172043e-05,
+      "loss": 0.2197,
+      "step": 107360
+    },
+    {
+      "epoch": 692.71,
+      "learning_rate": 1.7071684587813622e-05,
+      "loss": 0.2214,
+      "step": 107370
+    },
+    {
+      "epoch": 692.77,
+      "learning_rate": 1.706810035842294e-05,
+      "loss": 0.25,
+      "step": 107380
+    },
+    {
+      "epoch": 692.84,
+      "learning_rate": 1.706451612903226e-05,
+      "loss": 0.2619,
+      "step": 107390
+    },
+    {
+      "epoch": 692.9,
+      "learning_rate": 1.7060931899641577e-05,
+      "loss": 0.2549,
+      "step": 107400
+    },
+    {
+      "epoch": 692.96,
+      "learning_rate": 1.70573476702509e-05,
+      "loss": 0.2048,
+      "step": 107410
+    },
+    {
+      "epoch": 693.0,
+      "eval_accuracy": 0.736318407960199,
+      "eval_loss": 0.9555172324180603,
+      "eval_runtime": 0.8708,
+      "eval_samples_per_second": 230.832,
+      "eval_steps_per_second": 8.039,
+      "step": 107415
+    },
+    {
+      "epoch": 693.03,
+      "learning_rate": 1.7053763440860217e-05,
+      "loss": 0.2955,
+      "step": 107420
+    },
+    {
+      "epoch": 693.1,
+      "learning_rate": 1.7050179211469535e-05,
+      "loss": 0.1733,
+      "step": 107430
+    },
+    {
+      "epoch": 693.16,
+      "learning_rate": 1.7046594982078853e-05,
+      "loss": 0.2071,
+      "step": 107440
+    },
+    {
+      "epoch": 693.23,
+      "learning_rate": 1.7043010752688175e-05,
+      "loss": 0.2352,
+      "step": 107450
+    },
+    {
+      "epoch": 693.29,
+      "learning_rate": 1.7039426523297493e-05,
+      "loss": 0.2791,
+      "step": 107460
+    },
+    {
+      "epoch": 693.35,
+      "learning_rate": 1.703584229390681e-05,
+      "loss": 0.231,
+      "step": 107470
+    },
+    {
+      "epoch": 693.42,
+      "learning_rate": 1.703225806451613e-05,
+      "loss": 0.2734,
+      "step": 107480
+    },
+    {
+      "epoch": 693.48,
+      "learning_rate": 1.702867383512545e-05,
+      "loss": 0.2391,
+      "step": 107490
+    },
+    {
+      "epoch": 693.55,
+      "learning_rate": 1.702508960573477e-05,
+      "loss": 0.2077,
+      "step": 107500
+    },
+    {
+      "epoch": 693.61,
+      "learning_rate": 1.7021505376344087e-05,
+      "loss": 0.216,
+      "step": 107510
+    },
+    {
+      "epoch": 693.68,
+      "learning_rate": 1.7017921146953405e-05,
+      "loss": 0.1973,
+      "step": 107520
+    },
+    {
+      "epoch": 693.74,
+      "learning_rate": 1.7014336917562727e-05,
+      "loss": 0.2081,
+      "step": 107530
+    },
+    {
+      "epoch": 693.8,
+      "learning_rate": 1.7010752688172045e-05,
+      "loss": 0.2596,
+      "step": 107540
+    },
+    {
+      "epoch": 693.87,
+      "learning_rate": 1.7007168458781363e-05,
+      "loss": 0.2254,
+      "step": 107550
+    },
+    {
+      "epoch": 693.93,
+      "learning_rate": 1.700358422939068e-05,
+      "loss": 0.2471,
+      "step": 107560
+    },
+    {
+      "epoch": 694.0,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.265,
+      "step": 107570
+    },
+    {
+      "epoch": 694.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9945268630981445,
+      "eval_runtime": 0.8898,
+      "eval_samples_per_second": 225.9,
+      "eval_steps_per_second": 7.867,
+      "step": 107570
+    },
+    {
+      "epoch": 694.06,
+      "learning_rate": 1.6996415770609318e-05,
+      "loss": 0.1963,
+      "step": 107580
+    },
+    {
+      "epoch": 694.13,
+      "learning_rate": 1.699283154121864e-05,
+      "loss": 0.2175,
+      "step": 107590
+    },
+    {
+      "epoch": 694.19,
+      "learning_rate": 1.6989247311827958e-05,
+      "loss": 0.188,
+      "step": 107600
+    },
+    {
+      "epoch": 694.26,
+      "learning_rate": 1.698566308243728e-05,
+      "loss": 0.2558,
+      "step": 107610
+    },
+    {
+      "epoch": 694.32,
+      "learning_rate": 1.6982078853046594e-05,
+      "loss": 0.2176,
+      "step": 107620
+    },
+    {
+      "epoch": 694.39,
+      "learning_rate": 1.6978494623655916e-05,
+      "loss": 0.2109,
+      "step": 107630
+    },
+    {
+      "epoch": 694.45,
+      "learning_rate": 1.6974910394265234e-05,
+      "loss": 0.2147,
+      "step": 107640
+    },
+    {
+      "epoch": 694.51,
+      "learning_rate": 1.6971326164874552e-05,
+      "loss": 0.2375,
+      "step": 107650
+    },
+    {
+      "epoch": 694.58,
+      "learning_rate": 1.696774193548387e-05,
+      "loss": 0.1995,
+      "step": 107660
+    },
+    {
+      "epoch": 694.64,
+      "learning_rate": 1.696415770609319e-05,
+      "loss": 0.227,
+      "step": 107670
+    },
+    {
+      "epoch": 694.71,
+      "learning_rate": 1.696057347670251e-05,
+      "loss": 0.2302,
+      "step": 107680
+    },
+    {
+      "epoch": 694.77,
+      "learning_rate": 1.695698924731183e-05,
+      "loss": 0.2335,
+      "step": 107690
+    },
+    {
+      "epoch": 694.84,
+      "learning_rate": 1.6953405017921147e-05,
+      "loss": 0.2267,
+      "step": 107700
+    },
+    {
+      "epoch": 694.9,
+      "learning_rate": 1.6949820788530465e-05,
+      "loss": 0.2401,
+      "step": 107710
+    },
+    {
+      "epoch": 694.96,
+      "learning_rate": 1.6946236559139786e-05,
+      "loss": 0.1938,
+      "step": 107720
+    },
+    {
+      "epoch": 695.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8973069787025452,
+      "eval_runtime": 0.8683,
+      "eval_samples_per_second": 231.493,
+      "eval_steps_per_second": 8.062,
+      "step": 107725
+    },
+    {
+      "epoch": 695.03,
+      "learning_rate": 1.6942652329749105e-05,
+      "loss": 0.2397,
+      "step": 107730
+    },
+    {
+      "epoch": 695.1,
+      "learning_rate": 1.6939068100358423e-05,
+      "loss": 0.2309,
+      "step": 107740
+    },
+    {
+      "epoch": 695.16,
+      "learning_rate": 1.693548387096774e-05,
+      "loss": 0.2296,
+      "step": 107750
+    },
+    {
+      "epoch": 695.23,
+      "learning_rate": 1.6931899641577063e-05,
+      "loss": 0.2484,
+      "step": 107760
+    },
+    {
+      "epoch": 695.29,
+      "learning_rate": 1.692831541218638e-05,
+      "loss": 0.2623,
+      "step": 107770
+    },
+    {
+      "epoch": 695.35,
+      "learning_rate": 1.69247311827957e-05,
+      "loss": 0.2167,
+      "step": 107780
+    },
+    {
+      "epoch": 695.42,
+      "learning_rate": 1.6921146953405017e-05,
+      "loss": 0.2029,
+      "step": 107790
+    },
+    {
+      "epoch": 695.48,
+      "learning_rate": 1.691756272401434e-05,
+      "loss": 0.1749,
+      "step": 107800
+    },
+    {
+      "epoch": 695.55,
+      "learning_rate": 1.6913978494623657e-05,
+      "loss": 0.2548,
+      "step": 107810
+    },
+    {
+      "epoch": 695.61,
+      "learning_rate": 1.6910394265232975e-05,
+      "loss": 0.2188,
+      "step": 107820
+    },
+    {
+      "epoch": 695.68,
+      "learning_rate": 1.6906810035842294e-05,
+      "loss": 0.2961,
+      "step": 107830
+    },
+    {
+      "epoch": 695.74,
+      "learning_rate": 1.6903225806451615e-05,
+      "loss": 0.2385,
+      "step": 107840
+    },
+    {
+      "epoch": 695.8,
+      "learning_rate": 1.6899641577060933e-05,
+      "loss": 0.1858,
+      "step": 107850
+    },
+    {
+      "epoch": 695.87,
+      "learning_rate": 1.689605734767025e-05,
+      "loss": 0.188,
+      "step": 107860
+    },
+    {
+      "epoch": 695.93,
+      "learning_rate": 1.689247311827957e-05,
+      "loss": 0.2436,
+      "step": 107870
+    },
+    {
+      "epoch": 696.0,
+      "learning_rate": 1.688888888888889e-05,
+      "loss": 0.2216,
+      "step": 107880
+    },
+    {
+      "epoch": 696.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9257676005363464,
+      "eval_runtime": 0.8958,
+      "eval_samples_per_second": 224.381,
+      "eval_steps_per_second": 7.814,
+      "step": 107880
+    },
+    {
+      "epoch": 696.06,
+      "learning_rate": 1.6885304659498206e-05,
+      "loss": 0.2636,
+      "step": 107890
+    },
+    {
+      "epoch": 696.13,
+      "learning_rate": 1.6881720430107528e-05,
+      "loss": 0.2623,
+      "step": 107900
+    },
+    {
+      "epoch": 696.19,
+      "learning_rate": 1.6878136200716846e-05,
+      "loss": 0.1899,
+      "step": 107910
+    },
+    {
+      "epoch": 696.26,
+      "learning_rate": 1.6874551971326168e-05,
+      "loss": 0.2444,
+      "step": 107920
+    },
+    {
+      "epoch": 696.32,
+      "learning_rate": 1.6870967741935482e-05,
+      "loss": 0.25,
+      "step": 107930
+    },
+    {
+      "epoch": 696.39,
+      "learning_rate": 1.6867383512544804e-05,
+      "loss": 0.2246,
+      "step": 107940
+    },
+    {
+      "epoch": 696.45,
+      "learning_rate": 1.6863799283154122e-05,
+      "loss": 0.2043,
+      "step": 107950
+    },
+    {
+      "epoch": 696.51,
+      "learning_rate": 1.6860215053763444e-05,
+      "loss": 0.2169,
+      "step": 107960
+    },
+    {
+      "epoch": 696.58,
+      "learning_rate": 1.685663082437276e-05,
+      "loss": 0.2177,
+      "step": 107970
+    },
+    {
+      "epoch": 696.64,
+      "learning_rate": 1.685304659498208e-05,
+      "loss": 0.2154,
+      "step": 107980
+    },
+    {
+      "epoch": 696.71,
+      "learning_rate": 1.68494623655914e-05,
+      "loss": 0.26,
+      "step": 107990
+    },
+    {
+      "epoch": 696.77,
+      "learning_rate": 1.684587813620072e-05,
+      "loss": 0.2419,
+      "step": 108000
+    },
+    {
+      "epoch": 696.84,
+      "learning_rate": 1.6842293906810035e-05,
+      "loss": 0.2376,
+      "step": 108010
+    },
+    {
+      "epoch": 696.9,
+      "learning_rate": 1.6838709677419356e-05,
+      "loss": 0.2472,
+      "step": 108020
+    },
+    {
+      "epoch": 696.96,
+      "learning_rate": 1.6835125448028675e-05,
+      "loss": 0.1646,
+      "step": 108030
+    },
+    {
+      "epoch": 697.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.9232531785964966,
+      "eval_runtime": 0.8676,
+      "eval_samples_per_second": 231.681,
+      "eval_steps_per_second": 8.068,
+      "step": 108035
+    },
+    {
+      "epoch": 697.03,
+      "learning_rate": 1.6831541218637996e-05,
+      "loss": 0.238,
+      "step": 108040
+    },
+    {
+      "epoch": 697.1,
+      "learning_rate": 1.682795698924731e-05,
+      "loss": 0.2312,
+      "step": 108050
+    },
+    {
+      "epoch": 697.16,
+      "learning_rate": 1.6824372759856633e-05,
+      "loss": 0.2056,
+      "step": 108060
+    },
+    {
+      "epoch": 697.23,
+      "learning_rate": 1.682078853046595e-05,
+      "loss": 0.2435,
+      "step": 108070
+    },
+    {
+      "epoch": 697.29,
+      "learning_rate": 1.681720430107527e-05,
+      "loss": 0.2587,
+      "step": 108080
+    },
+    {
+      "epoch": 697.35,
+      "learning_rate": 1.6813620071684587e-05,
+      "loss": 0.2252,
+      "step": 108090
+    },
+    {
+      "epoch": 697.42,
+      "learning_rate": 1.681003584229391e-05,
+      "loss": 0.2097,
+      "step": 108100
+    },
+    {
+      "epoch": 697.48,
+      "learning_rate": 1.6806451612903227e-05,
+      "loss": 0.2742,
+      "step": 108110
+    },
+    {
+      "epoch": 697.55,
+      "learning_rate": 1.6802867383512545e-05,
+      "loss": 0.2582,
+      "step": 108120
+    },
+    {
+      "epoch": 697.61,
+      "learning_rate": 1.6799283154121863e-05,
+      "loss": 0.2414,
+      "step": 108130
+    },
+    {
+      "epoch": 697.68,
+      "learning_rate": 1.6795698924731185e-05,
+      "loss": 0.2292,
+      "step": 108140
+    },
+    {
+      "epoch": 697.74,
+      "learning_rate": 1.6792114695340503e-05,
+      "loss": 0.2713,
+      "step": 108150
+    },
+    {
+      "epoch": 697.8,
+      "learning_rate": 1.678853046594982e-05,
+      "loss": 0.1824,
+      "step": 108160
+    },
+    {
+      "epoch": 697.87,
+      "learning_rate": 1.678494623655914e-05,
+      "loss": 0.2249,
+      "step": 108170
+    },
+    {
+      "epoch": 697.93,
+      "learning_rate": 1.678136200716846e-05,
+      "loss": 0.2264,
+      "step": 108180
+    },
+    {
+      "epoch": 698.0,
+      "learning_rate": 1.677777777777778e-05,
+      "loss": 0.2226,
+      "step": 108190
+    },
+    {
+      "epoch": 698.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9463791251182556,
+      "eval_runtime": 0.8764,
+      "eval_samples_per_second": 229.343,
+      "eval_steps_per_second": 7.987,
+      "step": 108190
+    },
+    {
+      "epoch": 698.06,
+      "learning_rate": 1.6774193548387098e-05,
+      "loss": 0.2222,
+      "step": 108200
+    },
+    {
+      "epoch": 698.13,
+      "learning_rate": 1.6770609318996416e-05,
+      "loss": 0.2305,
+      "step": 108210
+    },
+    {
+      "epoch": 698.19,
+      "learning_rate": 1.6767025089605737e-05,
+      "loss": 0.2162,
+      "step": 108220
+    },
+    {
+      "epoch": 698.26,
+      "learning_rate": 1.6763440860215056e-05,
+      "loss": 0.1841,
+      "step": 108230
+    },
+    {
+      "epoch": 698.32,
+      "learning_rate": 1.6759856630824374e-05,
+      "loss": 0.2113,
+      "step": 108240
+    },
+    {
+      "epoch": 698.39,
+      "learning_rate": 1.6756272401433692e-05,
+      "loss": 0.261,
+      "step": 108250
+    },
+    {
+      "epoch": 698.45,
+      "learning_rate": 1.6752688172043014e-05,
+      "loss": 0.1951,
+      "step": 108260
+    },
+    {
+      "epoch": 698.51,
+      "learning_rate": 1.6749103942652332e-05,
+      "loss": 0.2197,
+      "step": 108270
+    },
+    {
+      "epoch": 698.58,
+      "learning_rate": 1.674551971326165e-05,
+      "loss": 0.1685,
+      "step": 108280
+    },
+    {
+      "epoch": 698.64,
+      "learning_rate": 1.6741935483870968e-05,
+      "loss": 0.2096,
+      "step": 108290
+    },
+    {
+      "epoch": 698.71,
+      "learning_rate": 1.6738351254480286e-05,
+      "loss": 0.2212,
+      "step": 108300
+    },
+    {
+      "epoch": 698.77,
+      "learning_rate": 1.6734767025089608e-05,
+      "loss": 0.2341,
+      "step": 108310
+    },
+    {
+      "epoch": 698.84,
+      "learning_rate": 1.6731182795698923e-05,
+      "loss": 0.2332,
+      "step": 108320
+    },
+    {
+      "epoch": 698.9,
+      "learning_rate": 1.6727598566308244e-05,
+      "loss": 0.2456,
+      "step": 108330
+    },
+    {
+      "epoch": 698.96,
+      "learning_rate": 1.6724014336917563e-05,
+      "loss": 0.258,
+      "step": 108340
+    },
+    {
+      "epoch": 699.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9094772934913635,
+      "eval_runtime": 0.892,
+      "eval_samples_per_second": 225.332,
+      "eval_steps_per_second": 7.847,
+      "step": 108345
+    },
+    {
+      "epoch": 699.03,
+      "learning_rate": 1.6720430107526884e-05,
+      "loss": 0.2489,
+      "step": 108350
+    },
+    {
+      "epoch": 699.1,
+      "learning_rate": 1.67168458781362e-05,
+      "loss": 0.2236,
+      "step": 108360
+    },
+    {
+      "epoch": 699.16,
+      "learning_rate": 1.671326164874552e-05,
+      "loss": 0.2302,
+      "step": 108370
+    },
+    {
+      "epoch": 699.23,
+      "learning_rate": 1.670967741935484e-05,
+      "loss": 0.2024,
+      "step": 108380
+    },
+    {
+      "epoch": 699.29,
+      "learning_rate": 1.6706093189964157e-05,
+      "loss": 0.228,
+      "step": 108390
+    },
+    {
+      "epoch": 699.35,
+      "learning_rate": 1.6702508960573475e-05,
+      "loss": 0.2674,
+      "step": 108400
+    },
+    {
+      "epoch": 699.42,
+      "learning_rate": 1.6698924731182797e-05,
+      "loss": 0.2328,
+      "step": 108410
+    },
+    {
+      "epoch": 699.48,
+      "learning_rate": 1.6695340501792115e-05,
+      "loss": 0.1956,
+      "step": 108420
+    },
+    {
+      "epoch": 699.55,
+      "learning_rate": 1.6691756272401433e-05,
+      "loss": 0.2052,
+      "step": 108430
+    },
+    {
+      "epoch": 699.61,
+      "learning_rate": 1.668817204301075e-05,
+      "loss": 0.2017,
+      "step": 108440
+    },
+    {
+      "epoch": 699.68,
+      "learning_rate": 1.6684587813620073e-05,
+      "loss": 0.1818,
+      "step": 108450
+    },
+    {
+      "epoch": 699.74,
+      "learning_rate": 1.668100358422939e-05,
+      "loss": 0.2623,
+      "step": 108460
+    },
+    {
+      "epoch": 699.8,
+      "learning_rate": 1.667741935483871e-05,
+      "loss": 0.1622,
+      "step": 108470
+    },
+    {
+      "epoch": 699.87,
+      "learning_rate": 1.6673835125448028e-05,
+      "loss": 0.2403,
+      "step": 108480
+    },
+    {
+      "epoch": 699.93,
+      "learning_rate": 1.667025089605735e-05,
+      "loss": 0.2552,
+      "step": 108490
+    },
+    {
+      "epoch": 700.0,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.2139,
+      "step": 108500
+    },
+    {
+      "epoch": 700.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9180562496185303,
+      "eval_runtime": 0.8761,
+      "eval_samples_per_second": 229.421,
+      "eval_steps_per_second": 7.99,
+      "step": 108500
+    },
+    {
+      "epoch": 700.06,
+      "learning_rate": 1.6663082437275986e-05,
+      "loss": 0.2355,
+      "step": 108510
+    },
+    {
+      "epoch": 700.13,
+      "learning_rate": 1.6659498207885304e-05,
+      "loss": 0.2045,
+      "step": 108520
+    },
+    {
+      "epoch": 700.19,
+      "learning_rate": 1.6655913978494625e-05,
+      "loss": 0.2057,
+      "step": 108530
+    },
+    {
+      "epoch": 700.26,
+      "learning_rate": 1.6652329749103944e-05,
+      "loss": 0.1799,
+      "step": 108540
+    },
+    {
+      "epoch": 700.32,
+      "learning_rate": 1.6648745519713262e-05,
+      "loss": 0.2074,
+      "step": 108550
+    },
+    {
+      "epoch": 700.39,
+      "learning_rate": 1.664516129032258e-05,
+      "loss": 0.2139,
+      "step": 108560
+    },
+    {
+      "epoch": 700.45,
+      "learning_rate": 1.66415770609319e-05,
+      "loss": 0.23,
+      "step": 108570
+    },
+    {
+      "epoch": 700.51,
+      "learning_rate": 1.663799283154122e-05,
+      "loss": 0.2327,
+      "step": 108580
+    },
+    {
+      "epoch": 700.58,
+      "learning_rate": 1.6634408602150538e-05,
+      "loss": 0.2361,
+      "step": 108590
+    },
+    {
+      "epoch": 700.64,
+      "learning_rate": 1.6630824372759856e-05,
+      "loss": 0.2339,
+      "step": 108600
+    },
+    {
+      "epoch": 700.71,
+      "learning_rate": 1.6627240143369178e-05,
+      "loss": 0.2226,
+      "step": 108610
+    },
+    {
+      "epoch": 700.77,
+      "learning_rate": 1.6623655913978496e-05,
+      "loss": 0.2122,
+      "step": 108620
+    },
+    {
+      "epoch": 700.84,
+      "learning_rate": 1.6620071684587814e-05,
+      "loss": 0.216,
+      "step": 108630
+    },
+    {
+      "epoch": 700.9,
+      "learning_rate": 1.6616487455197132e-05,
+      "loss": 0.2237,
+      "step": 108640
+    },
+    {
+      "epoch": 700.96,
+      "learning_rate": 1.6612903225806454e-05,
+      "loss": 0.2261,
+      "step": 108650
+    },
+    {
+      "epoch": 701.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9917797446250916,
+      "eval_runtime": 0.8658,
+      "eval_samples_per_second": 232.159,
+      "eval_steps_per_second": 8.085,
+      "step": 108655
+    },
+    {
+      "epoch": 701.03,
+      "learning_rate": 1.6609318996415772e-05,
+      "loss": 0.2251,
+      "step": 108660
+    },
+    {
+      "epoch": 701.1,
+      "learning_rate": 1.660573476702509e-05,
+      "loss": 0.2143,
+      "step": 108670
+    },
+    {
+      "epoch": 701.16,
+      "learning_rate": 1.660215053763441e-05,
+      "loss": 0.2471,
+      "step": 108680
+    },
+    {
+      "epoch": 701.23,
+      "learning_rate": 1.659856630824373e-05,
+      "loss": 0.2704,
+      "step": 108690
+    },
+    {
+      "epoch": 701.29,
+      "learning_rate": 1.659498207885305e-05,
+      "loss": 0.2315,
+      "step": 108700
+    },
+    {
+      "epoch": 701.35,
+      "learning_rate": 1.6591397849462367e-05,
+      "loss": 0.2265,
+      "step": 108710
+    },
+    {
+      "epoch": 701.42,
+      "learning_rate": 1.6587813620071685e-05,
+      "loss": 0.2636,
+      "step": 108720
+    },
+    {
+      "epoch": 701.48,
+      "learning_rate": 1.6584229390681007e-05,
+      "loss": 0.2595,
+      "step": 108730
+    },
+    {
+      "epoch": 701.55,
+      "learning_rate": 1.658064516129032e-05,
+      "loss": 0.2235,
+      "step": 108740
+    },
+    {
+      "epoch": 701.61,
+      "learning_rate": 1.6577060931899643e-05,
+      "loss": 0.2539,
+      "step": 108750
+    },
+    {
+      "epoch": 701.68,
+      "learning_rate": 1.657347670250896e-05,
+      "loss": 0.2393,
+      "step": 108760
+    },
+    {
+      "epoch": 701.74,
+      "learning_rate": 1.6569892473118283e-05,
+      "loss": 0.2039,
+      "step": 108770
+    },
+    {
+      "epoch": 701.8,
+      "learning_rate": 1.6566308243727598e-05,
+      "loss": 0.2643,
+      "step": 108780
+    },
+    {
+      "epoch": 701.87,
+      "learning_rate": 1.656272401433692e-05,
+      "loss": 0.2357,
+      "step": 108790
+    },
+    {
+      "epoch": 701.93,
+      "learning_rate": 1.6559139784946237e-05,
+      "loss": 0.1925,
+      "step": 108800
+    },
+    {
+      "epoch": 702.0,
+      "learning_rate": 1.655555555555556e-05,
+      "loss": 0.2598,
+      "step": 108810
+    },
+    {
+      "epoch": 702.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9659743905067444,
+      "eval_runtime": 0.8763,
+      "eval_samples_per_second": 229.362,
+      "eval_steps_per_second": 7.988,
+      "step": 108810
+    },
+    {
+      "epoch": 702.06,
+      "learning_rate": 1.6551971326164874e-05,
+      "loss": 0.2585,
+      "step": 108820
+    },
+    {
+      "epoch": 702.13,
+      "learning_rate": 1.6548387096774195e-05,
+      "loss": 0.2233,
+      "step": 108830
+    },
+    {
+      "epoch": 702.19,
+      "learning_rate": 1.6544802867383514e-05,
+      "loss": 0.193,
+      "step": 108840
+    },
+    {
+      "epoch": 702.26,
+      "learning_rate": 1.6541218637992835e-05,
+      "loss": 0.2038,
+      "step": 108850
+    },
+    {
+      "epoch": 702.32,
+      "learning_rate": 1.653763440860215e-05,
+      "loss": 0.2058,
+      "step": 108860
+    },
+    {
+      "epoch": 702.39,
+      "learning_rate": 1.653405017921147e-05,
+      "loss": 0.1994,
+      "step": 108870
+    },
+    {
+      "epoch": 702.45,
+      "learning_rate": 1.653046594982079e-05,
+      "loss": 0.2245,
+      "step": 108880
+    },
+    {
+      "epoch": 702.51,
+      "learning_rate": 1.6526881720430108e-05,
+      "loss": 0.2085,
+      "step": 108890
+    },
+    {
+      "epoch": 702.58,
+      "learning_rate": 1.6523297491039426e-05,
+      "loss": 0.2739,
+      "step": 108900
+    },
+    {
+      "epoch": 702.64,
+      "learning_rate": 1.6519713261648748e-05,
+      "loss": 0.2238,
+      "step": 108910
+    },
+    {
+      "epoch": 702.71,
+      "learning_rate": 1.6516129032258066e-05,
+      "loss": 0.2301,
+      "step": 108920
+    },
+    {
+      "epoch": 702.77,
+      "learning_rate": 1.6512544802867384e-05,
+      "loss": 0.2101,
+      "step": 108930
+    },
+    {
+      "epoch": 702.84,
+      "learning_rate": 1.6508960573476702e-05,
+      "loss": 0.2728,
+      "step": 108940
+    },
+    {
+      "epoch": 702.9,
+      "learning_rate": 1.650537634408602e-05,
+      "loss": 0.2414,
+      "step": 108950
+    },
+    {
+      "epoch": 702.96,
+      "learning_rate": 1.6501792114695342e-05,
+      "loss": 0.2152,
+      "step": 108960
+    },
+    {
+      "epoch": 703.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9704497456550598,
+      "eval_runtime": 0.8748,
+      "eval_samples_per_second": 229.762,
+      "eval_steps_per_second": 8.002,
+      "step": 108965
+    },
+    {
+      "epoch": 703.03,
+      "learning_rate": 1.649820788530466e-05,
+      "loss": 0.2031,
+      "step": 108970
+    },
+    {
+      "epoch": 703.1,
+      "learning_rate": 1.649462365591398e-05,
+      "loss": 0.1622,
+      "step": 108980
+    },
+    {
+      "epoch": 703.16,
+      "learning_rate": 1.6491039426523297e-05,
+      "loss": 0.2281,
+      "step": 108990
+    },
+    {
+      "epoch": 703.23,
+      "learning_rate": 1.648745519713262e-05,
+      "loss": 0.1883,
+      "step": 109000
+    },
+    {
+      "epoch": 703.29,
+      "learning_rate": 1.6483870967741937e-05,
+      "loss": 0.2615,
+      "step": 109010
+    },
+    {
+      "epoch": 703.35,
+      "learning_rate": 1.6480286738351255e-05,
+      "loss": 0.2393,
+      "step": 109020
+    },
+    {
+      "epoch": 703.42,
+      "learning_rate": 1.6476702508960573e-05,
+      "loss": 0.2328,
+      "step": 109030
+    },
+    {
+      "epoch": 703.48,
+      "learning_rate": 1.6473118279569895e-05,
+      "loss": 0.2212,
+      "step": 109040
+    },
+    {
+      "epoch": 703.55,
+      "learning_rate": 1.6469534050179213e-05,
+      "loss": 0.2055,
+      "step": 109050
+    },
+    {
+      "epoch": 703.61,
+      "learning_rate": 1.646594982078853e-05,
+      "loss": 0.2424,
+      "step": 109060
+    },
+    {
+      "epoch": 703.68,
+      "learning_rate": 1.646236559139785e-05,
+      "loss": 0.2307,
+      "step": 109070
+    },
+    {
+      "epoch": 703.74,
+      "learning_rate": 1.645878136200717e-05,
+      "loss": 0.1796,
+      "step": 109080
+    },
+    {
+      "epoch": 703.8,
+      "learning_rate": 1.6455197132616486e-05,
+      "loss": 0.2526,
+      "step": 109090
+    },
+    {
+      "epoch": 703.87,
+      "learning_rate": 1.6451612903225807e-05,
+      "loss": 0.2647,
+      "step": 109100
+    },
+    {
+      "epoch": 703.93,
+      "learning_rate": 1.6448028673835125e-05,
+      "loss": 0.2385,
+      "step": 109110
+    },
+    {
+      "epoch": 704.0,
+      "learning_rate": 1.6444444444444447e-05,
+      "loss": 0.2016,
+      "step": 109120
+    },
+    {
+      "epoch": 704.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9852574467658997,
+      "eval_runtime": 0.9035,
+      "eval_samples_per_second": 222.461,
+      "eval_steps_per_second": 7.747,
+      "step": 109120
+    },
+    {
+      "epoch": 704.06,
+      "learning_rate": 1.6440860215053762e-05,
+      "loss": 0.2347,
+      "step": 109130
+    },
+    {
+      "epoch": 704.13,
+      "learning_rate": 1.6437275985663083e-05,
+      "loss": 0.2199,
+      "step": 109140
+    },
+    {
+      "epoch": 704.19,
+      "learning_rate": 1.64336917562724e-05,
+      "loss": 0.2198,
+      "step": 109150
+    },
+    {
+      "epoch": 704.26,
+      "learning_rate": 1.6430107526881723e-05,
+      "loss": 0.2671,
+      "step": 109160
+    },
+    {
+      "epoch": 704.32,
+      "learning_rate": 1.6426523297491038e-05,
+      "loss": 0.2219,
+      "step": 109170
+    },
+    {
+      "epoch": 704.39,
+      "learning_rate": 1.642293906810036e-05,
+      "loss": 0.2573,
+      "step": 109180
+    },
+    {
+      "epoch": 704.45,
+      "learning_rate": 1.6419354838709678e-05,
+      "loss": 0.1901,
+      "step": 109190
+    },
+    {
+      "epoch": 704.51,
+      "learning_rate": 1.6415770609319e-05,
+      "loss": 0.282,
+      "step": 109200
+    },
+    {
+      "epoch": 704.58,
+      "learning_rate": 1.6412186379928314e-05,
+      "loss": 0.2662,
+      "step": 109210
+    },
+    {
+      "epoch": 704.64,
+      "learning_rate": 1.6408602150537636e-05,
+      "loss": 0.1978,
+      "step": 109220
+    },
+    {
+      "epoch": 704.71,
+      "learning_rate": 1.6405017921146954e-05,
+      "loss": 0.2425,
+      "step": 109230
+    },
+    {
+      "epoch": 704.77,
+      "learning_rate": 1.6401433691756272e-05,
+      "loss": 0.1838,
+      "step": 109240
+    },
+    {
+      "epoch": 704.84,
+      "learning_rate": 1.639784946236559e-05,
+      "loss": 0.2195,
+      "step": 109250
+    },
+    {
+      "epoch": 704.9,
+      "learning_rate": 1.6394265232974912e-05,
+      "loss": 0.2577,
+      "step": 109260
+    },
+    {
+      "epoch": 704.96,
+      "learning_rate": 1.639068100358423e-05,
+      "loss": 0.2534,
+      "step": 109270
+    },
+    {
+      "epoch": 705.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9954023361206055,
+      "eval_runtime": 0.862,
+      "eval_samples_per_second": 233.169,
+      "eval_steps_per_second": 8.12,
+      "step": 109275
+    },
+    {
+      "epoch": 705.03,
+      "learning_rate": 1.638709677419355e-05,
+      "loss": 0.2191,
+      "step": 109280
+    },
+    {
+      "epoch": 705.1,
+      "learning_rate": 1.6383512544802867e-05,
+      "loss": 0.2674,
+      "step": 109290
+    },
+    {
+      "epoch": 705.16,
+      "learning_rate": 1.6379928315412188e-05,
+      "loss": 0.2094,
+      "step": 109300
+    },
+    {
+      "epoch": 705.23,
+      "learning_rate": 1.6376344086021506e-05,
+      "loss": 0.2512,
+      "step": 109310
+    },
+    {
+      "epoch": 705.29,
+      "learning_rate": 1.6372759856630825e-05,
+      "loss": 0.1926,
+      "step": 109320
+    },
+    {
+      "epoch": 705.35,
+      "learning_rate": 1.6369175627240143e-05,
+      "loss": 0.2225,
+      "step": 109330
+    },
+    {
+      "epoch": 705.42,
+      "learning_rate": 1.6365591397849464e-05,
+      "loss": 0.1793,
+      "step": 109340
+    },
+    {
+      "epoch": 705.48,
+      "learning_rate": 1.6362007168458783e-05,
+      "loss": 0.2457,
+      "step": 109350
+    },
+    {
+      "epoch": 705.55,
+      "learning_rate": 1.63584229390681e-05,
+      "loss": 0.2372,
+      "step": 109360
+    },
+    {
+      "epoch": 705.61,
+      "learning_rate": 1.635483870967742e-05,
+      "loss": 0.2035,
+      "step": 109370
+    },
+    {
+      "epoch": 705.68,
+      "learning_rate": 1.635125448028674e-05,
+      "loss": 0.1695,
+      "step": 109380
+    },
+    {
+      "epoch": 705.74,
+      "learning_rate": 1.634767025089606e-05,
+      "loss": 0.2498,
+      "step": 109390
+    },
+    {
+      "epoch": 705.8,
+      "learning_rate": 1.6344086021505377e-05,
+      "loss": 0.2081,
+      "step": 109400
+    },
+    {
+      "epoch": 705.87,
+      "learning_rate": 1.6340501792114695e-05,
+      "loss": 0.2077,
+      "step": 109410
+    },
+    {
+      "epoch": 705.93,
+      "learning_rate": 1.6336917562724017e-05,
+      "loss": 0.2071,
+      "step": 109420
+    },
+    {
+      "epoch": 706.0,
+      "learning_rate": 1.6333333333333335e-05,
+      "loss": 0.2074,
+      "step": 109430
+    },
+    {
+      "epoch": 706.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9932730793952942,
+      "eval_runtime": 0.8667,
+      "eval_samples_per_second": 231.918,
+      "eval_steps_per_second": 8.077,
+      "step": 109430
+    },
+    {
+      "epoch": 706.06,
+      "learning_rate": 1.6329749103942653e-05,
+      "loss": 0.2595,
+      "step": 109440
+    },
+    {
+      "epoch": 706.13,
+      "learning_rate": 1.632616487455197e-05,
+      "loss": 0.2601,
+      "step": 109450
+    },
+    {
+      "epoch": 706.19,
+      "learning_rate": 1.6322580645161293e-05,
+      "loss": 0.2516,
+      "step": 109460
+    },
+    {
+      "epoch": 706.26,
+      "learning_rate": 1.631899641577061e-05,
+      "loss": 0.2434,
+      "step": 109470
+    },
+    {
+      "epoch": 706.32,
+      "learning_rate": 1.631541218637993e-05,
+      "loss": 0.2367,
+      "step": 109480
+    },
+    {
+      "epoch": 706.39,
+      "learning_rate": 1.6311827956989248e-05,
+      "loss": 0.2302,
+      "step": 109490
+    },
+    {
+      "epoch": 706.45,
+      "learning_rate": 1.630824372759857e-05,
+      "loss": 0.2027,
+      "step": 109500
+    },
+    {
+      "epoch": 706.51,
+      "learning_rate": 1.6304659498207887e-05,
+      "loss": 0.1952,
+      "step": 109510
+    },
+    {
+      "epoch": 706.58,
+      "learning_rate": 1.6301075268817206e-05,
+      "loss": 0.2261,
+      "step": 109520
+    },
+    {
+      "epoch": 706.64,
+      "learning_rate": 1.6297491039426524e-05,
+      "loss": 0.2272,
+      "step": 109530
+    },
+    {
+      "epoch": 706.71,
+      "learning_rate": 1.6293906810035845e-05,
+      "loss": 0.2382,
+      "step": 109540
+    },
+    {
+      "epoch": 706.77,
+      "learning_rate": 1.6290322580645164e-05,
+      "loss": 0.2099,
+      "step": 109550
+    },
+    {
+      "epoch": 706.84,
+      "learning_rate": 1.6286738351254482e-05,
+      "loss": 0.2383,
+      "step": 109560
+    },
+    {
+      "epoch": 706.9,
+      "learning_rate": 1.62831541218638e-05,
+      "loss": 0.2275,
+      "step": 109570
+    },
+    {
+      "epoch": 706.96,
+      "learning_rate": 1.6279569892473118e-05,
+      "loss": 0.2698,
+      "step": 109580
+    },
+    {
+      "epoch": 707.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9789064526557922,
+      "eval_runtime": 0.8632,
+      "eval_samples_per_second": 232.841,
+      "eval_steps_per_second": 8.109,
+      "step": 109585
+    },
+    {
+      "epoch": 707.03,
+      "learning_rate": 1.6275985663082437e-05,
+      "loss": 0.2062,
+      "step": 109590
+    },
+    {
+      "epoch": 707.1,
+      "learning_rate": 1.6272401433691755e-05,
+      "loss": 0.2265,
+      "step": 109600
+    },
+    {
+      "epoch": 707.16,
+      "learning_rate": 1.6268817204301076e-05,
+      "loss": 0.2002,
+      "step": 109610
+    },
+    {
+      "epoch": 707.23,
+      "learning_rate": 1.6265232974910395e-05,
+      "loss": 0.2313,
+      "step": 109620
+    },
+    {
+      "epoch": 707.29,
+      "learning_rate": 1.6261648745519713e-05,
+      "loss": 0.1902,
+      "step": 109630
+    },
+    {
+      "epoch": 707.35,
+      "learning_rate": 1.625806451612903e-05,
+      "loss": 0.2076,
+      "step": 109640
+    },
+    {
+      "epoch": 707.42,
+      "learning_rate": 1.6254480286738353e-05,
+      "loss": 0.2367,
+      "step": 109650
+    },
+    {
+      "epoch": 707.48,
+      "learning_rate": 1.625089605734767e-05,
+      "loss": 0.2226,
+      "step": 109660
+    },
+    {
+      "epoch": 707.55,
+      "learning_rate": 1.624731182795699e-05,
+      "loss": 0.2358,
+      "step": 109670
+    },
+    {
+      "epoch": 707.61,
+      "learning_rate": 1.6243727598566307e-05,
+      "loss": 0.2245,
+      "step": 109680
+    },
+    {
+      "epoch": 707.68,
+      "learning_rate": 1.624014336917563e-05,
+      "loss": 0.2585,
+      "step": 109690
+    },
+    {
+      "epoch": 707.74,
+      "learning_rate": 1.6236559139784947e-05,
+      "loss": 0.1778,
+      "step": 109700
+    },
+    {
+      "epoch": 707.8,
+      "learning_rate": 1.6232974910394265e-05,
+      "loss": 0.1997,
+      "step": 109710
+    },
+    {
+      "epoch": 707.87,
+      "learning_rate": 1.6229390681003583e-05,
+      "loss": 0.2486,
+      "step": 109720
+    },
+    {
+      "epoch": 707.93,
+      "learning_rate": 1.6225806451612905e-05,
+      "loss": 0.2036,
+      "step": 109730
+    },
+    {
+      "epoch": 708.0,
+      "learning_rate": 1.6222222222222223e-05,
+      "loss": 0.2327,
+      "step": 109740
+    },
+    {
+      "epoch": 708.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9705785512924194,
+      "eval_runtime": 0.8532,
+      "eval_samples_per_second": 235.588,
+      "eval_steps_per_second": 8.205,
+      "step": 109740
+    },
+    {
+      "epoch": 708.06,
+      "learning_rate": 1.621863799283154e-05,
+      "loss": 0.2425,
+      "step": 109750
+    },
+    {
+      "epoch": 708.13,
+      "learning_rate": 1.621505376344086e-05,
+      "loss": 0.2236,
+      "step": 109760
+    },
+    {
+      "epoch": 708.19,
+      "learning_rate": 1.621146953405018e-05,
+      "loss": 0.2192,
+      "step": 109770
+    },
+    {
+      "epoch": 708.26,
+      "learning_rate": 1.62078853046595e-05,
+      "loss": 0.2615,
+      "step": 109780
+    },
+    {
+      "epoch": 708.32,
+      "learning_rate": 1.6204301075268818e-05,
+      "loss": 0.2422,
+      "step": 109790
+    },
+    {
+      "epoch": 708.39,
+      "learning_rate": 1.6200716845878136e-05,
+      "loss": 0.2288,
+      "step": 109800
+    },
+    {
+      "epoch": 708.45,
+      "learning_rate": 1.6197132616487457e-05,
+      "loss": 0.2299,
+      "step": 109810
+    },
+    {
+      "epoch": 708.51,
+      "learning_rate": 1.6193548387096776e-05,
+      "loss": 0.2508,
+      "step": 109820
+    },
+    {
+      "epoch": 708.58,
+      "learning_rate": 1.6189964157706094e-05,
+      "loss": 0.2203,
+      "step": 109830
+    },
+    {
+      "epoch": 708.64,
+      "learning_rate": 1.6186379928315412e-05,
+      "loss": 0.2307,
+      "step": 109840
+    },
+    {
+      "epoch": 708.71,
+      "learning_rate": 1.6182795698924734e-05,
+      "loss": 0.1843,
+      "step": 109850
+    },
+    {
+      "epoch": 708.77,
+      "learning_rate": 1.6179211469534052e-05,
+      "loss": 0.2962,
+      "step": 109860
+    },
+    {
+      "epoch": 708.84,
+      "learning_rate": 1.617562724014337e-05,
+      "loss": 0.197,
+      "step": 109870
+    },
+    {
+      "epoch": 708.9,
+      "learning_rate": 1.6172043010752688e-05,
+      "loss": 0.2335,
+      "step": 109880
+    },
+    {
+      "epoch": 708.96,
+      "learning_rate": 1.616845878136201e-05,
+      "loss": 0.2122,
+      "step": 109890
+    },
+    {
+      "epoch": 709.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9446938037872314,
+      "eval_runtime": 0.8728,
+      "eval_samples_per_second": 230.285,
+      "eval_steps_per_second": 8.02,
+      "step": 109895
+    },
+    {
+      "epoch": 709.03,
+      "learning_rate": 1.6164874551971328e-05,
+      "loss": 0.2471,
+      "step": 109900
+    },
+    {
+      "epoch": 709.1,
+      "learning_rate": 1.6161290322580646e-05,
+      "loss": 0.2035,
+      "step": 109910
+    },
+    {
+      "epoch": 709.16,
+      "learning_rate": 1.6157706093189964e-05,
+      "loss": 0.2896,
+      "step": 109920
+    },
+    {
+      "epoch": 709.23,
+      "learning_rate": 1.6154121863799286e-05,
+      "loss": 0.2534,
+      "step": 109930
+    },
+    {
+      "epoch": 709.29,
+      "learning_rate": 1.61505376344086e-05,
+      "loss": 0.2315,
+      "step": 109940
+    },
+    {
+      "epoch": 709.35,
+      "learning_rate": 1.6146953405017922e-05,
+      "loss": 0.2218,
+      "step": 109950
+    },
+    {
+      "epoch": 709.42,
+      "learning_rate": 1.614336917562724e-05,
+      "loss": 0.1879,
+      "step": 109960
+    },
+    {
+      "epoch": 709.48,
+      "learning_rate": 1.6139784946236562e-05,
+      "loss": 0.2666,
+      "step": 109970
+    },
+    {
+      "epoch": 709.55,
+      "learning_rate": 1.6136200716845877e-05,
+      "loss": 0.2676,
+      "step": 109980
+    },
+    {
+      "epoch": 709.61,
+      "learning_rate": 1.61326164874552e-05,
+      "loss": 0.2383,
+      "step": 109990
+    },
+    {
+      "epoch": 709.68,
+      "learning_rate": 1.6129032258064517e-05,
+      "loss": 0.2248,
+      "step": 110000
+    },
+    {
+      "epoch": 709.74,
+      "learning_rate": 1.612544802867384e-05,
+      "loss": 0.2381,
+      "step": 110010
+    },
+    {
+      "epoch": 709.8,
+      "learning_rate": 1.6121863799283153e-05,
+      "loss": 0.231,
+      "step": 110020
+    },
+    {
+      "epoch": 709.87,
+      "learning_rate": 1.6118279569892475e-05,
+      "loss": 0.2029,
+      "step": 110030
+    },
+    {
+      "epoch": 709.93,
+      "learning_rate": 1.6114695340501793e-05,
+      "loss": 0.2386,
+      "step": 110040
+    },
+    {
+      "epoch": 710.0,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 0.2333,
+      "step": 110050
+    },
+    {
+      "epoch": 710.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9643989205360413,
+      "eval_runtime": 0.8582,
+      "eval_samples_per_second": 234.222,
+      "eval_steps_per_second": 8.157,
+      "step": 110050
+    },
+    {
+      "epoch": 710.06,
+      "learning_rate": 1.610752688172043e-05,
+      "loss": 0.2475,
+      "step": 110060
+    },
+    {
+      "epoch": 710.13,
+      "learning_rate": 1.610394265232975e-05,
+      "loss": 0.2455,
+      "step": 110070
+    },
+    {
+      "epoch": 710.19,
+      "learning_rate": 1.610035842293907e-05,
+      "loss": 0.1722,
+      "step": 110080
+    },
+    {
+      "epoch": 710.26,
+      "learning_rate": 1.6096774193548387e-05,
+      "loss": 0.2269,
+      "step": 110090
+    },
+    {
+      "epoch": 710.32,
+      "learning_rate": 1.6093189964157706e-05,
+      "loss": 0.2233,
+      "step": 110100
+    },
+    {
+      "epoch": 710.39,
+      "learning_rate": 1.6089605734767027e-05,
+      "loss": 0.2118,
+      "step": 110110
+    },
+    {
+      "epoch": 710.45,
+      "learning_rate": 1.6086021505376345e-05,
+      "loss": 0.2318,
+      "step": 110120
+    },
+    {
+      "epoch": 710.51,
+      "learning_rate": 1.6082437275985664e-05,
+      "loss": 0.2477,
+      "step": 110130
+    },
+    {
+      "epoch": 710.58,
+      "learning_rate": 1.6078853046594982e-05,
+      "loss": 0.252,
+      "step": 110140
+    },
+    {
+      "epoch": 710.64,
+      "learning_rate": 1.6075268817204303e-05,
+      "loss": 0.2316,
+      "step": 110150
+    },
+    {
+      "epoch": 710.71,
+      "learning_rate": 1.607168458781362e-05,
+      "loss": 0.2383,
+      "step": 110160
+    },
+    {
+      "epoch": 710.77,
+      "learning_rate": 1.606810035842294e-05,
+      "loss": 0.2161,
+      "step": 110170
+    },
+    {
+      "epoch": 710.84,
+      "learning_rate": 1.6064516129032258e-05,
+      "loss": 0.2464,
+      "step": 110180
+    },
+    {
+      "epoch": 710.9,
+      "learning_rate": 1.606093189964158e-05,
+      "loss": 0.2149,
+      "step": 110190
+    },
+    {
+      "epoch": 710.96,
+      "learning_rate": 1.6057347670250898e-05,
+      "loss": 0.2136,
+      "step": 110200
+    },
+    {
+      "epoch": 711.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9580843448638916,
+      "eval_runtime": 0.878,
+      "eval_samples_per_second": 228.928,
+      "eval_steps_per_second": 7.973,
+      "step": 110205
+    },
+    {
+      "epoch": 711.03,
+      "learning_rate": 1.6053763440860216e-05,
+      "loss": 0.2698,
+      "step": 110210
+    },
+    {
+      "epoch": 711.1,
+      "learning_rate": 1.6050179211469534e-05,
+      "loss": 0.2126,
+      "step": 110220
+    },
+    {
+      "epoch": 711.16,
+      "learning_rate": 1.6046594982078852e-05,
+      "loss": 0.1739,
+      "step": 110230
+    },
+    {
+      "epoch": 711.23,
+      "learning_rate": 1.6043010752688174e-05,
+      "loss": 0.2191,
+      "step": 110240
+    },
+    {
+      "epoch": 711.29,
+      "learning_rate": 1.603942652329749e-05,
+      "loss": 0.2574,
+      "step": 110250
+    },
+    {
+      "epoch": 711.35,
+      "learning_rate": 1.603584229390681e-05,
+      "loss": 0.2301,
+      "step": 110260
+    },
+    {
+      "epoch": 711.42,
+      "learning_rate": 1.603225806451613e-05,
+      "loss": 0.2271,
+      "step": 110270
+    },
+    {
+      "epoch": 711.48,
+      "learning_rate": 1.602867383512545e-05,
+      "loss": 0.2174,
+      "step": 110280
+    },
+    {
+      "epoch": 711.55,
+      "learning_rate": 1.6025089605734765e-05,
+      "loss": 0.2206,
+      "step": 110290
+    },
+    {
+      "epoch": 711.61,
+      "learning_rate": 1.6021505376344087e-05,
+      "loss": 0.2377,
+      "step": 110300
+    },
+    {
+      "epoch": 711.68,
+      "learning_rate": 1.6017921146953405e-05,
+      "loss": 0.1852,
+      "step": 110310
+    },
+    {
+      "epoch": 711.74,
+      "learning_rate": 1.6014336917562726e-05,
+      "loss": 0.283,
+      "step": 110320
+    },
+    {
+      "epoch": 711.8,
+      "learning_rate": 1.601075268817204e-05,
+      "loss": 0.2359,
+      "step": 110330
+    },
+    {
+      "epoch": 711.87,
+      "learning_rate": 1.6007168458781363e-05,
+      "loss": 0.2303,
+      "step": 110340
+    },
+    {
+      "epoch": 711.93,
+      "learning_rate": 1.600358422939068e-05,
+      "loss": 0.2558,
+      "step": 110350
+    },
+    {
+      "epoch": 712.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.196,
+      "step": 110360
+    },
+    {
+      "epoch": 712.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9823014736175537,
+      "eval_runtime": 0.8738,
+      "eval_samples_per_second": 230.028,
+      "eval_steps_per_second": 8.011,
+      "step": 110360
+    },
+    {
+      "epoch": 712.06,
+      "learning_rate": 1.5996415770609317e-05,
+      "loss": 0.197,
+      "step": 110370
+    },
+    {
+      "epoch": 712.13,
+      "learning_rate": 1.599283154121864e-05,
+      "loss": 0.1938,
+      "step": 110380
+    },
+    {
+      "epoch": 712.19,
+      "learning_rate": 1.5989247311827957e-05,
+      "loss": 0.2077,
+      "step": 110390
+    },
+    {
+      "epoch": 712.26,
+      "learning_rate": 1.5985663082437275e-05,
+      "loss": 0.2152,
+      "step": 110400
+    },
+    {
+      "epoch": 712.32,
+      "learning_rate": 1.5982078853046594e-05,
+      "loss": 0.2608,
+      "step": 110410
+    },
+    {
+      "epoch": 712.39,
+      "learning_rate": 1.5978494623655915e-05,
+      "loss": 0.2247,
+      "step": 110420
+    },
+    {
+      "epoch": 712.45,
+      "learning_rate": 1.5974910394265233e-05,
+      "loss": 0.2444,
+      "step": 110430
+    },
+    {
+      "epoch": 712.51,
+      "learning_rate": 1.597132616487455e-05,
+      "loss": 0.2415,
+      "step": 110440
+    },
+    {
+      "epoch": 712.58,
+      "learning_rate": 1.596774193548387e-05,
+      "loss": 0.2242,
+      "step": 110450
+    },
+    {
+      "epoch": 712.64,
+      "learning_rate": 1.596415770609319e-05,
+      "loss": 0.2132,
+      "step": 110460
+    },
+    {
+      "epoch": 712.71,
+      "learning_rate": 1.596057347670251e-05,
+      "loss": 0.1826,
+      "step": 110470
+    },
+    {
+      "epoch": 712.77,
+      "learning_rate": 1.5956989247311828e-05,
+      "loss": 0.2305,
+      "step": 110480
+    },
+    {
+      "epoch": 712.84,
+      "learning_rate": 1.5953405017921146e-05,
+      "loss": 0.2463,
+      "step": 110490
+    },
+    {
+      "epoch": 712.9,
+      "learning_rate": 1.5949820788530468e-05,
+      "loss": 0.2357,
+      "step": 110500
+    },
+    {
+      "epoch": 712.96,
+      "learning_rate": 1.5946236559139786e-05,
+      "loss": 0.2323,
+      "step": 110510
+    },
+    {
+      "epoch": 713.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9492032527923584,
+      "eval_runtime": 0.8694,
+      "eval_samples_per_second": 231.182,
+      "eval_steps_per_second": 8.051,
+      "step": 110515
+    },
+    {
+      "epoch": 713.03,
+      "learning_rate": 1.5942652329749104e-05,
+      "loss": 0.2188,
+      "step": 110520
+    },
+    {
+      "epoch": 713.1,
+      "learning_rate": 1.5939068100358422e-05,
+      "loss": 0.2139,
+      "step": 110530
+    },
+    {
+      "epoch": 713.16,
+      "learning_rate": 1.5935483870967744e-05,
+      "loss": 0.2499,
+      "step": 110540
+    },
+    {
+      "epoch": 713.23,
+      "learning_rate": 1.5931899641577062e-05,
+      "loss": 0.1864,
+      "step": 110550
+    },
+    {
+      "epoch": 713.29,
+      "learning_rate": 1.592831541218638e-05,
+      "loss": 0.264,
+      "step": 110560
+    },
+    {
+      "epoch": 713.35,
+      "learning_rate": 1.59247311827957e-05,
+      "loss": 0.2318,
+      "step": 110570
+    },
+    {
+      "epoch": 713.42,
+      "learning_rate": 1.592114695340502e-05,
+      "loss": 0.2399,
+      "step": 110580
+    },
+    {
+      "epoch": 713.48,
+      "learning_rate": 1.591756272401434e-05,
+      "loss": 0.237,
+      "step": 110590
+    },
+    {
+      "epoch": 713.55,
+      "learning_rate": 1.5913978494623657e-05,
+      "loss": 0.2398,
+      "step": 110600
+    },
+    {
+      "epoch": 713.61,
+      "learning_rate": 1.5910394265232975e-05,
+      "loss": 0.2333,
+      "step": 110610
+    },
+    {
+      "epoch": 713.68,
+      "learning_rate": 1.5906810035842296e-05,
+      "loss": 0.214,
+      "step": 110620
+    },
+    {
+      "epoch": 713.74,
+      "learning_rate": 1.5903225806451615e-05,
+      "loss": 0.2415,
+      "step": 110630
+    },
+    {
+      "epoch": 713.8,
+      "learning_rate": 1.5899641577060933e-05,
+      "loss": 0.2043,
+      "step": 110640
+    },
+    {
+      "epoch": 713.87,
+      "learning_rate": 1.589605734767025e-05,
+      "loss": 0.2236,
+      "step": 110650
+    },
+    {
+      "epoch": 713.93,
+      "learning_rate": 1.5892473118279573e-05,
+      "loss": 0.195,
+      "step": 110660
+    },
+    {
+      "epoch": 714.0,
+      "learning_rate": 1.588888888888889e-05,
+      "loss": 0.2865,
+      "step": 110670
+    },
+    {
+      "epoch": 714.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9239528775215149,
+      "eval_runtime": 0.8606,
+      "eval_samples_per_second": 233.545,
+      "eval_steps_per_second": 8.133,
+      "step": 110670
+    },
+    {
+      "epoch": 714.06,
+      "learning_rate": 1.588530465949821e-05,
+      "loss": 0.1782,
+      "step": 110680
+    },
+    {
+      "epoch": 714.13,
+      "learning_rate": 1.5881720430107527e-05,
+      "loss": 0.2022,
+      "step": 110690
+    },
+    {
+      "epoch": 714.19,
+      "learning_rate": 1.587813620071685e-05,
+      "loss": 0.236,
+      "step": 110700
+    },
+    {
+      "epoch": 714.26,
+      "learning_rate": 1.5874551971326167e-05,
+      "loss": 0.2008,
+      "step": 110710
+    },
+    {
+      "epoch": 714.32,
+      "learning_rate": 1.5870967741935485e-05,
+      "loss": 0.2354,
+      "step": 110720
+    },
+    {
+      "epoch": 714.39,
+      "learning_rate": 1.5867383512544803e-05,
+      "loss": 0.2544,
+      "step": 110730
+    },
+    {
+      "epoch": 714.45,
+      "learning_rate": 1.5863799283154125e-05,
+      "loss": 0.2355,
+      "step": 110740
+    },
+    {
+      "epoch": 714.51,
+      "learning_rate": 1.586021505376344e-05,
+      "loss": 0.2472,
+      "step": 110750
+    },
+    {
+      "epoch": 714.58,
+      "learning_rate": 1.585663082437276e-05,
+      "loss": 0.2604,
+      "step": 110760
+    },
+    {
+      "epoch": 714.64,
+      "learning_rate": 1.585304659498208e-05,
+      "loss": 0.2015,
+      "step": 110770
+    },
+    {
+      "epoch": 714.71,
+      "learning_rate": 1.58494623655914e-05,
+      "loss": 0.1812,
+      "step": 110780
+    },
+    {
+      "epoch": 714.77,
+      "learning_rate": 1.5845878136200716e-05,
+      "loss": 0.2568,
+      "step": 110790
+    },
+    {
+      "epoch": 714.84,
+      "learning_rate": 1.5842293906810038e-05,
+      "loss": 0.2004,
+      "step": 110800
+    },
+    {
+      "epoch": 714.9,
+      "learning_rate": 1.5838709677419356e-05,
+      "loss": 0.2567,
+      "step": 110810
+    },
+    {
+      "epoch": 714.96,
+      "learning_rate": 1.5835125448028677e-05,
+      "loss": 0.2094,
+      "step": 110820
+    },
+    {
+      "epoch": 715.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9182600975036621,
+      "eval_runtime": 0.8423,
+      "eval_samples_per_second": 238.637,
+      "eval_steps_per_second": 8.311,
+      "step": 110825
+    },
+    {
+      "epoch": 715.03,
+      "learning_rate": 1.5831541218637992e-05,
+      "loss": 0.1992,
+      "step": 110830
+    },
+    {
+      "epoch": 715.1,
+      "learning_rate": 1.5827956989247314e-05,
+      "loss": 0.2209,
+      "step": 110840
+    },
+    {
+      "epoch": 715.16,
+      "learning_rate": 1.5824372759856632e-05,
+      "loss": 0.2429,
+      "step": 110850
+    },
+    {
+      "epoch": 715.23,
+      "learning_rate": 1.582078853046595e-05,
+      "loss": 0.1976,
+      "step": 110860
+    },
+    {
+      "epoch": 715.29,
+      "learning_rate": 1.581720430107527e-05,
+      "loss": 0.1991,
+      "step": 110870
+    },
+    {
+      "epoch": 715.35,
+      "learning_rate": 1.5813620071684587e-05,
+      "loss": 0.239,
+      "step": 110880
+    },
+    {
+      "epoch": 715.42,
+      "learning_rate": 1.5810035842293908e-05,
+      "loss": 0.2485,
+      "step": 110890
+    },
+    {
+      "epoch": 715.48,
+      "learning_rate": 1.5806451612903226e-05,
+      "loss": 0.2327,
+      "step": 110900
+    },
+    {
+      "epoch": 715.55,
+      "learning_rate": 1.5802867383512545e-05,
+      "loss": 0.1973,
+      "step": 110910
+    },
+    {
+      "epoch": 715.61,
+      "learning_rate": 1.5799283154121863e-05,
+      "loss": 0.2572,
+      "step": 110920
+    },
+    {
+      "epoch": 715.68,
+      "learning_rate": 1.5795698924731184e-05,
+      "loss": 0.211,
+      "step": 110930
+    },
+    {
+      "epoch": 715.74,
+      "learning_rate": 1.5792114695340503e-05,
+      "loss": 0.1866,
+      "step": 110940
+    },
+    {
+      "epoch": 715.8,
+      "learning_rate": 1.578853046594982e-05,
+      "loss": 0.233,
+      "step": 110950
+    },
+    {
+      "epoch": 715.87,
+      "learning_rate": 1.578494623655914e-05,
+      "loss": 0.1943,
+      "step": 110960
+    },
+    {
+      "epoch": 715.93,
+      "learning_rate": 1.578136200716846e-05,
+      "loss": 0.1736,
+      "step": 110970
+    },
+    {
+      "epoch": 716.0,
+      "learning_rate": 1.577777777777778e-05,
+      "loss": 0.215,
+      "step": 110980
+    },
+    {
+      "epoch": 716.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.9011384844779968,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.831,
+      "eval_steps_per_second": 8.074,
+      "step": 110980
+    },
+    {
+      "epoch": 716.06,
+      "learning_rate": 1.5774193548387097e-05,
+      "loss": 0.2126,
+      "step": 110990
+    },
+    {
+      "epoch": 716.13,
+      "learning_rate": 1.5770609318996415e-05,
+      "loss": 0.2261,
+      "step": 111000
+    },
+    {
+      "epoch": 716.19,
+      "learning_rate": 1.5767025089605737e-05,
+      "loss": 0.2148,
+      "step": 111010
+    },
+    {
+      "epoch": 716.26,
+      "learning_rate": 1.5763440860215055e-05,
+      "loss": 0.1983,
+      "step": 111020
+    },
+    {
+      "epoch": 716.32,
+      "learning_rate": 1.5759856630824373e-05,
+      "loss": 0.2205,
+      "step": 111030
+    },
+    {
+      "epoch": 716.39,
+      "learning_rate": 1.575627240143369e-05,
+      "loss": 0.2292,
+      "step": 111040
+    },
+    {
+      "epoch": 716.45,
+      "learning_rate": 1.5752688172043013e-05,
+      "loss": 0.2175,
+      "step": 111050
+    },
+    {
+      "epoch": 716.51,
+      "learning_rate": 1.574910394265233e-05,
+      "loss": 0.207,
+      "step": 111060
+    },
+    {
+      "epoch": 716.58,
+      "learning_rate": 1.574551971326165e-05,
+      "loss": 0.2673,
+      "step": 111070
+    },
+    {
+      "epoch": 716.64,
+      "learning_rate": 1.5741935483870968e-05,
+      "loss": 0.1611,
+      "step": 111080
+    },
+    {
+      "epoch": 716.71,
+      "learning_rate": 1.573835125448029e-05,
+      "loss": 0.2469,
+      "step": 111090
+    },
+    {
+      "epoch": 716.77,
+      "learning_rate": 1.5734767025089604e-05,
+      "loss": 0.2081,
+      "step": 111100
+    },
+    {
+      "epoch": 716.84,
+      "learning_rate": 1.5731182795698926e-05,
+      "loss": 0.2587,
+      "step": 111110
+    },
+    {
+      "epoch": 716.9,
+      "learning_rate": 1.5727598566308244e-05,
+      "loss": 0.241,
+      "step": 111120
+    },
+    {
+      "epoch": 716.96,
+      "learning_rate": 1.5724014336917565e-05,
+      "loss": 0.2562,
+      "step": 111130
+    },
+    {
+      "epoch": 717.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8990073204040527,
+      "eval_runtime": 0.8825,
+      "eval_samples_per_second": 227.767,
+      "eval_steps_per_second": 7.932,
+      "step": 111135
+    },
+    {
+      "epoch": 717.03,
+      "learning_rate": 1.572043010752688e-05,
+      "loss": 0.2454,
+      "step": 111140
+    },
+    {
+      "epoch": 717.1,
+      "learning_rate": 1.5716845878136202e-05,
+      "loss": 0.2182,
+      "step": 111150
+    },
+    {
+      "epoch": 717.16,
+      "learning_rate": 1.571326164874552e-05,
+      "loss": 0.2164,
+      "step": 111160
+    },
+    {
+      "epoch": 717.23,
+      "learning_rate": 1.570967741935484e-05,
+      "loss": 0.1998,
+      "step": 111170
+    },
+    {
+      "epoch": 717.29,
+      "learning_rate": 1.5706093189964156e-05,
+      "loss": 0.2252,
+      "step": 111180
+    },
+    {
+      "epoch": 717.35,
+      "learning_rate": 1.5702508960573478e-05,
+      "loss": 0.1888,
+      "step": 111190
+    },
+    {
+      "epoch": 717.42,
+      "learning_rate": 1.5698924731182796e-05,
+      "loss": 0.2557,
+      "step": 111200
+    },
+    {
+      "epoch": 717.48,
+      "learning_rate": 1.5695340501792118e-05,
+      "loss": 0.2473,
+      "step": 111210
+    },
+    {
+      "epoch": 717.55,
+      "learning_rate": 1.5691756272401433e-05,
+      "loss": 0.198,
+      "step": 111220
+    },
+    {
+      "epoch": 717.61,
+      "learning_rate": 1.5688172043010754e-05,
+      "loss": 0.1941,
+      "step": 111230
+    },
+    {
+      "epoch": 717.68,
+      "learning_rate": 1.5684587813620072e-05,
+      "loss": 0.2563,
+      "step": 111240
+    },
+    {
+      "epoch": 717.74,
+      "learning_rate": 1.568100358422939e-05,
+      "loss": 0.2411,
+      "step": 111250
+    },
+    {
+      "epoch": 717.8,
+      "learning_rate": 1.567741935483871e-05,
+      "loss": 0.2455,
+      "step": 111260
+    },
+    {
+      "epoch": 717.87,
+      "learning_rate": 1.567383512544803e-05,
+      "loss": 0.1883,
+      "step": 111270
+    },
+    {
+      "epoch": 717.93,
+      "learning_rate": 1.567025089605735e-05,
+      "loss": 0.2923,
+      "step": 111280
+    },
+    {
+      "epoch": 718.0,
+      "learning_rate": 1.5666666666666667e-05,
+      "loss": 0.1979,
+      "step": 111290
+    },
+    {
+      "epoch": 718.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9187625050544739,
+      "eval_runtime": 0.8608,
+      "eval_samples_per_second": 233.503,
+      "eval_steps_per_second": 8.132,
+      "step": 111290
+    },
+    {
+      "epoch": 718.06,
+      "learning_rate": 1.5663082437275985e-05,
+      "loss": 0.2367,
+      "step": 111300
+    },
+    {
+      "epoch": 718.13,
+      "learning_rate": 1.5659498207885307e-05,
+      "loss": 0.2063,
+      "step": 111310
+    },
+    {
+      "epoch": 718.19,
+      "learning_rate": 1.5655913978494625e-05,
+      "loss": 0.2338,
+      "step": 111320
+    },
+    {
+      "epoch": 718.26,
+      "learning_rate": 1.5652329749103943e-05,
+      "loss": 0.259,
+      "step": 111330
+    },
+    {
+      "epoch": 718.32,
+      "learning_rate": 1.564874551971326e-05,
+      "loss": 0.2029,
+      "step": 111340
+    },
+    {
+      "epoch": 718.39,
+      "learning_rate": 1.5645161290322583e-05,
+      "loss": 0.2261,
+      "step": 111350
+    },
+    {
+      "epoch": 718.45,
+      "learning_rate": 1.56415770609319e-05,
+      "loss": 0.2448,
+      "step": 111360
+    },
+    {
+      "epoch": 718.51,
+      "learning_rate": 1.563799283154122e-05,
+      "loss": 0.2435,
+      "step": 111370
+    },
+    {
+      "epoch": 718.58,
+      "learning_rate": 1.5634408602150537e-05,
+      "loss": 0.2617,
+      "step": 111380
+    },
+    {
+      "epoch": 718.64,
+      "learning_rate": 1.563082437275986e-05,
+      "loss": 0.2203,
+      "step": 111390
+    },
+    {
+      "epoch": 718.71,
+      "learning_rate": 1.5627240143369177e-05,
+      "loss": 0.2086,
+      "step": 111400
+    },
+    {
+      "epoch": 718.77,
+      "learning_rate": 1.5623655913978495e-05,
+      "loss": 0.2094,
+      "step": 111410
+    },
+    {
+      "epoch": 718.84,
+      "learning_rate": 1.5620071684587814e-05,
+      "loss": 0.2191,
+      "step": 111420
+    },
+    {
+      "epoch": 718.9,
+      "learning_rate": 1.5616487455197135e-05,
+      "loss": 0.2051,
+      "step": 111430
+    },
+    {
+      "epoch": 718.96,
+      "learning_rate": 1.5612903225806454e-05,
+      "loss": 0.2076,
+      "step": 111440
+    },
+    {
+      "epoch": 719.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9486153721809387,
+      "eval_runtime": 0.8661,
+      "eval_samples_per_second": 232.079,
+      "eval_steps_per_second": 8.082,
+      "step": 111445
+    },
+    {
+      "epoch": 719.03,
+      "learning_rate": 1.5609318996415772e-05,
+      "loss": 0.1985,
+      "step": 111450
+    },
+    {
+      "epoch": 719.1,
+      "learning_rate": 1.560573476702509e-05,
+      "loss": 0.2523,
+      "step": 111460
+    },
+    {
+      "epoch": 719.16,
+      "learning_rate": 1.560215053763441e-05,
+      "loss": 0.2238,
+      "step": 111470
+    },
+    {
+      "epoch": 719.23,
+      "learning_rate": 1.559856630824373e-05,
+      "loss": 0.1972,
+      "step": 111480
+    },
+    {
+      "epoch": 719.29,
+      "learning_rate": 1.5594982078853045e-05,
+      "loss": 0.2204,
+      "step": 111490
+    },
+    {
+      "epoch": 719.35,
+      "learning_rate": 1.5591397849462366e-05,
+      "loss": 0.2289,
+      "step": 111500
+    },
+    {
+      "epoch": 719.42,
+      "learning_rate": 1.5587813620071684e-05,
+      "loss": 0.2394,
+      "step": 111510
+    },
+    {
+      "epoch": 719.48,
+      "learning_rate": 1.5584229390681006e-05,
+      "loss": 0.2191,
+      "step": 111520
+    },
+    {
+      "epoch": 719.55,
+      "learning_rate": 1.558064516129032e-05,
+      "loss": 0.1879,
+      "step": 111530
+    },
+    {
+      "epoch": 719.61,
+      "learning_rate": 1.5577060931899642e-05,
+      "loss": 0.1876,
+      "step": 111540
+    },
+    {
+      "epoch": 719.68,
+      "learning_rate": 1.557347670250896e-05,
+      "loss": 0.238,
+      "step": 111550
+    },
+    {
+      "epoch": 719.74,
+      "learning_rate": 1.5569892473118282e-05,
+      "loss": 0.2925,
+      "step": 111560
+    },
+    {
+      "epoch": 719.8,
+      "learning_rate": 1.5566308243727597e-05,
+      "loss": 0.2482,
+      "step": 111570
+    },
+    {
+      "epoch": 719.87,
+      "learning_rate": 1.556272401433692e-05,
+      "loss": 0.195,
+      "step": 111580
+    },
+    {
+      "epoch": 719.93,
+      "learning_rate": 1.5559139784946237e-05,
+      "loss": 0.2431,
+      "step": 111590
+    },
+    {
+      "epoch": 720.0,
+      "learning_rate": 1.5555555555555555e-05,
+      "loss": 0.2242,
+      "step": 111600
+    },
+    {
+      "epoch": 720.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9075886607170105,
+      "eval_runtime": 0.8933,
+      "eval_samples_per_second": 225.018,
+      "eval_steps_per_second": 7.836,
+      "step": 111600
+    },
+    {
+      "epoch": 720.06,
+      "learning_rate": 1.5551971326164873e-05,
+      "loss": 0.217,
+      "step": 111610
+    },
+    {
+      "epoch": 720.13,
+      "learning_rate": 1.5548387096774195e-05,
+      "loss": 0.2185,
+      "step": 111620
+    },
+    {
+      "epoch": 720.19,
+      "learning_rate": 1.5544802867383513e-05,
+      "loss": 0.2268,
+      "step": 111630
+    },
+    {
+      "epoch": 720.26,
+      "learning_rate": 1.554121863799283e-05,
+      "loss": 0.2248,
+      "step": 111640
+    },
+    {
+      "epoch": 720.32,
+      "learning_rate": 1.553763440860215e-05,
+      "loss": 0.2234,
+      "step": 111650
+    },
+    {
+      "epoch": 720.39,
+      "learning_rate": 1.553405017921147e-05,
+      "loss": 0.2391,
+      "step": 111660
+    },
+    {
+      "epoch": 720.45,
+      "learning_rate": 1.553046594982079e-05,
+      "loss": 0.2512,
+      "step": 111670
+    },
+    {
+      "epoch": 720.51,
+      "learning_rate": 1.5526881720430107e-05,
+      "loss": 0.2505,
+      "step": 111680
+    },
+    {
+      "epoch": 720.58,
+      "learning_rate": 1.5523297491039426e-05,
+      "loss": 0.1836,
+      "step": 111690
+    },
+    {
+      "epoch": 720.64,
+      "learning_rate": 1.5519713261648747e-05,
+      "loss": 0.2093,
+      "step": 111700
+    },
+    {
+      "epoch": 720.71,
+      "learning_rate": 1.5516129032258065e-05,
+      "loss": 0.2448,
+      "step": 111710
+    },
+    {
+      "epoch": 720.77,
+      "learning_rate": 1.5512544802867384e-05,
+      "loss": 0.2508,
+      "step": 111720
+    },
+    {
+      "epoch": 720.84,
+      "learning_rate": 1.5508960573476702e-05,
+      "loss": 0.235,
+      "step": 111730
+    },
+    {
+      "epoch": 720.9,
+      "learning_rate": 1.5505376344086023e-05,
+      "loss": 0.2331,
+      "step": 111740
+    },
+    {
+      "epoch": 720.96,
+      "learning_rate": 1.550179211469534e-05,
+      "loss": 0.2018,
+      "step": 111750
+    },
+    {
+      "epoch": 721.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9081218838691711,
+      "eval_runtime": 0.8654,
+      "eval_samples_per_second": 232.251,
+      "eval_steps_per_second": 8.088,
+      "step": 111755
+    },
+    {
+      "epoch": 721.03,
+      "learning_rate": 1.549820788530466e-05,
+      "loss": 0.3056,
+      "step": 111760
+    },
+    {
+      "epoch": 721.1,
+      "learning_rate": 1.5494623655913978e-05,
+      "loss": 0.1991,
+      "step": 111770
+    },
+    {
+      "epoch": 721.16,
+      "learning_rate": 1.54910394265233e-05,
+      "loss": 0.2355,
+      "step": 111780
+    },
+    {
+      "epoch": 721.23,
+      "learning_rate": 1.5487455197132618e-05,
+      "loss": 0.2436,
+      "step": 111790
+    },
+    {
+      "epoch": 721.29,
+      "learning_rate": 1.5483870967741936e-05,
+      "loss": 0.2188,
+      "step": 111800
+    },
+    {
+      "epoch": 721.35,
+      "learning_rate": 1.5480286738351254e-05,
+      "loss": 0.2402,
+      "step": 111810
+    },
+    {
+      "epoch": 721.42,
+      "learning_rate": 1.5476702508960576e-05,
+      "loss": 0.2071,
+      "step": 111820
+    },
+    {
+      "epoch": 721.48,
+      "learning_rate": 1.5473118279569894e-05,
+      "loss": 0.2033,
+      "step": 111830
+    },
+    {
+      "epoch": 721.55,
+      "learning_rate": 1.5469534050179212e-05,
+      "loss": 0.2199,
+      "step": 111840
+    },
+    {
+      "epoch": 721.61,
+      "learning_rate": 1.546594982078853e-05,
+      "loss": 0.22,
+      "step": 111850
+    },
+    {
+      "epoch": 721.68,
+      "learning_rate": 1.5462365591397852e-05,
+      "loss": 0.1968,
+      "step": 111860
+    },
+    {
+      "epoch": 721.74,
+      "learning_rate": 1.545878136200717e-05,
+      "loss": 0.2017,
+      "step": 111870
+    },
+    {
+      "epoch": 721.8,
+      "learning_rate": 1.545519713261649e-05,
+      "loss": 0.2014,
+      "step": 111880
+    },
+    {
+      "epoch": 721.87,
+      "learning_rate": 1.5451612903225807e-05,
+      "loss": 0.1935,
+      "step": 111890
+    },
+    {
+      "epoch": 721.93,
+      "learning_rate": 1.5448028673835128e-05,
+      "loss": 0.2428,
+      "step": 111900
+    },
+    {
+      "epoch": 722.0,
+      "learning_rate": 1.5444444444444446e-05,
+      "loss": 0.2423,
+      "step": 111910
+    },
+    {
+      "epoch": 722.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9936788082122803,
+      "eval_runtime": 0.8898,
+      "eval_samples_per_second": 225.904,
+      "eval_steps_per_second": 7.867,
+      "step": 111910
+    },
+    {
+      "epoch": 722.06,
+      "learning_rate": 1.5440860215053765e-05,
+      "loss": 0.1963,
+      "step": 111920
+    },
+    {
+      "epoch": 722.13,
+      "learning_rate": 1.5437275985663083e-05,
+      "loss": 0.1891,
+      "step": 111930
+    },
+    {
+      "epoch": 722.19,
+      "learning_rate": 1.5433691756272404e-05,
+      "loss": 0.2647,
+      "step": 111940
+    },
+    {
+      "epoch": 722.26,
+      "learning_rate": 1.543010752688172e-05,
+      "loss": 0.1915,
+      "step": 111950
+    },
+    {
+      "epoch": 722.32,
+      "learning_rate": 1.542652329749104e-05,
+      "loss": 0.2521,
+      "step": 111960
+    },
+    {
+      "epoch": 722.39,
+      "learning_rate": 1.542293906810036e-05,
+      "loss": 0.2127,
+      "step": 111970
+    },
+    {
+      "epoch": 722.45,
+      "learning_rate": 1.541935483870968e-05,
+      "loss": 0.2058,
+      "step": 111980
+    },
+    {
+      "epoch": 722.51,
+      "learning_rate": 1.5415770609318995e-05,
+      "loss": 0.2605,
+      "step": 111990
+    },
+    {
+      "epoch": 722.58,
+      "learning_rate": 1.5412186379928317e-05,
+      "loss": 0.1797,
+      "step": 112000
+    },
+    {
+      "epoch": 722.64,
+      "learning_rate": 1.5408602150537635e-05,
+      "loss": 0.2243,
+      "step": 112010
+    },
+    {
+      "epoch": 722.71,
+      "learning_rate": 1.5405017921146957e-05,
+      "loss": 0.2049,
+      "step": 112020
+    },
+    {
+      "epoch": 722.77,
+      "learning_rate": 1.540143369175627e-05,
+      "loss": 0.1831,
+      "step": 112030
+    },
+    {
+      "epoch": 722.84,
+      "learning_rate": 1.5397849462365593e-05,
+      "loss": 0.2372,
+      "step": 112040
+    },
+    {
+      "epoch": 722.9,
+      "learning_rate": 1.539426523297491e-05,
+      "loss": 0.2253,
+      "step": 112050
+    },
+    {
+      "epoch": 722.96,
+      "learning_rate": 1.5390681003584233e-05,
+      "loss": 0.1955,
+      "step": 112060
+    },
+    {
+      "epoch": 723.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9489098787307739,
+      "eval_runtime": 0.8571,
+      "eval_samples_per_second": 234.523,
+      "eval_steps_per_second": 8.167,
+      "step": 112065
+    },
+    {
+      "epoch": 723.03,
+      "learning_rate": 1.5387096774193548e-05,
+      "loss": 0.22,
+      "step": 112070
+    },
+    {
+      "epoch": 723.1,
+      "learning_rate": 1.538351254480287e-05,
+      "loss": 0.2269,
+      "step": 112080
+    },
+    {
+      "epoch": 723.16,
+      "learning_rate": 1.5379928315412188e-05,
+      "loss": 0.2237,
+      "step": 112090
+    },
+    {
+      "epoch": 723.23,
+      "learning_rate": 1.5376344086021506e-05,
+      "loss": 0.2122,
+      "step": 112100
+    },
+    {
+      "epoch": 723.29,
+      "learning_rate": 1.5372759856630824e-05,
+      "loss": 0.2144,
+      "step": 112110
+    },
+    {
+      "epoch": 723.35,
+      "learning_rate": 1.5369175627240146e-05,
+      "loss": 0.219,
+      "step": 112120
+    },
+    {
+      "epoch": 723.42,
+      "learning_rate": 1.5365591397849464e-05,
+      "loss": 0.23,
+      "step": 112130
+    },
+    {
+      "epoch": 723.48,
+      "learning_rate": 1.5362007168458782e-05,
+      "loss": 0.2042,
+      "step": 112140
+    },
+    {
+      "epoch": 723.55,
+      "learning_rate": 1.53584229390681e-05,
+      "loss": 0.2118,
+      "step": 112150
+    },
+    {
+      "epoch": 723.61,
+      "learning_rate": 1.535483870967742e-05,
+      "loss": 0.2111,
+      "step": 112160
+    },
+    {
+      "epoch": 723.68,
+      "learning_rate": 1.535125448028674e-05,
+      "loss": 0.2292,
+      "step": 112170
+    },
+    {
+      "epoch": 723.74,
+      "learning_rate": 1.5347670250896058e-05,
+      "loss": 0.1954,
+      "step": 112180
+    },
+    {
+      "epoch": 723.8,
+      "learning_rate": 1.5344086021505376e-05,
+      "loss": 0.2165,
+      "step": 112190
+    },
+    {
+      "epoch": 723.87,
+      "learning_rate": 1.5340501792114695e-05,
+      "loss": 0.2214,
+      "step": 112200
+    },
+    {
+      "epoch": 723.93,
+      "learning_rate": 1.5336917562724016e-05,
+      "loss": 0.2516,
+      "step": 112210
+    },
+    {
+      "epoch": 724.0,
+      "learning_rate": 1.5333333333333334e-05,
+      "loss": 0.2006,
+      "step": 112220
+    },
+    {
+      "epoch": 724.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9431391358375549,
+      "eval_runtime": 0.8654,
+      "eval_samples_per_second": 232.255,
+      "eval_steps_per_second": 8.088,
+      "step": 112220
+    },
+    {
+      "epoch": 724.06,
+      "learning_rate": 1.5329749103942653e-05,
+      "loss": 0.2378,
+      "step": 112230
+    },
+    {
+      "epoch": 724.13,
+      "learning_rate": 1.532616487455197e-05,
+      "loss": 0.2254,
+      "step": 112240
+    },
+    {
+      "epoch": 724.19,
+      "learning_rate": 1.5322580645161292e-05,
+      "loss": 0.253,
+      "step": 112250
+    },
+    {
+      "epoch": 724.26,
+      "learning_rate": 1.5318996415770607e-05,
+      "loss": 0.2481,
+      "step": 112260
+    },
+    {
+      "epoch": 724.32,
+      "learning_rate": 1.531541218637993e-05,
+      "loss": 0.211,
+      "step": 112270
+    },
+    {
+      "epoch": 724.39,
+      "learning_rate": 1.5311827956989247e-05,
+      "loss": 0.2093,
+      "step": 112280
+    },
+    {
+      "epoch": 724.45,
+      "learning_rate": 1.530824372759857e-05,
+      "loss": 0.1856,
+      "step": 112290
+    },
+    {
+      "epoch": 724.51,
+      "learning_rate": 1.5304659498207884e-05,
+      "loss": 0.1982,
+      "step": 112300
+    },
+    {
+      "epoch": 724.58,
+      "learning_rate": 1.5301075268817205e-05,
+      "loss": 0.2141,
+      "step": 112310
+    },
+    {
+      "epoch": 724.64,
+      "learning_rate": 1.5297491039426523e-05,
+      "loss": 0.1775,
+      "step": 112320
+    },
+    {
+      "epoch": 724.71,
+      "learning_rate": 1.5293906810035845e-05,
+      "loss": 0.2327,
+      "step": 112330
+    },
+    {
+      "epoch": 724.77,
+      "learning_rate": 1.529032258064516e-05,
+      "loss": 0.2115,
+      "step": 112340
+    },
+    {
+      "epoch": 724.84,
+      "learning_rate": 1.528673835125448e-05,
+      "loss": 0.2121,
+      "step": 112350
+    },
+    {
+      "epoch": 724.9,
+      "learning_rate": 1.52831541218638e-05,
+      "loss": 0.2329,
+      "step": 112360
+    },
+    {
+      "epoch": 724.96,
+      "learning_rate": 1.527956989247312e-05,
+      "loss": 0.2073,
+      "step": 112370
+    },
+    {
+      "epoch": 725.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9037317037582397,
+      "eval_runtime": 0.8775,
+      "eval_samples_per_second": 229.054,
+      "eval_steps_per_second": 7.977,
+      "step": 112375
+    },
+    {
+      "epoch": 725.03,
+      "learning_rate": 1.5275985663082436e-05,
+      "loss": 0.2331,
+      "step": 112380
+    },
+    {
+      "epoch": 725.1,
+      "learning_rate": 1.5272401433691758e-05,
+      "loss": 0.1826,
+      "step": 112390
+    },
+    {
+      "epoch": 725.16,
+      "learning_rate": 1.5268817204301076e-05,
+      "loss": 0.1645,
+      "step": 112400
+    },
+    {
+      "epoch": 725.23,
+      "learning_rate": 1.5265232974910394e-05,
+      "loss": 0.2089,
+      "step": 112410
+    },
+    {
+      "epoch": 725.29,
+      "learning_rate": 1.5261648745519712e-05,
+      "loss": 0.1922,
+      "step": 112420
+    },
+    {
+      "epoch": 725.35,
+      "learning_rate": 1.5258064516129034e-05,
+      "loss": 0.2537,
+      "step": 112430
+    },
+    {
+      "epoch": 725.42,
+      "learning_rate": 1.5254480286738352e-05,
+      "loss": 0.2214,
+      "step": 112440
+    },
+    {
+      "epoch": 725.48,
+      "learning_rate": 1.5250896057347672e-05,
+      "loss": 0.2154,
+      "step": 112450
+    },
+    {
+      "epoch": 725.55,
+      "learning_rate": 1.524731182795699e-05,
+      "loss": 0.191,
+      "step": 112460
+    },
+    {
+      "epoch": 725.61,
+      "learning_rate": 1.524372759856631e-05,
+      "loss": 0.2228,
+      "step": 112470
+    },
+    {
+      "epoch": 725.68,
+      "learning_rate": 1.5240143369175628e-05,
+      "loss": 0.2134,
+      "step": 112480
+    },
+    {
+      "epoch": 725.74,
+      "learning_rate": 1.5236559139784948e-05,
+      "loss": 0.2034,
+      "step": 112490
+    },
+    {
+      "epoch": 725.8,
+      "learning_rate": 1.5232974910394265e-05,
+      "loss": 0.1998,
+      "step": 112500
+    },
+    {
+      "epoch": 725.87,
+      "learning_rate": 1.5229390681003586e-05,
+      "loss": 0.2344,
+      "step": 112510
+    },
+    {
+      "epoch": 725.93,
+      "learning_rate": 1.5225806451612903e-05,
+      "loss": 0.2068,
+      "step": 112520
+    },
+    {
+      "epoch": 726.0,
+      "learning_rate": 1.5222222222222224e-05,
+      "loss": 0.2365,
+      "step": 112530
+    },
+    {
+      "epoch": 726.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9315109848976135,
+      "eval_runtime": 0.8675,
+      "eval_samples_per_second": 231.704,
+      "eval_steps_per_second": 8.069,
+      "step": 112530
+    },
+    {
+      "epoch": 726.06,
+      "learning_rate": 1.521863799283154e-05,
+      "loss": 0.2161,
+      "step": 112540
+    },
+    {
+      "epoch": 726.13,
+      "learning_rate": 1.5215053763440862e-05,
+      "loss": 0.1937,
+      "step": 112550
+    },
+    {
+      "epoch": 726.19,
+      "learning_rate": 1.5211469534050179e-05,
+      "loss": 0.2203,
+      "step": 112560
+    },
+    {
+      "epoch": 726.26,
+      "learning_rate": 1.52078853046595e-05,
+      "loss": 0.2237,
+      "step": 112570
+    },
+    {
+      "epoch": 726.32,
+      "learning_rate": 1.5204301075268817e-05,
+      "loss": 0.2615,
+      "step": 112580
+    },
+    {
+      "epoch": 726.39,
+      "learning_rate": 1.5200716845878139e-05,
+      "loss": 0.2108,
+      "step": 112590
+    },
+    {
+      "epoch": 726.45,
+      "learning_rate": 1.5197132616487455e-05,
+      "loss": 0.1987,
+      "step": 112600
+    },
+    {
+      "epoch": 726.51,
+      "learning_rate": 1.5193548387096777e-05,
+      "loss": 0.2813,
+      "step": 112610
+    },
+    {
+      "epoch": 726.58,
+      "learning_rate": 1.5189964157706093e-05,
+      "loss": 0.221,
+      "step": 112620
+    },
+    {
+      "epoch": 726.64,
+      "learning_rate": 1.5186379928315415e-05,
+      "loss": 0.1992,
+      "step": 112630
+    },
+    {
+      "epoch": 726.71,
+      "learning_rate": 1.5182795698924731e-05,
+      "loss": 0.2402,
+      "step": 112640
+    },
+    {
+      "epoch": 726.77,
+      "learning_rate": 1.5179211469534053e-05,
+      "loss": 0.2064,
+      "step": 112650
+    },
+    {
+      "epoch": 726.84,
+      "learning_rate": 1.517562724014337e-05,
+      "loss": 0.2436,
+      "step": 112660
+    },
+    {
+      "epoch": 726.9,
+      "learning_rate": 1.517204301075269e-05,
+      "loss": 0.2093,
+      "step": 112670
+    },
+    {
+      "epoch": 726.96,
+      "learning_rate": 1.5168458781362007e-05,
+      "loss": 0.2134,
+      "step": 112680
+    },
+    {
+      "epoch": 727.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9521178603172302,
+      "eval_runtime": 0.8893,
+      "eval_samples_per_second": 226.031,
+      "eval_steps_per_second": 7.872,
+      "step": 112685
+    },
+    {
+      "epoch": 727.03,
+      "learning_rate": 1.5164874551971327e-05,
+      "loss": 0.2636,
+      "step": 112690
+    },
+    {
+      "epoch": 727.1,
+      "learning_rate": 1.5161290322580646e-05,
+      "loss": 0.1817,
+      "step": 112700
+    },
+    {
+      "epoch": 727.16,
+      "learning_rate": 1.5157706093189965e-05,
+      "loss": 0.2332,
+      "step": 112710
+    },
+    {
+      "epoch": 727.23,
+      "learning_rate": 1.5154121863799284e-05,
+      "loss": 0.2408,
+      "step": 112720
+    },
+    {
+      "epoch": 727.29,
+      "learning_rate": 1.5150537634408604e-05,
+      "loss": 0.2112,
+      "step": 112730
+    },
+    {
+      "epoch": 727.35,
+      "learning_rate": 1.5146953405017922e-05,
+      "loss": 0.1944,
+      "step": 112740
+    },
+    {
+      "epoch": 727.42,
+      "learning_rate": 1.5143369175627242e-05,
+      "loss": 0.191,
+      "step": 112750
+    },
+    {
+      "epoch": 727.48,
+      "learning_rate": 1.513978494623656e-05,
+      "loss": 0.1958,
+      "step": 112760
+    },
+    {
+      "epoch": 727.55,
+      "learning_rate": 1.5136200716845878e-05,
+      "loss": 0.2112,
+      "step": 112770
+    },
+    {
+      "epoch": 727.61,
+      "learning_rate": 1.5132616487455198e-05,
+      "loss": 0.2362,
+      "step": 112780
+    },
+    {
+      "epoch": 727.68,
+      "learning_rate": 1.5129032258064516e-05,
+      "loss": 0.2589,
+      "step": 112790
+    },
+    {
+      "epoch": 727.74,
+      "learning_rate": 1.5125448028673836e-05,
+      "loss": 0.2662,
+      "step": 112800
+    },
+    {
+      "epoch": 727.8,
+      "learning_rate": 1.5121863799283154e-05,
+      "loss": 0.2217,
+      "step": 112810
+    },
+    {
+      "epoch": 727.87,
+      "learning_rate": 1.5118279569892474e-05,
+      "loss": 0.2823,
+      "step": 112820
+    },
+    {
+      "epoch": 727.93,
+      "learning_rate": 1.5114695340501792e-05,
+      "loss": 0.2352,
+      "step": 112830
+    },
+    {
+      "epoch": 728.0,
+      "learning_rate": 1.5111111111111112e-05,
+      "loss": 0.2083,
+      "step": 112840
+    },
+    {
+      "epoch": 728.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.966724157333374,
+      "eval_runtime": 0.8755,
+      "eval_samples_per_second": 229.58,
+      "eval_steps_per_second": 7.995,
+      "step": 112840
+    },
+    {
+      "epoch": 728.06,
+      "learning_rate": 1.5107526881720429e-05,
+      "loss": 0.2419,
+      "step": 112850
+    },
+    {
+      "epoch": 728.13,
+      "learning_rate": 1.510394265232975e-05,
+      "loss": 0.1945,
+      "step": 112860
+    },
+    {
+      "epoch": 728.19,
+      "learning_rate": 1.5100358422939067e-05,
+      "loss": 0.2121,
+      "step": 112870
+    },
+    {
+      "epoch": 728.26,
+      "learning_rate": 1.5096774193548389e-05,
+      "loss": 0.2108,
+      "step": 112880
+    },
+    {
+      "epoch": 728.32,
+      "learning_rate": 1.5093189964157705e-05,
+      "loss": 0.2048,
+      "step": 112890
+    },
+    {
+      "epoch": 728.39,
+      "learning_rate": 1.5089605734767027e-05,
+      "loss": 0.1945,
+      "step": 112900
+    },
+    {
+      "epoch": 728.45,
+      "learning_rate": 1.5086021505376343e-05,
+      "loss": 0.1925,
+      "step": 112910
+    },
+    {
+      "epoch": 728.51,
+      "learning_rate": 1.5082437275985665e-05,
+      "loss": 0.249,
+      "step": 112920
+    },
+    {
+      "epoch": 728.58,
+      "learning_rate": 1.5078853046594981e-05,
+      "loss": 0.2664,
+      "step": 112930
+    },
+    {
+      "epoch": 728.64,
+      "learning_rate": 1.5075268817204303e-05,
+      "loss": 0.2357,
+      "step": 112940
+    },
+    {
+      "epoch": 728.71,
+      "learning_rate": 1.507168458781362e-05,
+      "loss": 0.209,
+      "step": 112950
+    },
+    {
+      "epoch": 728.77,
+      "learning_rate": 1.5068100358422941e-05,
+      "loss": 0.2357,
+      "step": 112960
+    },
+    {
+      "epoch": 728.84,
+      "learning_rate": 1.5064516129032257e-05,
+      "loss": 0.2399,
+      "step": 112970
+    },
+    {
+      "epoch": 728.9,
+      "learning_rate": 1.5060931899641579e-05,
+      "loss": 0.2168,
+      "step": 112980
+    },
+    {
+      "epoch": 728.96,
+      "learning_rate": 1.5057347670250896e-05,
+      "loss": 0.1652,
+      "step": 112990
+    },
+    {
+      "epoch": 729.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.960196852684021,
+      "eval_runtime": 0.8606,
+      "eval_samples_per_second": 233.551,
+      "eval_steps_per_second": 8.134,
+      "step": 112995
+    },
+    {
+      "epoch": 729.03,
+      "learning_rate": 1.5053763440860215e-05,
+      "loss": 0.2349,
+      "step": 113000
+    },
+    {
+      "epoch": 729.1,
+      "learning_rate": 1.5050179211469534e-05,
+      "loss": 0.2606,
+      "step": 113010
+    },
+    {
+      "epoch": 729.16,
+      "learning_rate": 1.5046594982078854e-05,
+      "loss": 0.2289,
+      "step": 113020
+    },
+    {
+      "epoch": 729.23,
+      "learning_rate": 1.5043010752688172e-05,
+      "loss": 0.2056,
+      "step": 113030
+    },
+    {
+      "epoch": 729.29,
+      "learning_rate": 1.5039426523297492e-05,
+      "loss": 0.2323,
+      "step": 113040
+    },
+    {
+      "epoch": 729.35,
+      "learning_rate": 1.503584229390681e-05,
+      "loss": 0.2311,
+      "step": 113050
+    },
+    {
+      "epoch": 729.42,
+      "learning_rate": 1.503225806451613e-05,
+      "loss": 0.2807,
+      "step": 113060
+    },
+    {
+      "epoch": 729.48,
+      "learning_rate": 1.5028673835125448e-05,
+      "loss": 0.235,
+      "step": 113070
+    },
+    {
+      "epoch": 729.55,
+      "learning_rate": 1.5025089605734768e-05,
+      "loss": 0.2141,
+      "step": 113080
+    },
+    {
+      "epoch": 729.61,
+      "learning_rate": 1.5021505376344086e-05,
+      "loss": 0.2052,
+      "step": 113090
+    },
+    {
+      "epoch": 729.68,
+      "learning_rate": 1.5017921146953406e-05,
+      "loss": 0.2014,
+      "step": 113100
+    },
+    {
+      "epoch": 729.74,
+      "learning_rate": 1.5014336917562724e-05,
+      "loss": 0.2428,
+      "step": 113110
+    },
+    {
+      "epoch": 729.8,
+      "learning_rate": 1.5010752688172044e-05,
+      "loss": 0.1983,
+      "step": 113120
+    },
+    {
+      "epoch": 729.87,
+      "learning_rate": 1.5007168458781362e-05,
+      "loss": 0.2246,
+      "step": 113130
+    },
+    {
+      "epoch": 729.93,
+      "learning_rate": 1.5003584229390682e-05,
+      "loss": 0.2477,
+      "step": 113140
+    },
+    {
+      "epoch": 730.0,
+      "learning_rate": 1.5e-05,
+      "loss": 0.1904,
+      "step": 113150
+    },
+    {
+      "epoch": 730.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.980147659778595,
+      "eval_runtime": 0.9115,
+      "eval_samples_per_second": 220.518,
+      "eval_steps_per_second": 7.68,
+      "step": 113150
+    },
+    {
+      "epoch": 730.06,
+      "learning_rate": 1.499641577060932e-05,
+      "loss": 0.2326,
+      "step": 113160
+    },
+    {
+      "epoch": 730.13,
+      "learning_rate": 1.4992831541218638e-05,
+      "loss": 0.1679,
+      "step": 113170
+    },
+    {
+      "epoch": 730.19,
+      "learning_rate": 1.4989247311827958e-05,
+      "loss": 0.2388,
+      "step": 113180
+    },
+    {
+      "epoch": 730.26,
+      "learning_rate": 1.4985663082437277e-05,
+      "loss": 0.2238,
+      "step": 113190
+    },
+    {
+      "epoch": 730.32,
+      "learning_rate": 1.4982078853046596e-05,
+      "loss": 0.2565,
+      "step": 113200
+    },
+    {
+      "epoch": 730.39,
+      "learning_rate": 1.4978494623655915e-05,
+      "loss": 0.2501,
+      "step": 113210
+    },
+    {
+      "epoch": 730.45,
+      "learning_rate": 1.4974910394265235e-05,
+      "loss": 0.2025,
+      "step": 113220
+    },
+    {
+      "epoch": 730.51,
+      "learning_rate": 1.4971326164874553e-05,
+      "loss": 0.2434,
+      "step": 113230
+    },
+    {
+      "epoch": 730.58,
+      "learning_rate": 1.4967741935483873e-05,
+      "loss": 0.1971,
+      "step": 113240
+    },
+    {
+      "epoch": 730.64,
+      "learning_rate": 1.4964157706093191e-05,
+      "loss": 0.2098,
+      "step": 113250
+    },
+    {
+      "epoch": 730.71,
+      "learning_rate": 1.496057347670251e-05,
+      "loss": 0.2426,
+      "step": 113260
+    },
+    {
+      "epoch": 730.77,
+      "learning_rate": 1.4956989247311829e-05,
+      "loss": 0.2492,
+      "step": 113270
+    },
+    {
+      "epoch": 730.84,
+      "learning_rate": 1.4953405017921149e-05,
+      "loss": 0.2183,
+      "step": 113280
+    },
+    {
+      "epoch": 730.9,
+      "learning_rate": 1.4949820788530467e-05,
+      "loss": 0.2483,
+      "step": 113290
+    },
+    {
+      "epoch": 730.96,
+      "learning_rate": 1.4946236559139787e-05,
+      "loss": 0.2484,
+      "step": 113300
+    },
+    {
+      "epoch": 731.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.0081593990325928,
+      "eval_runtime": 0.8984,
+      "eval_samples_per_second": 223.725,
+      "eval_steps_per_second": 7.791,
+      "step": 113305
+    },
+    {
+      "epoch": 731.03,
+      "learning_rate": 1.4942652329749105e-05,
+      "loss": 0.2203,
+      "step": 113310
+    },
+    {
+      "epoch": 731.1,
+      "learning_rate": 1.4939068100358425e-05,
+      "loss": 0.224,
+      "step": 113320
+    },
+    {
+      "epoch": 731.16,
+      "learning_rate": 1.4935483870967743e-05,
+      "loss": 0.2447,
+      "step": 113330
+    },
+    {
+      "epoch": 731.23,
+      "learning_rate": 1.4931899641577063e-05,
+      "loss": 0.1763,
+      "step": 113340
+    },
+    {
+      "epoch": 731.29,
+      "learning_rate": 1.492831541218638e-05,
+      "loss": 0.261,
+      "step": 113350
+    },
+    {
+      "epoch": 731.35,
+      "learning_rate": 1.4924731182795701e-05,
+      "loss": 0.1954,
+      "step": 113360
+    },
+    {
+      "epoch": 731.42,
+      "learning_rate": 1.4921146953405018e-05,
+      "loss": 0.2408,
+      "step": 113370
+    },
+    {
+      "epoch": 731.48,
+      "learning_rate": 1.491756272401434e-05,
+      "loss": 0.1959,
+      "step": 113380
+    },
+    {
+      "epoch": 731.55,
+      "learning_rate": 1.4913978494623656e-05,
+      "loss": 0.2706,
+      "step": 113390
+    },
+    {
+      "epoch": 731.61,
+      "learning_rate": 1.4910394265232978e-05,
+      "loss": 0.2097,
+      "step": 113400
+    },
+    {
+      "epoch": 731.68,
+      "learning_rate": 1.4906810035842294e-05,
+      "loss": 0.2027,
+      "step": 113410
+    },
+    {
+      "epoch": 731.74,
+      "learning_rate": 1.4903225806451612e-05,
+      "loss": 0.2234,
+      "step": 113420
+    },
+    {
+      "epoch": 731.8,
+      "learning_rate": 1.4899641577060932e-05,
+      "loss": 0.2044,
+      "step": 113430
+    },
+    {
+      "epoch": 731.87,
+      "learning_rate": 1.489605734767025e-05,
+      "loss": 0.2663,
+      "step": 113440
+    },
+    {
+      "epoch": 731.93,
+      "learning_rate": 1.489247311827957e-05,
+      "loss": 0.2278,
+      "step": 113450
+    },
+    {
+      "epoch": 732.0,
+      "learning_rate": 1.4888888888888888e-05,
+      "loss": 0.2339,
+      "step": 113460
+    },
+    {
+      "epoch": 732.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9585430026054382,
+      "eval_runtime": 0.8978,
+      "eval_samples_per_second": 223.881,
+      "eval_steps_per_second": 7.797,
+      "step": 113460
+    },
+    {
+      "epoch": 732.06,
+      "learning_rate": 1.4885304659498208e-05,
+      "loss": 0.2434,
+      "step": 113470
+    },
+    {
+      "epoch": 732.13,
+      "learning_rate": 1.4881720430107527e-05,
+      "loss": 0.2051,
+      "step": 113480
+    },
+    {
+      "epoch": 732.19,
+      "learning_rate": 1.4878136200716846e-05,
+      "loss": 0.2276,
+      "step": 113490
+    },
+    {
+      "epoch": 732.26,
+      "learning_rate": 1.4874551971326165e-05,
+      "loss": 0.195,
+      "step": 113500
+    },
+    {
+      "epoch": 732.32,
+      "learning_rate": 1.4870967741935485e-05,
+      "loss": 0.2003,
+      "step": 113510
+    },
+    {
+      "epoch": 732.39,
+      "learning_rate": 1.4867383512544803e-05,
+      "loss": 0.2485,
+      "step": 113520
+    },
+    {
+      "epoch": 732.45,
+      "learning_rate": 1.4863799283154123e-05,
+      "loss": 0.2249,
+      "step": 113530
+    },
+    {
+      "epoch": 732.51,
+      "learning_rate": 1.4860215053763441e-05,
+      "loss": 0.1815,
+      "step": 113540
+    },
+    {
+      "epoch": 732.58,
+      "learning_rate": 1.485663082437276e-05,
+      "loss": 0.1676,
+      "step": 113550
+    },
+    {
+      "epoch": 732.64,
+      "learning_rate": 1.4853046594982079e-05,
+      "loss": 0.2024,
+      "step": 113560
+    },
+    {
+      "epoch": 732.71,
+      "learning_rate": 1.4849462365591399e-05,
+      "loss": 0.2172,
+      "step": 113570
+    },
+    {
+      "epoch": 732.77,
+      "learning_rate": 1.4845878136200717e-05,
+      "loss": 0.2002,
+      "step": 113580
+    },
+    {
+      "epoch": 732.84,
+      "learning_rate": 1.4842293906810037e-05,
+      "loss": 0.2054,
+      "step": 113590
+    },
+    {
+      "epoch": 732.9,
+      "learning_rate": 1.4838709677419355e-05,
+      "loss": 0.2365,
+      "step": 113600
+    },
+    {
+      "epoch": 732.96,
+      "learning_rate": 1.4835125448028675e-05,
+      "loss": 0.2075,
+      "step": 113610
+    },
+    {
+      "epoch": 733.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9733757376670837,
+      "eval_runtime": 0.8595,
+      "eval_samples_per_second": 233.852,
+      "eval_steps_per_second": 8.144,
+      "step": 113615
+    },
+    {
+      "epoch": 733.03,
+      "learning_rate": 1.4831541218637993e-05,
+      "loss": 0.1963,
+      "step": 113620
+    },
+    {
+      "epoch": 733.1,
+      "learning_rate": 1.4827956989247313e-05,
+      "loss": 0.2664,
+      "step": 113630
+    },
+    {
+      "epoch": 733.16,
+      "learning_rate": 1.4824372759856631e-05,
+      "loss": 0.1896,
+      "step": 113640
+    },
+    {
+      "epoch": 733.23,
+      "learning_rate": 1.4820788530465951e-05,
+      "loss": 0.2359,
+      "step": 113650
+    },
+    {
+      "epoch": 733.29,
+      "learning_rate": 1.481720430107527e-05,
+      "loss": 0.2234,
+      "step": 113660
+    },
+    {
+      "epoch": 733.35,
+      "learning_rate": 1.481362007168459e-05,
+      "loss": 0.2746,
+      "step": 113670
+    },
+    {
+      "epoch": 733.42,
+      "learning_rate": 1.4810035842293906e-05,
+      "loss": 0.2152,
+      "step": 113680
+    },
+    {
+      "epoch": 733.48,
+      "learning_rate": 1.4806451612903227e-05,
+      "loss": 0.1918,
+      "step": 113690
+    },
+    {
+      "epoch": 733.55,
+      "learning_rate": 1.4802867383512544e-05,
+      "loss": 0.2044,
+      "step": 113700
+    },
+    {
+      "epoch": 733.61,
+      "learning_rate": 1.4799283154121866e-05,
+      "loss": 0.1686,
+      "step": 113710
+    },
+    {
+      "epoch": 733.68,
+      "learning_rate": 1.4795698924731182e-05,
+      "loss": 0.2354,
+      "step": 113720
+    },
+    {
+      "epoch": 733.74,
+      "learning_rate": 1.4792114695340504e-05,
+      "loss": 0.2366,
+      "step": 113730
+    },
+    {
+      "epoch": 733.8,
+      "learning_rate": 1.478853046594982e-05,
+      "loss": 0.2425,
+      "step": 113740
+    },
+    {
+      "epoch": 733.87,
+      "learning_rate": 1.4784946236559142e-05,
+      "loss": 0.2338,
+      "step": 113750
+    },
+    {
+      "epoch": 733.93,
+      "learning_rate": 1.4781362007168458e-05,
+      "loss": 0.2223,
+      "step": 113760
+    },
+    {
+      "epoch": 734.0,
+      "learning_rate": 1.477777777777778e-05,
+      "loss": 0.219,
+      "step": 113770
+    },
+    {
+      "epoch": 734.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.02963387966156,
+      "eval_runtime": 0.8707,
+      "eval_samples_per_second": 230.854,
+      "eval_steps_per_second": 8.04,
+      "step": 113770
+    },
+    {
+      "epoch": 734.06,
+      "learning_rate": 1.4774193548387096e-05,
+      "loss": 0.2517,
+      "step": 113780
+    },
+    {
+      "epoch": 734.13,
+      "learning_rate": 1.4770609318996418e-05,
+      "loss": 0.2532,
+      "step": 113790
+    },
+    {
+      "epoch": 734.19,
+      "learning_rate": 1.4767025089605735e-05,
+      "loss": 0.1615,
+      "step": 113800
+    },
+    {
+      "epoch": 734.26,
+      "learning_rate": 1.4763440860215056e-05,
+      "loss": 0.1985,
+      "step": 113810
+    },
+    {
+      "epoch": 734.32,
+      "learning_rate": 1.4759856630824373e-05,
+      "loss": 0.2186,
+      "step": 113820
+    },
+    {
+      "epoch": 734.39,
+      "learning_rate": 1.4756272401433694e-05,
+      "loss": 0.2369,
+      "step": 113830
+    },
+    {
+      "epoch": 734.45,
+      "learning_rate": 1.475268817204301e-05,
+      "loss": 0.1717,
+      "step": 113840
+    },
+    {
+      "epoch": 734.51,
+      "learning_rate": 1.474910394265233e-05,
+      "loss": 0.2119,
+      "step": 113850
+    },
+    {
+      "epoch": 734.58,
+      "learning_rate": 1.4745519713261649e-05,
+      "loss": 0.213,
+      "step": 113860
+    },
+    {
+      "epoch": 734.64,
+      "learning_rate": 1.4741935483870969e-05,
+      "loss": 0.1769,
+      "step": 113870
+    },
+    {
+      "epoch": 734.71,
+      "learning_rate": 1.4738351254480287e-05,
+      "loss": 0.2136,
+      "step": 113880
+    },
+    {
+      "epoch": 734.77,
+      "learning_rate": 1.4734767025089607e-05,
+      "loss": 0.219,
+      "step": 113890
+    },
+    {
+      "epoch": 734.84,
+      "learning_rate": 1.4731182795698925e-05,
+      "loss": 0.1671,
+      "step": 113900
+    },
+    {
+      "epoch": 734.9,
+      "learning_rate": 1.4727598566308245e-05,
+      "loss": 0.2331,
+      "step": 113910
+    },
+    {
+      "epoch": 734.96,
+      "learning_rate": 1.4724014336917563e-05,
+      "loss": 0.2056,
+      "step": 113920
+    },
+    {
+      "epoch": 735.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 1.025156855583191,
+      "eval_runtime": 0.8822,
+      "eval_samples_per_second": 227.849,
+      "eval_steps_per_second": 7.935,
+      "step": 113925
+    },
+    {
+      "epoch": 735.03,
+      "learning_rate": 1.4720430107526883e-05,
+      "loss": 0.1842,
+      "step": 113930
+    },
+    {
+      "epoch": 735.1,
+      "learning_rate": 1.4716845878136201e-05,
+      "loss": 0.2237,
+      "step": 113940
+    },
+    {
+      "epoch": 735.16,
+      "learning_rate": 1.4713261648745521e-05,
+      "loss": 0.2313,
+      "step": 113950
+    },
+    {
+      "epoch": 735.23,
+      "learning_rate": 1.470967741935484e-05,
+      "loss": 0.223,
+      "step": 113960
+    },
+    {
+      "epoch": 735.29,
+      "learning_rate": 1.470609318996416e-05,
+      "loss": 0.2328,
+      "step": 113970
+    },
+    {
+      "epoch": 735.35,
+      "learning_rate": 1.4702508960573477e-05,
+      "loss": 0.1886,
+      "step": 113980
+    },
+    {
+      "epoch": 735.42,
+      "learning_rate": 1.4698924731182797e-05,
+      "loss": 0.2262,
+      "step": 113990
+    },
+    {
+      "epoch": 735.48,
+      "learning_rate": 1.4695340501792116e-05,
+      "loss": 0.2307,
+      "step": 114000
+    },
+    {
+      "epoch": 735.55,
+      "learning_rate": 1.4691756272401435e-05,
+      "loss": 0.2331,
+      "step": 114010
+    },
+    {
+      "epoch": 735.61,
+      "learning_rate": 1.4688172043010754e-05,
+      "loss": 0.1793,
+      "step": 114020
+    },
+    {
+      "epoch": 735.68,
+      "learning_rate": 1.4684587813620074e-05,
+      "loss": 0.2019,
+      "step": 114030
+    },
+    {
+      "epoch": 735.74,
+      "learning_rate": 1.4681003584229392e-05,
+      "loss": 0.2056,
+      "step": 114040
+    },
+    {
+      "epoch": 735.8,
+      "learning_rate": 1.4677419354838708e-05,
+      "loss": 0.2054,
+      "step": 114050
+    },
+    {
+      "epoch": 735.87,
+      "learning_rate": 1.467383512544803e-05,
+      "loss": 0.2603,
+      "step": 114060
+    },
+    {
+      "epoch": 735.93,
+      "learning_rate": 1.4670250896057346e-05,
+      "loss": 0.2395,
+      "step": 114070
+    },
+    {
+      "epoch": 736.0,
+      "learning_rate": 1.4666666666666668e-05,
+      "loss": 0.2367,
+      "step": 114080
+    },
+    {
+      "epoch": 736.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.013355016708374,
+      "eval_runtime": 0.9088,
+      "eval_samples_per_second": 221.182,
+      "eval_steps_per_second": 7.703,
+      "step": 114080
+    },
+    {
+      "epoch": 736.06,
+      "learning_rate": 1.4663082437275984e-05,
+      "loss": 0.2172,
+      "step": 114090
+    },
+    {
+      "epoch": 736.13,
+      "learning_rate": 1.4659498207885306e-05,
+      "loss": 0.1911,
+      "step": 114100
+    },
+    {
+      "epoch": 736.19,
+      "learning_rate": 1.4655913978494623e-05,
+      "loss": 0.2057,
+      "step": 114110
+    },
+    {
+      "epoch": 736.26,
+      "learning_rate": 1.4652329749103944e-05,
+      "loss": 0.2002,
+      "step": 114120
+    },
+    {
+      "epoch": 736.32,
+      "learning_rate": 1.464874551971326e-05,
+      "loss": 0.1852,
+      "step": 114130
+    },
+    {
+      "epoch": 736.39,
+      "learning_rate": 1.4645161290322582e-05,
+      "loss": 0.2164,
+      "step": 114140
+    },
+    {
+      "epoch": 736.45,
+      "learning_rate": 1.4641577060931899e-05,
+      "loss": 0.2227,
+      "step": 114150
+    },
+    {
+      "epoch": 736.51,
+      "learning_rate": 1.463799283154122e-05,
+      "loss": 0.2331,
+      "step": 114160
+    },
+    {
+      "epoch": 736.58,
+      "learning_rate": 1.4634408602150537e-05,
+      "loss": 0.2214,
+      "step": 114170
+    },
+    {
+      "epoch": 736.64,
+      "learning_rate": 1.4630824372759857e-05,
+      "loss": 0.2173,
+      "step": 114180
+    },
+    {
+      "epoch": 736.71,
+      "learning_rate": 1.4627240143369175e-05,
+      "loss": 0.2054,
+      "step": 114190
+    },
+    {
+      "epoch": 736.77,
+      "learning_rate": 1.4623655913978495e-05,
+      "loss": 0.2368,
+      "step": 114200
+    },
+    {
+      "epoch": 736.84,
+      "learning_rate": 1.4620071684587813e-05,
+      "loss": 0.2223,
+      "step": 114210
+    },
+    {
+      "epoch": 736.9,
+      "learning_rate": 1.4616487455197133e-05,
+      "loss": 0.2345,
+      "step": 114220
+    },
+    {
+      "epoch": 736.96,
+      "learning_rate": 1.4612903225806451e-05,
+      "loss": 0.1876,
+      "step": 114230
+    },
+    {
+      "epoch": 737.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 1.0350953340530396,
+      "eval_runtime": 0.8738,
+      "eval_samples_per_second": 230.029,
+      "eval_steps_per_second": 8.011,
+      "step": 114235
+    },
+    {
+      "epoch": 737.03,
+      "learning_rate": 1.4609318996415771e-05,
+      "loss": 0.2299,
+      "step": 114240
+    },
+    {
+      "epoch": 737.1,
+      "learning_rate": 1.460573476702509e-05,
+      "loss": 0.1919,
+      "step": 114250
+    },
+    {
+      "epoch": 737.16,
+      "learning_rate": 1.460215053763441e-05,
+      "loss": 0.1995,
+      "step": 114260
+    },
+    {
+      "epoch": 737.23,
+      "learning_rate": 1.4598566308243727e-05,
+      "loss": 0.2427,
+      "step": 114270
+    },
+    {
+      "epoch": 737.29,
+      "learning_rate": 1.4594982078853047e-05,
+      "loss": 0.2327,
+      "step": 114280
+    },
+    {
+      "epoch": 737.35,
+      "learning_rate": 1.4591397849462366e-05,
+      "loss": 0.2358,
+      "step": 114290
+    },
+    {
+      "epoch": 737.42,
+      "learning_rate": 1.4587813620071685e-05,
+      "loss": 0.2092,
+      "step": 114300
+    },
+    {
+      "epoch": 737.48,
+      "learning_rate": 1.4584229390681004e-05,
+      "loss": 0.2317,
+      "step": 114310
+    },
+    {
+      "epoch": 737.55,
+      "learning_rate": 1.4580645161290324e-05,
+      "loss": 0.2383,
+      "step": 114320
+    },
+    {
+      "epoch": 737.61,
+      "learning_rate": 1.4577060931899642e-05,
+      "loss": 0.2229,
+      "step": 114330
+    },
+    {
+      "epoch": 737.68,
+      "learning_rate": 1.4573476702508962e-05,
+      "loss": 0.179,
+      "step": 114340
+    },
+    {
+      "epoch": 737.74,
+      "learning_rate": 1.456989247311828e-05,
+      "loss": 0.2149,
+      "step": 114350
+    },
+    {
+      "epoch": 737.8,
+      "learning_rate": 1.45663082437276e-05,
+      "loss": 0.2152,
+      "step": 114360
+    },
+    {
+      "epoch": 737.87,
+      "learning_rate": 1.4562724014336918e-05,
+      "loss": 0.1939,
+      "step": 114370
+    },
+    {
+      "epoch": 737.93,
+      "learning_rate": 1.4559139784946238e-05,
+      "loss": 0.2259,
+      "step": 114380
+    },
+    {
+      "epoch": 738.0,
+      "learning_rate": 1.4555555555555556e-05,
+      "loss": 0.1978,
+      "step": 114390
+    },
+    {
+      "epoch": 738.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.961627185344696,
+      "eval_runtime": 0.8619,
+      "eval_samples_per_second": 233.195,
+      "eval_steps_per_second": 8.121,
+      "step": 114390
+    },
+    {
+      "epoch": 738.06,
+      "learning_rate": 1.4551971326164876e-05,
+      "loss": 0.2018,
+      "step": 114400
+    },
+    {
+      "epoch": 738.13,
+      "learning_rate": 1.4548387096774194e-05,
+      "loss": 0.1936,
+      "step": 114410
+    },
+    {
+      "epoch": 738.19,
+      "learning_rate": 1.4544802867383514e-05,
+      "loss": 0.229,
+      "step": 114420
+    },
+    {
+      "epoch": 738.26,
+      "learning_rate": 1.4541218637992832e-05,
+      "loss": 0.2541,
+      "step": 114430
+    },
+    {
+      "epoch": 738.32,
+      "learning_rate": 1.4537634408602152e-05,
+      "loss": 0.2486,
+      "step": 114440
+    },
+    {
+      "epoch": 738.39,
+      "learning_rate": 1.453405017921147e-05,
+      "loss": 0.2742,
+      "step": 114450
+    },
+    {
+      "epoch": 738.45,
+      "learning_rate": 1.453046594982079e-05,
+      "loss": 0.2258,
+      "step": 114460
+    },
+    {
+      "epoch": 738.51,
+      "learning_rate": 1.4526881720430108e-05,
+      "loss": 0.2425,
+      "step": 114470
+    },
+    {
+      "epoch": 738.58,
+      "learning_rate": 1.4523297491039428e-05,
+      "loss": 0.2207,
+      "step": 114480
+    },
+    {
+      "epoch": 738.64,
+      "learning_rate": 1.4519713261648747e-05,
+      "loss": 0.2099,
+      "step": 114490
+    },
+    {
+      "epoch": 738.71,
+      "learning_rate": 1.4516129032258066e-05,
+      "loss": 0.2039,
+      "step": 114500
+    },
+    {
+      "epoch": 738.77,
+      "learning_rate": 1.4512544802867385e-05,
+      "loss": 0.2391,
+      "step": 114510
+    },
+    {
+      "epoch": 738.84,
+      "learning_rate": 1.4508960573476705e-05,
+      "loss": 0.272,
+      "step": 114520
+    },
+    {
+      "epoch": 738.9,
+      "learning_rate": 1.4505376344086021e-05,
+      "loss": 0.221,
+      "step": 114530
+    },
+    {
+      "epoch": 738.96,
+      "learning_rate": 1.4501792114695343e-05,
+      "loss": 0.269,
+      "step": 114540
+    },
+    {
+      "epoch": 739.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.995594322681427,
+      "eval_runtime": 0.8904,
+      "eval_samples_per_second": 225.739,
+      "eval_steps_per_second": 7.862,
+      "step": 114545
+    },
+    {
+      "epoch": 739.03,
+      "learning_rate": 1.449820788530466e-05,
+      "loss": 0.2396,
+      "step": 114550
+    },
+    {
+      "epoch": 739.1,
+      "learning_rate": 1.449462365591398e-05,
+      "loss": 0.2494,
+      "step": 114560
+    },
+    {
+      "epoch": 739.16,
+      "learning_rate": 1.4491039426523297e-05,
+      "loss": 0.236,
+      "step": 114570
+    },
+    {
+      "epoch": 739.23,
+      "learning_rate": 1.4487455197132619e-05,
+      "loss": 0.2274,
+      "step": 114580
+    },
+    {
+      "epoch": 739.29,
+      "learning_rate": 1.4483870967741935e-05,
+      "loss": 0.207,
+      "step": 114590
+    },
+    {
+      "epoch": 739.35,
+      "learning_rate": 1.4480286738351257e-05,
+      "loss": 0.2223,
+      "step": 114600
+    },
+    {
+      "epoch": 739.42,
+      "learning_rate": 1.4476702508960573e-05,
+      "loss": 0.2115,
+      "step": 114610
+    },
+    {
+      "epoch": 739.48,
+      "learning_rate": 1.4473118279569895e-05,
+      "loss": 0.2165,
+      "step": 114620
+    },
+    {
+      "epoch": 739.55,
+      "learning_rate": 1.4469534050179212e-05,
+      "loss": 0.2333,
+      "step": 114630
+    },
+    {
+      "epoch": 739.61,
+      "learning_rate": 1.4465949820788533e-05,
+      "loss": 0.2182,
+      "step": 114640
+    },
+    {
+      "epoch": 739.68,
+      "learning_rate": 1.446236559139785e-05,
+      "loss": 0.2297,
+      "step": 114650
+    },
+    {
+      "epoch": 739.74,
+      "learning_rate": 1.4458781362007171e-05,
+      "loss": 0.2492,
+      "step": 114660
+    },
+    {
+      "epoch": 739.8,
+      "learning_rate": 1.4455197132616488e-05,
+      "loss": 0.2153,
+      "step": 114670
+    },
+    {
+      "epoch": 739.87,
+      "learning_rate": 1.4451612903225808e-05,
+      "loss": 0.2366,
+      "step": 114680
+    },
+    {
+      "epoch": 739.93,
+      "learning_rate": 1.4448028673835126e-05,
+      "loss": 0.1666,
+      "step": 114690
+    },
+    {
+      "epoch": 740.0,
+      "learning_rate": 1.4444444444444444e-05,
+      "loss": 0.2334,
+      "step": 114700
+    },
+    {
+      "epoch": 740.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9390358924865723,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.137,
+      "eval_steps_per_second": 8.154,
+      "step": 114700
+    },
+    {
+      "epoch": 740.06,
+      "learning_rate": 1.4440860215053764e-05,
+      "loss": 0.2225,
+      "step": 114710
+    },
+    {
+      "epoch": 740.13,
+      "learning_rate": 1.4437275985663082e-05,
+      "loss": 0.2163,
+      "step": 114720
+    },
+    {
+      "epoch": 740.19,
+      "learning_rate": 1.4433691756272402e-05,
+      "loss": 0.2169,
+      "step": 114730
+    },
+    {
+      "epoch": 740.26,
+      "learning_rate": 1.443010752688172e-05,
+      "loss": 0.1736,
+      "step": 114740
+    },
+    {
+      "epoch": 740.32,
+      "learning_rate": 1.442652329749104e-05,
+      "loss": 0.2046,
+      "step": 114750
+    },
+    {
+      "epoch": 740.39,
+      "learning_rate": 1.4422939068100358e-05,
+      "loss": 0.2147,
+      "step": 114760
+    },
+    {
+      "epoch": 740.45,
+      "learning_rate": 1.4419354838709678e-05,
+      "loss": 0.2467,
+      "step": 114770
+    },
+    {
+      "epoch": 740.51,
+      "learning_rate": 1.4415770609318997e-05,
+      "loss": 0.2228,
+      "step": 114780
+    },
+    {
+      "epoch": 740.58,
+      "learning_rate": 1.4412186379928316e-05,
+      "loss": 0.2165,
+      "step": 114790
+    },
+    {
+      "epoch": 740.64,
+      "learning_rate": 1.4408602150537635e-05,
+      "loss": 0.1514,
+      "step": 114800
+    },
+    {
+      "epoch": 740.71,
+      "learning_rate": 1.4405017921146955e-05,
+      "loss": 0.2096,
+      "step": 114810
+    },
+    {
+      "epoch": 740.77,
+      "learning_rate": 1.4401433691756273e-05,
+      "loss": 0.2394,
+      "step": 114820
+    },
+    {
+      "epoch": 740.84,
+      "learning_rate": 1.4397849462365593e-05,
+      "loss": 0.1836,
+      "step": 114830
+    },
+    {
+      "epoch": 740.9,
+      "learning_rate": 1.4394265232974911e-05,
+      "loss": 0.1978,
+      "step": 114840
+    },
+    {
+      "epoch": 740.96,
+      "learning_rate": 1.439068100358423e-05,
+      "loss": 0.2216,
+      "step": 114850
+    },
+    {
+      "epoch": 741.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.950566828250885,
+      "eval_runtime": 0.9006,
+      "eval_samples_per_second": 223.177,
+      "eval_steps_per_second": 7.772,
+      "step": 114855
+    },
+    {
+      "epoch": 741.03,
+      "learning_rate": 1.4387096774193547e-05,
+      "loss": 0.2386,
+      "step": 114860
+    },
+    {
+      "epoch": 741.1,
+      "learning_rate": 1.4383512544802869e-05,
+      "loss": 0.211,
+      "step": 114870
+    },
+    {
+      "epoch": 741.16,
+      "learning_rate": 1.4379928315412185e-05,
+      "loss": 0.2087,
+      "step": 114880
+    },
+    {
+      "epoch": 741.23,
+      "learning_rate": 1.4376344086021507e-05,
+      "loss": 0.2495,
+      "step": 114890
+    },
+    {
+      "epoch": 741.29,
+      "learning_rate": 1.4372759856630823e-05,
+      "loss": 0.2176,
+      "step": 114900
+    },
+    {
+      "epoch": 741.35,
+      "learning_rate": 1.4369175627240145e-05,
+      "loss": 0.2613,
+      "step": 114910
+    },
+    {
+      "epoch": 741.42,
+      "learning_rate": 1.4365591397849462e-05,
+      "loss": 0.2293,
+      "step": 114920
+    },
+    {
+      "epoch": 741.48,
+      "learning_rate": 1.4362007168458783e-05,
+      "loss": 0.1949,
+      "step": 114930
+    },
+    {
+      "epoch": 741.55,
+      "learning_rate": 1.43584229390681e-05,
+      "loss": 0.2193,
+      "step": 114940
+    },
+    {
+      "epoch": 741.61,
+      "learning_rate": 1.4354838709677421e-05,
+      "loss": 0.1934,
+      "step": 114950
+    },
+    {
+      "epoch": 741.68,
+      "learning_rate": 1.4351254480286738e-05,
+      "loss": 0.2773,
+      "step": 114960
+    },
+    {
+      "epoch": 741.74,
+      "learning_rate": 1.434767025089606e-05,
+      "loss": 0.2378,
+      "step": 114970
+    },
+    {
+      "epoch": 741.8,
+      "learning_rate": 1.4344086021505376e-05,
+      "loss": 0.1704,
+      "step": 114980
+    },
+    {
+      "epoch": 741.87,
+      "learning_rate": 1.4340501792114697e-05,
+      "loss": 0.2711,
+      "step": 114990
+    },
+    {
+      "epoch": 741.93,
+      "learning_rate": 1.4336917562724014e-05,
+      "loss": 0.1988,
+      "step": 115000
+    },
+    {
+      "epoch": 742.0,
+      "learning_rate": 1.4333333333333334e-05,
+      "loss": 0.2789,
+      "step": 115010
+    },
+    {
+      "epoch": 742.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9997456669807434,
+      "eval_runtime": 0.8677,
+      "eval_samples_per_second": 231.642,
+      "eval_steps_per_second": 8.067,
+      "step": 115010
+    },
+    {
+      "epoch": 742.06,
+      "learning_rate": 1.4329749103942652e-05,
+      "loss": 0.2757,
+      "step": 115020
+    },
+    {
+      "epoch": 742.13,
+      "learning_rate": 1.4326164874551972e-05,
+      "loss": 0.241,
+      "step": 115030
+    },
+    {
+      "epoch": 742.19,
+      "learning_rate": 1.432258064516129e-05,
+      "loss": 0.2434,
+      "step": 115040
+    },
+    {
+      "epoch": 742.26,
+      "learning_rate": 1.431899641577061e-05,
+      "loss": 0.1737,
+      "step": 115050
+    },
+    {
+      "epoch": 742.32,
+      "learning_rate": 1.4315412186379928e-05,
+      "loss": 0.2202,
+      "step": 115060
+    },
+    {
+      "epoch": 742.39,
+      "learning_rate": 1.4311827956989248e-05,
+      "loss": 0.2119,
+      "step": 115070
+    },
+    {
+      "epoch": 742.45,
+      "learning_rate": 1.4308243727598566e-05,
+      "loss": 0.1719,
+      "step": 115080
+    },
+    {
+      "epoch": 742.51,
+      "learning_rate": 1.4304659498207886e-05,
+      "loss": 0.185,
+      "step": 115090
+    },
+    {
+      "epoch": 742.58,
+      "learning_rate": 1.4301075268817205e-05,
+      "loss": 0.2568,
+      "step": 115100
+    },
+    {
+      "epoch": 742.64,
+      "learning_rate": 1.4297491039426524e-05,
+      "loss": 0.2133,
+      "step": 115110
+    },
+    {
+      "epoch": 742.71,
+      "learning_rate": 1.4293906810035843e-05,
+      "loss": 0.224,
+      "step": 115120
+    },
+    {
+      "epoch": 742.77,
+      "learning_rate": 1.4290322580645163e-05,
+      "loss": 0.2049,
+      "step": 115130
+    },
+    {
+      "epoch": 742.84,
+      "learning_rate": 1.428673835125448e-05,
+      "loss": 0.2286,
+      "step": 115140
+    },
+    {
+      "epoch": 742.9,
+      "learning_rate": 1.42831541218638e-05,
+      "loss": 0.2375,
+      "step": 115150
+    },
+    {
+      "epoch": 742.96,
+      "learning_rate": 1.4279569892473119e-05,
+      "loss": 0.2644,
+      "step": 115160
+    },
+    {
+      "epoch": 743.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9739050269126892,
+      "eval_runtime": 0.8717,
+      "eval_samples_per_second": 230.585,
+      "eval_steps_per_second": 8.03,
+      "step": 115165
+    },
+    {
+      "epoch": 743.03,
+      "learning_rate": 1.4275985663082439e-05,
+      "loss": 0.2192,
+      "step": 115170
+    },
+    {
+      "epoch": 743.1,
+      "learning_rate": 1.4272401433691757e-05,
+      "loss": 0.246,
+      "step": 115180
+    },
+    {
+      "epoch": 743.16,
+      "learning_rate": 1.4268817204301077e-05,
+      "loss": 0.2539,
+      "step": 115190
+    },
+    {
+      "epoch": 743.23,
+      "learning_rate": 1.4265232974910395e-05,
+      "loss": 0.2011,
+      "step": 115200
+    },
+    {
+      "epoch": 743.29,
+      "learning_rate": 1.4261648745519715e-05,
+      "loss": 0.166,
+      "step": 115210
+    },
+    {
+      "epoch": 743.35,
+      "learning_rate": 1.4258064516129033e-05,
+      "loss": 0.2464,
+      "step": 115220
+    },
+    {
+      "epoch": 743.42,
+      "learning_rate": 1.4254480286738353e-05,
+      "loss": 0.2409,
+      "step": 115230
+    },
+    {
+      "epoch": 743.48,
+      "learning_rate": 1.4250896057347671e-05,
+      "loss": 0.2376,
+      "step": 115240
+    },
+    {
+      "epoch": 743.55,
+      "learning_rate": 1.4247311827956991e-05,
+      "loss": 0.234,
+      "step": 115250
+    },
+    {
+      "epoch": 743.61,
+      "learning_rate": 1.424372759856631e-05,
+      "loss": 0.2318,
+      "step": 115260
+    },
+    {
+      "epoch": 743.68,
+      "learning_rate": 1.424014336917563e-05,
+      "loss": 0.1934,
+      "step": 115270
+    },
+    {
+      "epoch": 743.74,
+      "learning_rate": 1.4236559139784947e-05,
+      "loss": 0.2715,
+      "step": 115280
+    },
+    {
+      "epoch": 743.8,
+      "learning_rate": 1.4232974910394267e-05,
+      "loss": 0.2382,
+      "step": 115290
+    },
+    {
+      "epoch": 743.87,
+      "learning_rate": 1.4229390681003586e-05,
+      "loss": 0.1959,
+      "step": 115300
+    },
+    {
+      "epoch": 743.93,
+      "learning_rate": 1.4225806451612905e-05,
+      "loss": 0.2026,
+      "step": 115310
+    },
+    {
+      "epoch": 744.0,
+      "learning_rate": 1.4222222222222224e-05,
+      "loss": 0.2607,
+      "step": 115320
+    },
+    {
+      "epoch": 744.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9859427809715271,
+      "eval_runtime": 0.8988,
+      "eval_samples_per_second": 223.632,
+      "eval_steps_per_second": 7.788,
+      "step": 115320
+    },
+    {
+      "epoch": 744.06,
+      "learning_rate": 1.421863799283154e-05,
+      "loss": 0.3061,
+      "step": 115330
+    },
+    {
+      "epoch": 744.13,
+      "learning_rate": 1.4215053763440862e-05,
+      "loss": 0.1856,
+      "step": 115340
+    },
+    {
+      "epoch": 744.19,
+      "learning_rate": 1.4211469534050178e-05,
+      "loss": 0.2001,
+      "step": 115350
+    },
+    {
+      "epoch": 744.26,
+      "learning_rate": 1.4207885304659498e-05,
+      "loss": 0.2178,
+      "step": 115360
+    },
+    {
+      "epoch": 744.32,
+      "learning_rate": 1.4204301075268816e-05,
+      "loss": 0.2106,
+      "step": 115370
+    },
+    {
+      "epoch": 744.39,
+      "learning_rate": 1.4200716845878136e-05,
+      "loss": 0.2007,
+      "step": 115380
+    },
+    {
+      "epoch": 744.45,
+      "learning_rate": 1.4197132616487454e-05,
+      "loss": 0.23,
+      "step": 115390
+    },
+    {
+      "epoch": 744.51,
+      "learning_rate": 1.4193548387096774e-05,
+      "loss": 0.2296,
+      "step": 115400
+    },
+    {
+      "epoch": 744.58,
+      "learning_rate": 1.4189964157706093e-05,
+      "loss": 0.2182,
+      "step": 115410
+    },
+    {
+      "epoch": 744.64,
+      "learning_rate": 1.4186379928315412e-05,
+      "loss": 0.2112,
+      "step": 115420
+    },
+    {
+      "epoch": 744.71,
+      "learning_rate": 1.418279569892473e-05,
+      "loss": 0.2274,
+      "step": 115430
+    },
+    {
+      "epoch": 744.77,
+      "learning_rate": 1.417921146953405e-05,
+      "loss": 0.2311,
+      "step": 115440
+    },
+    {
+      "epoch": 744.84,
+      "learning_rate": 1.4175627240143369e-05,
+      "loss": 0.2361,
+      "step": 115450
+    },
+    {
+      "epoch": 744.9,
+      "learning_rate": 1.4172043010752689e-05,
+      "loss": 0.2173,
+      "step": 115460
+    },
+    {
+      "epoch": 744.96,
+      "learning_rate": 1.4168458781362007e-05,
+      "loss": 0.2158,
+      "step": 115470
+    },
+    {
+      "epoch": 745.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9541528224945068,
+      "eval_runtime": 0.8832,
+      "eval_samples_per_second": 227.576,
+      "eval_steps_per_second": 7.926,
+      "step": 115475
+    },
+    {
+      "epoch": 745.03,
+      "learning_rate": 1.4164874551971327e-05,
+      "loss": 0.2389,
+      "step": 115480
+    },
+    {
+      "epoch": 745.1,
+      "learning_rate": 1.4161290322580645e-05,
+      "loss": 0.2469,
+      "step": 115490
+    },
+    {
+      "epoch": 745.16,
+      "learning_rate": 1.4157706093189965e-05,
+      "loss": 0.2551,
+      "step": 115500
+    },
+    {
+      "epoch": 745.23,
+      "learning_rate": 1.4154121863799283e-05,
+      "loss": 0.2241,
+      "step": 115510
+    },
+    {
+      "epoch": 745.29,
+      "learning_rate": 1.4150537634408603e-05,
+      "loss": 0.2157,
+      "step": 115520
+    },
+    {
+      "epoch": 745.35,
+      "learning_rate": 1.4146953405017921e-05,
+      "loss": 0.2136,
+      "step": 115530
+    },
+    {
+      "epoch": 745.42,
+      "learning_rate": 1.4143369175627241e-05,
+      "loss": 0.2244,
+      "step": 115540
+    },
+    {
+      "epoch": 745.48,
+      "learning_rate": 1.413978494623656e-05,
+      "loss": 0.2205,
+      "step": 115550
+    },
+    {
+      "epoch": 745.55,
+      "learning_rate": 1.413620071684588e-05,
+      "loss": 0.2061,
+      "step": 115560
+    },
+    {
+      "epoch": 745.61,
+      "learning_rate": 1.4132616487455197e-05,
+      "loss": 0.2045,
+      "step": 115570
+    },
+    {
+      "epoch": 745.68,
+      "learning_rate": 1.4129032258064517e-05,
+      "loss": 0.2491,
+      "step": 115580
+    },
+    {
+      "epoch": 745.74,
+      "learning_rate": 1.4125448028673836e-05,
+      "loss": 0.2063,
+      "step": 115590
+    },
+    {
+      "epoch": 745.8,
+      "learning_rate": 1.4121863799283155e-05,
+      "loss": 0.195,
+      "step": 115600
+    },
+    {
+      "epoch": 745.87,
+      "learning_rate": 1.4118279569892474e-05,
+      "loss": 0.2376,
+      "step": 115610
+    },
+    {
+      "epoch": 745.93,
+      "learning_rate": 1.4114695340501794e-05,
+      "loss": 0.2533,
+      "step": 115620
+    },
+    {
+      "epoch": 746.0,
+      "learning_rate": 1.4111111111111112e-05,
+      "loss": 0.2863,
+      "step": 115630
+    },
+    {
+      "epoch": 746.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9409778714179993,
+      "eval_runtime": 0.9256,
+      "eval_samples_per_second": 217.167,
+      "eval_steps_per_second": 7.563,
+      "step": 115630
+    },
+    {
+      "epoch": 746.06,
+      "learning_rate": 1.4107526881720432e-05,
+      "loss": 0.2381,
+      "step": 115640
+    },
+    {
+      "epoch": 746.13,
+      "learning_rate": 1.410394265232975e-05,
+      "loss": 0.2193,
+      "step": 115650
+    },
+    {
+      "epoch": 746.19,
+      "learning_rate": 1.410035842293907e-05,
+      "loss": 0.1889,
+      "step": 115660
+    },
+    {
+      "epoch": 746.26,
+      "learning_rate": 1.4096774193548388e-05,
+      "loss": 0.2303,
+      "step": 115670
+    },
+    {
+      "epoch": 746.32,
+      "learning_rate": 1.4093189964157708e-05,
+      "loss": 0.2503,
+      "step": 115680
+    },
+    {
+      "epoch": 746.39,
+      "learning_rate": 1.4089605734767024e-05,
+      "loss": 0.2086,
+      "step": 115690
+    },
+    {
+      "epoch": 746.45,
+      "learning_rate": 1.4086021505376346e-05,
+      "loss": 0.1921,
+      "step": 115700
+    },
+    {
+      "epoch": 746.51,
+      "learning_rate": 1.4082437275985662e-05,
+      "loss": 0.2362,
+      "step": 115710
+    },
+    {
+      "epoch": 746.58,
+      "learning_rate": 1.4078853046594984e-05,
+      "loss": 0.2137,
+      "step": 115720
+    },
+    {
+      "epoch": 746.64,
+      "learning_rate": 1.40752688172043e-05,
+      "loss": 0.2063,
+      "step": 115730
+    },
+    {
+      "epoch": 746.71,
+      "learning_rate": 1.4071684587813622e-05,
+      "loss": 0.2517,
+      "step": 115740
+    },
+    {
+      "epoch": 746.77,
+      "learning_rate": 1.4068100358422939e-05,
+      "loss": 0.2281,
+      "step": 115750
+    },
+    {
+      "epoch": 746.84,
+      "learning_rate": 1.406451612903226e-05,
+      "loss": 0.1832,
+      "step": 115760
+    },
+    {
+      "epoch": 746.9,
+      "learning_rate": 1.4060931899641577e-05,
+      "loss": 0.2818,
+      "step": 115770
+    },
+    {
+      "epoch": 746.96,
+      "learning_rate": 1.4057347670250898e-05,
+      "loss": 0.2133,
+      "step": 115780
+    },
+    {
+      "epoch": 747.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9558022022247314,
+      "eval_runtime": 0.9003,
+      "eval_samples_per_second": 223.262,
+      "eval_steps_per_second": 7.775,
+      "step": 115785
+    },
+    {
+      "epoch": 747.03,
+      "learning_rate": 1.4053763440860215e-05,
+      "loss": 0.215,
+      "step": 115790
+    },
+    {
+      "epoch": 747.1,
+      "learning_rate": 1.4050179211469536e-05,
+      "loss": 0.1917,
+      "step": 115800
+    },
+    {
+      "epoch": 747.16,
+      "learning_rate": 1.4046594982078853e-05,
+      "loss": 0.2146,
+      "step": 115810
+    },
+    {
+      "epoch": 747.23,
+      "learning_rate": 1.4043010752688175e-05,
+      "loss": 0.1975,
+      "step": 115820
+    },
+    {
+      "epoch": 747.29,
+      "learning_rate": 1.4039426523297491e-05,
+      "loss": 0.2362,
+      "step": 115830
+    },
+    {
+      "epoch": 747.35,
+      "learning_rate": 1.4035842293906813e-05,
+      "loss": 0.2621,
+      "step": 115840
+    },
+    {
+      "epoch": 747.42,
+      "learning_rate": 1.403225806451613e-05,
+      "loss": 0.1847,
+      "step": 115850
+    },
+    {
+      "epoch": 747.48,
+      "learning_rate": 1.4028673835125449e-05,
+      "loss": 0.2111,
+      "step": 115860
+    },
+    {
+      "epoch": 747.55,
+      "learning_rate": 1.4025089605734767e-05,
+      "loss": 0.213,
+      "step": 115870
+    },
+    {
+      "epoch": 747.61,
+      "learning_rate": 1.4021505376344087e-05,
+      "loss": 0.2009,
+      "step": 115880
+    },
+    {
+      "epoch": 747.68,
+      "learning_rate": 1.4017921146953405e-05,
+      "loss": 0.2142,
+      "step": 115890
+    },
+    {
+      "epoch": 747.74,
+      "learning_rate": 1.4014336917562725e-05,
+      "loss": 0.1713,
+      "step": 115900
+    },
+    {
+      "epoch": 747.8,
+      "learning_rate": 1.4010752688172043e-05,
+      "loss": 0.2174,
+      "step": 115910
+    },
+    {
+      "epoch": 747.87,
+      "learning_rate": 1.4007168458781363e-05,
+      "loss": 0.2488,
+      "step": 115920
+    },
+    {
+      "epoch": 747.93,
+      "learning_rate": 1.4003584229390682e-05,
+      "loss": 0.199,
+      "step": 115930
+    },
+    {
+      "epoch": 748.0,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.2201,
+      "step": 115940
+    },
+    {
+      "epoch": 748.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9374572038650513,
+      "eval_runtime": 0.8709,
+      "eval_samples_per_second": 230.806,
+      "eval_steps_per_second": 8.038,
+      "step": 115940
+    },
+    {
+      "epoch": 748.06,
+      "learning_rate": 1.399641577060932e-05,
+      "loss": 0.2285,
+      "step": 115950
+    },
+    {
+      "epoch": 748.13,
+      "learning_rate": 1.399283154121864e-05,
+      "loss": 0.2314,
+      "step": 115960
+    },
+    {
+      "epoch": 748.19,
+      "learning_rate": 1.3989247311827958e-05,
+      "loss": 0.2054,
+      "step": 115970
+    },
+    {
+      "epoch": 748.26,
+      "learning_rate": 1.3985663082437276e-05,
+      "loss": 0.204,
+      "step": 115980
+    },
+    {
+      "epoch": 748.32,
+      "learning_rate": 1.3982078853046596e-05,
+      "loss": 0.216,
+      "step": 115990
+    },
+    {
+      "epoch": 748.39,
+      "learning_rate": 1.3978494623655914e-05,
+      "loss": 0.2056,
+      "step": 116000
+    },
+    {
+      "epoch": 748.45,
+      "learning_rate": 1.3974910394265234e-05,
+      "loss": 0.2096,
+      "step": 116010
+    },
+    {
+      "epoch": 748.51,
+      "learning_rate": 1.3971326164874552e-05,
+      "loss": 0.2049,
+      "step": 116020
+    },
+    {
+      "epoch": 748.58,
+      "learning_rate": 1.3967741935483872e-05,
+      "loss": 0.2164,
+      "step": 116030
+    },
+    {
+      "epoch": 748.64,
+      "learning_rate": 1.3964157706093189e-05,
+      "loss": 0.22,
+      "step": 116040
+    },
+    {
+      "epoch": 748.71,
+      "learning_rate": 1.396057347670251e-05,
+      "loss": 0.1913,
+      "step": 116050
+    },
+    {
+      "epoch": 748.77,
+      "learning_rate": 1.3956989247311827e-05,
+      "loss": 0.1992,
+      "step": 116060
+    },
+    {
+      "epoch": 748.84,
+      "learning_rate": 1.3953405017921148e-05,
+      "loss": 0.2456,
+      "step": 116070
+    },
+    {
+      "epoch": 748.9,
+      "learning_rate": 1.3949820788530465e-05,
+      "loss": 0.2132,
+      "step": 116080
+    },
+    {
+      "epoch": 748.96,
+      "learning_rate": 1.3946236559139786e-05,
+      "loss": 0.2338,
+      "step": 116090
+    },
+    {
+      "epoch": 749.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9793904423713684,
+      "eval_runtime": 0.8711,
+      "eval_samples_per_second": 230.755,
+      "eval_steps_per_second": 8.036,
+      "step": 116095
+    },
+    {
+      "epoch": 749.03,
+      "learning_rate": 1.3942652329749103e-05,
+      "loss": 0.258,
+      "step": 116100
+    },
+    {
+      "epoch": 749.1,
+      "learning_rate": 1.3939068100358425e-05,
+      "loss": 0.2259,
+      "step": 116110
+    },
+    {
+      "epoch": 749.16,
+      "learning_rate": 1.3935483870967741e-05,
+      "loss": 0.2168,
+      "step": 116120
+    },
+    {
+      "epoch": 749.23,
+      "learning_rate": 1.3931899641577063e-05,
+      "loss": 0.2366,
+      "step": 116130
+    },
+    {
+      "epoch": 749.29,
+      "learning_rate": 1.3928315412186379e-05,
+      "loss": 0.1973,
+      "step": 116140
+    },
+    {
+      "epoch": 749.35,
+      "learning_rate": 1.39247311827957e-05,
+      "loss": 0.239,
+      "step": 116150
+    },
+    {
+      "epoch": 749.42,
+      "learning_rate": 1.3921146953405017e-05,
+      "loss": 0.2304,
+      "step": 116160
+    },
+    {
+      "epoch": 749.48,
+      "learning_rate": 1.3917562724014339e-05,
+      "loss": 0.2108,
+      "step": 116170
+    },
+    {
+      "epoch": 749.55,
+      "learning_rate": 1.3913978494623655e-05,
+      "loss": 0.1966,
+      "step": 116180
+    },
+    {
+      "epoch": 749.61,
+      "learning_rate": 1.3910394265232975e-05,
+      "loss": 0.2619,
+      "step": 116190
+    },
+    {
+      "epoch": 749.68,
+      "learning_rate": 1.3906810035842293e-05,
+      "loss": 0.2049,
+      "step": 116200
+    },
+    {
+      "epoch": 749.74,
+      "learning_rate": 1.3903225806451613e-05,
+      "loss": 0.2863,
+      "step": 116210
+    },
+    {
+      "epoch": 749.8,
+      "learning_rate": 1.3899641577060932e-05,
+      "loss": 0.1821,
+      "step": 116220
+    },
+    {
+      "epoch": 749.87,
+      "learning_rate": 1.3896057347670251e-05,
+      "loss": 0.1961,
+      "step": 116230
+    },
+    {
+      "epoch": 749.93,
+      "learning_rate": 1.389247311827957e-05,
+      "loss": 0.2426,
+      "step": 116240
+    },
+    {
+      "epoch": 750.0,
+      "learning_rate": 1.388888888888889e-05,
+      "loss": 0.1997,
+      "step": 116250
+    },
+    {
+      "epoch": 750.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9624980092048645,
+      "eval_runtime": 0.8765,
+      "eval_samples_per_second": 229.326,
+      "eval_steps_per_second": 7.986,
+      "step": 116250
+    },
+    {
+      "epoch": 750.06,
+      "learning_rate": 1.3885304659498208e-05,
+      "loss": 0.23,
+      "step": 116260
+    },
+    {
+      "epoch": 750.13,
+      "learning_rate": 1.3881720430107528e-05,
+      "loss": 0.2474,
+      "step": 116270
+    },
+    {
+      "epoch": 750.19,
+      "learning_rate": 1.3878136200716846e-05,
+      "loss": 0.2852,
+      "step": 116280
+    },
+    {
+      "epoch": 750.26,
+      "learning_rate": 1.3874551971326166e-05,
+      "loss": 0.2581,
+      "step": 116290
+    },
+    {
+      "epoch": 750.32,
+      "learning_rate": 1.3870967741935484e-05,
+      "loss": 0.2561,
+      "step": 116300
+    },
+    {
+      "epoch": 750.39,
+      "learning_rate": 1.3867383512544804e-05,
+      "loss": 0.2409,
+      "step": 116310
+    },
+    {
+      "epoch": 750.45,
+      "learning_rate": 1.3863799283154122e-05,
+      "loss": 0.1532,
+      "step": 116320
+    },
+    {
+      "epoch": 750.51,
+      "learning_rate": 1.3860215053763442e-05,
+      "loss": 0.1907,
+      "step": 116330
+    },
+    {
+      "epoch": 750.58,
+      "learning_rate": 1.385663082437276e-05,
+      "loss": 0.1984,
+      "step": 116340
+    },
+    {
+      "epoch": 750.64,
+      "learning_rate": 1.385304659498208e-05,
+      "loss": 0.2382,
+      "step": 116350
+    },
+    {
+      "epoch": 750.71,
+      "learning_rate": 1.3849462365591398e-05,
+      "loss": 0.1909,
+      "step": 116360
+    },
+    {
+      "epoch": 750.77,
+      "learning_rate": 1.3845878136200718e-05,
+      "loss": 0.2452,
+      "step": 116370
+    },
+    {
+      "epoch": 750.84,
+      "learning_rate": 1.3842293906810036e-05,
+      "loss": 0.2181,
+      "step": 116380
+    },
+    {
+      "epoch": 750.9,
+      "learning_rate": 1.3838709677419356e-05,
+      "loss": 0.23,
+      "step": 116390
+    },
+    {
+      "epoch": 750.96,
+      "learning_rate": 1.3835125448028674e-05,
+      "loss": 0.2263,
+      "step": 116400
+    },
+    {
+      "epoch": 751.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9652277231216431,
+      "eval_runtime": 0.8789,
+      "eval_samples_per_second": 228.7,
+      "eval_steps_per_second": 7.965,
+      "step": 116405
+    },
+    {
+      "epoch": 751.03,
+      "learning_rate": 1.3831541218637994e-05,
+      "loss": 0.217,
+      "step": 116410
+    },
+    {
+      "epoch": 751.1,
+      "learning_rate": 1.3827956989247313e-05,
+      "loss": 0.1941,
+      "step": 116420
+    },
+    {
+      "epoch": 751.16,
+      "learning_rate": 1.3824372759856632e-05,
+      "loss": 0.2361,
+      "step": 116430
+    },
+    {
+      "epoch": 751.23,
+      "learning_rate": 1.382078853046595e-05,
+      "loss": 0.1677,
+      "step": 116440
+    },
+    {
+      "epoch": 751.29,
+      "learning_rate": 1.381720430107527e-05,
+      "loss": 0.2156,
+      "step": 116450
+    },
+    {
+      "epoch": 751.35,
+      "learning_rate": 1.3813620071684589e-05,
+      "loss": 0.2549,
+      "step": 116460
+    },
+    {
+      "epoch": 751.42,
+      "learning_rate": 1.3810035842293909e-05,
+      "loss": 0.2329,
+      "step": 116470
+    },
+    {
+      "epoch": 751.48,
+      "learning_rate": 1.3806451612903227e-05,
+      "loss": 0.2043,
+      "step": 116480
+    },
+    {
+      "epoch": 751.55,
+      "learning_rate": 1.3802867383512547e-05,
+      "loss": 0.2138,
+      "step": 116490
+    },
+    {
+      "epoch": 751.61,
+      "learning_rate": 1.3799283154121865e-05,
+      "loss": 0.2088,
+      "step": 116500
+    },
+    {
+      "epoch": 751.68,
+      "learning_rate": 1.3795698924731185e-05,
+      "loss": 0.1897,
+      "step": 116510
+    },
+    {
+      "epoch": 751.74,
+      "learning_rate": 1.3792114695340503e-05,
+      "loss": 0.2188,
+      "step": 116520
+    },
+    {
+      "epoch": 751.8,
+      "learning_rate": 1.3788530465949823e-05,
+      "loss": 0.1964,
+      "step": 116530
+    },
+    {
+      "epoch": 751.87,
+      "learning_rate": 1.378494623655914e-05,
+      "loss": 0.2092,
+      "step": 116540
+    },
+    {
+      "epoch": 751.93,
+      "learning_rate": 1.3781362007168461e-05,
+      "loss": 0.207,
+      "step": 116550
+    },
+    {
+      "epoch": 752.0,
+      "learning_rate": 1.3777777777777778e-05,
+      "loss": 0.1821,
+      "step": 116560
+    },
+    {
+      "epoch": 752.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9533801674842834,
+      "eval_runtime": 0.8949,
+      "eval_samples_per_second": 224.617,
+      "eval_steps_per_second": 7.822,
+      "step": 116560
+    },
+    {
+      "epoch": 752.06,
+      "learning_rate": 1.37741935483871e-05,
+      "loss": 0.2226,
+      "step": 116570
+    },
+    {
+      "epoch": 752.13,
+      "learning_rate": 1.3770609318996416e-05,
+      "loss": 0.238,
+      "step": 116580
+    },
+    {
+      "epoch": 752.19,
+      "learning_rate": 1.3767025089605737e-05,
+      "loss": 0.1909,
+      "step": 116590
+    },
+    {
+      "epoch": 752.26,
+      "learning_rate": 1.3763440860215054e-05,
+      "loss": 0.1985,
+      "step": 116600
+    },
+    {
+      "epoch": 752.32,
+      "learning_rate": 1.3759856630824372e-05,
+      "loss": 0.206,
+      "step": 116610
+    },
+    {
+      "epoch": 752.39,
+      "learning_rate": 1.3756272401433692e-05,
+      "loss": 0.2094,
+      "step": 116620
+    },
+    {
+      "epoch": 752.45,
+      "learning_rate": 1.375268817204301e-05,
+      "loss": 0.2638,
+      "step": 116630
+    },
+    {
+      "epoch": 752.51,
+      "learning_rate": 1.374910394265233e-05,
+      "loss": 0.2459,
+      "step": 116640
+    },
+    {
+      "epoch": 752.58,
+      "learning_rate": 1.3745519713261648e-05,
+      "loss": 0.2184,
+      "step": 116650
+    },
+    {
+      "epoch": 752.64,
+      "learning_rate": 1.3741935483870968e-05,
+      "loss": 0.2316,
+      "step": 116660
+    },
+    {
+      "epoch": 752.71,
+      "learning_rate": 1.3738351254480286e-05,
+      "loss": 0.1939,
+      "step": 116670
+    },
+    {
+      "epoch": 752.77,
+      "learning_rate": 1.3734767025089606e-05,
+      "loss": 0.2023,
+      "step": 116680
+    },
+    {
+      "epoch": 752.84,
+      "learning_rate": 1.3731182795698924e-05,
+      "loss": 0.2779,
+      "step": 116690
+    },
+    {
+      "epoch": 752.9,
+      "learning_rate": 1.3727598566308244e-05,
+      "loss": 0.2634,
+      "step": 116700
+    },
+    {
+      "epoch": 752.96,
+      "learning_rate": 1.3724014336917563e-05,
+      "loss": 0.1827,
+      "step": 116710
+    },
+    {
+      "epoch": 753.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9823763966560364,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.838,
+      "eval_steps_per_second": 8.074,
+      "step": 116715
+    },
+    {
+      "epoch": 753.03,
+      "learning_rate": 1.3720430107526882e-05,
+      "loss": 0.1727,
+      "step": 116720
+    },
+    {
+      "epoch": 753.1,
+      "learning_rate": 1.37168458781362e-05,
+      "loss": 0.2666,
+      "step": 116730
+    },
+    {
+      "epoch": 753.16,
+      "learning_rate": 1.371326164874552e-05,
+      "loss": 0.2298,
+      "step": 116740
+    },
+    {
+      "epoch": 753.23,
+      "learning_rate": 1.3709677419354839e-05,
+      "loss": 0.2568,
+      "step": 116750
+    },
+    {
+      "epoch": 753.29,
+      "learning_rate": 1.3706093189964159e-05,
+      "loss": 0.2029,
+      "step": 116760
+    },
+    {
+      "epoch": 753.35,
+      "learning_rate": 1.3702508960573477e-05,
+      "loss": 0.2156,
+      "step": 116770
+    },
+    {
+      "epoch": 753.42,
+      "learning_rate": 1.3698924731182797e-05,
+      "loss": 0.2506,
+      "step": 116780
+    },
+    {
+      "epoch": 753.48,
+      "learning_rate": 1.3695340501792115e-05,
+      "loss": 0.1861,
+      "step": 116790
+    },
+    {
+      "epoch": 753.55,
+      "learning_rate": 1.3691756272401435e-05,
+      "loss": 0.2448,
+      "step": 116800
+    },
+    {
+      "epoch": 753.61,
+      "learning_rate": 1.3688172043010753e-05,
+      "loss": 0.2148,
+      "step": 116810
+    },
+    {
+      "epoch": 753.68,
+      "learning_rate": 1.3684587813620073e-05,
+      "loss": 0.2455,
+      "step": 116820
+    },
+    {
+      "epoch": 753.74,
+      "learning_rate": 1.3681003584229391e-05,
+      "loss": 0.2248,
+      "step": 116830
+    },
+    {
+      "epoch": 753.8,
+      "learning_rate": 1.3677419354838711e-05,
+      "loss": 0.163,
+      "step": 116840
+    },
+    {
+      "epoch": 753.87,
+      "learning_rate": 1.367383512544803e-05,
+      "loss": 0.2068,
+      "step": 116850
+    },
+    {
+      "epoch": 753.93,
+      "learning_rate": 1.367025089605735e-05,
+      "loss": 0.2307,
+      "step": 116860
+    },
+    {
+      "epoch": 754.0,
+      "learning_rate": 1.3666666666666666e-05,
+      "loss": 0.2406,
+      "step": 116870
+    },
+    {
+      "epoch": 754.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.971406102180481,
+      "eval_runtime": 0.8988,
+      "eval_samples_per_second": 223.622,
+      "eval_steps_per_second": 7.788,
+      "step": 116870
+    },
+    {
+      "epoch": 754.06,
+      "learning_rate": 1.3663082437275987e-05,
+      "loss": 0.2128,
+      "step": 116880
+    },
+    {
+      "epoch": 754.13,
+      "learning_rate": 1.3659498207885304e-05,
+      "loss": 0.2224,
+      "step": 116890
+    },
+    {
+      "epoch": 754.19,
+      "learning_rate": 1.3655913978494625e-05,
+      "loss": 0.2567,
+      "step": 116900
+    },
+    {
+      "epoch": 754.26,
+      "learning_rate": 1.3652329749103942e-05,
+      "loss": 0.1995,
+      "step": 116910
+    },
+    {
+      "epoch": 754.32,
+      "learning_rate": 1.3648745519713263e-05,
+      "loss": 0.2476,
+      "step": 116920
+    },
+    {
+      "epoch": 754.39,
+      "learning_rate": 1.364516129032258e-05,
+      "loss": 0.2476,
+      "step": 116930
+    },
+    {
+      "epoch": 754.45,
+      "learning_rate": 1.3641577060931902e-05,
+      "loss": 0.2398,
+      "step": 116940
+    },
+    {
+      "epoch": 754.51,
+      "learning_rate": 1.3637992831541218e-05,
+      "loss": 0.1572,
+      "step": 116950
+    },
+    {
+      "epoch": 754.58,
+      "learning_rate": 1.363440860215054e-05,
+      "loss": 0.2308,
+      "step": 116960
+    },
+    {
+      "epoch": 754.64,
+      "learning_rate": 1.3630824372759856e-05,
+      "loss": 0.2037,
+      "step": 116970
+    },
+    {
+      "epoch": 754.71,
+      "learning_rate": 1.3627240143369178e-05,
+      "loss": 0.2213,
+      "step": 116980
+    },
+    {
+      "epoch": 754.77,
+      "learning_rate": 1.3623655913978494e-05,
+      "loss": 0.213,
+      "step": 116990
+    },
+    {
+      "epoch": 754.84,
+      "learning_rate": 1.3620071684587816e-05,
+      "loss": 0.2081,
+      "step": 117000
+    },
+    {
+      "epoch": 754.9,
+      "learning_rate": 1.3616487455197132e-05,
+      "loss": 0.2191,
+      "step": 117010
+    },
+    {
+      "epoch": 754.96,
+      "learning_rate": 1.3612903225806452e-05,
+      "loss": 0.1884,
+      "step": 117020
+    },
+    {
+      "epoch": 755.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0164926052093506,
+      "eval_runtime": 0.847,
+      "eval_samples_per_second": 237.316,
+      "eval_steps_per_second": 8.265,
+      "step": 117025
+    },
+    {
+      "epoch": 755.03,
+      "learning_rate": 1.360931899641577e-05,
+      "loss": 0.2799,
+      "step": 117030
+    },
+    {
+      "epoch": 755.1,
+      "learning_rate": 1.360573476702509e-05,
+      "loss": 0.1992,
+      "step": 117040
+    },
+    {
+      "epoch": 755.16,
+      "learning_rate": 1.3602150537634409e-05,
+      "loss": 0.2008,
+      "step": 117050
+    },
+    {
+      "epoch": 755.23,
+      "learning_rate": 1.3598566308243729e-05,
+      "loss": 0.2263,
+      "step": 117060
+    },
+    {
+      "epoch": 755.29,
+      "learning_rate": 1.3594982078853047e-05,
+      "loss": 0.2286,
+      "step": 117070
+    },
+    {
+      "epoch": 755.35,
+      "learning_rate": 1.3591397849462367e-05,
+      "loss": 0.1909,
+      "step": 117080
+    },
+    {
+      "epoch": 755.42,
+      "learning_rate": 1.3587813620071685e-05,
+      "loss": 0.1834,
+      "step": 117090
+    },
+    {
+      "epoch": 755.48,
+      "learning_rate": 1.3584229390681005e-05,
+      "loss": 0.2571,
+      "step": 117100
+    },
+    {
+      "epoch": 755.55,
+      "learning_rate": 1.3580645161290323e-05,
+      "loss": 0.209,
+      "step": 117110
+    },
+    {
+      "epoch": 755.61,
+      "learning_rate": 1.3577060931899643e-05,
+      "loss": 0.2235,
+      "step": 117120
+    },
+    {
+      "epoch": 755.68,
+      "learning_rate": 1.3573476702508961e-05,
+      "loss": 0.2464,
+      "step": 117130
+    },
+    {
+      "epoch": 755.74,
+      "learning_rate": 1.3569892473118281e-05,
+      "loss": 0.2277,
+      "step": 117140
+    },
+    {
+      "epoch": 755.8,
+      "learning_rate": 1.3566308243727599e-05,
+      "loss": 0.2036,
+      "step": 117150
+    },
+    {
+      "epoch": 755.87,
+      "learning_rate": 1.3562724014336919e-05,
+      "loss": 0.2203,
+      "step": 117160
+    },
+    {
+      "epoch": 755.93,
+      "learning_rate": 1.3559139784946237e-05,
+      "loss": 0.2318,
+      "step": 117170
+    },
+    {
+      "epoch": 756.0,
+      "learning_rate": 1.3555555555555557e-05,
+      "loss": 0.2282,
+      "step": 117180
+    },
+    {
+      "epoch": 756.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.975602388381958,
+      "eval_runtime": 0.9074,
+      "eval_samples_per_second": 221.51,
+      "eval_steps_per_second": 7.714,
+      "step": 117180
+    },
+    {
+      "epoch": 756.06,
+      "learning_rate": 1.3551971326164875e-05,
+      "loss": 0.2154,
+      "step": 117190
+    },
+    {
+      "epoch": 756.13,
+      "learning_rate": 1.3548387096774195e-05,
+      "loss": 0.2768,
+      "step": 117200
+    },
+    {
+      "epoch": 756.19,
+      "learning_rate": 1.3544802867383513e-05,
+      "loss": 0.218,
+      "step": 117210
+    },
+    {
+      "epoch": 756.26,
+      "learning_rate": 1.3541218637992833e-05,
+      "loss": 0.2158,
+      "step": 117220
+    },
+    {
+      "epoch": 756.32,
+      "learning_rate": 1.3537634408602152e-05,
+      "loss": 0.2344,
+      "step": 117230
+    },
+    {
+      "epoch": 756.39,
+      "learning_rate": 1.3534050179211471e-05,
+      "loss": 0.2444,
+      "step": 117240
+    },
+    {
+      "epoch": 756.45,
+      "learning_rate": 1.353046594982079e-05,
+      "loss": 0.2148,
+      "step": 117250
+    },
+    {
+      "epoch": 756.51,
+      "learning_rate": 1.3526881720430106e-05,
+      "loss": 0.2075,
+      "step": 117260
+    },
+    {
+      "epoch": 756.58,
+      "learning_rate": 1.3523297491039428e-05,
+      "loss": 0.2526,
+      "step": 117270
+    },
+    {
+      "epoch": 756.64,
+      "learning_rate": 1.3519713261648744e-05,
+      "loss": 0.1874,
+      "step": 117280
+    },
+    {
+      "epoch": 756.71,
+      "learning_rate": 1.3516129032258066e-05,
+      "loss": 0.2556,
+      "step": 117290
+    },
+    {
+      "epoch": 756.77,
+      "learning_rate": 1.3512544802867382e-05,
+      "loss": 0.221,
+      "step": 117300
+    },
+    {
+      "epoch": 756.84,
+      "learning_rate": 1.3508960573476704e-05,
+      "loss": 0.2437,
+      "step": 117310
+    },
+    {
+      "epoch": 756.9,
+      "learning_rate": 1.350537634408602e-05,
+      "loss": 0.2248,
+      "step": 117320
+    },
+    {
+      "epoch": 756.96,
+      "learning_rate": 1.3501792114695342e-05,
+      "loss": 0.2591,
+      "step": 117330
+    },
+    {
+      "epoch": 757.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 1.0033679008483887,
+      "eval_runtime": 0.8738,
+      "eval_samples_per_second": 230.025,
+      "eval_steps_per_second": 8.011,
+      "step": 117335
+    },
+    {
+      "epoch": 757.03,
+      "learning_rate": 1.3498207885304659e-05,
+      "loss": 0.1953,
+      "step": 117340
+    },
+    {
+      "epoch": 757.1,
+      "learning_rate": 1.349462365591398e-05,
+      "loss": 0.2146,
+      "step": 117350
+    },
+    {
+      "epoch": 757.16,
+      "learning_rate": 1.3491039426523297e-05,
+      "loss": 0.1883,
+      "step": 117360
+    },
+    {
+      "epoch": 757.23,
+      "learning_rate": 1.3487455197132617e-05,
+      "loss": 0.2184,
+      "step": 117370
+    },
+    {
+      "epoch": 757.29,
+      "learning_rate": 1.3483870967741935e-05,
+      "loss": 0.2066,
+      "step": 117380
+    },
+    {
+      "epoch": 757.35,
+      "learning_rate": 1.3480286738351255e-05,
+      "loss": 0.2088,
+      "step": 117390
+    },
+    {
+      "epoch": 757.42,
+      "learning_rate": 1.3476702508960573e-05,
+      "loss": 0.2346,
+      "step": 117400
+    },
+    {
+      "epoch": 757.48,
+      "learning_rate": 1.3473118279569893e-05,
+      "loss": 0.2557,
+      "step": 117410
+    },
+    {
+      "epoch": 757.55,
+      "learning_rate": 1.3469534050179211e-05,
+      "loss": 0.2134,
+      "step": 117420
+    },
+    {
+      "epoch": 757.61,
+      "learning_rate": 1.3465949820788531e-05,
+      "loss": 0.2367,
+      "step": 117430
+    },
+    {
+      "epoch": 757.68,
+      "learning_rate": 1.3462365591397849e-05,
+      "loss": 0.2596,
+      "step": 117440
+    },
+    {
+      "epoch": 757.74,
+      "learning_rate": 1.3458781362007169e-05,
+      "loss": 0.187,
+      "step": 117450
+    },
+    {
+      "epoch": 757.8,
+      "learning_rate": 1.3455197132616487e-05,
+      "loss": 0.2028,
+      "step": 117460
+    },
+    {
+      "epoch": 757.87,
+      "learning_rate": 1.3451612903225807e-05,
+      "loss": 0.2139,
+      "step": 117470
+    },
+    {
+      "epoch": 757.93,
+      "learning_rate": 1.3448028673835125e-05,
+      "loss": 0.2259,
+      "step": 117480
+    },
+    {
+      "epoch": 758.0,
+      "learning_rate": 1.3444444444444445e-05,
+      "loss": 0.1872,
+      "step": 117490
+    },
+    {
+      "epoch": 758.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9783754348754883,
+      "eval_runtime": 0.8841,
+      "eval_samples_per_second": 227.36,
+      "eval_steps_per_second": 7.918,
+      "step": 117490
+    },
+    {
+      "epoch": 758.06,
+      "learning_rate": 1.3440860215053763e-05,
+      "loss": 0.2664,
+      "step": 117500
+    },
+    {
+      "epoch": 758.13,
+      "learning_rate": 1.3437275985663083e-05,
+      "loss": 0.2324,
+      "step": 117510
+    },
+    {
+      "epoch": 758.19,
+      "learning_rate": 1.3433691756272402e-05,
+      "loss": 0.2224,
+      "step": 117520
+    },
+    {
+      "epoch": 758.26,
+      "learning_rate": 1.3430107526881721e-05,
+      "loss": 0.2681,
+      "step": 117530
+    },
+    {
+      "epoch": 758.32,
+      "learning_rate": 1.342652329749104e-05,
+      "loss": 0.2421,
+      "step": 117540
+    },
+    {
+      "epoch": 758.39,
+      "learning_rate": 1.342293906810036e-05,
+      "loss": 0.2361,
+      "step": 117550
+    },
+    {
+      "epoch": 758.45,
+      "learning_rate": 1.3419354838709678e-05,
+      "loss": 0.2252,
+      "step": 117560
+    },
+    {
+      "epoch": 758.51,
+      "learning_rate": 1.3415770609318998e-05,
+      "loss": 0.22,
+      "step": 117570
+    },
+    {
+      "epoch": 758.58,
+      "learning_rate": 1.3412186379928316e-05,
+      "loss": 0.1901,
+      "step": 117580
+    },
+    {
+      "epoch": 758.64,
+      "learning_rate": 1.3408602150537636e-05,
+      "loss": 0.259,
+      "step": 117590
+    },
+    {
+      "epoch": 758.71,
+      "learning_rate": 1.3405017921146954e-05,
+      "loss": 0.2402,
+      "step": 117600
+    },
+    {
+      "epoch": 758.77,
+      "learning_rate": 1.3401433691756274e-05,
+      "loss": 0.2329,
+      "step": 117610
+    },
+    {
+      "epoch": 758.84,
+      "learning_rate": 1.3397849462365592e-05,
+      "loss": 0.193,
+      "step": 117620
+    },
+    {
+      "epoch": 758.9,
+      "learning_rate": 1.3394265232974912e-05,
+      "loss": 0.2033,
+      "step": 117630
+    },
+    {
+      "epoch": 758.96,
+      "learning_rate": 1.339068100358423e-05,
+      "loss": 0.259,
+      "step": 117640
+    },
+    {
+      "epoch": 759.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9592739343643188,
+      "eval_runtime": 0.9084,
+      "eval_samples_per_second": 221.274,
+      "eval_steps_per_second": 7.706,
+      "step": 117645
+    },
+    {
+      "epoch": 759.03,
+      "learning_rate": 1.338709677419355e-05,
+      "loss": 0.217,
+      "step": 117650
+    },
+    {
+      "epoch": 759.1,
+      "learning_rate": 1.3383512544802868e-05,
+      "loss": 0.2038,
+      "step": 117660
+    },
+    {
+      "epoch": 759.16,
+      "learning_rate": 1.3379928315412188e-05,
+      "loss": 0.2346,
+      "step": 117670
+    },
+    {
+      "epoch": 759.23,
+      "learning_rate": 1.3376344086021506e-05,
+      "loss": 0.2114,
+      "step": 117680
+    },
+    {
+      "epoch": 759.29,
+      "learning_rate": 1.3372759856630826e-05,
+      "loss": 0.1803,
+      "step": 117690
+    },
+    {
+      "epoch": 759.35,
+      "learning_rate": 1.3369175627240143e-05,
+      "loss": 0.1896,
+      "step": 117700
+    },
+    {
+      "epoch": 759.42,
+      "learning_rate": 1.3365591397849464e-05,
+      "loss": 0.2287,
+      "step": 117710
+    },
+    {
+      "epoch": 759.48,
+      "learning_rate": 1.3362007168458781e-05,
+      "loss": 0.2635,
+      "step": 117720
+    },
+    {
+      "epoch": 759.55,
+      "learning_rate": 1.3358422939068102e-05,
+      "loss": 0.2113,
+      "step": 117730
+    },
+    {
+      "epoch": 759.61,
+      "learning_rate": 1.3354838709677419e-05,
+      "loss": 0.1936,
+      "step": 117740
+    },
+    {
+      "epoch": 759.68,
+      "learning_rate": 1.335125448028674e-05,
+      "loss": 0.2411,
+      "step": 117750
+    },
+    {
+      "epoch": 759.74,
+      "learning_rate": 1.3347670250896057e-05,
+      "loss": 0.2213,
+      "step": 117760
+    },
+    {
+      "epoch": 759.8,
+      "learning_rate": 1.3344086021505379e-05,
+      "loss": 0.1321,
+      "step": 117770
+    },
+    {
+      "epoch": 759.87,
+      "learning_rate": 1.3340501792114695e-05,
+      "loss": 0.2309,
+      "step": 117780
+    },
+    {
+      "epoch": 759.93,
+      "learning_rate": 1.3336917562724017e-05,
+      "loss": 0.2167,
+      "step": 117790
+    },
+    {
+      "epoch": 760.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.1556,
+      "step": 117800
+    },
+    {
+      "epoch": 760.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9590680599212646,
+      "eval_runtime": 0.8585,
+      "eval_samples_per_second": 234.118,
+      "eval_steps_per_second": 8.153,
+      "step": 117800
+    },
+    {
+      "epoch": 760.06,
+      "learning_rate": 1.3329749103942655e-05,
+      "loss": 0.227,
+      "step": 117810
+    },
+    {
+      "epoch": 760.13,
+      "learning_rate": 1.3326164874551971e-05,
+      "loss": 0.2188,
+      "step": 117820
+    },
+    {
+      "epoch": 760.19,
+      "learning_rate": 1.3322580645161293e-05,
+      "loss": 0.2157,
+      "step": 117830
+    },
+    {
+      "epoch": 760.26,
+      "learning_rate": 1.331899641577061e-05,
+      "loss": 0.2385,
+      "step": 117840
+    },
+    {
+      "epoch": 760.32,
+      "learning_rate": 1.3315412186379931e-05,
+      "loss": 0.1969,
+      "step": 117850
+    },
+    {
+      "epoch": 760.39,
+      "learning_rate": 1.3311827956989248e-05,
+      "loss": 0.2614,
+      "step": 117860
+    },
+    {
+      "epoch": 760.45,
+      "learning_rate": 1.3308243727598568e-05,
+      "loss": 0.2005,
+      "step": 117870
+    },
+    {
+      "epoch": 760.51,
+      "learning_rate": 1.3304659498207886e-05,
+      "loss": 0.2422,
+      "step": 117880
+    },
+    {
+      "epoch": 760.58,
+      "learning_rate": 1.3301075268817204e-05,
+      "loss": 0.2275,
+      "step": 117890
+    },
+    {
+      "epoch": 760.64,
+      "learning_rate": 1.3297491039426524e-05,
+      "loss": 0.1994,
+      "step": 117900
+    },
+    {
+      "epoch": 760.71,
+      "learning_rate": 1.3293906810035842e-05,
+      "loss": 0.2305,
+      "step": 117910
+    },
+    {
+      "epoch": 760.77,
+      "learning_rate": 1.3290322580645162e-05,
+      "loss": 0.2604,
+      "step": 117920
+    },
+    {
+      "epoch": 760.84,
+      "learning_rate": 1.328673835125448e-05,
+      "loss": 0.2247,
+      "step": 117930
+    },
+    {
+      "epoch": 760.9,
+      "learning_rate": 1.32831541218638e-05,
+      "loss": 0.2066,
+      "step": 117940
+    },
+    {
+      "epoch": 760.96,
+      "learning_rate": 1.3279569892473118e-05,
+      "loss": 0.2095,
+      "step": 117950
+    },
+    {
+      "epoch": 761.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9515822529792786,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.913,
+      "eval_steps_per_second": 7.937,
+      "step": 117955
+    },
+    {
+      "epoch": 761.03,
+      "learning_rate": 1.3275985663082438e-05,
+      "loss": 0.2547,
+      "step": 117960
+    },
+    {
+      "epoch": 761.1,
+      "learning_rate": 1.3272401433691756e-05,
+      "loss": 0.2212,
+      "step": 117970
+    },
+    {
+      "epoch": 761.16,
+      "learning_rate": 1.3268817204301076e-05,
+      "loss": 0.2051,
+      "step": 117980
+    },
+    {
+      "epoch": 761.23,
+      "learning_rate": 1.3265232974910394e-05,
+      "loss": 0.2126,
+      "step": 117990
+    },
+    {
+      "epoch": 761.29,
+      "learning_rate": 1.3261648745519714e-05,
+      "loss": 0.2094,
+      "step": 118000
+    },
+    {
+      "epoch": 761.35,
+      "learning_rate": 1.3258064516129033e-05,
+      "loss": 0.2449,
+      "step": 118010
+    },
+    {
+      "epoch": 761.42,
+      "learning_rate": 1.3254480286738352e-05,
+      "loss": 0.2101,
+      "step": 118020
+    },
+    {
+      "epoch": 761.48,
+      "learning_rate": 1.325089605734767e-05,
+      "loss": 0.2007,
+      "step": 118030
+    },
+    {
+      "epoch": 761.55,
+      "learning_rate": 1.324731182795699e-05,
+      "loss": 0.1716,
+      "step": 118040
+    },
+    {
+      "epoch": 761.61,
+      "learning_rate": 1.3243727598566307e-05,
+      "loss": 0.2077,
+      "step": 118050
+    },
+    {
+      "epoch": 761.68,
+      "learning_rate": 1.3240143369175629e-05,
+      "loss": 0.1968,
+      "step": 118060
+    },
+    {
+      "epoch": 761.74,
+      "learning_rate": 1.3236559139784945e-05,
+      "loss": 0.1902,
+      "step": 118070
+    },
+    {
+      "epoch": 761.8,
+      "learning_rate": 1.3232974910394267e-05,
+      "loss": 0.2107,
+      "step": 118080
+    },
+    {
+      "epoch": 761.87,
+      "learning_rate": 1.3229390681003583e-05,
+      "loss": 0.2096,
+      "step": 118090
+    },
+    {
+      "epoch": 761.93,
+      "learning_rate": 1.3225806451612905e-05,
+      "loss": 0.193,
+      "step": 118100
+    },
+    {
+      "epoch": 762.0,
+      "learning_rate": 1.3222222222222221e-05,
+      "loss": 0.2307,
+      "step": 118110
+    },
+    {
+      "epoch": 762.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.9455006122589111,
+      "eval_runtime": 0.8884,
+      "eval_samples_per_second": 226.246,
+      "eval_steps_per_second": 7.879,
+      "step": 118110
+    },
+    {
+      "epoch": 762.06,
+      "learning_rate": 1.3218637992831543e-05,
+      "loss": 0.2198,
+      "step": 118120
+    },
+    {
+      "epoch": 762.13,
+      "learning_rate": 1.321505376344086e-05,
+      "loss": 0.1622,
+      "step": 118130
+    },
+    {
+      "epoch": 762.19,
+      "learning_rate": 1.3211469534050181e-05,
+      "loss": 0.2304,
+      "step": 118140
+    },
+    {
+      "epoch": 762.26,
+      "learning_rate": 1.3207885304659498e-05,
+      "loss": 0.2054,
+      "step": 118150
+    },
+    {
+      "epoch": 762.32,
+      "learning_rate": 1.320430107526882e-05,
+      "loss": 0.2255,
+      "step": 118160
+    },
+    {
+      "epoch": 762.39,
+      "learning_rate": 1.3200716845878136e-05,
+      "loss": 0.21,
+      "step": 118170
+    },
+    {
+      "epoch": 762.45,
+      "learning_rate": 1.3197132616487457e-05,
+      "loss": 0.2127,
+      "step": 118180
+    },
+    {
+      "epoch": 762.51,
+      "learning_rate": 1.3193548387096774e-05,
+      "loss": 0.1855,
+      "step": 118190
+    },
+    {
+      "epoch": 762.58,
+      "learning_rate": 1.3189964157706094e-05,
+      "loss": 0.232,
+      "step": 118200
+    },
+    {
+      "epoch": 762.64,
+      "learning_rate": 1.3186379928315412e-05,
+      "loss": 0.1915,
+      "step": 118210
+    },
+    {
+      "epoch": 762.71,
+      "learning_rate": 1.3182795698924732e-05,
+      "loss": 0.2133,
+      "step": 118220
+    },
+    {
+      "epoch": 762.77,
+      "learning_rate": 1.317921146953405e-05,
+      "loss": 0.1823,
+      "step": 118230
+    },
+    {
+      "epoch": 762.84,
+      "learning_rate": 1.317562724014337e-05,
+      "loss": 0.2266,
+      "step": 118240
+    },
+    {
+      "epoch": 762.9,
+      "learning_rate": 1.3172043010752688e-05,
+      "loss": 0.1991,
+      "step": 118250
+    },
+    {
+      "epoch": 762.96,
+      "learning_rate": 1.3168458781362008e-05,
+      "loss": 0.2411,
+      "step": 118260
+    },
+    {
+      "epoch": 763.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9675697088241577,
+      "eval_runtime": 0.8755,
+      "eval_samples_per_second": 229.585,
+      "eval_steps_per_second": 7.996,
+      "step": 118265
+    },
+    {
+      "epoch": 763.03,
+      "learning_rate": 1.3164874551971326e-05,
+      "loss": 0.1937,
+      "step": 118270
+    },
+    {
+      "epoch": 763.1,
+      "learning_rate": 1.3161290322580646e-05,
+      "loss": 0.1878,
+      "step": 118280
+    },
+    {
+      "epoch": 763.16,
+      "learning_rate": 1.3157706093189964e-05,
+      "loss": 0.2101,
+      "step": 118290
+    },
+    {
+      "epoch": 763.23,
+      "learning_rate": 1.3154121863799284e-05,
+      "loss": 0.2434,
+      "step": 118300
+    },
+    {
+      "epoch": 763.29,
+      "learning_rate": 1.3150537634408602e-05,
+      "loss": 0.2265,
+      "step": 118310
+    },
+    {
+      "epoch": 763.35,
+      "learning_rate": 1.3146953405017922e-05,
+      "loss": 0.2298,
+      "step": 118320
+    },
+    {
+      "epoch": 763.42,
+      "learning_rate": 1.314336917562724e-05,
+      "loss": 0.2145,
+      "step": 118330
+    },
+    {
+      "epoch": 763.48,
+      "learning_rate": 1.313978494623656e-05,
+      "loss": 0.1943,
+      "step": 118340
+    },
+    {
+      "epoch": 763.55,
+      "learning_rate": 1.3136200716845879e-05,
+      "loss": 0.1899,
+      "step": 118350
+    },
+    {
+      "epoch": 763.61,
+      "learning_rate": 1.3132616487455199e-05,
+      "loss": 0.194,
+      "step": 118360
+    },
+    {
+      "epoch": 763.68,
+      "learning_rate": 1.3129032258064517e-05,
+      "loss": 0.2613,
+      "step": 118370
+    },
+    {
+      "epoch": 763.74,
+      "learning_rate": 1.3125448028673837e-05,
+      "loss": 0.2521,
+      "step": 118380
+    },
+    {
+      "epoch": 763.8,
+      "learning_rate": 1.3121863799283155e-05,
+      "loss": 0.2301,
+      "step": 118390
+    },
+    {
+      "epoch": 763.87,
+      "learning_rate": 1.3118279569892475e-05,
+      "loss": 0.1899,
+      "step": 118400
+    },
+    {
+      "epoch": 763.93,
+      "learning_rate": 1.3114695340501793e-05,
+      "loss": 0.2088,
+      "step": 118410
+    },
+    {
+      "epoch": 764.0,
+      "learning_rate": 1.3111111111111113e-05,
+      "loss": 0.1836,
+      "step": 118420
+    },
+    {
+      "epoch": 764.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9801154732704163,
+      "eval_runtime": 0.8822,
+      "eval_samples_per_second": 227.836,
+      "eval_steps_per_second": 7.935,
+      "step": 118420
+    },
+    {
+      "epoch": 764.06,
+      "learning_rate": 1.3107526881720431e-05,
+      "loss": 0.2493,
+      "step": 118430
+    },
+    {
+      "epoch": 764.13,
+      "learning_rate": 1.3103942652329751e-05,
+      "loss": 0.2328,
+      "step": 118440
+    },
+    {
+      "epoch": 764.19,
+      "learning_rate": 1.3100358422939069e-05,
+      "loss": 0.1768,
+      "step": 118450
+    },
+    {
+      "epoch": 764.26,
+      "learning_rate": 1.3096774193548389e-05,
+      "loss": 0.2225,
+      "step": 118460
+    },
+    {
+      "epoch": 764.32,
+      "learning_rate": 1.3093189964157707e-05,
+      "loss": 0.2365,
+      "step": 118470
+    },
+    {
+      "epoch": 764.39,
+      "learning_rate": 1.3089605734767027e-05,
+      "loss": 0.2114,
+      "step": 118480
+    },
+    {
+      "epoch": 764.45,
+      "learning_rate": 1.3086021505376345e-05,
+      "loss": 0.2226,
+      "step": 118490
+    },
+    {
+      "epoch": 764.51,
+      "learning_rate": 1.3082437275985665e-05,
+      "loss": 0.2455,
+      "step": 118500
+    },
+    {
+      "epoch": 764.58,
+      "learning_rate": 1.3078853046594983e-05,
+      "loss": 0.2571,
+      "step": 118510
+    },
+    {
+      "epoch": 764.64,
+      "learning_rate": 1.3075268817204303e-05,
+      "loss": 0.2447,
+      "step": 118520
+    },
+    {
+      "epoch": 764.71,
+      "learning_rate": 1.3071684587813622e-05,
+      "loss": 0.2119,
+      "step": 118530
+    },
+    {
+      "epoch": 764.77,
+      "learning_rate": 1.3068100358422938e-05,
+      "loss": 0.1898,
+      "step": 118540
+    },
+    {
+      "epoch": 764.84,
+      "learning_rate": 1.3064516129032258e-05,
+      "loss": 0.2098,
+      "step": 118550
+    },
+    {
+      "epoch": 764.9,
+      "learning_rate": 1.3060931899641576e-05,
+      "loss": 0.2239,
+      "step": 118560
+    },
+    {
+      "epoch": 764.96,
+      "learning_rate": 1.3057347670250896e-05,
+      "loss": 0.2667,
+      "step": 118570
+    },
+    {
+      "epoch": 765.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 1.0422730445861816,
+      "eval_runtime": 0.8615,
+      "eval_samples_per_second": 233.304,
+      "eval_steps_per_second": 8.125,
+      "step": 118575
+    },
+    {
+      "epoch": 765.03,
+      "learning_rate": 1.3053763440860214e-05,
+      "loss": 0.2129,
+      "step": 118580
+    },
+    {
+      "epoch": 765.1,
+      "learning_rate": 1.3050179211469534e-05,
+      "loss": 0.1899,
+      "step": 118590
+    },
+    {
+      "epoch": 765.16,
+      "learning_rate": 1.3046594982078852e-05,
+      "loss": 0.1815,
+      "step": 118600
+    },
+    {
+      "epoch": 765.23,
+      "learning_rate": 1.3043010752688172e-05,
+      "loss": 0.2366,
+      "step": 118610
+    },
+    {
+      "epoch": 765.29,
+      "learning_rate": 1.303942652329749e-05,
+      "loss": 0.1736,
+      "step": 118620
+    },
+    {
+      "epoch": 765.35,
+      "learning_rate": 1.303584229390681e-05,
+      "loss": 0.2092,
+      "step": 118630
+    },
+    {
+      "epoch": 765.42,
+      "learning_rate": 1.3032258064516129e-05,
+      "loss": 0.1998,
+      "step": 118640
+    },
+    {
+      "epoch": 765.48,
+      "learning_rate": 1.3028673835125448e-05,
+      "loss": 0.2232,
+      "step": 118650
+    },
+    {
+      "epoch": 765.55,
+      "learning_rate": 1.3025089605734767e-05,
+      "loss": 0.2286,
+      "step": 118660
+    },
+    {
+      "epoch": 765.61,
+      "learning_rate": 1.3021505376344087e-05,
+      "loss": 0.2435,
+      "step": 118670
+    },
+    {
+      "epoch": 765.68,
+      "learning_rate": 1.3017921146953405e-05,
+      "loss": 0.2361,
+      "step": 118680
+    },
+    {
+      "epoch": 765.74,
+      "learning_rate": 1.3014336917562725e-05,
+      "loss": 0.2022,
+      "step": 118690
+    },
+    {
+      "epoch": 765.8,
+      "learning_rate": 1.3010752688172043e-05,
+      "loss": 0.242,
+      "step": 118700
+    },
+    {
+      "epoch": 765.87,
+      "learning_rate": 1.3007168458781363e-05,
+      "loss": 0.2596,
+      "step": 118710
+    },
+    {
+      "epoch": 765.93,
+      "learning_rate": 1.3003584229390681e-05,
+      "loss": 0.1905,
+      "step": 118720
+    },
+    {
+      "epoch": 766.0,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.2377,
+      "step": 118730
+    },
+    {
+      "epoch": 766.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9618730545043945,
+      "eval_runtime": 0.8745,
+      "eval_samples_per_second": 229.853,
+      "eval_steps_per_second": 8.005,
+      "step": 118730
+    },
+    {
+      "epoch": 766.06,
+      "learning_rate": 1.2996415770609319e-05,
+      "loss": 0.2017,
+      "step": 118740
+    },
+    {
+      "epoch": 766.13,
+      "learning_rate": 1.2992831541218639e-05,
+      "loss": 0.2246,
+      "step": 118750
+    },
+    {
+      "epoch": 766.19,
+      "learning_rate": 1.2989247311827957e-05,
+      "loss": 0.2105,
+      "step": 118760
+    },
+    {
+      "epoch": 766.26,
+      "learning_rate": 1.2985663082437277e-05,
+      "loss": 0.211,
+      "step": 118770
+    },
+    {
+      "epoch": 766.32,
+      "learning_rate": 1.2982078853046595e-05,
+      "loss": 0.1933,
+      "step": 118780
+    },
+    {
+      "epoch": 766.39,
+      "learning_rate": 1.2978494623655915e-05,
+      "loss": 0.2382,
+      "step": 118790
+    },
+    {
+      "epoch": 766.45,
+      "learning_rate": 1.2974910394265233e-05,
+      "loss": 0.2072,
+      "step": 118800
+    },
+    {
+      "epoch": 766.51,
+      "learning_rate": 1.2971326164874553e-05,
+      "loss": 0.21,
+      "step": 118810
+    },
+    {
+      "epoch": 766.58,
+      "learning_rate": 1.2967741935483872e-05,
+      "loss": 0.2496,
+      "step": 118820
+    },
+    {
+      "epoch": 766.64,
+      "learning_rate": 1.2964157706093191e-05,
+      "loss": 0.2124,
+      "step": 118830
+    },
+    {
+      "epoch": 766.71,
+      "learning_rate": 1.296057347670251e-05,
+      "loss": 0.2204,
+      "step": 118840
+    },
+    {
+      "epoch": 766.77,
+      "learning_rate": 1.295698924731183e-05,
+      "loss": 0.2409,
+      "step": 118850
+    },
+    {
+      "epoch": 766.84,
+      "learning_rate": 1.2953405017921148e-05,
+      "loss": 0.1815,
+      "step": 118860
+    },
+    {
+      "epoch": 766.9,
+      "learning_rate": 1.2949820788530468e-05,
+      "loss": 0.1975,
+      "step": 118870
+    },
+    {
+      "epoch": 766.96,
+      "learning_rate": 1.2946236559139784e-05,
+      "loss": 0.1874,
+      "step": 118880
+    },
+    {
+      "epoch": 767.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9244812726974487,
+      "eval_runtime": 0.9108,
+      "eval_samples_per_second": 220.69,
+      "eval_steps_per_second": 7.686,
+      "step": 118885
+    },
+    {
+      "epoch": 767.03,
+      "learning_rate": 1.2942652329749106e-05,
+      "loss": 0.2527,
+      "step": 118890
+    },
+    {
+      "epoch": 767.1,
+      "learning_rate": 1.2939068100358422e-05,
+      "loss": 0.2042,
+      "step": 118900
+    },
+    {
+      "epoch": 767.16,
+      "learning_rate": 1.2935483870967744e-05,
+      "loss": 0.2651,
+      "step": 118910
+    },
+    {
+      "epoch": 767.23,
+      "learning_rate": 1.293189964157706e-05,
+      "loss": 0.2117,
+      "step": 118920
+    },
+    {
+      "epoch": 767.29,
+      "learning_rate": 1.2928315412186382e-05,
+      "loss": 0.2126,
+      "step": 118930
+    },
+    {
+      "epoch": 767.35,
+      "learning_rate": 1.2924731182795698e-05,
+      "loss": 0.2512,
+      "step": 118940
+    },
+    {
+      "epoch": 767.42,
+      "learning_rate": 1.292114695340502e-05,
+      "loss": 0.1991,
+      "step": 118950
+    },
+    {
+      "epoch": 767.48,
+      "learning_rate": 1.2917562724014337e-05,
+      "loss": 0.2044,
+      "step": 118960
+    },
+    {
+      "epoch": 767.55,
+      "learning_rate": 1.2913978494623658e-05,
+      "loss": 0.2095,
+      "step": 118970
+    },
+    {
+      "epoch": 767.61,
+      "learning_rate": 1.2910394265232975e-05,
+      "loss": 0.2365,
+      "step": 118980
+    },
+    {
+      "epoch": 767.68,
+      "learning_rate": 1.2906810035842296e-05,
+      "loss": 0.2077,
+      "step": 118990
+    },
+    {
+      "epoch": 767.74,
+      "learning_rate": 1.2903225806451613e-05,
+      "loss": 0.2279,
+      "step": 119000
+    },
+    {
+      "epoch": 767.8,
+      "learning_rate": 1.2899641577060934e-05,
+      "loss": 0.2155,
+      "step": 119010
+    },
+    {
+      "epoch": 767.87,
+      "learning_rate": 1.2896057347670251e-05,
+      "loss": 0.1915,
+      "step": 119020
+    },
+    {
+      "epoch": 767.93,
+      "learning_rate": 1.289247311827957e-05,
+      "loss": 0.2643,
+      "step": 119030
+    },
+    {
+      "epoch": 768.0,
+      "learning_rate": 1.2888888888888889e-05,
+      "loss": 0.1887,
+      "step": 119040
+    },
+    {
+      "epoch": 768.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9586871862411499,
+      "eval_runtime": 0.8886,
+      "eval_samples_per_second": 226.198,
+      "eval_steps_per_second": 7.878,
+      "step": 119040
+    },
+    {
+      "epoch": 768.06,
+      "learning_rate": 1.2885304659498209e-05,
+      "loss": 0.2943,
+      "step": 119050
+    },
+    {
+      "epoch": 768.13,
+      "learning_rate": 1.2881720430107527e-05,
+      "loss": 0.2049,
+      "step": 119060
+    },
+    {
+      "epoch": 768.19,
+      "learning_rate": 1.2878136200716847e-05,
+      "loss": 0.1856,
+      "step": 119070
+    },
+    {
+      "epoch": 768.26,
+      "learning_rate": 1.2874551971326165e-05,
+      "loss": 0.2539,
+      "step": 119080
+    },
+    {
+      "epoch": 768.32,
+      "learning_rate": 1.2870967741935485e-05,
+      "loss": 0.2087,
+      "step": 119090
+    },
+    {
+      "epoch": 768.39,
+      "learning_rate": 1.2867383512544803e-05,
+      "loss": 0.2459,
+      "step": 119100
+    },
+    {
+      "epoch": 768.45,
+      "learning_rate": 1.2863799283154123e-05,
+      "loss": 0.2104,
+      "step": 119110
+    },
+    {
+      "epoch": 768.51,
+      "learning_rate": 1.2860215053763441e-05,
+      "loss": 0.1888,
+      "step": 119120
+    },
+    {
+      "epoch": 768.58,
+      "learning_rate": 1.2856630824372761e-05,
+      "loss": 0.2211,
+      "step": 119130
+    },
+    {
+      "epoch": 768.64,
+      "learning_rate": 1.285304659498208e-05,
+      "loss": 0.2378,
+      "step": 119140
+    },
+    {
+      "epoch": 768.71,
+      "learning_rate": 1.28494623655914e-05,
+      "loss": 0.2019,
+      "step": 119150
+    },
+    {
+      "epoch": 768.77,
+      "learning_rate": 1.2845878136200718e-05,
+      "loss": 0.2602,
+      "step": 119160
+    },
+    {
+      "epoch": 768.84,
+      "learning_rate": 1.2842293906810036e-05,
+      "loss": 0.2336,
+      "step": 119170
+    },
+    {
+      "epoch": 768.9,
+      "learning_rate": 1.2838709677419356e-05,
+      "loss": 0.2075,
+      "step": 119180
+    },
+    {
+      "epoch": 768.96,
+      "learning_rate": 1.2835125448028674e-05,
+      "loss": 0.1849,
+      "step": 119190
+    },
+    {
+      "epoch": 769.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9971348643302917,
+      "eval_runtime": 0.8656,
+      "eval_samples_per_second": 232.213,
+      "eval_steps_per_second": 8.087,
+      "step": 119195
+    },
+    {
+      "epoch": 769.03,
+      "learning_rate": 1.2831541218637994e-05,
+      "loss": 0.2082,
+      "step": 119200
+    },
+    {
+      "epoch": 769.1,
+      "learning_rate": 1.282795698924731e-05,
+      "loss": 0.2784,
+      "step": 119210
+    },
+    {
+      "epoch": 769.16,
+      "learning_rate": 1.2824372759856632e-05,
+      "loss": 0.2071,
+      "step": 119220
+    },
+    {
+      "epoch": 769.23,
+      "learning_rate": 1.2820788530465948e-05,
+      "loss": 0.2358,
+      "step": 119230
+    },
+    {
+      "epoch": 769.29,
+      "learning_rate": 1.281720430107527e-05,
+      "loss": 0.2445,
+      "step": 119240
+    },
+    {
+      "epoch": 769.35,
+      "learning_rate": 1.2813620071684587e-05,
+      "loss": 0.1699,
+      "step": 119250
+    },
+    {
+      "epoch": 769.42,
+      "learning_rate": 1.2810035842293908e-05,
+      "loss": 0.2167,
+      "step": 119260
+    },
+    {
+      "epoch": 769.48,
+      "learning_rate": 1.2806451612903225e-05,
+      "loss": 0.2192,
+      "step": 119270
+    },
+    {
+      "epoch": 769.55,
+      "learning_rate": 1.2802867383512546e-05,
+      "loss": 0.225,
+      "step": 119280
+    },
+    {
+      "epoch": 769.61,
+      "learning_rate": 1.2799283154121863e-05,
+      "loss": 0.1672,
+      "step": 119290
+    },
+    {
+      "epoch": 769.68,
+      "learning_rate": 1.2795698924731184e-05,
+      "loss": 0.2602,
+      "step": 119300
+    },
+    {
+      "epoch": 769.74,
+      "learning_rate": 1.27921146953405e-05,
+      "loss": 0.1857,
+      "step": 119310
+    },
+    {
+      "epoch": 769.8,
+      "learning_rate": 1.2788530465949822e-05,
+      "loss": 0.2401,
+      "step": 119320
+    },
+    {
+      "epoch": 769.87,
+      "learning_rate": 1.2784946236559139e-05,
+      "loss": 0.1983,
+      "step": 119330
+    },
+    {
+      "epoch": 769.93,
+      "learning_rate": 1.278136200716846e-05,
+      "loss": 0.1976,
+      "step": 119340
+    },
+    {
+      "epoch": 770.0,
+      "learning_rate": 1.2777777777777777e-05,
+      "loss": 0.2332,
+      "step": 119350
+    },
+    {
+      "epoch": 770.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9644247889518738,
+      "eval_runtime": 0.8719,
+      "eval_samples_per_second": 230.543,
+      "eval_steps_per_second": 8.029,
+      "step": 119350
+    },
+    {
+      "epoch": 770.06,
+      "learning_rate": 1.2774193548387099e-05,
+      "loss": 0.2435,
+      "step": 119360
+    },
+    {
+      "epoch": 770.13,
+      "learning_rate": 1.2770609318996415e-05,
+      "loss": 0.2122,
+      "step": 119370
+    },
+    {
+      "epoch": 770.19,
+      "learning_rate": 1.2767025089605735e-05,
+      "loss": 0.2539,
+      "step": 119380
+    },
+    {
+      "epoch": 770.26,
+      "learning_rate": 1.2763440860215053e-05,
+      "loss": 0.2485,
+      "step": 119390
+    },
+    {
+      "epoch": 770.32,
+      "learning_rate": 1.2759856630824373e-05,
+      "loss": 0.224,
+      "step": 119400
+    },
+    {
+      "epoch": 770.39,
+      "learning_rate": 1.2756272401433691e-05,
+      "loss": 0.251,
+      "step": 119410
+    },
+    {
+      "epoch": 770.45,
+      "learning_rate": 1.2752688172043011e-05,
+      "loss": 0.1996,
+      "step": 119420
+    },
+    {
+      "epoch": 770.51,
+      "learning_rate": 1.274910394265233e-05,
+      "loss": 0.2221,
+      "step": 119430
+    },
+    {
+      "epoch": 770.58,
+      "learning_rate": 1.274551971326165e-05,
+      "loss": 0.2417,
+      "step": 119440
+    },
+    {
+      "epoch": 770.64,
+      "learning_rate": 1.2741935483870968e-05,
+      "loss": 0.2758,
+      "step": 119450
+    },
+    {
+      "epoch": 770.71,
+      "learning_rate": 1.2738351254480287e-05,
+      "loss": 0.2456,
+      "step": 119460
+    },
+    {
+      "epoch": 770.77,
+      "learning_rate": 1.2734767025089606e-05,
+      "loss": 0.2123,
+      "step": 119470
+    },
+    {
+      "epoch": 770.84,
+      "learning_rate": 1.2731182795698926e-05,
+      "loss": 0.2241,
+      "step": 119480
+    },
+    {
+      "epoch": 770.9,
+      "learning_rate": 1.2727598566308244e-05,
+      "loss": 0.2177,
+      "step": 119490
+    },
+    {
+      "epoch": 770.96,
+      "learning_rate": 1.2724014336917564e-05,
+      "loss": 0.2602,
+      "step": 119500
+    },
+    {
+      "epoch": 771.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 1.0203444957733154,
+      "eval_runtime": 0.8818,
+      "eval_samples_per_second": 227.953,
+      "eval_steps_per_second": 7.939,
+      "step": 119505
+    },
+    {
+      "epoch": 771.03,
+      "learning_rate": 1.2720430107526882e-05,
+      "loss": 0.3024,
+      "step": 119510
+    },
+    {
+      "epoch": 771.1,
+      "learning_rate": 1.2716845878136202e-05,
+      "loss": 0.2485,
+      "step": 119520
+    },
+    {
+      "epoch": 771.16,
+      "learning_rate": 1.271326164874552e-05,
+      "loss": 0.2451,
+      "step": 119530
+    },
+    {
+      "epoch": 771.23,
+      "learning_rate": 1.270967741935484e-05,
+      "loss": 0.2291,
+      "step": 119540
+    },
+    {
+      "epoch": 771.29,
+      "learning_rate": 1.2706093189964158e-05,
+      "loss": 0.2324,
+      "step": 119550
+    },
+    {
+      "epoch": 771.35,
+      "learning_rate": 1.2702508960573478e-05,
+      "loss": 0.2279,
+      "step": 119560
+    },
+    {
+      "epoch": 771.42,
+      "learning_rate": 1.2698924731182796e-05,
+      "loss": 0.2357,
+      "step": 119570
+    },
+    {
+      "epoch": 771.48,
+      "learning_rate": 1.2695340501792116e-05,
+      "loss": 0.2271,
+      "step": 119580
+    },
+    {
+      "epoch": 771.55,
+      "learning_rate": 1.2691756272401434e-05,
+      "loss": 0.2015,
+      "step": 119590
+    },
+    {
+      "epoch": 771.61,
+      "learning_rate": 1.2688172043010754e-05,
+      "loss": 0.2261,
+      "step": 119600
+    },
+    {
+      "epoch": 771.68,
+      "learning_rate": 1.2684587813620072e-05,
+      "loss": 0.2146,
+      "step": 119610
+    },
+    {
+      "epoch": 771.74,
+      "learning_rate": 1.2681003584229392e-05,
+      "loss": 0.2437,
+      "step": 119620
+    },
+    {
+      "epoch": 771.8,
+      "learning_rate": 1.267741935483871e-05,
+      "loss": 0.2342,
+      "step": 119630
+    },
+    {
+      "epoch": 771.87,
+      "learning_rate": 1.267383512544803e-05,
+      "loss": 0.2418,
+      "step": 119640
+    },
+    {
+      "epoch": 771.93,
+      "learning_rate": 1.2670250896057349e-05,
+      "loss": 0.2709,
+      "step": 119650
+    },
+    {
+      "epoch": 772.0,
+      "learning_rate": 1.2666666666666668e-05,
+      "loss": 0.196,
+      "step": 119660
+    },
+    {
+      "epoch": 772.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9487534165382385,
+      "eval_runtime": 0.881,
+      "eval_samples_per_second": 228.152,
+      "eval_steps_per_second": 7.946,
+      "step": 119660
+    },
+    {
+      "epoch": 772.06,
+      "learning_rate": 1.2663082437275987e-05,
+      "loss": 0.2502,
+      "step": 119670
+    },
+    {
+      "epoch": 772.13,
+      "learning_rate": 1.2659498207885307e-05,
+      "loss": 0.2118,
+      "step": 119680
+    },
+    {
+      "epoch": 772.19,
+      "learning_rate": 1.2655913978494625e-05,
+      "loss": 0.191,
+      "step": 119690
+    },
+    {
+      "epoch": 772.26,
+      "learning_rate": 1.2652329749103945e-05,
+      "loss": 0.1847,
+      "step": 119700
+    },
+    {
+      "epoch": 772.32,
+      "learning_rate": 1.2648745519713261e-05,
+      "loss": 0.1925,
+      "step": 119710
+    },
+    {
+      "epoch": 772.39,
+      "learning_rate": 1.2645161290322583e-05,
+      "loss": 0.1857,
+      "step": 119720
+    },
+    {
+      "epoch": 772.45,
+      "learning_rate": 1.26415770609319e-05,
+      "loss": 0.1937,
+      "step": 119730
+    },
+    {
+      "epoch": 772.51,
+      "learning_rate": 1.2637992831541221e-05,
+      "loss": 0.2404,
+      "step": 119740
+    },
+    {
+      "epoch": 772.58,
+      "learning_rate": 1.2634408602150537e-05,
+      "loss": 0.2251,
+      "step": 119750
+    },
+    {
+      "epoch": 772.64,
+      "learning_rate": 1.2630824372759859e-05,
+      "loss": 0.2557,
+      "step": 119760
+    },
+    {
+      "epoch": 772.71,
+      "learning_rate": 1.2627240143369176e-05,
+      "loss": 0.2228,
+      "step": 119770
+    },
+    {
+      "epoch": 772.77,
+      "learning_rate": 1.2623655913978497e-05,
+      "loss": 0.2272,
+      "step": 119780
+    },
+    {
+      "epoch": 772.84,
+      "learning_rate": 1.2620071684587814e-05,
+      "loss": 0.2532,
+      "step": 119790
+    },
+    {
+      "epoch": 772.9,
+      "learning_rate": 1.2616487455197135e-05,
+      "loss": 0.207,
+      "step": 119800
+    },
+    {
+      "epoch": 772.96,
+      "learning_rate": 1.2612903225806452e-05,
+      "loss": 0.2221,
+      "step": 119810
+    },
+    {
+      "epoch": 773.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9159120917320251,
+      "eval_runtime": 0.8702,
+      "eval_samples_per_second": 230.988,
+      "eval_steps_per_second": 8.044,
+      "step": 119815
+    },
+    {
+      "epoch": 773.03,
+      "learning_rate": 1.260931899641577e-05,
+      "loss": 0.1977,
+      "step": 119820
+    },
+    {
+      "epoch": 773.1,
+      "learning_rate": 1.260573476702509e-05,
+      "loss": 0.191,
+      "step": 119830
+    },
+    {
+      "epoch": 773.16,
+      "learning_rate": 1.2602150537634408e-05,
+      "loss": 0.1973,
+      "step": 119840
+    },
+    {
+      "epoch": 773.23,
+      "learning_rate": 1.2598566308243728e-05,
+      "loss": 0.1689,
+      "step": 119850
+    },
+    {
+      "epoch": 773.29,
+      "learning_rate": 1.2594982078853046e-05,
+      "loss": 0.1936,
+      "step": 119860
+    },
+    {
+      "epoch": 773.35,
+      "learning_rate": 1.2591397849462366e-05,
+      "loss": 0.2088,
+      "step": 119870
+    },
+    {
+      "epoch": 773.42,
+      "learning_rate": 1.2587813620071684e-05,
+      "loss": 0.1996,
+      "step": 119880
+    },
+    {
+      "epoch": 773.48,
+      "learning_rate": 1.2584229390681004e-05,
+      "loss": 0.2091,
+      "step": 119890
+    },
+    {
+      "epoch": 773.55,
+      "learning_rate": 1.2580645161290322e-05,
+      "loss": 0.2232,
+      "step": 119900
+    },
+    {
+      "epoch": 773.61,
+      "learning_rate": 1.2577060931899642e-05,
+      "loss": 0.2235,
+      "step": 119910
+    },
+    {
+      "epoch": 773.68,
+      "learning_rate": 1.257347670250896e-05,
+      "loss": 0.215,
+      "step": 119920
+    },
+    {
+      "epoch": 773.74,
+      "learning_rate": 1.256989247311828e-05,
+      "loss": 0.2473,
+      "step": 119930
+    },
+    {
+      "epoch": 773.8,
+      "learning_rate": 1.2566308243727599e-05,
+      "loss": 0.2407,
+      "step": 119940
+    },
+    {
+      "epoch": 773.87,
+      "learning_rate": 1.2562724014336918e-05,
+      "loss": 0.2387,
+      "step": 119950
+    },
+    {
+      "epoch": 773.93,
+      "learning_rate": 1.2559139784946237e-05,
+      "loss": 0.2194,
+      "step": 119960
+    },
+    {
+      "epoch": 774.0,
+      "learning_rate": 1.2555555555555557e-05,
+      "loss": 0.2148,
+      "step": 119970
+    },
+    {
+      "epoch": 774.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9201505184173584,
+      "eval_runtime": 0.9068,
+      "eval_samples_per_second": 221.657,
+      "eval_steps_per_second": 7.719,
+      "step": 119970
+    },
+    {
+      "epoch": 774.06,
+      "learning_rate": 1.2551971326164875e-05,
+      "loss": 0.2126,
+      "step": 119980
+    },
+    {
+      "epoch": 774.13,
+      "learning_rate": 1.2548387096774195e-05,
+      "loss": 0.2374,
+      "step": 119990
+    },
+    {
+      "epoch": 774.19,
+      "learning_rate": 1.2544802867383513e-05,
+      "loss": 0.2022,
+      "step": 120000
+    },
+    {
+      "epoch": 774.26,
+      "learning_rate": 1.2541218637992833e-05,
+      "loss": 0.2471,
+      "step": 120010
+    },
+    {
+      "epoch": 774.32,
+      "learning_rate": 1.2537634408602151e-05,
+      "loss": 0.1926,
+      "step": 120020
+    },
+    {
+      "epoch": 774.39,
+      "learning_rate": 1.2534050179211471e-05,
+      "loss": 0.2563,
+      "step": 120030
+    },
+    {
+      "epoch": 774.45,
+      "learning_rate": 1.2530465949820789e-05,
+      "loss": 0.1933,
+      "step": 120040
+    },
+    {
+      "epoch": 774.51,
+      "learning_rate": 1.2526881720430109e-05,
+      "loss": 0.2267,
+      "step": 120050
+    },
+    {
+      "epoch": 774.58,
+      "learning_rate": 1.2523297491039425e-05,
+      "loss": 0.2209,
+      "step": 120060
+    },
+    {
+      "epoch": 774.64,
+      "learning_rate": 1.2519713261648747e-05,
+      "loss": 0.1957,
+      "step": 120070
+    },
+    {
+      "epoch": 774.71,
+      "learning_rate": 1.2516129032258064e-05,
+      "loss": 0.2375,
+      "step": 120080
+    },
+    {
+      "epoch": 774.77,
+      "learning_rate": 1.2512544802867385e-05,
+      "loss": 0.2176,
+      "step": 120090
+    },
+    {
+      "epoch": 774.84,
+      "learning_rate": 1.2508960573476702e-05,
+      "loss": 0.1861,
+      "step": 120100
+    },
+    {
+      "epoch": 774.9,
+      "learning_rate": 1.2505376344086023e-05,
+      "loss": 0.208,
+      "step": 120110
+    },
+    {
+      "epoch": 774.96,
+      "learning_rate": 1.250179211469534e-05,
+      "loss": 0.1886,
+      "step": 120120
+    },
+    {
+      "epoch": 775.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9218161106109619,
+      "eval_runtime": 0.9471,
+      "eval_samples_per_second": 212.231,
+      "eval_steps_per_second": 7.391,
+      "step": 120125
+    },
+    {
+      "epoch": 775.03,
+      "learning_rate": 1.249820788530466e-05,
+      "loss": 0.2028,
+      "step": 120130
+    },
+    {
+      "epoch": 775.1,
+      "learning_rate": 1.2494623655913978e-05,
+      "loss": 0.2162,
+      "step": 120140
+    },
+    {
+      "epoch": 775.16,
+      "learning_rate": 1.2491039426523298e-05,
+      "loss": 0.2237,
+      "step": 120150
+    },
+    {
+      "epoch": 775.23,
+      "learning_rate": 1.2487455197132616e-05,
+      "loss": 0.2323,
+      "step": 120160
+    },
+    {
+      "epoch": 775.29,
+      "learning_rate": 1.2483870967741936e-05,
+      "loss": 0.1838,
+      "step": 120170
+    },
+    {
+      "epoch": 775.35,
+      "learning_rate": 1.2480286738351254e-05,
+      "loss": 0.2006,
+      "step": 120180
+    },
+    {
+      "epoch": 775.42,
+      "learning_rate": 1.2476702508960574e-05,
+      "loss": 0.1979,
+      "step": 120190
+    },
+    {
+      "epoch": 775.48,
+      "learning_rate": 1.2473118279569892e-05,
+      "loss": 0.2537,
+      "step": 120200
+    },
+    {
+      "epoch": 775.55,
+      "learning_rate": 1.2469534050179212e-05,
+      "loss": 0.1835,
+      "step": 120210
+    },
+    {
+      "epoch": 775.61,
+      "learning_rate": 1.246594982078853e-05,
+      "loss": 0.2017,
+      "step": 120220
+    },
+    {
+      "epoch": 775.68,
+      "learning_rate": 1.246236559139785e-05,
+      "loss": 0.2517,
+      "step": 120230
+    },
+    {
+      "epoch": 775.74,
+      "learning_rate": 1.2458781362007168e-05,
+      "loss": 0.2352,
+      "step": 120240
+    },
+    {
+      "epoch": 775.8,
+      "learning_rate": 1.2455197132616488e-05,
+      "loss": 0.2552,
+      "step": 120250
+    },
+    {
+      "epoch": 775.87,
+      "learning_rate": 1.2451612903225807e-05,
+      "loss": 0.2187,
+      "step": 120260
+    },
+    {
+      "epoch": 775.93,
+      "learning_rate": 1.2448028673835126e-05,
+      "loss": 0.1826,
+      "step": 120270
+    },
+    {
+      "epoch": 776.0,
+      "learning_rate": 1.2444444444444445e-05,
+      "loss": 0.2409,
+      "step": 120280
+    },
+    {
+      "epoch": 776.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9627440571784973,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.915,
+      "eval_steps_per_second": 7.937,
+      "step": 120280
+    },
+    {
+      "epoch": 776.06,
+      "learning_rate": 1.2440860215053765e-05,
+      "loss": 0.2032,
+      "step": 120290
+    },
+    {
+      "epoch": 776.13,
+      "learning_rate": 1.2437275985663083e-05,
+      "loss": 0.2426,
+      "step": 120300
+    },
+    {
+      "epoch": 776.19,
+      "learning_rate": 1.2433691756272401e-05,
+      "loss": 0.1905,
+      "step": 120310
+    },
+    {
+      "epoch": 776.26,
+      "learning_rate": 1.243010752688172e-05,
+      "loss": 0.2283,
+      "step": 120320
+    },
+    {
+      "epoch": 776.32,
+      "learning_rate": 1.2426523297491039e-05,
+      "loss": 0.2187,
+      "step": 120330
+    },
+    {
+      "epoch": 776.39,
+      "learning_rate": 1.2422939068100359e-05,
+      "loss": 0.2069,
+      "step": 120340
+    },
+    {
+      "epoch": 776.45,
+      "learning_rate": 1.2419354838709677e-05,
+      "loss": 0.1566,
+      "step": 120350
+    },
+    {
+      "epoch": 776.51,
+      "learning_rate": 1.2415770609318997e-05,
+      "loss": 0.2203,
+      "step": 120360
+    },
+    {
+      "epoch": 776.58,
+      "learning_rate": 1.2412186379928315e-05,
+      "loss": 0.1895,
+      "step": 120370
+    },
+    {
+      "epoch": 776.64,
+      "learning_rate": 1.2408602150537635e-05,
+      "loss": 0.2055,
+      "step": 120380
+    },
+    {
+      "epoch": 776.71,
+      "learning_rate": 1.2405017921146953e-05,
+      "loss": 0.2091,
+      "step": 120390
+    },
+    {
+      "epoch": 776.77,
+      "learning_rate": 1.2401433691756273e-05,
+      "loss": 0.2021,
+      "step": 120400
+    },
+    {
+      "epoch": 776.84,
+      "learning_rate": 1.2397849462365591e-05,
+      "loss": 0.2395,
+      "step": 120410
+    },
+    {
+      "epoch": 776.9,
+      "learning_rate": 1.2394265232974911e-05,
+      "loss": 0.2432,
+      "step": 120420
+    },
+    {
+      "epoch": 776.96,
+      "learning_rate": 1.239068100358423e-05,
+      "loss": 0.1873,
+      "step": 120430
+    },
+    {
+      "epoch": 777.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9212930202484131,
+      "eval_runtime": 0.8685,
+      "eval_samples_per_second": 231.429,
+      "eval_steps_per_second": 8.06,
+      "step": 120435
+    },
+    {
+      "epoch": 777.03,
+      "learning_rate": 1.238709677419355e-05,
+      "loss": 0.2606,
+      "step": 120440
+    },
+    {
+      "epoch": 777.1,
+      "learning_rate": 1.2383512544802868e-05,
+      "loss": 0.2151,
+      "step": 120450
+    },
+    {
+      "epoch": 777.16,
+      "learning_rate": 1.2379928315412188e-05,
+      "loss": 0.1865,
+      "step": 120460
+    },
+    {
+      "epoch": 777.23,
+      "learning_rate": 1.2376344086021506e-05,
+      "loss": 0.218,
+      "step": 120470
+    },
+    {
+      "epoch": 777.29,
+      "learning_rate": 1.2372759856630826e-05,
+      "loss": 0.2073,
+      "step": 120480
+    },
+    {
+      "epoch": 777.35,
+      "learning_rate": 1.2369175627240144e-05,
+      "loss": 0.2326,
+      "step": 120490
+    },
+    {
+      "epoch": 777.42,
+      "learning_rate": 1.2365591397849464e-05,
+      "loss": 0.2166,
+      "step": 120500
+    },
+    {
+      "epoch": 777.48,
+      "learning_rate": 1.2362007168458782e-05,
+      "loss": 0.2299,
+      "step": 120510
+    },
+    {
+      "epoch": 777.55,
+      "learning_rate": 1.2358422939068102e-05,
+      "loss": 0.2483,
+      "step": 120520
+    },
+    {
+      "epoch": 777.61,
+      "learning_rate": 1.235483870967742e-05,
+      "loss": 0.2192,
+      "step": 120530
+    },
+    {
+      "epoch": 777.68,
+      "learning_rate": 1.235125448028674e-05,
+      "loss": 0.1894,
+      "step": 120540
+    },
+    {
+      "epoch": 777.74,
+      "learning_rate": 1.2347670250896058e-05,
+      "loss": 0.1807,
+      "step": 120550
+    },
+    {
+      "epoch": 777.8,
+      "learning_rate": 1.2344086021505376e-05,
+      "loss": 0.1372,
+      "step": 120560
+    },
+    {
+      "epoch": 777.87,
+      "learning_rate": 1.2340501792114696e-05,
+      "loss": 0.2498,
+      "step": 120570
+    },
+    {
+      "epoch": 777.93,
+      "learning_rate": 1.2336917562724014e-05,
+      "loss": 0.2124,
+      "step": 120580
+    },
+    {
+      "epoch": 778.0,
+      "learning_rate": 1.2333333333333334e-05,
+      "loss": 0.171,
+      "step": 120590
+    },
+    {
+      "epoch": 778.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9274424314498901,
+      "eval_runtime": 0.8734,
+      "eval_samples_per_second": 230.138,
+      "eval_steps_per_second": 8.015,
+      "step": 120590
+    },
+    {
+      "epoch": 778.06,
+      "learning_rate": 1.2329749103942653e-05,
+      "loss": 0.239,
+      "step": 120600
+    },
+    {
+      "epoch": 778.13,
+      "learning_rate": 1.2326164874551973e-05,
+      "loss": 0.1803,
+      "step": 120610
+    },
+    {
+      "epoch": 778.19,
+      "learning_rate": 1.232258064516129e-05,
+      "loss": 0.223,
+      "step": 120620
+    },
+    {
+      "epoch": 778.26,
+      "learning_rate": 1.231899641577061e-05,
+      "loss": 0.2306,
+      "step": 120630
+    },
+    {
+      "epoch": 778.32,
+      "learning_rate": 1.2315412186379929e-05,
+      "loss": 0.2113,
+      "step": 120640
+    },
+    {
+      "epoch": 778.39,
+      "learning_rate": 1.2311827956989249e-05,
+      "loss": 0.2256,
+      "step": 120650
+    },
+    {
+      "epoch": 778.45,
+      "learning_rate": 1.2308243727598567e-05,
+      "loss": 0.1948,
+      "step": 120660
+    },
+    {
+      "epoch": 778.51,
+      "learning_rate": 1.2304659498207887e-05,
+      "loss": 0.1828,
+      "step": 120670
+    },
+    {
+      "epoch": 778.58,
+      "learning_rate": 1.2301075268817205e-05,
+      "loss": 0.2421,
+      "step": 120680
+    },
+    {
+      "epoch": 778.64,
+      "learning_rate": 1.2297491039426525e-05,
+      "loss": 0.2408,
+      "step": 120690
+    },
+    {
+      "epoch": 778.71,
+      "learning_rate": 1.2293906810035843e-05,
+      "loss": 0.2067,
+      "step": 120700
+    },
+    {
+      "epoch": 778.77,
+      "learning_rate": 1.2290322580645163e-05,
+      "loss": 0.2407,
+      "step": 120710
+    },
+    {
+      "epoch": 778.84,
+      "learning_rate": 1.2286738351254481e-05,
+      "loss": 0.2167,
+      "step": 120720
+    },
+    {
+      "epoch": 778.9,
+      "learning_rate": 1.2283154121863801e-05,
+      "loss": 0.1893,
+      "step": 120730
+    },
+    {
+      "epoch": 778.96,
+      "learning_rate": 1.227956989247312e-05,
+      "loss": 0.2169,
+      "step": 120740
+    },
+    {
+      "epoch": 779.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9699149131774902,
+      "eval_runtime": 0.8678,
+      "eval_samples_per_second": 231.622,
+      "eval_steps_per_second": 8.066,
+      "step": 120745
+    },
+    {
+      "epoch": 779.03,
+      "learning_rate": 1.227598566308244e-05,
+      "loss": 0.2394,
+      "step": 120750
+    },
+    {
+      "epoch": 779.1,
+      "learning_rate": 1.2272401433691757e-05,
+      "loss": 0.1782,
+      "step": 120760
+    },
+    {
+      "epoch": 779.16,
+      "learning_rate": 1.2268817204301076e-05,
+      "loss": 0.2212,
+      "step": 120770
+    },
+    {
+      "epoch": 779.23,
+      "learning_rate": 1.2265232974910394e-05,
+      "loss": 0.202,
+      "step": 120780
+    },
+    {
+      "epoch": 779.29,
+      "learning_rate": 1.2261648745519714e-05,
+      "loss": 0.234,
+      "step": 120790
+    },
+    {
+      "epoch": 779.35,
+      "learning_rate": 1.2258064516129032e-05,
+      "loss": 0.2092,
+      "step": 120800
+    },
+    {
+      "epoch": 779.42,
+      "learning_rate": 1.2254480286738352e-05,
+      "loss": 0.1913,
+      "step": 120810
+    },
+    {
+      "epoch": 779.48,
+      "learning_rate": 1.225089605734767e-05,
+      "loss": 0.1655,
+      "step": 120820
+    },
+    {
+      "epoch": 779.55,
+      "learning_rate": 1.224731182795699e-05,
+      "loss": 0.199,
+      "step": 120830
+    },
+    {
+      "epoch": 779.61,
+      "learning_rate": 1.2243727598566308e-05,
+      "loss": 0.2136,
+      "step": 120840
+    },
+    {
+      "epoch": 779.68,
+      "learning_rate": 1.2240143369175628e-05,
+      "loss": 0.2319,
+      "step": 120850
+    },
+    {
+      "epoch": 779.74,
+      "learning_rate": 1.2236559139784946e-05,
+      "loss": 0.2832,
+      "step": 120860
+    },
+    {
+      "epoch": 779.8,
+      "learning_rate": 1.2232974910394266e-05,
+      "loss": 0.2302,
+      "step": 120870
+    },
+    {
+      "epoch": 779.87,
+      "learning_rate": 1.2229390681003584e-05,
+      "loss": 0.1967,
+      "step": 120880
+    },
+    {
+      "epoch": 779.93,
+      "learning_rate": 1.2225806451612903e-05,
+      "loss": 0.2135,
+      "step": 120890
+    },
+    {
+      "epoch": 780.0,
+      "learning_rate": 1.2222222222222222e-05,
+      "loss": 0.2277,
+      "step": 120900
+    },
+    {
+      "epoch": 780.0,
+      "eval_accuracy": 0.7960199004975125,
+      "eval_loss": 0.8898761868476868,
+      "eval_runtime": 0.8764,
+      "eval_samples_per_second": 229.355,
+      "eval_steps_per_second": 7.987,
+      "step": 120900
+    },
+    {
+      "epoch": 780.06,
+      "learning_rate": 1.221863799283154e-05,
+      "loss": 0.2397,
+      "step": 120910
+    },
+    {
+      "epoch": 780.13,
+      "learning_rate": 1.221505376344086e-05,
+      "loss": 0.2388,
+      "step": 120920
+    },
+    {
+      "epoch": 780.19,
+      "learning_rate": 1.2211469534050179e-05,
+      "loss": 0.2618,
+      "step": 120930
+    },
+    {
+      "epoch": 780.26,
+      "learning_rate": 1.2207885304659499e-05,
+      "loss": 0.201,
+      "step": 120940
+    },
+    {
+      "epoch": 780.32,
+      "learning_rate": 1.2204301075268817e-05,
+      "loss": 0.1896,
+      "step": 120950
+    },
+    {
+      "epoch": 780.39,
+      "learning_rate": 1.2200716845878137e-05,
+      "loss": 0.1928,
+      "step": 120960
+    },
+    {
+      "epoch": 780.45,
+      "learning_rate": 1.2197132616487455e-05,
+      "loss": 0.2011,
+      "step": 120970
+    },
+    {
+      "epoch": 780.51,
+      "learning_rate": 1.2193548387096775e-05,
+      "loss": 0.2177,
+      "step": 120980
+    },
+    {
+      "epoch": 780.58,
+      "learning_rate": 1.2189964157706093e-05,
+      "loss": 0.2278,
+      "step": 120990
+    },
+    {
+      "epoch": 780.64,
+      "learning_rate": 1.2186379928315413e-05,
+      "loss": 0.1866,
+      "step": 121000
+    },
+    {
+      "epoch": 780.71,
+      "learning_rate": 1.2182795698924731e-05,
+      "loss": 0.2313,
+      "step": 121010
+    },
+    {
+      "epoch": 780.77,
+      "learning_rate": 1.2179211469534051e-05,
+      "loss": 0.239,
+      "step": 121020
+    },
+    {
+      "epoch": 780.84,
+      "learning_rate": 1.217562724014337e-05,
+      "loss": 0.2161,
+      "step": 121030
+    },
+    {
+      "epoch": 780.9,
+      "learning_rate": 1.217204301075269e-05,
+      "loss": 0.1935,
+      "step": 121040
+    },
+    {
+      "epoch": 780.96,
+      "learning_rate": 1.2168458781362007e-05,
+      "loss": 0.2212,
+      "step": 121050
+    },
+    {
+      "epoch": 781.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8936700224876404,
+      "eval_runtime": 0.8849,
+      "eval_samples_per_second": 227.145,
+      "eval_steps_per_second": 7.911,
+      "step": 121055
+    },
+    {
+      "epoch": 781.03,
+      "learning_rate": 1.2164874551971327e-05,
+      "loss": 0.2114,
+      "step": 121060
+    },
+    {
+      "epoch": 781.1,
+      "learning_rate": 1.2161290322580646e-05,
+      "loss": 0.2367,
+      "step": 121070
+    },
+    {
+      "epoch": 781.16,
+      "learning_rate": 1.2157706093189965e-05,
+      "loss": 0.2602,
+      "step": 121080
+    },
+    {
+      "epoch": 781.23,
+      "learning_rate": 1.2154121863799284e-05,
+      "loss": 0.2097,
+      "step": 121090
+    },
+    {
+      "epoch": 781.29,
+      "learning_rate": 1.2150537634408604e-05,
+      "loss": 0.2307,
+      "step": 121100
+    },
+    {
+      "epoch": 781.35,
+      "learning_rate": 1.2146953405017922e-05,
+      "loss": 0.2114,
+      "step": 121110
+    },
+    {
+      "epoch": 781.42,
+      "learning_rate": 1.2143369175627242e-05,
+      "loss": 0.2001,
+      "step": 121120
+    },
+    {
+      "epoch": 781.48,
+      "learning_rate": 1.213978494623656e-05,
+      "loss": 0.2,
+      "step": 121130
+    },
+    {
+      "epoch": 781.55,
+      "learning_rate": 1.2136200716845878e-05,
+      "loss": 0.2567,
+      "step": 121140
+    },
+    {
+      "epoch": 781.61,
+      "learning_rate": 1.2132616487455198e-05,
+      "loss": 0.2207,
+      "step": 121150
+    },
+    {
+      "epoch": 781.68,
+      "learning_rate": 1.2129032258064516e-05,
+      "loss": 0.24,
+      "step": 121160
+    },
+    {
+      "epoch": 781.74,
+      "learning_rate": 1.2125448028673836e-05,
+      "loss": 0.2015,
+      "step": 121170
+    },
+    {
+      "epoch": 781.8,
+      "learning_rate": 1.2121863799283154e-05,
+      "loss": 0.2247,
+      "step": 121180
+    },
+    {
+      "epoch": 781.87,
+      "learning_rate": 1.2118279569892474e-05,
+      "loss": 0.2228,
+      "step": 121190
+    },
+    {
+      "epoch": 781.93,
+      "learning_rate": 1.2114695340501792e-05,
+      "loss": 0.2236,
+      "step": 121200
+    },
+    {
+      "epoch": 782.0,
+      "learning_rate": 1.2111111111111112e-05,
+      "loss": 0.2539,
+      "step": 121210
+    },
+    {
+      "epoch": 782.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.8647252321243286,
+      "eval_runtime": 0.8769,
+      "eval_samples_per_second": 229.205,
+      "eval_steps_per_second": 7.982,
+      "step": 121210
+    },
+    {
+      "epoch": 782.06,
+      "learning_rate": 1.210752688172043e-05,
+      "loss": 0.2327,
+      "step": 121220
+    },
+    {
+      "epoch": 782.13,
+      "learning_rate": 1.210394265232975e-05,
+      "loss": 0.2383,
+      "step": 121230
+    },
+    {
+      "epoch": 782.19,
+      "learning_rate": 1.2100358422939069e-05,
+      "loss": 0.2498,
+      "step": 121240
+    },
+    {
+      "epoch": 782.26,
+      "learning_rate": 1.2096774193548388e-05,
+      "loss": 0.2164,
+      "step": 121250
+    },
+    {
+      "epoch": 782.32,
+      "learning_rate": 1.2093189964157707e-05,
+      "loss": 0.2069,
+      "step": 121260
+    },
+    {
+      "epoch": 782.39,
+      "learning_rate": 1.2089605734767027e-05,
+      "loss": 0.2175,
+      "step": 121270
+    },
+    {
+      "epoch": 782.45,
+      "learning_rate": 1.2086021505376345e-05,
+      "loss": 0.234,
+      "step": 121280
+    },
+    {
+      "epoch": 782.51,
+      "learning_rate": 1.2082437275985665e-05,
+      "loss": 0.1707,
+      "step": 121290
+    },
+    {
+      "epoch": 782.58,
+      "learning_rate": 1.2078853046594983e-05,
+      "loss": 0.2424,
+      "step": 121300
+    },
+    {
+      "epoch": 782.64,
+      "learning_rate": 1.2075268817204303e-05,
+      "loss": 0.2043,
+      "step": 121310
+    },
+    {
+      "epoch": 782.71,
+      "learning_rate": 1.2071684587813621e-05,
+      "loss": 0.2373,
+      "step": 121320
+    },
+    {
+      "epoch": 782.77,
+      "learning_rate": 1.2068100358422941e-05,
+      "loss": 0.2374,
+      "step": 121330
+    },
+    {
+      "epoch": 782.84,
+      "learning_rate": 1.2064516129032259e-05,
+      "loss": 0.2212,
+      "step": 121340
+    },
+    {
+      "epoch": 782.9,
+      "learning_rate": 1.2060931899641579e-05,
+      "loss": 0.2438,
+      "step": 121350
+    },
+    {
+      "epoch": 782.96,
+      "learning_rate": 1.2057347670250897e-05,
+      "loss": 0.2106,
+      "step": 121360
+    },
+    {
+      "epoch": 783.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8972164988517761,
+      "eval_runtime": 0.8698,
+      "eval_samples_per_second": 231.085,
+      "eval_steps_per_second": 8.048,
+      "step": 121365
+    },
+    {
+      "epoch": 783.03,
+      "learning_rate": 1.2053763440860217e-05,
+      "loss": 0.1959,
+      "step": 121370
+    },
+    {
+      "epoch": 783.1,
+      "learning_rate": 1.2050179211469535e-05,
+      "loss": 0.2381,
+      "step": 121380
+    },
+    {
+      "epoch": 783.16,
+      "learning_rate": 1.2046594982078853e-05,
+      "loss": 0.191,
+      "step": 121390
+    },
+    {
+      "epoch": 783.23,
+      "learning_rate": 1.2043010752688173e-05,
+      "loss": 0.1739,
+      "step": 121400
+    },
+    {
+      "epoch": 783.29,
+      "learning_rate": 1.2039426523297492e-05,
+      "loss": 0.2113,
+      "step": 121410
+    },
+    {
+      "epoch": 783.35,
+      "learning_rate": 1.203584229390681e-05,
+      "loss": 0.2185,
+      "step": 121420
+    },
+    {
+      "epoch": 783.42,
+      "learning_rate": 1.203225806451613e-05,
+      "loss": 0.2244,
+      "step": 121430
+    },
+    {
+      "epoch": 783.48,
+      "learning_rate": 1.2028673835125448e-05,
+      "loss": 0.2078,
+      "step": 121440
+    },
+    {
+      "epoch": 783.55,
+      "learning_rate": 1.2025089605734768e-05,
+      "loss": 0.2647,
+      "step": 121450
+    },
+    {
+      "epoch": 783.61,
+      "learning_rate": 1.2021505376344086e-05,
+      "loss": 0.1826,
+      "step": 121460
+    },
+    {
+      "epoch": 783.68,
+      "learning_rate": 1.2017921146953404e-05,
+      "loss": 0.1932,
+      "step": 121470
+    },
+    {
+      "epoch": 783.74,
+      "learning_rate": 1.2014336917562724e-05,
+      "loss": 0.2394,
+      "step": 121480
+    },
+    {
+      "epoch": 783.8,
+      "learning_rate": 1.2010752688172042e-05,
+      "loss": 0.2986,
+      "step": 121490
+    },
+    {
+      "epoch": 783.87,
+      "learning_rate": 1.2007168458781362e-05,
+      "loss": 0.2136,
+      "step": 121500
+    },
+    {
+      "epoch": 783.93,
+      "learning_rate": 1.200358422939068e-05,
+      "loss": 0.2564,
+      "step": 121510
+    },
+    {
+      "epoch": 784.0,
+      "learning_rate": 1.2e-05,
+      "loss": 0.198,
+      "step": 121520
+    },
+    {
+      "epoch": 784.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9181596040725708,
+      "eval_runtime": 0.8707,
+      "eval_samples_per_second": 230.841,
+      "eval_steps_per_second": 8.039,
+      "step": 121520
+    },
+    {
+      "epoch": 784.06,
+      "learning_rate": 1.1996415770609319e-05,
+      "loss": 0.2169,
+      "step": 121530
+    },
+    {
+      "epoch": 784.13,
+      "learning_rate": 1.1992831541218638e-05,
+      "loss": 0.2156,
+      "step": 121540
+    },
+    {
+      "epoch": 784.19,
+      "learning_rate": 1.1989247311827957e-05,
+      "loss": 0.2009,
+      "step": 121550
+    },
+    {
+      "epoch": 784.26,
+      "learning_rate": 1.1985663082437277e-05,
+      "loss": 0.2092,
+      "step": 121560
+    },
+    {
+      "epoch": 784.32,
+      "learning_rate": 1.1982078853046595e-05,
+      "loss": 0.2435,
+      "step": 121570
+    },
+    {
+      "epoch": 784.39,
+      "learning_rate": 1.1978494623655915e-05,
+      "loss": 0.2082,
+      "step": 121580
+    },
+    {
+      "epoch": 784.45,
+      "learning_rate": 1.1974910394265233e-05,
+      "loss": 0.2775,
+      "step": 121590
+    },
+    {
+      "epoch": 784.51,
+      "learning_rate": 1.1971326164874553e-05,
+      "loss": 0.2514,
+      "step": 121600
+    },
+    {
+      "epoch": 784.58,
+      "learning_rate": 1.1967741935483871e-05,
+      "loss": 0.2467,
+      "step": 121610
+    },
+    {
+      "epoch": 784.64,
+      "learning_rate": 1.196415770609319e-05,
+      "loss": 0.1979,
+      "step": 121620
+    },
+    {
+      "epoch": 784.71,
+      "learning_rate": 1.1960573476702509e-05,
+      "loss": 0.2594,
+      "step": 121630
+    },
+    {
+      "epoch": 784.77,
+      "learning_rate": 1.1956989247311829e-05,
+      "loss": 0.1716,
+      "step": 121640
+    },
+    {
+      "epoch": 784.84,
+      "learning_rate": 1.1953405017921147e-05,
+      "loss": 0.2421,
+      "step": 121650
+    },
+    {
+      "epoch": 784.9,
+      "learning_rate": 1.1949820788530467e-05,
+      "loss": 0.2194,
+      "step": 121660
+    },
+    {
+      "epoch": 784.96,
+      "learning_rate": 1.1946236559139785e-05,
+      "loss": 0.2123,
+      "step": 121670
+    },
+    {
+      "epoch": 785.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9015013575553894,
+      "eval_runtime": 0.8649,
+      "eval_samples_per_second": 232.394,
+      "eval_steps_per_second": 8.093,
+      "step": 121675
+    },
+    {
+      "epoch": 785.03,
+      "learning_rate": 1.1942652329749105e-05,
+      "loss": 0.244,
+      "step": 121680
+    },
+    {
+      "epoch": 785.1,
+      "learning_rate": 1.1939068100358423e-05,
+      "loss": 0.2751,
+      "step": 121690
+    },
+    {
+      "epoch": 785.16,
+      "learning_rate": 1.1935483870967743e-05,
+      "loss": 0.2004,
+      "step": 121700
+    },
+    {
+      "epoch": 785.23,
+      "learning_rate": 1.1931899641577061e-05,
+      "loss": 0.1978,
+      "step": 121710
+    },
+    {
+      "epoch": 785.29,
+      "learning_rate": 1.192831541218638e-05,
+      "loss": 0.1679,
+      "step": 121720
+    },
+    {
+      "epoch": 785.35,
+      "learning_rate": 1.19247311827957e-05,
+      "loss": 0.2266,
+      "step": 121730
+    },
+    {
+      "epoch": 785.42,
+      "learning_rate": 1.1921146953405018e-05,
+      "loss": 0.2079,
+      "step": 121740
+    },
+    {
+      "epoch": 785.48,
+      "learning_rate": 1.1917562724014338e-05,
+      "loss": 0.1669,
+      "step": 121750
+    },
+    {
+      "epoch": 785.55,
+      "learning_rate": 1.1913978494623656e-05,
+      "loss": 0.1704,
+      "step": 121760
+    },
+    {
+      "epoch": 785.61,
+      "learning_rate": 1.1910394265232976e-05,
+      "loss": 0.1529,
+      "step": 121770
+    },
+    {
+      "epoch": 785.68,
+      "learning_rate": 1.1906810035842294e-05,
+      "loss": 0.2659,
+      "step": 121780
+    },
+    {
+      "epoch": 785.74,
+      "learning_rate": 1.1903225806451614e-05,
+      "loss": 0.2523,
+      "step": 121790
+    },
+    {
+      "epoch": 785.8,
+      "learning_rate": 1.1899641577060932e-05,
+      "loss": 0.2078,
+      "step": 121800
+    },
+    {
+      "epoch": 785.87,
+      "learning_rate": 1.1896057347670252e-05,
+      "loss": 0.1645,
+      "step": 121810
+    },
+    {
+      "epoch": 785.93,
+      "learning_rate": 1.189247311827957e-05,
+      "loss": 0.1751,
+      "step": 121820
+    },
+    {
+      "epoch": 786.0,
+      "learning_rate": 1.188888888888889e-05,
+      "loss": 0.1706,
+      "step": 121830
+    },
+    {
+      "epoch": 786.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9181354641914368,
+      "eval_runtime": 0.8791,
+      "eval_samples_per_second": 228.651,
+      "eval_steps_per_second": 7.963,
+      "step": 121830
+    },
+    {
+      "epoch": 786.06,
+      "learning_rate": 1.1885304659498208e-05,
+      "loss": 0.246,
+      "step": 121840
+    },
+    {
+      "epoch": 786.13,
+      "learning_rate": 1.1881720430107528e-05,
+      "loss": 0.2146,
+      "step": 121850
+    },
+    {
+      "epoch": 786.19,
+      "learning_rate": 1.1878136200716846e-05,
+      "loss": 0.1938,
+      "step": 121860
+    },
+    {
+      "epoch": 786.26,
+      "learning_rate": 1.1874551971326166e-05,
+      "loss": 0.2414,
+      "step": 121870
+    },
+    {
+      "epoch": 786.32,
+      "learning_rate": 1.1870967741935484e-05,
+      "loss": 0.1884,
+      "step": 121880
+    },
+    {
+      "epoch": 786.39,
+      "learning_rate": 1.1867383512544804e-05,
+      "loss": 0.1974,
+      "step": 121890
+    },
+    {
+      "epoch": 786.45,
+      "learning_rate": 1.1863799283154123e-05,
+      "loss": 0.2377,
+      "step": 121900
+    },
+    {
+      "epoch": 786.51,
+      "learning_rate": 1.1860215053763442e-05,
+      "loss": 0.2242,
+      "step": 121910
+    },
+    {
+      "epoch": 786.58,
+      "learning_rate": 1.185663082437276e-05,
+      "loss": 0.2005,
+      "step": 121920
+    },
+    {
+      "epoch": 786.64,
+      "learning_rate": 1.185304659498208e-05,
+      "loss": 0.2232,
+      "step": 121930
+    },
+    {
+      "epoch": 786.71,
+      "learning_rate": 1.1849462365591399e-05,
+      "loss": 0.196,
+      "step": 121940
+    },
+    {
+      "epoch": 786.77,
+      "learning_rate": 1.1845878136200719e-05,
+      "loss": 0.2254,
+      "step": 121950
+    },
+    {
+      "epoch": 786.84,
+      "learning_rate": 1.1842293906810037e-05,
+      "loss": 0.1796,
+      "step": 121960
+    },
+    {
+      "epoch": 786.9,
+      "learning_rate": 1.1838709677419355e-05,
+      "loss": 0.2236,
+      "step": 121970
+    },
+    {
+      "epoch": 786.96,
+      "learning_rate": 1.1835125448028675e-05,
+      "loss": 0.2443,
+      "step": 121980
+    },
+    {
+      "epoch": 787.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8912036418914795,
+      "eval_runtime": 0.8645,
+      "eval_samples_per_second": 232.497,
+      "eval_steps_per_second": 8.097,
+      "step": 121985
+    },
+    {
+      "epoch": 787.03,
+      "learning_rate": 1.1831541218637993e-05,
+      "loss": 0.1749,
+      "step": 121990
+    },
+    {
+      "epoch": 787.1,
+      "learning_rate": 1.1827956989247313e-05,
+      "loss": 0.2262,
+      "step": 122000
+    },
+    {
+      "epoch": 787.16,
+      "learning_rate": 1.1824372759856631e-05,
+      "loss": 0.2239,
+      "step": 122010
+    },
+    {
+      "epoch": 787.23,
+      "learning_rate": 1.1820788530465951e-05,
+      "loss": 0.2576,
+      "step": 122020
+    },
+    {
+      "epoch": 787.29,
+      "learning_rate": 1.181720430107527e-05,
+      "loss": 0.2303,
+      "step": 122030
+    },
+    {
+      "epoch": 787.35,
+      "learning_rate": 1.181362007168459e-05,
+      "loss": 0.2104,
+      "step": 122040
+    },
+    {
+      "epoch": 787.42,
+      "learning_rate": 1.1810035842293908e-05,
+      "loss": 0.1802,
+      "step": 122050
+    },
+    {
+      "epoch": 787.48,
+      "learning_rate": 1.1806451612903226e-05,
+      "loss": 0.204,
+      "step": 122060
+    },
+    {
+      "epoch": 787.55,
+      "learning_rate": 1.1802867383512544e-05,
+      "loss": 0.2202,
+      "step": 122070
+    },
+    {
+      "epoch": 787.61,
+      "learning_rate": 1.1799283154121864e-05,
+      "loss": 0.2394,
+      "step": 122080
+    },
+    {
+      "epoch": 787.68,
+      "learning_rate": 1.1795698924731182e-05,
+      "loss": 0.1834,
+      "step": 122090
+    },
+    {
+      "epoch": 787.74,
+      "learning_rate": 1.1792114695340502e-05,
+      "loss": 0.1784,
+      "step": 122100
+    },
+    {
+      "epoch": 787.8,
+      "learning_rate": 1.178853046594982e-05,
+      "loss": 0.1956,
+      "step": 122110
+    },
+    {
+      "epoch": 787.87,
+      "learning_rate": 1.178494623655914e-05,
+      "loss": 0.1798,
+      "step": 122120
+    },
+    {
+      "epoch": 787.93,
+      "learning_rate": 1.1781362007168458e-05,
+      "loss": 0.2363,
+      "step": 122130
+    },
+    {
+      "epoch": 788.0,
+      "learning_rate": 1.1777777777777778e-05,
+      "loss": 0.235,
+      "step": 122140
+    },
+    {
+      "epoch": 788.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9347364902496338,
+      "eval_runtime": 0.8631,
+      "eval_samples_per_second": 232.884,
+      "eval_steps_per_second": 8.11,
+      "step": 122140
+    },
+    {
+      "epoch": 788.06,
+      "learning_rate": 1.1774193548387096e-05,
+      "loss": 0.2128,
+      "step": 122150
+    },
+    {
+      "epoch": 788.13,
+      "learning_rate": 1.1770609318996416e-05,
+      "loss": 0.1918,
+      "step": 122160
+    },
+    {
+      "epoch": 788.19,
+      "learning_rate": 1.1767025089605734e-05,
+      "loss": 0.2139,
+      "step": 122170
+    },
+    {
+      "epoch": 788.26,
+      "learning_rate": 1.1763440860215054e-05,
+      "loss": 0.2602,
+      "step": 122180
+    },
+    {
+      "epoch": 788.32,
+      "learning_rate": 1.1759856630824373e-05,
+      "loss": 0.1693,
+      "step": 122190
+    },
+    {
+      "epoch": 788.39,
+      "learning_rate": 1.1756272401433692e-05,
+      "loss": 0.2068,
+      "step": 122200
+    },
+    {
+      "epoch": 788.45,
+      "learning_rate": 1.175268817204301e-05,
+      "loss": 0.1569,
+      "step": 122210
+    },
+    {
+      "epoch": 788.51,
+      "learning_rate": 1.174910394265233e-05,
+      "loss": 0.2272,
+      "step": 122220
+    },
+    {
+      "epoch": 788.58,
+      "learning_rate": 1.1745519713261649e-05,
+      "loss": 0.2067,
+      "step": 122230
+    },
+    {
+      "epoch": 788.64,
+      "learning_rate": 1.1741935483870969e-05,
+      "loss": 0.2483,
+      "step": 122240
+    },
+    {
+      "epoch": 788.71,
+      "learning_rate": 1.1738351254480287e-05,
+      "loss": 0.1982,
+      "step": 122250
+    },
+    {
+      "epoch": 788.77,
+      "learning_rate": 1.1734767025089607e-05,
+      "loss": 0.2356,
+      "step": 122260
+    },
+    {
+      "epoch": 788.84,
+      "learning_rate": 1.1731182795698925e-05,
+      "loss": 0.2297,
+      "step": 122270
+    },
+    {
+      "epoch": 788.9,
+      "learning_rate": 1.1727598566308245e-05,
+      "loss": 0.2041,
+      "step": 122280
+    },
+    {
+      "epoch": 788.96,
+      "learning_rate": 1.1724014336917563e-05,
+      "loss": 0.1957,
+      "step": 122290
+    },
+    {
+      "epoch": 789.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9655798077583313,
+      "eval_runtime": 0.8633,
+      "eval_samples_per_second": 232.817,
+      "eval_steps_per_second": 8.108,
+      "step": 122295
+    },
+    {
+      "epoch": 789.03,
+      "learning_rate": 1.1720430107526883e-05,
+      "loss": 0.1858,
+      "step": 122300
+    },
+    {
+      "epoch": 789.1,
+      "learning_rate": 1.1716845878136201e-05,
+      "loss": 0.1984,
+      "step": 122310
+    },
+    {
+      "epoch": 789.16,
+      "learning_rate": 1.171326164874552e-05,
+      "loss": 0.2145,
+      "step": 122320
+    },
+    {
+      "epoch": 789.23,
+      "learning_rate": 1.170967741935484e-05,
+      "loss": 0.2205,
+      "step": 122330
+    },
+    {
+      "epoch": 789.29,
+      "learning_rate": 1.1706093189964157e-05,
+      "loss": 0.2056,
+      "step": 122340
+    },
+    {
+      "epoch": 789.35,
+      "learning_rate": 1.1702508960573477e-05,
+      "loss": 0.2456,
+      "step": 122350
+    },
+    {
+      "epoch": 789.42,
+      "learning_rate": 1.1698924731182796e-05,
+      "loss": 0.2224,
+      "step": 122360
+    },
+    {
+      "epoch": 789.48,
+      "learning_rate": 1.1695340501792115e-05,
+      "loss": 0.1799,
+      "step": 122370
+    },
+    {
+      "epoch": 789.55,
+      "learning_rate": 1.1691756272401434e-05,
+      "loss": 0.2034,
+      "step": 122380
+    },
+    {
+      "epoch": 789.61,
+      "learning_rate": 1.1688172043010754e-05,
+      "loss": 0.2102,
+      "step": 122390
+    },
+    {
+      "epoch": 789.68,
+      "learning_rate": 1.1684587813620072e-05,
+      "loss": 0.2794,
+      "step": 122400
+    },
+    {
+      "epoch": 789.74,
+      "learning_rate": 1.1681003584229392e-05,
+      "loss": 0.1763,
+      "step": 122410
+    },
+    {
+      "epoch": 789.8,
+      "learning_rate": 1.167741935483871e-05,
+      "loss": 0.2364,
+      "step": 122420
+    },
+    {
+      "epoch": 789.87,
+      "learning_rate": 1.167383512544803e-05,
+      "loss": 0.2538,
+      "step": 122430
+    },
+    {
+      "epoch": 789.93,
+      "learning_rate": 1.1670250896057348e-05,
+      "loss": 0.2082,
+      "step": 122440
+    },
+    {
+      "epoch": 790.0,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 0.2381,
+      "step": 122450
+    },
+    {
+      "epoch": 790.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9534384608268738,
+      "eval_runtime": 0.8834,
+      "eval_samples_per_second": 227.529,
+      "eval_steps_per_second": 7.924,
+      "step": 122450
+    },
+    {
+      "epoch": 790.06,
+      "learning_rate": 1.1663082437275986e-05,
+      "loss": 0.2074,
+      "step": 122460
+    },
+    {
+      "epoch": 790.13,
+      "learning_rate": 1.1659498207885306e-05,
+      "loss": 0.2516,
+      "step": 122470
+    },
+    {
+      "epoch": 790.19,
+      "learning_rate": 1.1655913978494624e-05,
+      "loss": 0.2413,
+      "step": 122480
+    },
+    {
+      "epoch": 790.26,
+      "learning_rate": 1.1652329749103944e-05,
+      "loss": 0.177,
+      "step": 122490
+    },
+    {
+      "epoch": 790.32,
+      "learning_rate": 1.1648745519713262e-05,
+      "loss": 0.2193,
+      "step": 122500
+    },
+    {
+      "epoch": 790.39,
+      "learning_rate": 1.1645161290322582e-05,
+      "loss": 0.2102,
+      "step": 122510
+    },
+    {
+      "epoch": 790.45,
+      "learning_rate": 1.16415770609319e-05,
+      "loss": 0.2598,
+      "step": 122520
+    },
+    {
+      "epoch": 790.51,
+      "learning_rate": 1.163799283154122e-05,
+      "loss": 0.2095,
+      "step": 122530
+    },
+    {
+      "epoch": 790.58,
+      "learning_rate": 1.1634408602150539e-05,
+      "loss": 0.2047,
+      "step": 122540
+    },
+    {
+      "epoch": 790.64,
+      "learning_rate": 1.1630824372759858e-05,
+      "loss": 0.192,
+      "step": 122550
+    },
+    {
+      "epoch": 790.71,
+      "learning_rate": 1.1627240143369177e-05,
+      "loss": 0.1996,
+      "step": 122560
+    },
+    {
+      "epoch": 790.77,
+      "learning_rate": 1.1623655913978495e-05,
+      "loss": 0.2331,
+      "step": 122570
+    },
+    {
+      "epoch": 790.84,
+      "learning_rate": 1.1620071684587815e-05,
+      "loss": 0.1702,
+      "step": 122580
+    },
+    {
+      "epoch": 790.9,
+      "learning_rate": 1.1616487455197133e-05,
+      "loss": 0.2168,
+      "step": 122590
+    },
+    {
+      "epoch": 790.96,
+      "learning_rate": 1.1612903225806453e-05,
+      "loss": 0.214,
+      "step": 122600
+    },
+    {
+      "epoch": 791.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9115102291107178,
+      "eval_runtime": 0.8634,
+      "eval_samples_per_second": 232.8,
+      "eval_steps_per_second": 8.107,
+      "step": 122605
+    },
+    {
+      "epoch": 791.03,
+      "learning_rate": 1.1609318996415771e-05,
+      "loss": 0.2142,
+      "step": 122610
+    },
+    {
+      "epoch": 791.1,
+      "learning_rate": 1.1605734767025091e-05,
+      "loss": 0.2338,
+      "step": 122620
+    },
+    {
+      "epoch": 791.16,
+      "learning_rate": 1.1602150537634409e-05,
+      "loss": 0.1757,
+      "step": 122630
+    },
+    {
+      "epoch": 791.23,
+      "learning_rate": 1.1598566308243729e-05,
+      "loss": 0.186,
+      "step": 122640
+    },
+    {
+      "epoch": 791.29,
+      "learning_rate": 1.1594982078853047e-05,
+      "loss": 0.2336,
+      "step": 122650
+    },
+    {
+      "epoch": 791.35,
+      "learning_rate": 1.1591397849462367e-05,
+      "loss": 0.2335,
+      "step": 122660
+    },
+    {
+      "epoch": 791.42,
+      "learning_rate": 1.1587813620071685e-05,
+      "loss": 0.2118,
+      "step": 122670
+    },
+    {
+      "epoch": 791.48,
+      "learning_rate": 1.1584229390681005e-05,
+      "loss": 0.2311,
+      "step": 122680
+    },
+    {
+      "epoch": 791.55,
+      "learning_rate": 1.1580645161290322e-05,
+      "loss": 0.1955,
+      "step": 122690
+    },
+    {
+      "epoch": 791.61,
+      "learning_rate": 1.1577060931899642e-05,
+      "loss": 0.2272,
+      "step": 122700
+    },
+    {
+      "epoch": 791.68,
+      "learning_rate": 1.157347670250896e-05,
+      "loss": 0.2624,
+      "step": 122710
+    },
+    {
+      "epoch": 791.74,
+      "learning_rate": 1.156989247311828e-05,
+      "loss": 0.2191,
+      "step": 122720
+    },
+    {
+      "epoch": 791.8,
+      "learning_rate": 1.1566308243727598e-05,
+      "loss": 0.2092,
+      "step": 122730
+    },
+    {
+      "epoch": 791.87,
+      "learning_rate": 1.1562724014336918e-05,
+      "loss": 0.2413,
+      "step": 122740
+    },
+    {
+      "epoch": 791.93,
+      "learning_rate": 1.1559139784946236e-05,
+      "loss": 0.2344,
+      "step": 122750
+    },
+    {
+      "epoch": 792.0,
+      "learning_rate": 1.1555555555555556e-05,
+      "loss": 0.2448,
+      "step": 122760
+    },
+    {
+      "epoch": 792.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9192383885383606,
+      "eval_runtime": 0.8685,
+      "eval_samples_per_second": 231.43,
+      "eval_steps_per_second": 8.06,
+      "step": 122760
+    },
+    {
+      "epoch": 792.06,
+      "learning_rate": 1.1551971326164874e-05,
+      "loss": 0.1918,
+      "step": 122770
+    },
+    {
+      "epoch": 792.13,
+      "learning_rate": 1.1548387096774194e-05,
+      "loss": 0.2233,
+      "step": 122780
+    },
+    {
+      "epoch": 792.19,
+      "learning_rate": 1.1544802867383512e-05,
+      "loss": 0.2186,
+      "step": 122790
+    },
+    {
+      "epoch": 792.26,
+      "learning_rate": 1.1541218637992832e-05,
+      "loss": 0.1597,
+      "step": 122800
+    },
+    {
+      "epoch": 792.32,
+      "learning_rate": 1.153763440860215e-05,
+      "loss": 0.1625,
+      "step": 122810
+    },
+    {
+      "epoch": 792.39,
+      "learning_rate": 1.153405017921147e-05,
+      "loss": 0.2759,
+      "step": 122820
+    },
+    {
+      "epoch": 792.45,
+      "learning_rate": 1.1530465949820788e-05,
+      "loss": 0.1641,
+      "step": 122830
+    },
+    {
+      "epoch": 792.51,
+      "learning_rate": 1.1526881720430108e-05,
+      "loss": 0.2292,
+      "step": 122840
+    },
+    {
+      "epoch": 792.58,
+      "learning_rate": 1.1523297491039427e-05,
+      "loss": 0.2027,
+      "step": 122850
+    },
+    {
+      "epoch": 792.64,
+      "learning_rate": 1.1519713261648746e-05,
+      "loss": 0.2496,
+      "step": 122860
+    },
+    {
+      "epoch": 792.71,
+      "learning_rate": 1.1516129032258065e-05,
+      "loss": 0.2258,
+      "step": 122870
+    },
+    {
+      "epoch": 792.77,
+      "learning_rate": 1.1512544802867385e-05,
+      "loss": 0.2107,
+      "step": 122880
+    },
+    {
+      "epoch": 792.84,
+      "learning_rate": 1.1508960573476703e-05,
+      "loss": 0.2103,
+      "step": 122890
+    },
+    {
+      "epoch": 792.9,
+      "learning_rate": 1.1505376344086021e-05,
+      "loss": 0.2804,
+      "step": 122900
+    },
+    {
+      "epoch": 792.96,
+      "learning_rate": 1.1501792114695341e-05,
+      "loss": 0.2515,
+      "step": 122910
+    },
+    {
+      "epoch": 793.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9097501039505005,
+      "eval_runtime": 0.9591,
+      "eval_samples_per_second": 209.568,
+      "eval_steps_per_second": 7.298,
+      "step": 122915
+    },
+    {
+      "epoch": 793.03,
+      "learning_rate": 1.1498207885304659e-05,
+      "loss": 0.2241,
+      "step": 122920
+    },
+    {
+      "epoch": 793.1,
+      "learning_rate": 1.1494623655913979e-05,
+      "loss": 0.2226,
+      "step": 122930
+    },
+    {
+      "epoch": 793.16,
+      "learning_rate": 1.1491039426523297e-05,
+      "loss": 0.1984,
+      "step": 122940
+    },
+    {
+      "epoch": 793.23,
+      "learning_rate": 1.1487455197132617e-05,
+      "loss": 0.2809,
+      "step": 122950
+    },
+    {
+      "epoch": 793.29,
+      "learning_rate": 1.1483870967741935e-05,
+      "loss": 0.1954,
+      "step": 122960
+    },
+    {
+      "epoch": 793.35,
+      "learning_rate": 1.1480286738351255e-05,
+      "loss": 0.2304,
+      "step": 122970
+    },
+    {
+      "epoch": 793.42,
+      "learning_rate": 1.1476702508960573e-05,
+      "loss": 0.1606,
+      "step": 122980
+    },
+    {
+      "epoch": 793.48,
+      "learning_rate": 1.1473118279569893e-05,
+      "loss": 0.2411,
+      "step": 122990
+    },
+    {
+      "epoch": 793.55,
+      "learning_rate": 1.1469534050179212e-05,
+      "loss": 0.1936,
+      "step": 123000
+    },
+    {
+      "epoch": 793.61,
+      "learning_rate": 1.1465949820788531e-05,
+      "loss": 0.2193,
+      "step": 123010
+    },
+    {
+      "epoch": 793.68,
+      "learning_rate": 1.146236559139785e-05,
+      "loss": 0.2247,
+      "step": 123020
+    },
+    {
+      "epoch": 793.74,
+      "learning_rate": 1.145878136200717e-05,
+      "loss": 0.2216,
+      "step": 123030
+    },
+    {
+      "epoch": 793.8,
+      "learning_rate": 1.1455197132616488e-05,
+      "loss": 0.1904,
+      "step": 123040
+    },
+    {
+      "epoch": 793.87,
+      "learning_rate": 1.1451612903225808e-05,
+      "loss": 0.2435,
+      "step": 123050
+    },
+    {
+      "epoch": 793.93,
+      "learning_rate": 1.1448028673835126e-05,
+      "loss": 0.1954,
+      "step": 123060
+    },
+    {
+      "epoch": 794.0,
+      "learning_rate": 1.1444444444444446e-05,
+      "loss": 0.2457,
+      "step": 123070
+    },
+    {
+      "epoch": 794.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9127843976020813,
+      "eval_runtime": 0.8806,
+      "eval_samples_per_second": 228.241,
+      "eval_steps_per_second": 7.949,
+      "step": 123070
+    },
+    {
+      "epoch": 794.06,
+      "learning_rate": 1.1440860215053764e-05,
+      "loss": 0.2283,
+      "step": 123080
+    },
+    {
+      "epoch": 794.13,
+      "learning_rate": 1.1437275985663084e-05,
+      "loss": 0.2151,
+      "step": 123090
+    },
+    {
+      "epoch": 794.19,
+      "learning_rate": 1.1433691756272402e-05,
+      "loss": 0.2387,
+      "step": 123100
+    },
+    {
+      "epoch": 794.26,
+      "learning_rate": 1.1430107526881722e-05,
+      "loss": 0.2265,
+      "step": 123110
+    },
+    {
+      "epoch": 794.32,
+      "learning_rate": 1.142652329749104e-05,
+      "loss": 0.2138,
+      "step": 123120
+    },
+    {
+      "epoch": 794.39,
+      "learning_rate": 1.142293906810036e-05,
+      "loss": 0.2202,
+      "step": 123130
+    },
+    {
+      "epoch": 794.45,
+      "learning_rate": 1.1419354838709678e-05,
+      "loss": 0.1879,
+      "step": 123140
+    },
+    {
+      "epoch": 794.51,
+      "learning_rate": 1.1415770609318996e-05,
+      "loss": 0.2005,
+      "step": 123150
+    },
+    {
+      "epoch": 794.58,
+      "learning_rate": 1.1412186379928316e-05,
+      "loss": 0.1819,
+      "step": 123160
+    },
+    {
+      "epoch": 794.64,
+      "learning_rate": 1.1408602150537635e-05,
+      "loss": 0.2288,
+      "step": 123170
+    },
+    {
+      "epoch": 794.71,
+      "learning_rate": 1.1405017921146954e-05,
+      "loss": 0.199,
+      "step": 123180
+    },
+    {
+      "epoch": 794.77,
+      "learning_rate": 1.1401433691756273e-05,
+      "loss": 0.2223,
+      "step": 123190
+    },
+    {
+      "epoch": 794.84,
+      "learning_rate": 1.1397849462365593e-05,
+      "loss": 0.202,
+      "step": 123200
+    },
+    {
+      "epoch": 794.9,
+      "learning_rate": 1.139426523297491e-05,
+      "loss": 0.2692,
+      "step": 123210
+    },
+    {
+      "epoch": 794.96,
+      "learning_rate": 1.139068100358423e-05,
+      "loss": 0.2036,
+      "step": 123220
+    },
+    {
+      "epoch": 795.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9622470736503601,
+      "eval_runtime": 0.9486,
+      "eval_samples_per_second": 211.881,
+      "eval_steps_per_second": 7.379,
+      "step": 123225
+    },
+    {
+      "epoch": 795.03,
+      "learning_rate": 1.1387096774193549e-05,
+      "loss": 0.2048,
+      "step": 123230
+    },
+    {
+      "epoch": 795.1,
+      "learning_rate": 1.1383512544802869e-05,
+      "loss": 0.198,
+      "step": 123240
+    },
+    {
+      "epoch": 795.16,
+      "learning_rate": 1.1379928315412187e-05,
+      "loss": 0.2241,
+      "step": 123250
+    },
+    {
+      "epoch": 795.23,
+      "learning_rate": 1.1376344086021507e-05,
+      "loss": 0.2149,
+      "step": 123260
+    },
+    {
+      "epoch": 795.29,
+      "learning_rate": 1.1372759856630825e-05,
+      "loss": 0.2666,
+      "step": 123270
+    },
+    {
+      "epoch": 795.35,
+      "learning_rate": 1.1369175627240145e-05,
+      "loss": 0.2463,
+      "step": 123280
+    },
+    {
+      "epoch": 795.42,
+      "learning_rate": 1.1365591397849463e-05,
+      "loss": 0.1789,
+      "step": 123290
+    },
+    {
+      "epoch": 795.48,
+      "learning_rate": 1.1362007168458783e-05,
+      "loss": 0.2507,
+      "step": 123300
+    },
+    {
+      "epoch": 795.55,
+      "learning_rate": 1.1358422939068101e-05,
+      "loss": 0.197,
+      "step": 123310
+    },
+    {
+      "epoch": 795.61,
+      "learning_rate": 1.1354838709677421e-05,
+      "loss": 0.2249,
+      "step": 123320
+    },
+    {
+      "epoch": 795.68,
+      "learning_rate": 1.1351254480286738e-05,
+      "loss": 0.2436,
+      "step": 123330
+    },
+    {
+      "epoch": 795.74,
+      "learning_rate": 1.1347670250896058e-05,
+      "loss": 0.2202,
+      "step": 123340
+    },
+    {
+      "epoch": 795.8,
+      "learning_rate": 1.1344086021505376e-05,
+      "loss": 0.2038,
+      "step": 123350
+    },
+    {
+      "epoch": 795.87,
+      "learning_rate": 1.1340501792114696e-05,
+      "loss": 0.2184,
+      "step": 123360
+    },
+    {
+      "epoch": 795.93,
+      "learning_rate": 1.1336917562724014e-05,
+      "loss": 0.2067,
+      "step": 123370
+    },
+    {
+      "epoch": 796.0,
+      "learning_rate": 1.1333333333333334e-05,
+      "loss": 0.2094,
+      "step": 123380
+    },
+    {
+      "epoch": 796.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9725513458251953,
+      "eval_runtime": 0.9064,
+      "eval_samples_per_second": 221.765,
+      "eval_steps_per_second": 7.723,
+      "step": 123380
+    },
+    {
+      "epoch": 796.06,
+      "learning_rate": 1.1329749103942652e-05,
+      "loss": 0.2047,
+      "step": 123390
+    },
+    {
+      "epoch": 796.13,
+      "learning_rate": 1.1326164874551972e-05,
+      "loss": 0.2273,
+      "step": 123400
+    },
+    {
+      "epoch": 796.19,
+      "learning_rate": 1.132258064516129e-05,
+      "loss": 0.2242,
+      "step": 123410
+    },
+    {
+      "epoch": 796.26,
+      "learning_rate": 1.131899641577061e-05,
+      "loss": 0.2655,
+      "step": 123420
+    },
+    {
+      "epoch": 796.32,
+      "learning_rate": 1.1315412186379928e-05,
+      "loss": 0.1821,
+      "step": 123430
+    },
+    {
+      "epoch": 796.39,
+      "learning_rate": 1.1311827956989248e-05,
+      "loss": 0.1583,
+      "step": 123440
+    },
+    {
+      "epoch": 796.45,
+      "learning_rate": 1.1308243727598566e-05,
+      "loss": 0.2185,
+      "step": 123450
+    },
+    {
+      "epoch": 796.51,
+      "learning_rate": 1.1304659498207886e-05,
+      "loss": 0.2611,
+      "step": 123460
+    },
+    {
+      "epoch": 796.58,
+      "learning_rate": 1.1301075268817204e-05,
+      "loss": 0.2374,
+      "step": 123470
+    },
+    {
+      "epoch": 796.64,
+      "learning_rate": 1.1297491039426524e-05,
+      "loss": 0.2835,
+      "step": 123480
+    },
+    {
+      "epoch": 796.71,
+      "learning_rate": 1.1293906810035843e-05,
+      "loss": 0.2305,
+      "step": 123490
+    },
+    {
+      "epoch": 796.77,
+      "learning_rate": 1.129032258064516e-05,
+      "loss": 0.193,
+      "step": 123500
+    },
+    {
+      "epoch": 796.84,
+      "learning_rate": 1.128673835125448e-05,
+      "loss": 0.2467,
+      "step": 123510
+    },
+    {
+      "epoch": 796.9,
+      "learning_rate": 1.1283154121863799e-05,
+      "loss": 0.2056,
+      "step": 123520
+    },
+    {
+      "epoch": 796.96,
+      "learning_rate": 1.1279569892473119e-05,
+      "loss": 0.2007,
+      "step": 123530
+    },
+    {
+      "epoch": 797.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.946071207523346,
+      "eval_runtime": 0.884,
+      "eval_samples_per_second": 227.381,
+      "eval_steps_per_second": 7.919,
+      "step": 123535
+    },
+    {
+      "epoch": 797.03,
+      "learning_rate": 1.1275985663082437e-05,
+      "loss": 0.212,
+      "step": 123540
+    },
+    {
+      "epoch": 797.1,
+      "learning_rate": 1.1272401433691757e-05,
+      "loss": 0.2066,
+      "step": 123550
+    },
+    {
+      "epoch": 797.16,
+      "learning_rate": 1.1268817204301075e-05,
+      "loss": 0.2496,
+      "step": 123560
+    },
+    {
+      "epoch": 797.23,
+      "learning_rate": 1.1265232974910395e-05,
+      "loss": 0.1921,
+      "step": 123570
+    },
+    {
+      "epoch": 797.29,
+      "learning_rate": 1.1261648745519713e-05,
+      "loss": 0.264,
+      "step": 123580
+    },
+    {
+      "epoch": 797.35,
+      "learning_rate": 1.1258064516129033e-05,
+      "loss": 0.22,
+      "step": 123590
+    },
+    {
+      "epoch": 797.42,
+      "learning_rate": 1.1254480286738351e-05,
+      "loss": 0.2149,
+      "step": 123600
+    },
+    {
+      "epoch": 797.48,
+      "learning_rate": 1.1250896057347671e-05,
+      "loss": 0.1958,
+      "step": 123610
+    },
+    {
+      "epoch": 797.55,
+      "learning_rate": 1.124731182795699e-05,
+      "loss": 0.1506,
+      "step": 123620
+    },
+    {
+      "epoch": 797.61,
+      "learning_rate": 1.124372759856631e-05,
+      "loss": 0.2027,
+      "step": 123630
+    },
+    {
+      "epoch": 797.68,
+      "learning_rate": 1.1240143369175627e-05,
+      "loss": 0.2284,
+      "step": 123640
+    },
+    {
+      "epoch": 797.74,
+      "learning_rate": 1.1236559139784947e-05,
+      "loss": 0.2286,
+      "step": 123650
+    },
+    {
+      "epoch": 797.8,
+      "learning_rate": 1.1232974910394266e-05,
+      "loss": 0.236,
+      "step": 123660
+    },
+    {
+      "epoch": 797.87,
+      "learning_rate": 1.1229390681003585e-05,
+      "loss": 0.1612,
+      "step": 123670
+    },
+    {
+      "epoch": 797.93,
+      "learning_rate": 1.1225806451612904e-05,
+      "loss": 0.1897,
+      "step": 123680
+    },
+    {
+      "epoch": 798.0,
+      "learning_rate": 1.1222222222222224e-05,
+      "loss": 0.2156,
+      "step": 123690
+    },
+    {
+      "epoch": 798.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9193282127380371,
+      "eval_runtime": 0.8809,
+      "eval_samples_per_second": 228.166,
+      "eval_steps_per_second": 7.946,
+      "step": 123690
+    },
+    {
+      "epoch": 798.06,
+      "learning_rate": 1.1218637992831542e-05,
+      "loss": 0.2177,
+      "step": 123700
+    },
+    {
+      "epoch": 798.13,
+      "learning_rate": 1.1215053763440862e-05,
+      "loss": 0.2195,
+      "step": 123710
+    },
+    {
+      "epoch": 798.19,
+      "learning_rate": 1.121146953405018e-05,
+      "loss": 0.1956,
+      "step": 123720
+    },
+    {
+      "epoch": 798.26,
+      "learning_rate": 1.1207885304659498e-05,
+      "loss": 0.2191,
+      "step": 123730
+    },
+    {
+      "epoch": 798.32,
+      "learning_rate": 1.1204301075268818e-05,
+      "loss": 0.2551,
+      "step": 123740
+    },
+    {
+      "epoch": 798.39,
+      "learning_rate": 1.1200716845878136e-05,
+      "loss": 0.1554,
+      "step": 123750
+    },
+    {
+      "epoch": 798.45,
+      "learning_rate": 1.1197132616487456e-05,
+      "loss": 0.2326,
+      "step": 123760
+    },
+    {
+      "epoch": 798.51,
+      "learning_rate": 1.1193548387096774e-05,
+      "loss": 0.1611,
+      "step": 123770
+    },
+    {
+      "epoch": 798.58,
+      "learning_rate": 1.1189964157706094e-05,
+      "loss": 0.2378,
+      "step": 123780
+    },
+    {
+      "epoch": 798.64,
+      "learning_rate": 1.1186379928315412e-05,
+      "loss": 0.2275,
+      "step": 123790
+    },
+    {
+      "epoch": 798.71,
+      "learning_rate": 1.1182795698924732e-05,
+      "loss": 0.1619,
+      "step": 123800
+    },
+    {
+      "epoch": 798.77,
+      "learning_rate": 1.117921146953405e-05,
+      "loss": 0.2277,
+      "step": 123810
+    },
+    {
+      "epoch": 798.84,
+      "learning_rate": 1.117562724014337e-05,
+      "loss": 0.2667,
+      "step": 123820
+    },
+    {
+      "epoch": 798.9,
+      "learning_rate": 1.1172043010752689e-05,
+      "loss": 0.2477,
+      "step": 123830
+    },
+    {
+      "epoch": 798.96,
+      "learning_rate": 1.1168458781362009e-05,
+      "loss": 0.2407,
+      "step": 123840
+    },
+    {
+      "epoch": 799.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9262321591377258,
+      "eval_runtime": 0.907,
+      "eval_samples_per_second": 221.598,
+      "eval_steps_per_second": 7.717,
+      "step": 123845
+    },
+    {
+      "epoch": 799.03,
+      "learning_rate": 1.1164874551971327e-05,
+      "loss": 0.2612,
+      "step": 123850
+    },
+    {
+      "epoch": 799.1,
+      "learning_rate": 1.1161290322580647e-05,
+      "loss": 0.2486,
+      "step": 123860
+    },
+    {
+      "epoch": 799.16,
+      "learning_rate": 1.1157706093189965e-05,
+      "loss": 0.1572,
+      "step": 123870
+    },
+    {
+      "epoch": 799.23,
+      "learning_rate": 1.1154121863799285e-05,
+      "loss": 0.1878,
+      "step": 123880
+    },
+    {
+      "epoch": 799.29,
+      "learning_rate": 1.1150537634408603e-05,
+      "loss": 0.2024,
+      "step": 123890
+    },
+    {
+      "epoch": 799.35,
+      "learning_rate": 1.1146953405017923e-05,
+      "loss": 0.2456,
+      "step": 123900
+    },
+    {
+      "epoch": 799.42,
+      "learning_rate": 1.1143369175627241e-05,
+      "loss": 0.2416,
+      "step": 123910
+    },
+    {
+      "epoch": 799.48,
+      "learning_rate": 1.1139784946236561e-05,
+      "loss": 0.1941,
+      "step": 123920
+    },
+    {
+      "epoch": 799.55,
+      "learning_rate": 1.1136200716845879e-05,
+      "loss": 0.2212,
+      "step": 123930
+    },
+    {
+      "epoch": 799.61,
+      "learning_rate": 1.1132616487455199e-05,
+      "loss": 0.2839,
+      "step": 123940
+    },
+    {
+      "epoch": 799.68,
+      "learning_rate": 1.1129032258064517e-05,
+      "loss": 0.2125,
+      "step": 123950
+    },
+    {
+      "epoch": 799.74,
+      "learning_rate": 1.1125448028673837e-05,
+      "loss": 0.171,
+      "step": 123960
+    },
+    {
+      "epoch": 799.8,
+      "learning_rate": 1.1121863799283154e-05,
+      "loss": 0.2062,
+      "step": 123970
+    },
+    {
+      "epoch": 799.87,
+      "learning_rate": 1.1118279569892474e-05,
+      "loss": 0.1822,
+      "step": 123980
+    },
+    {
+      "epoch": 799.93,
+      "learning_rate": 1.1114695340501792e-05,
+      "loss": 0.1828,
+      "step": 123990
+    },
+    {
+      "epoch": 800.0,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 0.1948,
+      "step": 124000
+    },
+    {
+      "epoch": 800.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9411609172821045,
+      "eval_runtime": 0.8654,
+      "eval_samples_per_second": 232.261,
+      "eval_steps_per_second": 8.089,
+      "step": 124000
+    },
+    {
+      "epoch": 800.06,
+      "learning_rate": 1.110752688172043e-05,
+      "loss": 0.1989,
+      "step": 124010
+    },
+    {
+      "epoch": 800.13,
+      "learning_rate": 1.110394265232975e-05,
+      "loss": 0.2288,
+      "step": 124020
+    },
+    {
+      "epoch": 800.19,
+      "learning_rate": 1.1100358422939068e-05,
+      "loss": 0.1972,
+      "step": 124030
+    },
+    {
+      "epoch": 800.26,
+      "learning_rate": 1.1096774193548388e-05,
+      "loss": 0.2591,
+      "step": 124040
+    },
+    {
+      "epoch": 800.32,
+      "learning_rate": 1.1093189964157706e-05,
+      "loss": 0.1777,
+      "step": 124050
+    },
+    {
+      "epoch": 800.39,
+      "learning_rate": 1.1089605734767026e-05,
+      "loss": 0.2416,
+      "step": 124060
+    },
+    {
+      "epoch": 800.45,
+      "learning_rate": 1.1086021505376344e-05,
+      "loss": 0.2359,
+      "step": 124070
+    },
+    {
+      "epoch": 800.51,
+      "learning_rate": 1.1082437275985662e-05,
+      "loss": 0.2469,
+      "step": 124080
+    },
+    {
+      "epoch": 800.58,
+      "learning_rate": 1.1078853046594982e-05,
+      "loss": 0.2323,
+      "step": 124090
+    },
+    {
+      "epoch": 800.64,
+      "learning_rate": 1.10752688172043e-05,
+      "loss": 0.2345,
+      "step": 124100
+    },
+    {
+      "epoch": 800.71,
+      "learning_rate": 1.107168458781362e-05,
+      "loss": 0.2872,
+      "step": 124110
+    },
+    {
+      "epoch": 800.77,
+      "learning_rate": 1.1068100358422939e-05,
+      "loss": 0.2169,
+      "step": 124120
+    },
+    {
+      "epoch": 800.84,
+      "learning_rate": 1.1064516129032258e-05,
+      "loss": 0.2343,
+      "step": 124130
+    },
+    {
+      "epoch": 800.9,
+      "learning_rate": 1.1060931899641577e-05,
+      "loss": 0.212,
+      "step": 124140
+    },
+    {
+      "epoch": 800.96,
+      "learning_rate": 1.1057347670250897e-05,
+      "loss": 0.2269,
+      "step": 124150
+    },
+    {
+      "epoch": 801.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9304038286209106,
+      "eval_runtime": 0.8729,
+      "eval_samples_per_second": 230.272,
+      "eval_steps_per_second": 8.019,
+      "step": 124155
+    },
+    {
+      "epoch": 801.03,
+      "learning_rate": 1.1053763440860215e-05,
+      "loss": 0.2239,
+      "step": 124160
+    },
+    {
+      "epoch": 801.1,
+      "learning_rate": 1.1050179211469535e-05,
+      "loss": 0.204,
+      "step": 124170
+    },
+    {
+      "epoch": 801.16,
+      "learning_rate": 1.1046594982078853e-05,
+      "loss": 0.2501,
+      "step": 124180
+    },
+    {
+      "epoch": 801.23,
+      "learning_rate": 1.1043010752688173e-05,
+      "loss": 0.2359,
+      "step": 124190
+    },
+    {
+      "epoch": 801.29,
+      "learning_rate": 1.1039426523297491e-05,
+      "loss": 0.2447,
+      "step": 124200
+    },
+    {
+      "epoch": 801.35,
+      "learning_rate": 1.1035842293906811e-05,
+      "loss": 0.1954,
+      "step": 124210
+    },
+    {
+      "epoch": 801.42,
+      "learning_rate": 1.1032258064516129e-05,
+      "loss": 0.1938,
+      "step": 124220
+    },
+    {
+      "epoch": 801.48,
+      "learning_rate": 1.1028673835125449e-05,
+      "loss": 0.2143,
+      "step": 124230
+    },
+    {
+      "epoch": 801.55,
+      "learning_rate": 1.1025089605734767e-05,
+      "loss": 0.1716,
+      "step": 124240
+    },
+    {
+      "epoch": 801.61,
+      "learning_rate": 1.1021505376344087e-05,
+      "loss": 0.1984,
+      "step": 124250
+    },
+    {
+      "epoch": 801.68,
+      "learning_rate": 1.1017921146953405e-05,
+      "loss": 0.2637,
+      "step": 124260
+    },
+    {
+      "epoch": 801.74,
+      "learning_rate": 1.1014336917562725e-05,
+      "loss": 0.2084,
+      "step": 124270
+    },
+    {
+      "epoch": 801.8,
+      "learning_rate": 1.1010752688172043e-05,
+      "loss": 0.2086,
+      "step": 124280
+    },
+    {
+      "epoch": 801.87,
+      "learning_rate": 1.1007168458781363e-05,
+      "loss": 0.2183,
+      "step": 124290
+    },
+    {
+      "epoch": 801.93,
+      "learning_rate": 1.1003584229390682e-05,
+      "loss": 0.1971,
+      "step": 124300
+    },
+    {
+      "epoch": 802.0,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.2158,
+      "step": 124310
+    },
+    {
+      "epoch": 802.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9251276254653931,
+      "eval_runtime": 0.8667,
+      "eval_samples_per_second": 231.908,
+      "eval_steps_per_second": 8.076,
+      "step": 124310
+    },
+    {
+      "epoch": 802.06,
+      "learning_rate": 1.099641577060932e-05,
+      "loss": 0.2528,
+      "step": 124320
+    },
+    {
+      "epoch": 802.13,
+      "learning_rate": 1.0992831541218638e-05,
+      "loss": 0.2065,
+      "step": 124330
+    },
+    {
+      "epoch": 802.19,
+      "learning_rate": 1.0989247311827958e-05,
+      "loss": 0.1931,
+      "step": 124340
+    },
+    {
+      "epoch": 802.26,
+      "learning_rate": 1.0985663082437276e-05,
+      "loss": 0.2233,
+      "step": 124350
+    },
+    {
+      "epoch": 802.32,
+      "learning_rate": 1.0982078853046596e-05,
+      "loss": 0.2322,
+      "step": 124360
+    },
+    {
+      "epoch": 802.39,
+      "learning_rate": 1.0978494623655914e-05,
+      "loss": 0.2025,
+      "step": 124370
+    },
+    {
+      "epoch": 802.45,
+      "learning_rate": 1.0974910394265234e-05,
+      "loss": 0.2453,
+      "step": 124380
+    },
+    {
+      "epoch": 802.51,
+      "learning_rate": 1.0971326164874552e-05,
+      "loss": 0.2074,
+      "step": 124390
+    },
+    {
+      "epoch": 802.58,
+      "learning_rate": 1.0967741935483872e-05,
+      "loss": 0.199,
+      "step": 124400
+    },
+    {
+      "epoch": 802.64,
+      "learning_rate": 1.096415770609319e-05,
+      "loss": 0.2246,
+      "step": 124410
+    },
+    {
+      "epoch": 802.71,
+      "learning_rate": 1.096057347670251e-05,
+      "loss": 0.2602,
+      "step": 124420
+    },
+    {
+      "epoch": 802.77,
+      "learning_rate": 1.0956989247311828e-05,
+      "loss": 0.2138,
+      "step": 124430
+    },
+    {
+      "epoch": 802.84,
+      "learning_rate": 1.0953405017921148e-05,
+      "loss": 0.1973,
+      "step": 124440
+    },
+    {
+      "epoch": 802.9,
+      "learning_rate": 1.0949820788530466e-05,
+      "loss": 0.2002,
+      "step": 124450
+    },
+    {
+      "epoch": 802.96,
+      "learning_rate": 1.0946236559139786e-05,
+      "loss": 0.257,
+      "step": 124460
+    },
+    {
+      "epoch": 803.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9244642853736877,
+      "eval_runtime": 0.8745,
+      "eval_samples_per_second": 229.856,
+      "eval_steps_per_second": 8.005,
+      "step": 124465
+    },
+    {
+      "epoch": 803.03,
+      "learning_rate": 1.0942652329749105e-05,
+      "loss": 0.2083,
+      "step": 124470
+    },
+    {
+      "epoch": 803.1,
+      "learning_rate": 1.0939068100358424e-05,
+      "loss": 0.1896,
+      "step": 124480
+    },
+    {
+      "epoch": 803.16,
+      "learning_rate": 1.0935483870967743e-05,
+      "loss": 0.1782,
+      "step": 124490
+    },
+    {
+      "epoch": 803.23,
+      "learning_rate": 1.0931899641577063e-05,
+      "loss": 0.2309,
+      "step": 124500
+    },
+    {
+      "epoch": 803.29,
+      "learning_rate": 1.092831541218638e-05,
+      "loss": 0.1593,
+      "step": 124510
+    },
+    {
+      "epoch": 803.35,
+      "learning_rate": 1.09247311827957e-05,
+      "loss": 0.1889,
+      "step": 124520
+    },
+    {
+      "epoch": 803.42,
+      "learning_rate": 1.0921146953405019e-05,
+      "loss": 0.1833,
+      "step": 124530
+    },
+    {
+      "epoch": 803.48,
+      "learning_rate": 1.0917562724014339e-05,
+      "loss": 0.1845,
+      "step": 124540
+    },
+    {
+      "epoch": 803.55,
+      "learning_rate": 1.0913978494623657e-05,
+      "loss": 0.2049,
+      "step": 124550
+    },
+    {
+      "epoch": 803.61,
+      "learning_rate": 1.0910394265232977e-05,
+      "loss": 0.1822,
+      "step": 124560
+    },
+    {
+      "epoch": 803.68,
+      "learning_rate": 1.0906810035842295e-05,
+      "loss": 0.1926,
+      "step": 124570
+    },
+    {
+      "epoch": 803.74,
+      "learning_rate": 1.0903225806451613e-05,
+      "loss": 0.2019,
+      "step": 124580
+    },
+    {
+      "epoch": 803.8,
+      "learning_rate": 1.0899641577060933e-05,
+      "loss": 0.2104,
+      "step": 124590
+    },
+    {
+      "epoch": 803.87,
+      "learning_rate": 1.0896057347670251e-05,
+      "loss": 0.2369,
+      "step": 124600
+    },
+    {
+      "epoch": 803.93,
+      "learning_rate": 1.089247311827957e-05,
+      "loss": 0.2122,
+      "step": 124610
+    },
+    {
+      "epoch": 804.0,
+      "learning_rate": 1.088888888888889e-05,
+      "loss": 0.2163,
+      "step": 124620
+    },
+    {
+      "epoch": 804.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9462780356407166,
+      "eval_runtime": 0.8925,
+      "eval_samples_per_second": 225.216,
+      "eval_steps_per_second": 7.843,
+      "step": 124620
+    },
+    {
+      "epoch": 804.06,
+      "learning_rate": 1.0885304659498208e-05,
+      "loss": 0.226,
+      "step": 124630
+    },
+    {
+      "epoch": 804.13,
+      "learning_rate": 1.0881720430107528e-05,
+      "loss": 0.2478,
+      "step": 124640
+    },
+    {
+      "epoch": 804.19,
+      "learning_rate": 1.0878136200716846e-05,
+      "loss": 0.2341,
+      "step": 124650
+    },
+    {
+      "epoch": 804.26,
+      "learning_rate": 1.0874551971326164e-05,
+      "loss": 0.2196,
+      "step": 124660
+    },
+    {
+      "epoch": 804.32,
+      "learning_rate": 1.0870967741935484e-05,
+      "loss": 0.2361,
+      "step": 124670
+    },
+    {
+      "epoch": 804.39,
+      "learning_rate": 1.0867383512544802e-05,
+      "loss": 0.191,
+      "step": 124680
+    },
+    {
+      "epoch": 804.45,
+      "learning_rate": 1.0863799283154122e-05,
+      "loss": 0.2402,
+      "step": 124690
+    },
+    {
+      "epoch": 804.51,
+      "learning_rate": 1.086021505376344e-05,
+      "loss": 0.2082,
+      "step": 124700
+    },
+    {
+      "epoch": 804.58,
+      "learning_rate": 1.085663082437276e-05,
+      "loss": 0.2083,
+      "step": 124710
+    },
+    {
+      "epoch": 804.64,
+      "learning_rate": 1.0853046594982078e-05,
+      "loss": 0.1844,
+      "step": 124720
+    },
+    {
+      "epoch": 804.71,
+      "learning_rate": 1.0849462365591398e-05,
+      "loss": 0.225,
+      "step": 124730
+    },
+    {
+      "epoch": 804.77,
+      "learning_rate": 1.0845878136200716e-05,
+      "loss": 0.2338,
+      "step": 124740
+    },
+    {
+      "epoch": 804.84,
+      "learning_rate": 1.0842293906810036e-05,
+      "loss": 0.2485,
+      "step": 124750
+    },
+    {
+      "epoch": 804.9,
+      "learning_rate": 1.0838709677419355e-05,
+      "loss": 0.2003,
+      "step": 124760
+    },
+    {
+      "epoch": 804.96,
+      "learning_rate": 1.0835125448028674e-05,
+      "loss": 0.2504,
+      "step": 124770
+    },
+    {
+      "epoch": 805.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9168118238449097,
+      "eval_runtime": 0.903,
+      "eval_samples_per_second": 222.594,
+      "eval_steps_per_second": 7.752,
+      "step": 124775
+    },
+    {
+      "epoch": 805.03,
+      "learning_rate": 1.0831541218637993e-05,
+      "loss": 0.236,
+      "step": 124780
+    },
+    {
+      "epoch": 805.1,
+      "learning_rate": 1.0827956989247313e-05,
+      "loss": 0.2038,
+      "step": 124790
+    },
+    {
+      "epoch": 805.16,
+      "learning_rate": 1.082437275985663e-05,
+      "loss": 0.2275,
+      "step": 124800
+    },
+    {
+      "epoch": 805.23,
+      "learning_rate": 1.082078853046595e-05,
+      "loss": 0.2352,
+      "step": 124810
+    },
+    {
+      "epoch": 805.29,
+      "learning_rate": 1.0817204301075269e-05,
+      "loss": 0.2343,
+      "step": 124820
+    },
+    {
+      "epoch": 805.35,
+      "learning_rate": 1.0813620071684589e-05,
+      "loss": 0.2315,
+      "step": 124830
+    },
+    {
+      "epoch": 805.42,
+      "learning_rate": 1.0810035842293907e-05,
+      "loss": 0.1797,
+      "step": 124840
+    },
+    {
+      "epoch": 805.48,
+      "learning_rate": 1.0806451612903227e-05,
+      "loss": 0.2055,
+      "step": 124850
+    },
+    {
+      "epoch": 805.55,
+      "learning_rate": 1.0802867383512545e-05,
+      "loss": 0.2352,
+      "step": 124860
+    },
+    {
+      "epoch": 805.61,
+      "learning_rate": 1.0799283154121865e-05,
+      "loss": 0.1754,
+      "step": 124870
+    },
+    {
+      "epoch": 805.68,
+      "learning_rate": 1.0795698924731183e-05,
+      "loss": 0.1953,
+      "step": 124880
+    },
+    {
+      "epoch": 805.74,
+      "learning_rate": 1.0792114695340503e-05,
+      "loss": 0.2148,
+      "step": 124890
+    },
+    {
+      "epoch": 805.8,
+      "learning_rate": 1.0788530465949821e-05,
+      "loss": 0.2001,
+      "step": 124900
+    },
+    {
+      "epoch": 805.87,
+      "learning_rate": 1.078494623655914e-05,
+      "loss": 0.194,
+      "step": 124910
+    },
+    {
+      "epoch": 805.93,
+      "learning_rate": 1.078136200716846e-05,
+      "loss": 0.2494,
+      "step": 124920
+    },
+    {
+      "epoch": 806.0,
+      "learning_rate": 1.0777777777777778e-05,
+      "loss": 0.1921,
+      "step": 124930
+    },
+    {
+      "epoch": 806.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9356933236122131,
+      "eval_runtime": 0.8766,
+      "eval_samples_per_second": 229.293,
+      "eval_steps_per_second": 7.985,
+      "step": 124930
+    },
+    {
+      "epoch": 806.06,
+      "learning_rate": 1.0774193548387097e-05,
+      "loss": 0.2298,
+      "step": 124940
+    },
+    {
+      "epoch": 806.13,
+      "learning_rate": 1.0770609318996416e-05,
+      "loss": 0.211,
+      "step": 124950
+    },
+    {
+      "epoch": 806.19,
+      "learning_rate": 1.0767025089605736e-05,
+      "loss": 0.237,
+      "step": 124960
+    },
+    {
+      "epoch": 806.26,
+      "learning_rate": 1.0763440860215054e-05,
+      "loss": 0.1989,
+      "step": 124970
+    },
+    {
+      "epoch": 806.32,
+      "learning_rate": 1.0759856630824374e-05,
+      "loss": 0.2198,
+      "step": 124980
+    },
+    {
+      "epoch": 806.39,
+      "learning_rate": 1.0756272401433692e-05,
+      "loss": 0.2399,
+      "step": 124990
+    },
+    {
+      "epoch": 806.45,
+      "learning_rate": 1.0752688172043012e-05,
+      "loss": 0.2125,
+      "step": 125000
+    },
+    {
+      "epoch": 806.51,
+      "learning_rate": 1.074910394265233e-05,
+      "loss": 0.2441,
+      "step": 125010
+    },
+    {
+      "epoch": 806.58,
+      "learning_rate": 1.074551971326165e-05,
+      "loss": 0.2489,
+      "step": 125020
+    },
+    {
+      "epoch": 806.64,
+      "learning_rate": 1.0741935483870968e-05,
+      "loss": 0.2209,
+      "step": 125030
+    },
+    {
+      "epoch": 806.71,
+      "learning_rate": 1.0738351254480288e-05,
+      "loss": 0.2582,
+      "step": 125040
+    },
+    {
+      "epoch": 806.77,
+      "learning_rate": 1.0734767025089606e-05,
+      "loss": 0.2057,
+      "step": 125050
+    },
+    {
+      "epoch": 806.84,
+      "learning_rate": 1.0731182795698926e-05,
+      "loss": 0.2498,
+      "step": 125060
+    },
+    {
+      "epoch": 806.9,
+      "learning_rate": 1.0727598566308244e-05,
+      "loss": 0.2122,
+      "step": 125070
+    },
+    {
+      "epoch": 806.96,
+      "learning_rate": 1.0724014336917564e-05,
+      "loss": 0.1785,
+      "step": 125080
+    },
+    {
+      "epoch": 807.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9418792128562927,
+      "eval_runtime": 0.862,
+      "eval_samples_per_second": 233.188,
+      "eval_steps_per_second": 8.121,
+      "step": 125085
+    },
+    {
+      "epoch": 807.03,
+      "learning_rate": 1.0720430107526882e-05,
+      "loss": 0.2512,
+      "step": 125090
+    },
+    {
+      "epoch": 807.1,
+      "learning_rate": 1.0716845878136202e-05,
+      "loss": 0.2264,
+      "step": 125100
+    },
+    {
+      "epoch": 807.16,
+      "learning_rate": 1.071326164874552e-05,
+      "loss": 0.2056,
+      "step": 125110
+    },
+    {
+      "epoch": 807.23,
+      "learning_rate": 1.070967741935484e-05,
+      "loss": 0.2041,
+      "step": 125120
+    },
+    {
+      "epoch": 807.29,
+      "learning_rate": 1.0706093189964159e-05,
+      "loss": 0.2182,
+      "step": 125130
+    },
+    {
+      "epoch": 807.35,
+      "learning_rate": 1.0702508960573478e-05,
+      "loss": 0.2294,
+      "step": 125140
+    },
+    {
+      "epoch": 807.42,
+      "learning_rate": 1.0698924731182797e-05,
+      "loss": 0.226,
+      "step": 125150
+    },
+    {
+      "epoch": 807.48,
+      "learning_rate": 1.0695340501792115e-05,
+      "loss": 0.2114,
+      "step": 125160
+    },
+    {
+      "epoch": 807.55,
+      "learning_rate": 1.0691756272401435e-05,
+      "loss": 0.1914,
+      "step": 125170
+    },
+    {
+      "epoch": 807.61,
+      "learning_rate": 1.0688172043010753e-05,
+      "loss": 0.1635,
+      "step": 125180
+    },
+    {
+      "epoch": 807.68,
+      "learning_rate": 1.0684587813620073e-05,
+      "loss": 0.2019,
+      "step": 125190
+    },
+    {
+      "epoch": 807.74,
+      "learning_rate": 1.0681003584229391e-05,
+      "loss": 0.1937,
+      "step": 125200
+    },
+    {
+      "epoch": 807.8,
+      "learning_rate": 1.0677419354838711e-05,
+      "loss": 0.1693,
+      "step": 125210
+    },
+    {
+      "epoch": 807.87,
+      "learning_rate": 1.067383512544803e-05,
+      "loss": 0.22,
+      "step": 125220
+    },
+    {
+      "epoch": 807.93,
+      "learning_rate": 1.0670250896057349e-05,
+      "loss": 0.2004,
+      "step": 125230
+    },
+    {
+      "epoch": 808.0,
+      "learning_rate": 1.0666666666666667e-05,
+      "loss": 0.1997,
+      "step": 125240
+    },
+    {
+      "epoch": 808.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9386534690856934,
+      "eval_runtime": 0.8875,
+      "eval_samples_per_second": 226.481,
+      "eval_steps_per_second": 7.887,
+      "step": 125240
+    },
+    {
+      "epoch": 808.06,
+      "learning_rate": 1.0663082437275986e-05,
+      "loss": 0.2382,
+      "step": 125250
+    },
+    {
+      "epoch": 808.13,
+      "learning_rate": 1.0659498207885304e-05,
+      "loss": 0.2411,
+      "step": 125260
+    },
+    {
+      "epoch": 808.19,
+      "learning_rate": 1.0655913978494624e-05,
+      "loss": 0.211,
+      "step": 125270
+    },
+    {
+      "epoch": 808.26,
+      "learning_rate": 1.0652329749103942e-05,
+      "loss": 0.2059,
+      "step": 125280
+    },
+    {
+      "epoch": 808.32,
+      "learning_rate": 1.0648745519713262e-05,
+      "loss": 0.1933,
+      "step": 125290
+    },
+    {
+      "epoch": 808.39,
+      "learning_rate": 1.064516129032258e-05,
+      "loss": 0.2186,
+      "step": 125300
+    },
+    {
+      "epoch": 808.45,
+      "learning_rate": 1.06415770609319e-05,
+      "loss": 0.2157,
+      "step": 125310
+    },
+    {
+      "epoch": 808.51,
+      "learning_rate": 1.0637992831541218e-05,
+      "loss": 0.2141,
+      "step": 125320
+    },
+    {
+      "epoch": 808.58,
+      "learning_rate": 1.0634408602150538e-05,
+      "loss": 0.2032,
+      "step": 125330
+    },
+    {
+      "epoch": 808.64,
+      "learning_rate": 1.0630824372759856e-05,
+      "loss": 0.2131,
+      "step": 125340
+    },
+    {
+      "epoch": 808.71,
+      "learning_rate": 1.0627240143369176e-05,
+      "loss": 0.1936,
+      "step": 125350
+    },
+    {
+      "epoch": 808.77,
+      "learning_rate": 1.0623655913978494e-05,
+      "loss": 0.2348,
+      "step": 125360
+    },
+    {
+      "epoch": 808.84,
+      "learning_rate": 1.0620071684587814e-05,
+      "loss": 0.1991,
+      "step": 125370
+    },
+    {
+      "epoch": 808.9,
+      "learning_rate": 1.0616487455197132e-05,
+      "loss": 0.2328,
+      "step": 125380
+    },
+    {
+      "epoch": 808.96,
+      "learning_rate": 1.0612903225806452e-05,
+      "loss": 0.1861,
+      "step": 125390
+    },
+    {
+      "epoch": 809.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9183788299560547,
+      "eval_runtime": 0.8472,
+      "eval_samples_per_second": 237.24,
+      "eval_steps_per_second": 8.262,
+      "step": 125395
+    },
+    {
+      "epoch": 809.03,
+      "learning_rate": 1.060931899641577e-05,
+      "loss": 0.2058,
+      "step": 125400
+    },
+    {
+      "epoch": 809.1,
+      "learning_rate": 1.060573476702509e-05,
+      "loss": 0.1561,
+      "step": 125410
+    },
+    {
+      "epoch": 809.16,
+      "learning_rate": 1.0602150537634409e-05,
+      "loss": 0.1956,
+      "step": 125420
+    },
+    {
+      "epoch": 809.23,
+      "learning_rate": 1.0598566308243728e-05,
+      "loss": 0.2059,
+      "step": 125430
+    },
+    {
+      "epoch": 809.29,
+      "learning_rate": 1.0594982078853047e-05,
+      "loss": 0.2486,
+      "step": 125440
+    },
+    {
+      "epoch": 809.35,
+      "learning_rate": 1.0591397849462367e-05,
+      "loss": 0.1983,
+      "step": 125450
+    },
+    {
+      "epoch": 809.42,
+      "learning_rate": 1.0587813620071685e-05,
+      "loss": 0.2076,
+      "step": 125460
+    },
+    {
+      "epoch": 809.48,
+      "learning_rate": 1.0584229390681005e-05,
+      "loss": 0.1995,
+      "step": 125470
+    },
+    {
+      "epoch": 809.55,
+      "learning_rate": 1.0580645161290323e-05,
+      "loss": 0.2002,
+      "step": 125480
+    },
+    {
+      "epoch": 809.61,
+      "learning_rate": 1.0577060931899643e-05,
+      "loss": 0.234,
+      "step": 125490
+    },
+    {
+      "epoch": 809.68,
+      "learning_rate": 1.0573476702508961e-05,
+      "loss": 0.2015,
+      "step": 125500
+    },
+    {
+      "epoch": 809.74,
+      "learning_rate": 1.056989247311828e-05,
+      "loss": 0.1774,
+      "step": 125510
+    },
+    {
+      "epoch": 809.8,
+      "learning_rate": 1.0566308243727599e-05,
+      "loss": 0.1751,
+      "step": 125520
+    },
+    {
+      "epoch": 809.87,
+      "learning_rate": 1.0562724014336917e-05,
+      "loss": 0.1733,
+      "step": 125530
+    },
+    {
+      "epoch": 809.93,
+      "learning_rate": 1.0559139784946237e-05,
+      "loss": 0.1962,
+      "step": 125540
+    },
+    {
+      "epoch": 810.0,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 0.2004,
+      "step": 125550
+    },
+    {
+      "epoch": 810.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9644691944122314,
+      "eval_runtime": 0.8645,
+      "eval_samples_per_second": 232.514,
+      "eval_steps_per_second": 8.098,
+      "step": 125550
+    },
+    {
+      "epoch": 810.06,
+      "learning_rate": 1.0551971326164875e-05,
+      "loss": 0.2287,
+      "step": 125560
+    },
+    {
+      "epoch": 810.13,
+      "learning_rate": 1.0548387096774193e-05,
+      "loss": 0.1773,
+      "step": 125570
+    },
+    {
+      "epoch": 810.19,
+      "learning_rate": 1.0544802867383513e-05,
+      "loss": 0.2004,
+      "step": 125580
+    },
+    {
+      "epoch": 810.26,
+      "learning_rate": 1.0541218637992832e-05,
+      "loss": 0.1967,
+      "step": 125590
+    },
+    {
+      "epoch": 810.32,
+      "learning_rate": 1.0537634408602151e-05,
+      "loss": 0.1865,
+      "step": 125600
+    },
+    {
+      "epoch": 810.39,
+      "learning_rate": 1.053405017921147e-05,
+      "loss": 0.1845,
+      "step": 125610
+    },
+    {
+      "epoch": 810.45,
+      "learning_rate": 1.053046594982079e-05,
+      "loss": 0.2327,
+      "step": 125620
+    },
+    {
+      "epoch": 810.51,
+      "learning_rate": 1.0526881720430108e-05,
+      "loss": 0.2269,
+      "step": 125630
+    },
+    {
+      "epoch": 810.58,
+      "learning_rate": 1.0523297491039428e-05,
+      "loss": 0.2024,
+      "step": 125640
+    },
+    {
+      "epoch": 810.64,
+      "learning_rate": 1.0519713261648746e-05,
+      "loss": 0.2406,
+      "step": 125650
+    },
+    {
+      "epoch": 810.71,
+      "learning_rate": 1.0516129032258066e-05,
+      "loss": 0.2057,
+      "step": 125660
+    },
+    {
+      "epoch": 810.77,
+      "learning_rate": 1.0512544802867384e-05,
+      "loss": 0.2287,
+      "step": 125670
+    },
+    {
+      "epoch": 810.84,
+      "learning_rate": 1.0508960573476704e-05,
+      "loss": 0.2506,
+      "step": 125680
+    },
+    {
+      "epoch": 810.9,
+      "learning_rate": 1.0505376344086022e-05,
+      "loss": 0.194,
+      "step": 125690
+    },
+    {
+      "epoch": 810.96,
+      "learning_rate": 1.0501792114695342e-05,
+      "loss": 0.1798,
+      "step": 125700
+    },
+    {
+      "epoch": 811.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9562624096870422,
+      "eval_runtime": 0.8855,
+      "eval_samples_per_second": 226.997,
+      "eval_steps_per_second": 7.905,
+      "step": 125705
+    },
+    {
+      "epoch": 811.03,
+      "learning_rate": 1.049820788530466e-05,
+      "loss": 0.2519,
+      "step": 125710
+    },
+    {
+      "epoch": 811.1,
+      "learning_rate": 1.049462365591398e-05,
+      "loss": 0.2904,
+      "step": 125720
+    },
+    {
+      "epoch": 811.16,
+      "learning_rate": 1.0491039426523298e-05,
+      "loss": 0.1553,
+      "step": 125730
+    },
+    {
+      "epoch": 811.23,
+      "learning_rate": 1.0487455197132617e-05,
+      "loss": 0.2291,
+      "step": 125740
+    },
+    {
+      "epoch": 811.29,
+      "learning_rate": 1.0483870967741936e-05,
+      "loss": 0.1734,
+      "step": 125750
+    },
+    {
+      "epoch": 811.35,
+      "learning_rate": 1.0480286738351255e-05,
+      "loss": 0.2266,
+      "step": 125760
+    },
+    {
+      "epoch": 811.42,
+      "learning_rate": 1.0476702508960575e-05,
+      "loss": 0.2121,
+      "step": 125770
+    },
+    {
+      "epoch": 811.48,
+      "learning_rate": 1.0473118279569893e-05,
+      "loss": 0.2139,
+      "step": 125780
+    },
+    {
+      "epoch": 811.55,
+      "learning_rate": 1.0469534050179213e-05,
+      "loss": 0.2544,
+      "step": 125790
+    },
+    {
+      "epoch": 811.61,
+      "learning_rate": 1.046594982078853e-05,
+      "loss": 0.193,
+      "step": 125800
+    },
+    {
+      "epoch": 811.68,
+      "learning_rate": 1.046236559139785e-05,
+      "loss": 0.2076,
+      "step": 125810
+    },
+    {
+      "epoch": 811.74,
+      "learning_rate": 1.0458781362007169e-05,
+      "loss": 0.2295,
+      "step": 125820
+    },
+    {
+      "epoch": 811.8,
+      "learning_rate": 1.0455197132616489e-05,
+      "loss": 0.1981,
+      "step": 125830
+    },
+    {
+      "epoch": 811.87,
+      "learning_rate": 1.0451612903225807e-05,
+      "loss": 0.1978,
+      "step": 125840
+    },
+    {
+      "epoch": 811.93,
+      "learning_rate": 1.0448028673835127e-05,
+      "loss": 0.2231,
+      "step": 125850
+    },
+    {
+      "epoch": 812.0,
+      "learning_rate": 1.0444444444444445e-05,
+      "loss": 0.1802,
+      "step": 125860
+    },
+    {
+      "epoch": 812.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9369176030158997,
+      "eval_runtime": 0.8647,
+      "eval_samples_per_second": 232.449,
+      "eval_steps_per_second": 8.095,
+      "step": 125860
+    },
+    {
+      "epoch": 812.06,
+      "learning_rate": 1.0440860215053765e-05,
+      "loss": 0.2121,
+      "step": 125870
+    },
+    {
+      "epoch": 812.13,
+      "learning_rate": 1.0437275985663083e-05,
+      "loss": 0.2133,
+      "step": 125880
+    },
+    {
+      "epoch": 812.19,
+      "learning_rate": 1.0433691756272401e-05,
+      "loss": 0.2025,
+      "step": 125890
+    },
+    {
+      "epoch": 812.26,
+      "learning_rate": 1.043010752688172e-05,
+      "loss": 0.2315,
+      "step": 125900
+    },
+    {
+      "epoch": 812.32,
+      "learning_rate": 1.042652329749104e-05,
+      "loss": 0.2375,
+      "step": 125910
+    },
+    {
+      "epoch": 812.39,
+      "learning_rate": 1.0422939068100358e-05,
+      "loss": 0.1788,
+      "step": 125920
+    },
+    {
+      "epoch": 812.45,
+      "learning_rate": 1.0419354838709678e-05,
+      "loss": 0.1953,
+      "step": 125930
+    },
+    {
+      "epoch": 812.51,
+      "learning_rate": 1.0415770609318996e-05,
+      "loss": 0.2087,
+      "step": 125940
+    },
+    {
+      "epoch": 812.58,
+      "learning_rate": 1.0412186379928316e-05,
+      "loss": 0.2195,
+      "step": 125950
+    },
+    {
+      "epoch": 812.64,
+      "learning_rate": 1.0408602150537634e-05,
+      "loss": 0.1892,
+      "step": 125960
+    },
+    {
+      "epoch": 812.71,
+      "learning_rate": 1.0405017921146954e-05,
+      "loss": 0.2438,
+      "step": 125970
+    },
+    {
+      "epoch": 812.77,
+      "learning_rate": 1.0401433691756272e-05,
+      "loss": 0.2154,
+      "step": 125980
+    },
+    {
+      "epoch": 812.84,
+      "learning_rate": 1.0397849462365592e-05,
+      "loss": 0.2075,
+      "step": 125990
+    },
+    {
+      "epoch": 812.9,
+      "learning_rate": 1.039426523297491e-05,
+      "loss": 0.2225,
+      "step": 126000
+    },
+    {
+      "epoch": 812.96,
+      "learning_rate": 1.039068100358423e-05,
+      "loss": 0.2069,
+      "step": 126010
+    },
+    {
+      "epoch": 813.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9340156316757202,
+      "eval_runtime": 0.8808,
+      "eval_samples_per_second": 228.201,
+      "eval_steps_per_second": 7.947,
+      "step": 126015
+    },
+    {
+      "epoch": 813.03,
+      "learning_rate": 1.0387096774193548e-05,
+      "loss": 0.2161,
+      "step": 126020
+    },
+    {
+      "epoch": 813.1,
+      "learning_rate": 1.0383512544802868e-05,
+      "loss": 0.2475,
+      "step": 126030
+    },
+    {
+      "epoch": 813.16,
+      "learning_rate": 1.0379928315412186e-05,
+      "loss": 0.2065,
+      "step": 126040
+    },
+    {
+      "epoch": 813.23,
+      "learning_rate": 1.0376344086021506e-05,
+      "loss": 0.1695,
+      "step": 126050
+    },
+    {
+      "epoch": 813.29,
+      "learning_rate": 1.0372759856630824e-05,
+      "loss": 0.1582,
+      "step": 126060
+    },
+    {
+      "epoch": 813.35,
+      "learning_rate": 1.0369175627240144e-05,
+      "loss": 0.2435,
+      "step": 126070
+    },
+    {
+      "epoch": 813.42,
+      "learning_rate": 1.0365591397849463e-05,
+      "loss": 0.1802,
+      "step": 126080
+    },
+    {
+      "epoch": 813.48,
+      "learning_rate": 1.036200716845878e-05,
+      "loss": 0.2135,
+      "step": 126090
+    },
+    {
+      "epoch": 813.55,
+      "learning_rate": 1.03584229390681e-05,
+      "loss": 0.2199,
+      "step": 126100
+    },
+    {
+      "epoch": 813.61,
+      "learning_rate": 1.0354838709677419e-05,
+      "loss": 0.2185,
+      "step": 126110
+    },
+    {
+      "epoch": 813.68,
+      "learning_rate": 1.0351254480286739e-05,
+      "loss": 0.2109,
+      "step": 126120
+    },
+    {
+      "epoch": 813.74,
+      "learning_rate": 1.0347670250896057e-05,
+      "loss": 0.2237,
+      "step": 126130
+    },
+    {
+      "epoch": 813.8,
+      "learning_rate": 1.0344086021505377e-05,
+      "loss": 0.1962,
+      "step": 126140
+    },
+    {
+      "epoch": 813.87,
+      "learning_rate": 1.0340501792114695e-05,
+      "loss": 0.2185,
+      "step": 126150
+    },
+    {
+      "epoch": 813.93,
+      "learning_rate": 1.0336917562724015e-05,
+      "loss": 0.2377,
+      "step": 126160
+    },
+    {
+      "epoch": 814.0,
+      "learning_rate": 1.0333333333333333e-05,
+      "loss": 0.2047,
+      "step": 126170
+    },
+    {
+      "epoch": 814.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9062806963920593,
+      "eval_runtime": 0.8978,
+      "eval_samples_per_second": 223.876,
+      "eval_steps_per_second": 7.797,
+      "step": 126170
+    },
+    {
+      "epoch": 814.06,
+      "learning_rate": 1.0329749103942653e-05,
+      "loss": 0.213,
+      "step": 126180
+    },
+    {
+      "epoch": 814.13,
+      "learning_rate": 1.0326164874551971e-05,
+      "loss": 0.2302,
+      "step": 126190
+    },
+    {
+      "epoch": 814.19,
+      "learning_rate": 1.0322580645161291e-05,
+      "loss": 0.22,
+      "step": 126200
+    },
+    {
+      "epoch": 814.26,
+      "learning_rate": 1.031899641577061e-05,
+      "loss": 0.1773,
+      "step": 126210
+    },
+    {
+      "epoch": 814.32,
+      "learning_rate": 1.031541218637993e-05,
+      "loss": 0.1935,
+      "step": 126220
+    },
+    {
+      "epoch": 814.39,
+      "learning_rate": 1.0311827956989248e-05,
+      "loss": 0.2407,
+      "step": 126230
+    },
+    {
+      "epoch": 814.45,
+      "learning_rate": 1.0308243727598567e-05,
+      "loss": 0.2007,
+      "step": 126240
+    },
+    {
+      "epoch": 814.51,
+      "learning_rate": 1.0304659498207886e-05,
+      "loss": 0.2091,
+      "step": 126250
+    },
+    {
+      "epoch": 814.58,
+      "learning_rate": 1.0301075268817206e-05,
+      "loss": 0.2423,
+      "step": 126260
+    },
+    {
+      "epoch": 814.64,
+      "learning_rate": 1.0297491039426524e-05,
+      "loss": 0.176,
+      "step": 126270
+    },
+    {
+      "epoch": 814.71,
+      "learning_rate": 1.0293906810035844e-05,
+      "loss": 0.2193,
+      "step": 126280
+    },
+    {
+      "epoch": 814.77,
+      "learning_rate": 1.0290322580645162e-05,
+      "loss": 0.2071,
+      "step": 126290
+    },
+    {
+      "epoch": 814.84,
+      "learning_rate": 1.0286738351254482e-05,
+      "loss": 0.2232,
+      "step": 126300
+    },
+    {
+      "epoch": 814.9,
+      "learning_rate": 1.02831541218638e-05,
+      "loss": 0.2415,
+      "step": 126310
+    },
+    {
+      "epoch": 814.96,
+      "learning_rate": 1.027956989247312e-05,
+      "loss": 0.188,
+      "step": 126320
+    },
+    {
+      "epoch": 815.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9132570624351501,
+      "eval_runtime": 0.8565,
+      "eval_samples_per_second": 234.671,
+      "eval_steps_per_second": 8.173,
+      "step": 126325
+    },
+    {
+      "epoch": 815.03,
+      "learning_rate": 1.0275985663082438e-05,
+      "loss": 0.2172,
+      "step": 126330
+    },
+    {
+      "epoch": 815.1,
+      "learning_rate": 1.0272401433691756e-05,
+      "loss": 0.2098,
+      "step": 126340
+    },
+    {
+      "epoch": 815.16,
+      "learning_rate": 1.0268817204301076e-05,
+      "loss": 0.1996,
+      "step": 126350
+    },
+    {
+      "epoch": 815.23,
+      "learning_rate": 1.0265232974910394e-05,
+      "loss": 0.1938,
+      "step": 126360
+    },
+    {
+      "epoch": 815.29,
+      "learning_rate": 1.0261648745519714e-05,
+      "loss": 0.1742,
+      "step": 126370
+    },
+    {
+      "epoch": 815.35,
+      "learning_rate": 1.0258064516129032e-05,
+      "loss": 0.2125,
+      "step": 126380
+    },
+    {
+      "epoch": 815.42,
+      "learning_rate": 1.0254480286738352e-05,
+      "loss": 0.1877,
+      "step": 126390
+    },
+    {
+      "epoch": 815.48,
+      "learning_rate": 1.025089605734767e-05,
+      "loss": 0.2315,
+      "step": 126400
+    },
+    {
+      "epoch": 815.55,
+      "learning_rate": 1.024731182795699e-05,
+      "loss": 0.2316,
+      "step": 126410
+    },
+    {
+      "epoch": 815.61,
+      "learning_rate": 1.0243727598566309e-05,
+      "loss": 0.2091,
+      "step": 126420
+    },
+    {
+      "epoch": 815.68,
+      "learning_rate": 1.0240143369175629e-05,
+      "loss": 0.2472,
+      "step": 126430
+    },
+    {
+      "epoch": 815.74,
+      "learning_rate": 1.0236559139784947e-05,
+      "loss": 0.2098,
+      "step": 126440
+    },
+    {
+      "epoch": 815.8,
+      "learning_rate": 1.0232974910394267e-05,
+      "loss": 0.1715,
+      "step": 126450
+    },
+    {
+      "epoch": 815.87,
+      "learning_rate": 1.0229390681003585e-05,
+      "loss": 0.2527,
+      "step": 126460
+    },
+    {
+      "epoch": 815.93,
+      "learning_rate": 1.0225806451612905e-05,
+      "loss": 0.1716,
+      "step": 126470
+    },
+    {
+      "epoch": 816.0,
+      "learning_rate": 1.0222222222222223e-05,
+      "loss": 0.232,
+      "step": 126480
+    },
+    {
+      "epoch": 816.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9304980635643005,
+      "eval_runtime": 0.8678,
+      "eval_samples_per_second": 231.614,
+      "eval_steps_per_second": 8.066,
+      "step": 126480
+    },
+    {
+      "epoch": 816.06,
+      "learning_rate": 1.0218637992831543e-05,
+      "loss": 0.2033,
+      "step": 126490
+    },
+    {
+      "epoch": 816.13,
+      "learning_rate": 1.0215053763440861e-05,
+      "loss": 0.2355,
+      "step": 126500
+    },
+    {
+      "epoch": 816.19,
+      "learning_rate": 1.0211469534050181e-05,
+      "loss": 0.2149,
+      "step": 126510
+    },
+    {
+      "epoch": 816.26,
+      "learning_rate": 1.02078853046595e-05,
+      "loss": 0.2522,
+      "step": 126520
+    },
+    {
+      "epoch": 816.32,
+      "learning_rate": 1.0204301075268817e-05,
+      "loss": 0.2074,
+      "step": 126530
+    },
+    {
+      "epoch": 816.39,
+      "learning_rate": 1.0200716845878136e-05,
+      "loss": 0.2519,
+      "step": 126540
+    },
+    {
+      "epoch": 816.45,
+      "learning_rate": 1.0197132616487455e-05,
+      "loss": 0.2025,
+      "step": 126550
+    },
+    {
+      "epoch": 816.51,
+      "learning_rate": 1.0193548387096774e-05,
+      "loss": 0.2172,
+      "step": 126560
+    },
+    {
+      "epoch": 816.58,
+      "learning_rate": 1.0189964157706094e-05,
+      "loss": 0.2051,
+      "step": 126570
+    },
+    {
+      "epoch": 816.64,
+      "learning_rate": 1.0186379928315412e-05,
+      "loss": 0.1941,
+      "step": 126580
+    },
+    {
+      "epoch": 816.71,
+      "learning_rate": 1.0182795698924732e-05,
+      "loss": 0.197,
+      "step": 126590
+    },
+    {
+      "epoch": 816.77,
+      "learning_rate": 1.017921146953405e-05,
+      "loss": 0.2007,
+      "step": 126600
+    },
+    {
+      "epoch": 816.84,
+      "learning_rate": 1.017562724014337e-05,
+      "loss": 0.1931,
+      "step": 126610
+    },
+    {
+      "epoch": 816.9,
+      "learning_rate": 1.0172043010752688e-05,
+      "loss": 0.1925,
+      "step": 126620
+    },
+    {
+      "epoch": 816.96,
+      "learning_rate": 1.0168458781362008e-05,
+      "loss": 0.1892,
+      "step": 126630
+    },
+    {
+      "epoch": 817.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9293238520622253,
+      "eval_runtime": 0.8871,
+      "eval_samples_per_second": 226.577,
+      "eval_steps_per_second": 7.891,
+      "step": 126635
+    },
+    {
+      "epoch": 817.03,
+      "learning_rate": 1.0164874551971326e-05,
+      "loss": 0.2331,
+      "step": 126640
+    },
+    {
+      "epoch": 817.1,
+      "learning_rate": 1.0161290322580646e-05,
+      "loss": 0.164,
+      "step": 126650
+    },
+    {
+      "epoch": 817.16,
+      "learning_rate": 1.0157706093189964e-05,
+      "loss": 0.253,
+      "step": 126660
+    },
+    {
+      "epoch": 817.23,
+      "learning_rate": 1.0154121863799282e-05,
+      "loss": 0.1995,
+      "step": 126670
+    },
+    {
+      "epoch": 817.29,
+      "learning_rate": 1.0150537634408602e-05,
+      "loss": 0.2538,
+      "step": 126680
+    },
+    {
+      "epoch": 817.35,
+      "learning_rate": 1.014695340501792e-05,
+      "loss": 0.2026,
+      "step": 126690
+    },
+    {
+      "epoch": 817.42,
+      "learning_rate": 1.014336917562724e-05,
+      "loss": 0.2211,
+      "step": 126700
+    },
+    {
+      "epoch": 817.48,
+      "learning_rate": 1.0139784946236559e-05,
+      "loss": 0.2379,
+      "step": 126710
+    },
+    {
+      "epoch": 817.55,
+      "learning_rate": 1.0136200716845879e-05,
+      "loss": 0.2543,
+      "step": 126720
+    },
+    {
+      "epoch": 817.61,
+      "learning_rate": 1.0132616487455197e-05,
+      "loss": 0.2338,
+      "step": 126730
+    },
+    {
+      "epoch": 817.68,
+      "learning_rate": 1.0129032258064517e-05,
+      "loss": 0.1904,
+      "step": 126740
+    },
+    {
+      "epoch": 817.74,
+      "learning_rate": 1.0125448028673835e-05,
+      "loss": 0.1488,
+      "step": 126750
+    },
+    {
+      "epoch": 817.8,
+      "learning_rate": 1.0121863799283155e-05,
+      "loss": 0.1789,
+      "step": 126760
+    },
+    {
+      "epoch": 817.87,
+      "learning_rate": 1.0118279569892473e-05,
+      "loss": 0.1959,
+      "step": 126770
+    },
+    {
+      "epoch": 817.93,
+      "learning_rate": 1.0114695340501793e-05,
+      "loss": 0.1999,
+      "step": 126780
+    },
+    {
+      "epoch": 818.0,
+      "learning_rate": 1.0111111111111111e-05,
+      "loss": 0.2001,
+      "step": 126790
+    },
+    {
+      "epoch": 818.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9063965678215027,
+      "eval_runtime": 0.8829,
+      "eval_samples_per_second": 227.648,
+      "eval_steps_per_second": 7.928,
+      "step": 126790
+    },
+    {
+      "epoch": 818.06,
+      "learning_rate": 1.0107526881720431e-05,
+      "loss": 0.1869,
+      "step": 126800
+    },
+    {
+      "epoch": 818.13,
+      "learning_rate": 1.010394265232975e-05,
+      "loss": 0.2603,
+      "step": 126810
+    },
+    {
+      "epoch": 818.19,
+      "learning_rate": 1.0100358422939069e-05,
+      "loss": 0.2017,
+      "step": 126820
+    },
+    {
+      "epoch": 818.26,
+      "learning_rate": 1.0096774193548387e-05,
+      "loss": 0.1674,
+      "step": 126830
+    },
+    {
+      "epoch": 818.32,
+      "learning_rate": 1.0093189964157707e-05,
+      "loss": 0.2337,
+      "step": 126840
+    },
+    {
+      "epoch": 818.39,
+      "learning_rate": 1.0089605734767025e-05,
+      "loss": 0.1743,
+      "step": 126850
+    },
+    {
+      "epoch": 818.45,
+      "learning_rate": 1.0086021505376345e-05,
+      "loss": 0.2386,
+      "step": 126860
+    },
+    {
+      "epoch": 818.51,
+      "learning_rate": 1.0082437275985663e-05,
+      "loss": 0.2014,
+      "step": 126870
+    },
+    {
+      "epoch": 818.58,
+      "learning_rate": 1.0078853046594983e-05,
+      "loss": 0.1956,
+      "step": 126880
+    },
+    {
+      "epoch": 818.64,
+      "learning_rate": 1.0075268817204302e-05,
+      "loss": 0.2427,
+      "step": 126890
+    },
+    {
+      "epoch": 818.71,
+      "learning_rate": 1.0071684587813621e-05,
+      "loss": 0.1945,
+      "step": 126900
+    },
+    {
+      "epoch": 818.77,
+      "learning_rate": 1.006810035842294e-05,
+      "loss": 0.241,
+      "step": 126910
+    },
+    {
+      "epoch": 818.84,
+      "learning_rate": 1.0064516129032258e-05,
+      "loss": 0.1908,
+      "step": 126920
+    },
+    {
+      "epoch": 818.9,
+      "learning_rate": 1.0060931899641578e-05,
+      "loss": 0.2492,
+      "step": 126930
+    },
+    {
+      "epoch": 818.96,
+      "learning_rate": 1.0057347670250896e-05,
+      "loss": 0.1663,
+      "step": 126940
+    },
+    {
+      "epoch": 819.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9274577498435974,
+      "eval_runtime": 0.8646,
+      "eval_samples_per_second": 232.49,
+      "eval_steps_per_second": 8.097,
+      "step": 126945
+    },
+    {
+      "epoch": 819.03,
+      "learning_rate": 1.0053763440860216e-05,
+      "loss": 0.2366,
+      "step": 126950
+    },
+    {
+      "epoch": 819.1,
+      "learning_rate": 1.0050179211469534e-05,
+      "loss": 0.2081,
+      "step": 126960
+    },
+    {
+      "epoch": 819.16,
+      "learning_rate": 1.0046594982078854e-05,
+      "loss": 0.2231,
+      "step": 126970
+    },
+    {
+      "epoch": 819.23,
+      "learning_rate": 1.0043010752688172e-05,
+      "loss": 0.22,
+      "step": 126980
+    },
+    {
+      "epoch": 819.29,
+      "learning_rate": 1.0039426523297492e-05,
+      "loss": 0.2209,
+      "step": 126990
+    },
+    {
+      "epoch": 819.35,
+      "learning_rate": 1.003584229390681e-05,
+      "loss": 0.1997,
+      "step": 127000
+    },
+    {
+      "epoch": 819.42,
+      "learning_rate": 1.003225806451613e-05,
+      "loss": 0.2104,
+      "step": 127010
+    },
+    {
+      "epoch": 819.48,
+      "learning_rate": 1.0028673835125448e-05,
+      "loss": 0.2453,
+      "step": 127020
+    },
+    {
+      "epoch": 819.55,
+      "learning_rate": 1.0025089605734768e-05,
+      "loss": 0.285,
+      "step": 127030
+    },
+    {
+      "epoch": 819.61,
+      "learning_rate": 1.0021505376344087e-05,
+      "loss": 0.1534,
+      "step": 127040
+    },
+    {
+      "epoch": 819.68,
+      "learning_rate": 1.0017921146953406e-05,
+      "loss": 0.2422,
+      "step": 127050
+    },
+    {
+      "epoch": 819.74,
+      "learning_rate": 1.0014336917562725e-05,
+      "loss": 0.1864,
+      "step": 127060
+    },
+    {
+      "epoch": 819.8,
+      "learning_rate": 1.0010752688172045e-05,
+      "loss": 0.2277,
+      "step": 127070
+    },
+    {
+      "epoch": 819.87,
+      "learning_rate": 1.0007168458781363e-05,
+      "loss": 0.229,
+      "step": 127080
+    },
+    {
+      "epoch": 819.93,
+      "learning_rate": 1.0003584229390683e-05,
+      "loss": 0.2335,
+      "step": 127090
+    },
+    {
+      "epoch": 820.0,
+      "learning_rate": 1e-05,
+      "loss": 0.1669,
+      "step": 127100
+    },
+    {
+      "epoch": 820.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9403781890869141,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 230.716,
+      "eval_steps_per_second": 8.035,
+      "step": 127100
+    },
+    {
+      "epoch": 820.06,
+      "learning_rate": 9.99641577060932e-06,
+      "loss": 0.2023,
+      "step": 127110
+    },
+    {
+      "epoch": 820.13,
+      "learning_rate": 9.992831541218639e-06,
+      "loss": 0.2242,
+      "step": 127120
+    },
+    {
+      "epoch": 820.19,
+      "learning_rate": 9.989247311827959e-06,
+      "loss": 0.2066,
+      "step": 127130
+    },
+    {
+      "epoch": 820.26,
+      "learning_rate": 9.985663082437277e-06,
+      "loss": 0.1884,
+      "step": 127140
+    },
+    {
+      "epoch": 820.32,
+      "learning_rate": 9.982078853046597e-06,
+      "loss": 0.1949,
+      "step": 127150
+    },
+    {
+      "epoch": 820.39,
+      "learning_rate": 9.978494623655915e-06,
+      "loss": 0.2197,
+      "step": 127160
+    },
+    {
+      "epoch": 820.45,
+      "learning_rate": 9.974910394265233e-06,
+      "loss": 0.2302,
+      "step": 127170
+    },
+    {
+      "epoch": 820.51,
+      "learning_rate": 9.971326164874552e-06,
+      "loss": 0.2148,
+      "step": 127180
+    },
+    {
+      "epoch": 820.58,
+      "learning_rate": 9.967741935483871e-06,
+      "loss": 0.2151,
+      "step": 127190
+    },
+    {
+      "epoch": 820.64,
+      "learning_rate": 9.96415770609319e-06,
+      "loss": 0.1837,
+      "step": 127200
+    },
+    {
+      "epoch": 820.71,
+      "learning_rate": 9.96057347670251e-06,
+      "loss": 0.2118,
+      "step": 127210
+    },
+    {
+      "epoch": 820.77,
+      "learning_rate": 9.956989247311828e-06,
+      "loss": 0.2191,
+      "step": 127220
+    },
+    {
+      "epoch": 820.84,
+      "learning_rate": 9.953405017921148e-06,
+      "loss": 0.1759,
+      "step": 127230
+    },
+    {
+      "epoch": 820.9,
+      "learning_rate": 9.949820788530466e-06,
+      "loss": 0.2356,
+      "step": 127240
+    },
+    {
+      "epoch": 820.96,
+      "learning_rate": 9.946236559139786e-06,
+      "loss": 0.2649,
+      "step": 127250
+    },
+    {
+      "epoch": 821.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8865975141525269,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.929,
+      "eval_steps_per_second": 7.938,
+      "step": 127255
+    },
+    {
+      "epoch": 821.03,
+      "learning_rate": 9.942652329749104e-06,
+      "loss": 0.1882,
+      "step": 127260
+    },
+    {
+      "epoch": 821.1,
+      "learning_rate": 9.939068100358422e-06,
+      "loss": 0.1848,
+      "step": 127270
+    },
+    {
+      "epoch": 821.16,
+      "learning_rate": 9.935483870967742e-06,
+      "loss": 0.2098,
+      "step": 127280
+    },
+    {
+      "epoch": 821.23,
+      "learning_rate": 9.93189964157706e-06,
+      "loss": 0.173,
+      "step": 127290
+    },
+    {
+      "epoch": 821.29,
+      "learning_rate": 9.92831541218638e-06,
+      "loss": 0.2055,
+      "step": 127300
+    },
+    {
+      "epoch": 821.35,
+      "learning_rate": 9.924731182795698e-06,
+      "loss": 0.2177,
+      "step": 127310
+    },
+    {
+      "epoch": 821.42,
+      "learning_rate": 9.921146953405018e-06,
+      "loss": 0.1878,
+      "step": 127320
+    },
+    {
+      "epoch": 821.48,
+      "learning_rate": 9.917562724014336e-06,
+      "loss": 0.192,
+      "step": 127330
+    },
+    {
+      "epoch": 821.55,
+      "learning_rate": 9.913978494623656e-06,
+      "loss": 0.2163,
+      "step": 127340
+    },
+    {
+      "epoch": 821.61,
+      "learning_rate": 9.910394265232975e-06,
+      "loss": 0.1905,
+      "step": 127350
+    },
+    {
+      "epoch": 821.68,
+      "learning_rate": 9.906810035842294e-06,
+      "loss": 0.2256,
+      "step": 127360
+    },
+    {
+      "epoch": 821.74,
+      "learning_rate": 9.903225806451613e-06,
+      "loss": 0.1995,
+      "step": 127370
+    },
+    {
+      "epoch": 821.8,
+      "learning_rate": 9.899641577060933e-06,
+      "loss": 0.2043,
+      "step": 127380
+    },
+    {
+      "epoch": 821.87,
+      "learning_rate": 9.89605734767025e-06,
+      "loss": 0.2117,
+      "step": 127390
+    },
+    {
+      "epoch": 821.93,
+      "learning_rate": 9.89247311827957e-06,
+      "loss": 0.2501,
+      "step": 127400
+    },
+    {
+      "epoch": 822.0,
+      "learning_rate": 9.888888888888889e-06,
+      "loss": 0.1949,
+      "step": 127410
+    },
+    {
+      "epoch": 822.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9019115567207336,
+      "eval_runtime": 0.8699,
+      "eval_samples_per_second": 231.057,
+      "eval_steps_per_second": 8.047,
+      "step": 127410
+    },
+    {
+      "epoch": 822.06,
+      "learning_rate": 9.885304659498209e-06,
+      "loss": 0.2409,
+      "step": 127420
+    },
+    {
+      "epoch": 822.13,
+      "learning_rate": 9.881720430107527e-06,
+      "loss": 0.1999,
+      "step": 127430
+    },
+    {
+      "epoch": 822.19,
+      "learning_rate": 9.878136200716847e-06,
+      "loss": 0.1936,
+      "step": 127440
+    },
+    {
+      "epoch": 822.26,
+      "learning_rate": 9.874551971326165e-06,
+      "loss": 0.2354,
+      "step": 127450
+    },
+    {
+      "epoch": 822.32,
+      "learning_rate": 9.870967741935485e-06,
+      "loss": 0.1784,
+      "step": 127460
+    },
+    {
+      "epoch": 822.39,
+      "learning_rate": 9.867383512544803e-06,
+      "loss": 0.2213,
+      "step": 127470
+    },
+    {
+      "epoch": 822.45,
+      "learning_rate": 9.863799283154123e-06,
+      "loss": 0.2059,
+      "step": 127480
+    },
+    {
+      "epoch": 822.51,
+      "learning_rate": 9.860215053763441e-06,
+      "loss": 0.1946,
+      "step": 127490
+    },
+    {
+      "epoch": 822.58,
+      "learning_rate": 9.856630824372761e-06,
+      "loss": 0.2243,
+      "step": 127500
+    },
+    {
+      "epoch": 822.64,
+      "learning_rate": 9.85304659498208e-06,
+      "loss": 0.2357,
+      "step": 127510
+    },
+    {
+      "epoch": 822.71,
+      "learning_rate": 9.849462365591398e-06,
+      "loss": 0.211,
+      "step": 127520
+    },
+    {
+      "epoch": 822.77,
+      "learning_rate": 9.845878136200718e-06,
+      "loss": 0.2312,
+      "step": 127530
+    },
+    {
+      "epoch": 822.84,
+      "learning_rate": 9.842293906810036e-06,
+      "loss": 0.2319,
+      "step": 127540
+    },
+    {
+      "epoch": 822.9,
+      "learning_rate": 9.838709677419356e-06,
+      "loss": 0.1941,
+      "step": 127550
+    },
+    {
+      "epoch": 822.96,
+      "learning_rate": 9.835125448028674e-06,
+      "loss": 0.1619,
+      "step": 127560
+    },
+    {
+      "epoch": 823.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9351931214332581,
+      "eval_runtime": 0.8762,
+      "eval_samples_per_second": 229.388,
+      "eval_steps_per_second": 7.989,
+      "step": 127565
+    },
+    {
+      "epoch": 823.03,
+      "learning_rate": 9.831541218637994e-06,
+      "loss": 0.2284,
+      "step": 127570
+    },
+    {
+      "epoch": 823.1,
+      "learning_rate": 9.827956989247312e-06,
+      "loss": 0.2438,
+      "step": 127580
+    },
+    {
+      "epoch": 823.16,
+      "learning_rate": 9.824372759856632e-06,
+      "loss": 0.2375,
+      "step": 127590
+    },
+    {
+      "epoch": 823.23,
+      "learning_rate": 9.82078853046595e-06,
+      "loss": 0.2006,
+      "step": 127600
+    },
+    {
+      "epoch": 823.29,
+      "learning_rate": 9.81720430107527e-06,
+      "loss": 0.1652,
+      "step": 127610
+    },
+    {
+      "epoch": 823.35,
+      "learning_rate": 9.813620071684588e-06,
+      "loss": 0.2027,
+      "step": 127620
+    },
+    {
+      "epoch": 823.42,
+      "learning_rate": 9.810035842293908e-06,
+      "loss": 0.2207,
+      "step": 127630
+    },
+    {
+      "epoch": 823.48,
+      "learning_rate": 9.806451612903226e-06,
+      "loss": 0.26,
+      "step": 127640
+    },
+    {
+      "epoch": 823.55,
+      "learning_rate": 9.802867383512546e-06,
+      "loss": 0.2018,
+      "step": 127650
+    },
+    {
+      "epoch": 823.61,
+      "learning_rate": 9.799283154121864e-06,
+      "loss": 0.2189,
+      "step": 127660
+    },
+    {
+      "epoch": 823.68,
+      "learning_rate": 9.795698924731184e-06,
+      "loss": 0.2681,
+      "step": 127670
+    },
+    {
+      "epoch": 823.74,
+      "learning_rate": 9.792114695340502e-06,
+      "loss": 0.2242,
+      "step": 127680
+    },
+    {
+      "epoch": 823.8,
+      "learning_rate": 9.788530465949822e-06,
+      "loss": 0.242,
+      "step": 127690
+    },
+    {
+      "epoch": 823.87,
+      "learning_rate": 9.78494623655914e-06,
+      "loss": 0.2227,
+      "step": 127700
+    },
+    {
+      "epoch": 823.93,
+      "learning_rate": 9.78136200716846e-06,
+      "loss": 0.1621,
+      "step": 127710
+    },
+    {
+      "epoch": 824.0,
+      "learning_rate": 9.777777777777779e-06,
+      "loss": 0.2375,
+      "step": 127720
+    },
+    {
+      "epoch": 824.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9354615211486816,
+      "eval_runtime": 0.8887,
+      "eval_samples_per_second": 226.184,
+      "eval_steps_per_second": 7.877,
+      "step": 127720
+    },
+    {
+      "epoch": 824.06,
+      "learning_rate": 9.774193548387099e-06,
+      "loss": 0.26,
+      "step": 127730
+    },
+    {
+      "epoch": 824.13,
+      "learning_rate": 9.770609318996417e-06,
+      "loss": 0.2046,
+      "step": 127740
+    },
+    {
+      "epoch": 824.19,
+      "learning_rate": 9.767025089605737e-06,
+      "loss": 0.2146,
+      "step": 127750
+    },
+    {
+      "epoch": 824.26,
+      "learning_rate": 9.763440860215055e-06,
+      "loss": 0.2234,
+      "step": 127760
+    },
+    {
+      "epoch": 824.32,
+      "learning_rate": 9.759856630824373e-06,
+      "loss": 0.2357,
+      "step": 127770
+    },
+    {
+      "epoch": 824.39,
+      "learning_rate": 9.756272401433693e-06,
+      "loss": 0.2096,
+      "step": 127780
+    },
+    {
+      "epoch": 824.45,
+      "learning_rate": 9.752688172043011e-06,
+      "loss": 0.1821,
+      "step": 127790
+    },
+    {
+      "epoch": 824.51,
+      "learning_rate": 9.749103942652331e-06,
+      "loss": 0.2491,
+      "step": 127800
+    },
+    {
+      "epoch": 824.58,
+      "learning_rate": 9.74551971326165e-06,
+      "loss": 0.1896,
+      "step": 127810
+    },
+    {
+      "epoch": 824.64,
+      "learning_rate": 9.741935483870967e-06,
+      "loss": 0.2083,
+      "step": 127820
+    },
+    {
+      "epoch": 824.71,
+      "learning_rate": 9.738351254480287e-06,
+      "loss": 0.2295,
+      "step": 127830
+    },
+    {
+      "epoch": 824.77,
+      "learning_rate": 9.734767025089606e-06,
+      "loss": 0.2005,
+      "step": 127840
+    },
+    {
+      "epoch": 824.84,
+      "learning_rate": 9.731182795698924e-06,
+      "loss": 0.2005,
+      "step": 127850
+    },
+    {
+      "epoch": 824.9,
+      "learning_rate": 9.727598566308244e-06,
+      "loss": 0.2155,
+      "step": 127860
+    },
+    {
+      "epoch": 824.96,
+      "learning_rate": 9.724014336917562e-06,
+      "loss": 0.2177,
+      "step": 127870
+    },
+    {
+      "epoch": 825.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9461024403572083,
+      "eval_runtime": 0.8643,
+      "eval_samples_per_second": 232.546,
+      "eval_steps_per_second": 8.099,
+      "step": 127875
+    },
+    {
+      "epoch": 825.03,
+      "learning_rate": 9.720430107526882e-06,
+      "loss": 0.1992,
+      "step": 127880
+    },
+    {
+      "epoch": 825.1,
+      "learning_rate": 9.7168458781362e-06,
+      "loss": 0.1961,
+      "step": 127890
+    },
+    {
+      "epoch": 825.16,
+      "learning_rate": 9.71326164874552e-06,
+      "loss": 0.1854,
+      "step": 127900
+    },
+    {
+      "epoch": 825.23,
+      "learning_rate": 9.709677419354838e-06,
+      "loss": 0.1957,
+      "step": 127910
+    },
+    {
+      "epoch": 825.29,
+      "learning_rate": 9.706093189964158e-06,
+      "loss": 0.1768,
+      "step": 127920
+    },
+    {
+      "epoch": 825.35,
+      "learning_rate": 9.702508960573476e-06,
+      "loss": 0.2276,
+      "step": 127930
+    },
+    {
+      "epoch": 825.42,
+      "learning_rate": 9.698924731182796e-06,
+      "loss": 0.1961,
+      "step": 127940
+    },
+    {
+      "epoch": 825.48,
+      "learning_rate": 9.695340501792114e-06,
+      "loss": 0.2399,
+      "step": 127950
+    },
+    {
+      "epoch": 825.55,
+      "learning_rate": 9.691756272401434e-06,
+      "loss": 0.1909,
+      "step": 127960
+    },
+    {
+      "epoch": 825.61,
+      "learning_rate": 9.688172043010752e-06,
+      "loss": 0.2055,
+      "step": 127970
+    },
+    {
+      "epoch": 825.68,
+      "learning_rate": 9.684587813620072e-06,
+      "loss": 0.2261,
+      "step": 127980
+    },
+    {
+      "epoch": 825.74,
+      "learning_rate": 9.68100358422939e-06,
+      "loss": 0.2169,
+      "step": 127990
+    },
+    {
+      "epoch": 825.8,
+      "learning_rate": 9.67741935483871e-06,
+      "loss": 0.2665,
+      "step": 128000
+    },
+    {
+      "epoch": 825.87,
+      "learning_rate": 9.673835125448029e-06,
+      "loss": 0.2549,
+      "step": 128010
+    },
+    {
+      "epoch": 825.93,
+      "learning_rate": 9.670250896057349e-06,
+      "loss": 0.2574,
+      "step": 128020
+    },
+    {
+      "epoch": 826.0,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 0.1787,
+      "step": 128030
+    },
+    {
+      "epoch": 826.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9322907328605652,
+      "eval_runtime": 0.8642,
+      "eval_samples_per_second": 232.591,
+      "eval_steps_per_second": 8.1,
+      "step": 128030
+    },
+    {
+      "epoch": 826.06,
+      "learning_rate": 9.663082437275987e-06,
+      "loss": 0.2355,
+      "step": 128040
+    },
+    {
+      "epoch": 826.13,
+      "learning_rate": 9.659498207885305e-06,
+      "loss": 0.2408,
+      "step": 128050
+    },
+    {
+      "epoch": 826.19,
+      "learning_rate": 9.655913978494625e-06,
+      "loss": 0.2165,
+      "step": 128060
+    },
+    {
+      "epoch": 826.26,
+      "learning_rate": 9.652329749103943e-06,
+      "loss": 0.2587,
+      "step": 128070
+    },
+    {
+      "epoch": 826.32,
+      "learning_rate": 9.648745519713263e-06,
+      "loss": 0.2206,
+      "step": 128080
+    },
+    {
+      "epoch": 826.39,
+      "learning_rate": 9.645161290322581e-06,
+      "loss": 0.2278,
+      "step": 128090
+    },
+    {
+      "epoch": 826.45,
+      "learning_rate": 9.6415770609319e-06,
+      "loss": 0.2171,
+      "step": 128100
+    },
+    {
+      "epoch": 826.51,
+      "learning_rate": 9.637992831541219e-06,
+      "loss": 0.2367,
+      "step": 128110
+    },
+    {
+      "epoch": 826.58,
+      "learning_rate": 9.634408602150537e-06,
+      "loss": 0.2313,
+      "step": 128120
+    },
+    {
+      "epoch": 826.64,
+      "learning_rate": 9.630824372759857e-06,
+      "loss": 0.165,
+      "step": 128130
+    },
+    {
+      "epoch": 826.71,
+      "learning_rate": 9.627240143369175e-06,
+      "loss": 0.2158,
+      "step": 128140
+    },
+    {
+      "epoch": 826.77,
+      "learning_rate": 9.623655913978495e-06,
+      "loss": 0.207,
+      "step": 128150
+    },
+    {
+      "epoch": 826.84,
+      "learning_rate": 9.620071684587814e-06,
+      "loss": 0.2185,
+      "step": 128160
+    },
+    {
+      "epoch": 826.9,
+      "learning_rate": 9.616487455197133e-06,
+      "loss": 0.1965,
+      "step": 128170
+    },
+    {
+      "epoch": 826.96,
+      "learning_rate": 9.612903225806452e-06,
+      "loss": 0.2593,
+      "step": 128180
+    },
+    {
+      "epoch": 827.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9222093224525452,
+      "eval_runtime": 0.8822,
+      "eval_samples_per_second": 227.85,
+      "eval_steps_per_second": 7.935,
+      "step": 128185
+    },
+    {
+      "epoch": 827.03,
+      "learning_rate": 9.609318996415772e-06,
+      "loss": 0.2113,
+      "step": 128190
+    },
+    {
+      "epoch": 827.1,
+      "learning_rate": 9.60573476702509e-06,
+      "loss": 0.2156,
+      "step": 128200
+    },
+    {
+      "epoch": 827.16,
+      "learning_rate": 9.60215053763441e-06,
+      "loss": 0.1981,
+      "step": 128210
+    },
+    {
+      "epoch": 827.23,
+      "learning_rate": 9.598566308243728e-06,
+      "loss": 0.1889,
+      "step": 128220
+    },
+    {
+      "epoch": 827.29,
+      "learning_rate": 9.594982078853048e-06,
+      "loss": 0.2038,
+      "step": 128230
+    },
+    {
+      "epoch": 827.35,
+      "learning_rate": 9.591397849462366e-06,
+      "loss": 0.1928,
+      "step": 128240
+    },
+    {
+      "epoch": 827.42,
+      "learning_rate": 9.587813620071686e-06,
+      "loss": 0.2076,
+      "step": 128250
+    },
+    {
+      "epoch": 827.48,
+      "learning_rate": 9.584229390681004e-06,
+      "loss": 0.2243,
+      "step": 128260
+    },
+    {
+      "epoch": 827.55,
+      "learning_rate": 9.580645161290324e-06,
+      "loss": 0.2252,
+      "step": 128270
+    },
+    {
+      "epoch": 827.61,
+      "learning_rate": 9.577060931899642e-06,
+      "loss": 0.1779,
+      "step": 128280
+    },
+    {
+      "epoch": 827.68,
+      "learning_rate": 9.573476702508962e-06,
+      "loss": 0.2273,
+      "step": 128290
+    },
+    {
+      "epoch": 827.74,
+      "learning_rate": 9.56989247311828e-06,
+      "loss": 0.221,
+      "step": 128300
+    },
+    {
+      "epoch": 827.8,
+      "learning_rate": 9.5663082437276e-06,
+      "loss": 0.2384,
+      "step": 128310
+    },
+    {
+      "epoch": 827.87,
+      "learning_rate": 9.562724014336918e-06,
+      "loss": 0.22,
+      "step": 128320
+    },
+    {
+      "epoch": 827.93,
+      "learning_rate": 9.559139784946238e-06,
+      "loss": 0.195,
+      "step": 128330
+    },
+    {
+      "epoch": 828.0,
+      "learning_rate": 9.555555555555556e-06,
+      "loss": 0.2075,
+      "step": 128340
+    },
+    {
+      "epoch": 828.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.8877363204956055,
+      "eval_runtime": 0.8635,
+      "eval_samples_per_second": 232.768,
+      "eval_steps_per_second": 8.106,
+      "step": 128340
+    },
+    {
+      "epoch": 828.06,
+      "learning_rate": 9.551971326164875e-06,
+      "loss": 0.194,
+      "step": 128350
+    },
+    {
+      "epoch": 828.13,
+      "learning_rate": 9.548387096774195e-06,
+      "loss": 0.2154,
+      "step": 128360
+    },
+    {
+      "epoch": 828.19,
+      "learning_rate": 9.544802867383513e-06,
+      "loss": 0.1923,
+      "step": 128370
+    },
+    {
+      "epoch": 828.26,
+      "learning_rate": 9.541218637992833e-06,
+      "loss": 0.1959,
+      "step": 128380
+    },
+    {
+      "epoch": 828.32,
+      "learning_rate": 9.537634408602151e-06,
+      "loss": 0.1862,
+      "step": 128390
+    },
+    {
+      "epoch": 828.39,
+      "learning_rate": 9.53405017921147e-06,
+      "loss": 0.2027,
+      "step": 128400
+    },
+    {
+      "epoch": 828.45,
+      "learning_rate": 9.530465949820789e-06,
+      "loss": 0.1888,
+      "step": 128410
+    },
+    {
+      "epoch": 828.51,
+      "learning_rate": 9.526881720430109e-06,
+      "loss": 0.1878,
+      "step": 128420
+    },
+    {
+      "epoch": 828.58,
+      "learning_rate": 9.523297491039427e-06,
+      "loss": 0.2246,
+      "step": 128430
+    },
+    {
+      "epoch": 828.64,
+      "learning_rate": 9.519713261648747e-06,
+      "loss": 0.1937,
+      "step": 128440
+    },
+    {
+      "epoch": 828.71,
+      "learning_rate": 9.516129032258064e-06,
+      "loss": 0.1865,
+      "step": 128450
+    },
+    {
+      "epoch": 828.77,
+      "learning_rate": 9.512544802867383e-06,
+      "loss": 0.1746,
+      "step": 128460
+    },
+    {
+      "epoch": 828.84,
+      "learning_rate": 9.508960573476702e-06,
+      "loss": 0.2425,
+      "step": 128470
+    },
+    {
+      "epoch": 828.9,
+      "learning_rate": 9.505376344086022e-06,
+      "loss": 0.1747,
+      "step": 128480
+    },
+    {
+      "epoch": 828.96,
+      "learning_rate": 9.50179211469534e-06,
+      "loss": 0.1961,
+      "step": 128490
+    },
+    {
+      "epoch": 829.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.904450535774231,
+      "eval_runtime": 0.9252,
+      "eval_samples_per_second": 217.24,
+      "eval_steps_per_second": 7.566,
+      "step": 128495
+    },
+    {
+      "epoch": 829.03,
+      "learning_rate": 9.49820788530466e-06,
+      "loss": 0.2288,
+      "step": 128500
+    },
+    {
+      "epoch": 829.1,
+      "learning_rate": 9.494623655913978e-06,
+      "loss": 0.2248,
+      "step": 128510
+    },
+    {
+      "epoch": 829.16,
+      "learning_rate": 9.491039426523298e-06,
+      "loss": 0.2273,
+      "step": 128520
+    },
+    {
+      "epoch": 829.23,
+      "learning_rate": 9.487455197132616e-06,
+      "loss": 0.2047,
+      "step": 128530
+    },
+    {
+      "epoch": 829.29,
+      "learning_rate": 9.483870967741936e-06,
+      "loss": 0.2315,
+      "step": 128540
+    },
+    {
+      "epoch": 829.35,
+      "learning_rate": 9.480286738351254e-06,
+      "loss": 0.1822,
+      "step": 128550
+    },
+    {
+      "epoch": 829.42,
+      "learning_rate": 9.476702508960574e-06,
+      "loss": 0.2062,
+      "step": 128560
+    },
+    {
+      "epoch": 829.48,
+      "learning_rate": 9.473118279569892e-06,
+      "loss": 0.2162,
+      "step": 128570
+    },
+    {
+      "epoch": 829.55,
+      "learning_rate": 9.469534050179212e-06,
+      "loss": 0.2433,
+      "step": 128580
+    },
+    {
+      "epoch": 829.61,
+      "learning_rate": 9.46594982078853e-06,
+      "loss": 0.1906,
+      "step": 128590
+    },
+    {
+      "epoch": 829.68,
+      "learning_rate": 9.46236559139785e-06,
+      "loss": 0.2206,
+      "step": 128600
+    },
+    {
+      "epoch": 829.74,
+      "learning_rate": 9.458781362007168e-06,
+      "loss": 0.1996,
+      "step": 128610
+    },
+    {
+      "epoch": 829.8,
+      "learning_rate": 9.455197132616488e-06,
+      "loss": 0.2304,
+      "step": 128620
+    },
+    {
+      "epoch": 829.87,
+      "learning_rate": 9.451612903225806e-06,
+      "loss": 0.1985,
+      "step": 128630
+    },
+    {
+      "epoch": 829.93,
+      "learning_rate": 9.448028673835126e-06,
+      "loss": 0.1868,
+      "step": 128640
+    },
+    {
+      "epoch": 830.0,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 0.1809,
+      "step": 128650
+    },
+    {
+      "epoch": 830.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9055208563804626,
+      "eval_runtime": 0.868,
+      "eval_samples_per_second": 231.562,
+      "eval_steps_per_second": 8.064,
+      "step": 128650
+    },
+    {
+      "epoch": 830.06,
+      "learning_rate": 9.440860215053764e-06,
+      "loss": 0.2617,
+      "step": 128660
+    },
+    {
+      "epoch": 830.13,
+      "learning_rate": 9.437275985663083e-06,
+      "loss": 0.2206,
+      "step": 128670
+    },
+    {
+      "epoch": 830.19,
+      "learning_rate": 9.433691756272401e-06,
+      "loss": 0.1778,
+      "step": 128680
+    },
+    {
+      "epoch": 830.26,
+      "learning_rate": 9.43010752688172e-06,
+      "loss": 0.1828,
+      "step": 128690
+    },
+    {
+      "epoch": 830.32,
+      "learning_rate": 9.426523297491039e-06,
+      "loss": 0.2031,
+      "step": 128700
+    },
+    {
+      "epoch": 830.39,
+      "learning_rate": 9.422939068100359e-06,
+      "loss": 0.1725,
+      "step": 128710
+    },
+    {
+      "epoch": 830.45,
+      "learning_rate": 9.419354838709677e-06,
+      "loss": 0.2174,
+      "step": 128720
+    },
+    {
+      "epoch": 830.51,
+      "learning_rate": 9.415770609318997e-06,
+      "loss": 0.2078,
+      "step": 128730
+    },
+    {
+      "epoch": 830.58,
+      "learning_rate": 9.412186379928315e-06,
+      "loss": 0.1439,
+      "step": 128740
+    },
+    {
+      "epoch": 830.64,
+      "learning_rate": 9.408602150537635e-06,
+      "loss": 0.2324,
+      "step": 128750
+    },
+    {
+      "epoch": 830.71,
+      "learning_rate": 9.405017921146953e-06,
+      "loss": 0.2231,
+      "step": 128760
+    },
+    {
+      "epoch": 830.77,
+      "learning_rate": 9.401433691756273e-06,
+      "loss": 0.2336,
+      "step": 128770
+    },
+    {
+      "epoch": 830.84,
+      "learning_rate": 9.397849462365591e-06,
+      "loss": 0.2647,
+      "step": 128780
+    },
+    {
+      "epoch": 830.9,
+      "learning_rate": 9.394265232974911e-06,
+      "loss": 0.1994,
+      "step": 128790
+    },
+    {
+      "epoch": 830.96,
+      "learning_rate": 9.39068100358423e-06,
+      "loss": 0.1982,
+      "step": 128800
+    },
+    {
+      "epoch": 831.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9110960960388184,
+      "eval_runtime": 0.8779,
+      "eval_samples_per_second": 228.959,
+      "eval_steps_per_second": 7.974,
+      "step": 128805
+    },
+    {
+      "epoch": 831.03,
+      "learning_rate": 9.38709677419355e-06,
+      "loss": 0.2471,
+      "step": 128810
+    },
+    {
+      "epoch": 831.1,
+      "learning_rate": 9.383512544802868e-06,
+      "loss": 0.247,
+      "step": 128820
+    },
+    {
+      "epoch": 831.16,
+      "learning_rate": 9.379928315412187e-06,
+      "loss": 0.2121,
+      "step": 128830
+    },
+    {
+      "epoch": 831.23,
+      "learning_rate": 9.376344086021506e-06,
+      "loss": 0.169,
+      "step": 128840
+    },
+    {
+      "epoch": 831.29,
+      "learning_rate": 9.372759856630826e-06,
+      "loss": 0.2353,
+      "step": 128850
+    },
+    {
+      "epoch": 831.35,
+      "learning_rate": 9.369175627240144e-06,
+      "loss": 0.2403,
+      "step": 128860
+    },
+    {
+      "epoch": 831.42,
+      "learning_rate": 9.365591397849464e-06,
+      "loss": 0.1782,
+      "step": 128870
+    },
+    {
+      "epoch": 831.48,
+      "learning_rate": 9.362007168458782e-06,
+      "loss": 0.1863,
+      "step": 128880
+    },
+    {
+      "epoch": 831.55,
+      "learning_rate": 9.358422939068102e-06,
+      "loss": 0.2167,
+      "step": 128890
+    },
+    {
+      "epoch": 831.61,
+      "learning_rate": 9.35483870967742e-06,
+      "loss": 0.1779,
+      "step": 128900
+    },
+    {
+      "epoch": 831.68,
+      "learning_rate": 9.35125448028674e-06,
+      "loss": 0.2028,
+      "step": 128910
+    },
+    {
+      "epoch": 831.74,
+      "learning_rate": 9.347670250896058e-06,
+      "loss": 0.2002,
+      "step": 128920
+    },
+    {
+      "epoch": 831.8,
+      "learning_rate": 9.344086021505376e-06,
+      "loss": 0.2118,
+      "step": 128930
+    },
+    {
+      "epoch": 831.87,
+      "learning_rate": 9.340501792114696e-06,
+      "loss": 0.2439,
+      "step": 128940
+    },
+    {
+      "epoch": 831.93,
+      "learning_rate": 9.336917562724014e-06,
+      "loss": 0.1851,
+      "step": 128950
+    },
+    {
+      "epoch": 832.0,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 0.1722,
+      "step": 128960
+    },
+    {
+      "epoch": 832.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9586401581764221,
+      "eval_runtime": 0.8976,
+      "eval_samples_per_second": 223.939,
+      "eval_steps_per_second": 7.799,
+      "step": 128960
+    },
+    {
+      "epoch": 832.06,
+      "learning_rate": 9.329749103942653e-06,
+      "loss": 0.2164,
+      "step": 128970
+    },
+    {
+      "epoch": 832.13,
+      "learning_rate": 9.326164874551972e-06,
+      "loss": 0.168,
+      "step": 128980
+    },
+    {
+      "epoch": 832.19,
+      "learning_rate": 9.32258064516129e-06,
+      "loss": 0.1905,
+      "step": 128990
+    },
+    {
+      "epoch": 832.26,
+      "learning_rate": 9.31899641577061e-06,
+      "loss": 0.1987,
+      "step": 129000
+    },
+    {
+      "epoch": 832.32,
+      "learning_rate": 9.315412186379929e-06,
+      "loss": 0.2396,
+      "step": 129010
+    },
+    {
+      "epoch": 832.39,
+      "learning_rate": 9.311827956989249e-06,
+      "loss": 0.2021,
+      "step": 129020
+    },
+    {
+      "epoch": 832.45,
+      "learning_rate": 9.308243727598567e-06,
+      "loss": 0.1989,
+      "step": 129030
+    },
+    {
+      "epoch": 832.51,
+      "learning_rate": 9.304659498207887e-06,
+      "loss": 0.2037,
+      "step": 129040
+    },
+    {
+      "epoch": 832.58,
+      "learning_rate": 9.301075268817205e-06,
+      "loss": 0.2175,
+      "step": 129050
+    },
+    {
+      "epoch": 832.64,
+      "learning_rate": 9.297491039426525e-06,
+      "loss": 0.1967,
+      "step": 129060
+    },
+    {
+      "epoch": 832.71,
+      "learning_rate": 9.293906810035843e-06,
+      "loss": 0.169,
+      "step": 129070
+    },
+    {
+      "epoch": 832.77,
+      "learning_rate": 9.290322580645163e-06,
+      "loss": 0.2276,
+      "step": 129080
+    },
+    {
+      "epoch": 832.84,
+      "learning_rate": 9.28673835125448e-06,
+      "loss": 0.1731,
+      "step": 129090
+    },
+    {
+      "epoch": 832.9,
+      "learning_rate": 9.2831541218638e-06,
+      "loss": 0.1779,
+      "step": 129100
+    },
+    {
+      "epoch": 832.96,
+      "learning_rate": 9.279569892473118e-06,
+      "loss": 0.2069,
+      "step": 129110
+    },
+    {
+      "epoch": 833.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9229941964149475,
+      "eval_runtime": 0.8624,
+      "eval_samples_per_second": 233.083,
+      "eval_steps_per_second": 8.117,
+      "step": 129115
+    },
+    {
+      "epoch": 833.03,
+      "learning_rate": 9.275985663082437e-06,
+      "loss": 0.1998,
+      "step": 129120
+    },
+    {
+      "epoch": 833.1,
+      "learning_rate": 9.272401433691756e-06,
+      "loss": 0.1837,
+      "step": 129130
+    },
+    {
+      "epoch": 833.16,
+      "learning_rate": 9.268817204301076e-06,
+      "loss": 0.2,
+      "step": 129140
+    },
+    {
+      "epoch": 833.23,
+      "learning_rate": 9.265232974910394e-06,
+      "loss": 0.214,
+      "step": 129150
+    },
+    {
+      "epoch": 833.29,
+      "learning_rate": 9.261648745519714e-06,
+      "loss": 0.2082,
+      "step": 129160
+    },
+    {
+      "epoch": 833.35,
+      "learning_rate": 9.258064516129032e-06,
+      "loss": 0.2075,
+      "step": 129170
+    },
+    {
+      "epoch": 833.42,
+      "learning_rate": 9.254480286738352e-06,
+      "loss": 0.1892,
+      "step": 129180
+    },
+    {
+      "epoch": 833.48,
+      "learning_rate": 9.25089605734767e-06,
+      "loss": 0.2794,
+      "step": 129190
+    },
+    {
+      "epoch": 833.55,
+      "learning_rate": 9.24731182795699e-06,
+      "loss": 0.1738,
+      "step": 129200
+    },
+    {
+      "epoch": 833.61,
+      "learning_rate": 9.243727598566308e-06,
+      "loss": 0.1748,
+      "step": 129210
+    },
+    {
+      "epoch": 833.68,
+      "learning_rate": 9.240143369175628e-06,
+      "loss": 0.1844,
+      "step": 129220
+    },
+    {
+      "epoch": 833.74,
+      "learning_rate": 9.236559139784946e-06,
+      "loss": 0.2081,
+      "step": 129230
+    },
+    {
+      "epoch": 833.8,
+      "learning_rate": 9.232974910394266e-06,
+      "loss": 0.2194,
+      "step": 129240
+    },
+    {
+      "epoch": 833.87,
+      "learning_rate": 9.229390681003584e-06,
+      "loss": 0.2822,
+      "step": 129250
+    },
+    {
+      "epoch": 833.93,
+      "learning_rate": 9.225806451612904e-06,
+      "loss": 0.2119,
+      "step": 129260
+    },
+    {
+      "epoch": 834.0,
+      "learning_rate": 9.222222222222222e-06,
+      "loss": 0.2018,
+      "step": 129270
+    },
+    {
+      "epoch": 834.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9177015423774719,
+      "eval_runtime": 0.8804,
+      "eval_samples_per_second": 228.296,
+      "eval_steps_per_second": 7.951,
+      "step": 129270
+    },
+    {
+      "epoch": 834.06,
+      "learning_rate": 9.21863799283154e-06,
+      "loss": 0.2078,
+      "step": 129280
+    },
+    {
+      "epoch": 834.13,
+      "learning_rate": 9.21505376344086e-06,
+      "loss": 0.2337,
+      "step": 129290
+    },
+    {
+      "epoch": 834.19,
+      "learning_rate": 9.211469534050179e-06,
+      "loss": 0.2335,
+      "step": 129300
+    },
+    {
+      "epoch": 834.26,
+      "learning_rate": 9.207885304659499e-06,
+      "loss": 0.2205,
+      "step": 129310
+    },
+    {
+      "epoch": 834.32,
+      "learning_rate": 9.204301075268817e-06,
+      "loss": 0.1979,
+      "step": 129320
+    },
+    {
+      "epoch": 834.39,
+      "learning_rate": 9.200716845878137e-06,
+      "loss": 0.1988,
+      "step": 129330
+    },
+    {
+      "epoch": 834.45,
+      "learning_rate": 9.197132616487455e-06,
+      "loss": 0.1852,
+      "step": 129340
+    },
+    {
+      "epoch": 834.51,
+      "learning_rate": 9.193548387096775e-06,
+      "loss": 0.176,
+      "step": 129350
+    },
+    {
+      "epoch": 834.58,
+      "learning_rate": 9.189964157706093e-06,
+      "loss": 0.2277,
+      "step": 129360
+    },
+    {
+      "epoch": 834.64,
+      "learning_rate": 9.186379928315413e-06,
+      "loss": 0.2062,
+      "step": 129370
+    },
+    {
+      "epoch": 834.71,
+      "learning_rate": 9.182795698924731e-06,
+      "loss": 0.2376,
+      "step": 129380
+    },
+    {
+      "epoch": 834.77,
+      "learning_rate": 9.179211469534051e-06,
+      "loss": 0.1919,
+      "step": 129390
+    },
+    {
+      "epoch": 834.84,
+      "learning_rate": 9.17562724014337e-06,
+      "loss": 0.2025,
+      "step": 129400
+    },
+    {
+      "epoch": 834.9,
+      "learning_rate": 9.172043010752689e-06,
+      "loss": 0.217,
+      "step": 129410
+    },
+    {
+      "epoch": 834.96,
+      "learning_rate": 9.168458781362007e-06,
+      "loss": 0.2669,
+      "step": 129420
+    },
+    {
+      "epoch": 835.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.941429615020752,
+      "eval_runtime": 0.8572,
+      "eval_samples_per_second": 234.485,
+      "eval_steps_per_second": 8.166,
+      "step": 129425
+    },
+    {
+      "epoch": 835.03,
+      "learning_rate": 9.164874551971327e-06,
+      "loss": 0.2169,
+      "step": 129430
+    },
+    {
+      "epoch": 835.1,
+      "learning_rate": 9.161290322580645e-06,
+      "loss": 0.1856,
+      "step": 129440
+    },
+    {
+      "epoch": 835.16,
+      "learning_rate": 9.157706093189965e-06,
+      "loss": 0.1579,
+      "step": 129450
+    },
+    {
+      "epoch": 835.23,
+      "learning_rate": 9.154121863799284e-06,
+      "loss": 0.1971,
+      "step": 129460
+    },
+    {
+      "epoch": 835.29,
+      "learning_rate": 9.150537634408603e-06,
+      "loss": 0.1941,
+      "step": 129470
+    },
+    {
+      "epoch": 835.35,
+      "learning_rate": 9.146953405017922e-06,
+      "loss": 0.1998,
+      "step": 129480
+    },
+    {
+      "epoch": 835.42,
+      "learning_rate": 9.143369175627242e-06,
+      "loss": 0.228,
+      "step": 129490
+    },
+    {
+      "epoch": 835.48,
+      "learning_rate": 9.13978494623656e-06,
+      "loss": 0.1548,
+      "step": 129500
+    },
+    {
+      "epoch": 835.55,
+      "learning_rate": 9.13620071684588e-06,
+      "loss": 0.2254,
+      "step": 129510
+    },
+    {
+      "epoch": 835.61,
+      "learning_rate": 9.132616487455198e-06,
+      "loss": 0.1492,
+      "step": 129520
+    },
+    {
+      "epoch": 835.68,
+      "learning_rate": 9.129032258064516e-06,
+      "loss": 0.1947,
+      "step": 129530
+    },
+    {
+      "epoch": 835.74,
+      "learning_rate": 9.125448028673836e-06,
+      "loss": 0.2174,
+      "step": 129540
+    },
+    {
+      "epoch": 835.8,
+      "learning_rate": 9.121863799283154e-06,
+      "loss": 0.1831,
+      "step": 129550
+    },
+    {
+      "epoch": 835.87,
+      "learning_rate": 9.118279569892474e-06,
+      "loss": 0.2027,
+      "step": 129560
+    },
+    {
+      "epoch": 835.93,
+      "learning_rate": 9.114695340501792e-06,
+      "loss": 0.2514,
+      "step": 129570
+    },
+    {
+      "epoch": 836.0,
+      "learning_rate": 9.111111111111112e-06,
+      "loss": 0.2014,
+      "step": 129580
+    },
+    {
+      "epoch": 836.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9567084908485413,
+      "eval_runtime": 0.8689,
+      "eval_samples_per_second": 231.332,
+      "eval_steps_per_second": 8.056,
+      "step": 129580
+    },
+    {
+      "epoch": 836.06,
+      "learning_rate": 9.10752688172043e-06,
+      "loss": 0.1979,
+      "step": 129590
+    },
+    {
+      "epoch": 836.13,
+      "learning_rate": 9.10394265232975e-06,
+      "loss": 0.1912,
+      "step": 129600
+    },
+    {
+      "epoch": 836.19,
+      "learning_rate": 9.100358422939068e-06,
+      "loss": 0.1978,
+      "step": 129610
+    },
+    {
+      "epoch": 836.26,
+      "learning_rate": 9.096774193548388e-06,
+      "loss": 0.2037,
+      "step": 129620
+    },
+    {
+      "epoch": 836.32,
+      "learning_rate": 9.093189964157707e-06,
+      "loss": 0.2087,
+      "step": 129630
+    },
+    {
+      "epoch": 836.39,
+      "learning_rate": 9.089605734767026e-06,
+      "loss": 0.2112,
+      "step": 129640
+    },
+    {
+      "epoch": 836.45,
+      "learning_rate": 9.086021505376345e-06,
+      "loss": 0.2047,
+      "step": 129650
+    },
+    {
+      "epoch": 836.51,
+      "learning_rate": 9.082437275985665e-06,
+      "loss": 0.1887,
+      "step": 129660
+    },
+    {
+      "epoch": 836.58,
+      "learning_rate": 9.078853046594983e-06,
+      "loss": 0.192,
+      "step": 129670
+    },
+    {
+      "epoch": 836.64,
+      "learning_rate": 9.075268817204303e-06,
+      "loss": 0.2378,
+      "step": 129680
+    },
+    {
+      "epoch": 836.71,
+      "learning_rate": 9.071684587813621e-06,
+      "loss": 0.1761,
+      "step": 129690
+    },
+    {
+      "epoch": 836.77,
+      "learning_rate": 9.06810035842294e-06,
+      "loss": 0.2329,
+      "step": 129700
+    },
+    {
+      "epoch": 836.84,
+      "learning_rate": 9.064516129032259e-06,
+      "loss": 0.2117,
+      "step": 129710
+    },
+    {
+      "epoch": 836.9,
+      "learning_rate": 9.060931899641579e-06,
+      "loss": 0.2254,
+      "step": 129720
+    },
+    {
+      "epoch": 836.96,
+      "learning_rate": 9.057347670250895e-06,
+      "loss": 0.2137,
+      "step": 129730
+    },
+    {
+      "epoch": 837.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9374553561210632,
+      "eval_runtime": 0.8868,
+      "eval_samples_per_second": 226.648,
+      "eval_steps_per_second": 7.893,
+      "step": 129735
+    },
+    {
+      "epoch": 837.03,
+      "learning_rate": 9.053763440860215e-06,
+      "loss": 0.2278,
+      "step": 129740
+    },
+    {
+      "epoch": 837.1,
+      "learning_rate": 9.050179211469534e-06,
+      "loss": 0.1928,
+      "step": 129750
+    },
+    {
+      "epoch": 837.16,
+      "learning_rate": 9.046594982078853e-06,
+      "loss": 0.1978,
+      "step": 129760
+    },
+    {
+      "epoch": 837.23,
+      "learning_rate": 9.043010752688172e-06,
+      "loss": 0.2579,
+      "step": 129770
+    },
+    {
+      "epoch": 837.29,
+      "learning_rate": 9.039426523297492e-06,
+      "loss": 0.1904,
+      "step": 129780
+    },
+    {
+      "epoch": 837.35,
+      "learning_rate": 9.03584229390681e-06,
+      "loss": 0.2066,
+      "step": 129790
+    },
+    {
+      "epoch": 837.42,
+      "learning_rate": 9.03225806451613e-06,
+      "loss": 0.1872,
+      "step": 129800
+    },
+    {
+      "epoch": 837.48,
+      "learning_rate": 9.028673835125448e-06,
+      "loss": 0.1661,
+      "step": 129810
+    },
+    {
+      "epoch": 837.55,
+      "learning_rate": 9.025089605734768e-06,
+      "loss": 0.1841,
+      "step": 129820
+    },
+    {
+      "epoch": 837.61,
+      "learning_rate": 9.021505376344086e-06,
+      "loss": 0.1659,
+      "step": 129830
+    },
+    {
+      "epoch": 837.68,
+      "learning_rate": 9.017921146953406e-06,
+      "loss": 0.1814,
+      "step": 129840
+    },
+    {
+      "epoch": 837.74,
+      "learning_rate": 9.014336917562724e-06,
+      "loss": 0.1897,
+      "step": 129850
+    },
+    {
+      "epoch": 837.8,
+      "learning_rate": 9.010752688172042e-06,
+      "loss": 0.203,
+      "step": 129860
+    },
+    {
+      "epoch": 837.87,
+      "learning_rate": 9.007168458781362e-06,
+      "loss": 0.2286,
+      "step": 129870
+    },
+    {
+      "epoch": 837.93,
+      "learning_rate": 9.00358422939068e-06,
+      "loss": 0.1932,
+      "step": 129880
+    },
+    {
+      "epoch": 838.0,
+      "learning_rate": 9e-06,
+      "loss": 0.235,
+      "step": 129890
+    },
+    {
+      "epoch": 838.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9333366751670837,
+      "eval_runtime": 0.866,
+      "eval_samples_per_second": 232.113,
+      "eval_steps_per_second": 8.084,
+      "step": 129890
+    },
+    {
+      "epoch": 838.06,
+      "learning_rate": 8.996415770609318e-06,
+      "loss": 0.1941,
+      "step": 129900
+    },
+    {
+      "epoch": 838.13,
+      "learning_rate": 8.992831541218638e-06,
+      "loss": 0.23,
+      "step": 129910
+    },
+    {
+      "epoch": 838.19,
+      "learning_rate": 8.989247311827957e-06,
+      "loss": 0.1906,
+      "step": 129920
+    },
+    {
+      "epoch": 838.26,
+      "learning_rate": 8.985663082437276e-06,
+      "loss": 0.1946,
+      "step": 129930
+    },
+    {
+      "epoch": 838.32,
+      "learning_rate": 8.982078853046595e-06,
+      "loss": 0.192,
+      "step": 129940
+    },
+    {
+      "epoch": 838.39,
+      "learning_rate": 8.978494623655915e-06,
+      "loss": 0.1929,
+      "step": 129950
+    },
+    {
+      "epoch": 838.45,
+      "learning_rate": 8.974910394265233e-06,
+      "loss": 0.2533,
+      "step": 129960
+    },
+    {
+      "epoch": 838.51,
+      "learning_rate": 8.971326164874553e-06,
+      "loss": 0.2034,
+      "step": 129970
+    },
+    {
+      "epoch": 838.58,
+      "learning_rate": 8.967741935483871e-06,
+      "loss": 0.2293,
+      "step": 129980
+    },
+    {
+      "epoch": 838.64,
+      "learning_rate": 8.96415770609319e-06,
+      "loss": 0.2579,
+      "step": 129990
+    },
+    {
+      "epoch": 838.71,
+      "learning_rate": 8.960573476702509e-06,
+      "loss": 0.2049,
+      "step": 130000
+    },
+    {
+      "epoch": 838.77,
+      "learning_rate": 8.956989247311829e-06,
+      "loss": 0.2185,
+      "step": 130010
+    },
+    {
+      "epoch": 838.84,
+      "learning_rate": 8.953405017921147e-06,
+      "loss": 0.2249,
+      "step": 130020
+    },
+    {
+      "epoch": 838.9,
+      "learning_rate": 8.949820788530467e-06,
+      "loss": 0.1662,
+      "step": 130030
+    },
+    {
+      "epoch": 838.96,
+      "learning_rate": 8.946236559139785e-06,
+      "loss": 0.2414,
+      "step": 130040
+    },
+    {
+      "epoch": 839.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9146257638931274,
+      "eval_runtime": 0.9003,
+      "eval_samples_per_second": 223.264,
+      "eval_steps_per_second": 7.775,
+      "step": 130045
+    },
+    {
+      "epoch": 839.03,
+      "learning_rate": 8.942652329749105e-06,
+      "loss": 0.1987,
+      "step": 130050
+    },
+    {
+      "epoch": 839.1,
+      "learning_rate": 8.939068100358423e-06,
+      "loss": 0.1991,
+      "step": 130060
+    },
+    {
+      "epoch": 839.16,
+      "learning_rate": 8.935483870967743e-06,
+      "loss": 0.2203,
+      "step": 130070
+    },
+    {
+      "epoch": 839.23,
+      "learning_rate": 8.931899641577061e-06,
+      "loss": 0.208,
+      "step": 130080
+    },
+    {
+      "epoch": 839.29,
+      "learning_rate": 8.928315412186381e-06,
+      "loss": 0.1978,
+      "step": 130090
+    },
+    {
+      "epoch": 839.35,
+      "learning_rate": 8.9247311827957e-06,
+      "loss": 0.1779,
+      "step": 130100
+    },
+    {
+      "epoch": 839.42,
+      "learning_rate": 8.921146953405018e-06,
+      "loss": 0.1862,
+      "step": 130110
+    },
+    {
+      "epoch": 839.48,
+      "learning_rate": 8.917562724014338e-06,
+      "loss": 0.1868,
+      "step": 130120
+    },
+    {
+      "epoch": 839.55,
+      "learning_rate": 8.913978494623656e-06,
+      "loss": 0.2627,
+      "step": 130130
+    },
+    {
+      "epoch": 839.61,
+      "learning_rate": 8.910394265232976e-06,
+      "loss": 0.2199,
+      "step": 130140
+    },
+    {
+      "epoch": 839.68,
+      "learning_rate": 8.906810035842294e-06,
+      "loss": 0.1965,
+      "step": 130150
+    },
+    {
+      "epoch": 839.74,
+      "learning_rate": 8.903225806451614e-06,
+      "loss": 0.2613,
+      "step": 130160
+    },
+    {
+      "epoch": 839.8,
+      "learning_rate": 8.899641577060932e-06,
+      "loss": 0.1931,
+      "step": 130170
+    },
+    {
+      "epoch": 839.87,
+      "learning_rate": 8.896057347670252e-06,
+      "loss": 0.1625,
+      "step": 130180
+    },
+    {
+      "epoch": 839.93,
+      "learning_rate": 8.89247311827957e-06,
+      "loss": 0.2096,
+      "step": 130190
+    },
+    {
+      "epoch": 840.0,
+      "learning_rate": 8.88888888888889e-06,
+      "loss": 0.1687,
+      "step": 130200
+    },
+    {
+      "epoch": 840.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9051613807678223,
+      "eval_runtime": 0.8821,
+      "eval_samples_per_second": 227.869,
+      "eval_steps_per_second": 7.936,
+      "step": 130200
+    },
+    {
+      "epoch": 840.06,
+      "learning_rate": 8.885304659498208e-06,
+      "loss": 0.2348,
+      "step": 130210
+    },
+    {
+      "epoch": 840.13,
+      "learning_rate": 8.881720430107528e-06,
+      "loss": 0.231,
+      "step": 130220
+    },
+    {
+      "epoch": 840.19,
+      "learning_rate": 8.878136200716846e-06,
+      "loss": 0.1674,
+      "step": 130230
+    },
+    {
+      "epoch": 840.26,
+      "learning_rate": 8.874551971326166e-06,
+      "loss": 0.2045,
+      "step": 130240
+    },
+    {
+      "epoch": 840.32,
+      "learning_rate": 8.870967741935484e-06,
+      "loss": 0.2302,
+      "step": 130250
+    },
+    {
+      "epoch": 840.39,
+      "learning_rate": 8.867383512544804e-06,
+      "loss": 0.2476,
+      "step": 130260
+    },
+    {
+      "epoch": 840.45,
+      "learning_rate": 8.863799283154123e-06,
+      "loss": 0.1823,
+      "step": 130270
+    },
+    {
+      "epoch": 840.51,
+      "learning_rate": 8.860215053763442e-06,
+      "loss": 0.2115,
+      "step": 130280
+    },
+    {
+      "epoch": 840.58,
+      "learning_rate": 8.85663082437276e-06,
+      "loss": 0.2298,
+      "step": 130290
+    },
+    {
+      "epoch": 840.64,
+      "learning_rate": 8.85304659498208e-06,
+      "loss": 0.2374,
+      "step": 130300
+    },
+    {
+      "epoch": 840.71,
+      "learning_rate": 8.849462365591399e-06,
+      "loss": 0.2125,
+      "step": 130310
+    },
+    {
+      "epoch": 840.77,
+      "learning_rate": 8.845878136200719e-06,
+      "loss": 0.2205,
+      "step": 130320
+    },
+    {
+      "epoch": 840.84,
+      "learning_rate": 8.842293906810037e-06,
+      "loss": 0.1818,
+      "step": 130330
+    },
+    {
+      "epoch": 840.9,
+      "learning_rate": 8.838709677419357e-06,
+      "loss": 0.199,
+      "step": 130340
+    },
+    {
+      "epoch": 840.96,
+      "learning_rate": 8.835125448028675e-06,
+      "loss": 0.1929,
+      "step": 130350
+    },
+    {
+      "epoch": 841.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.903295636177063,
+      "eval_runtime": 0.8542,
+      "eval_samples_per_second": 235.313,
+      "eval_steps_per_second": 8.195,
+      "step": 130355
+    },
+    {
+      "epoch": 841.03,
+      "learning_rate": 8.831541218637993e-06,
+      "loss": 0.2259,
+      "step": 130360
+    },
+    {
+      "epoch": 841.1,
+      "learning_rate": 8.827956989247311e-06,
+      "loss": 0.2272,
+      "step": 130370
+    },
+    {
+      "epoch": 841.16,
+      "learning_rate": 8.824372759856631e-06,
+      "loss": 0.2124,
+      "step": 130380
+    },
+    {
+      "epoch": 841.23,
+      "learning_rate": 8.82078853046595e-06,
+      "loss": 0.1949,
+      "step": 130390
+    },
+    {
+      "epoch": 841.29,
+      "learning_rate": 8.81720430107527e-06,
+      "loss": 0.2067,
+      "step": 130400
+    },
+    {
+      "epoch": 841.35,
+      "learning_rate": 8.813620071684588e-06,
+      "loss": 0.2243,
+      "step": 130410
+    },
+    {
+      "epoch": 841.42,
+      "learning_rate": 8.810035842293907e-06,
+      "loss": 0.1664,
+      "step": 130420
+    },
+    {
+      "epoch": 841.48,
+      "learning_rate": 8.806451612903226e-06,
+      "loss": 0.1758,
+      "step": 130430
+    },
+    {
+      "epoch": 841.55,
+      "learning_rate": 8.802867383512544e-06,
+      "loss": 0.2862,
+      "step": 130440
+    },
+    {
+      "epoch": 841.61,
+      "learning_rate": 8.799283154121864e-06,
+      "loss": 0.24,
+      "step": 130450
+    },
+    {
+      "epoch": 841.68,
+      "learning_rate": 8.795698924731182e-06,
+      "loss": 0.2114,
+      "step": 130460
+    },
+    {
+      "epoch": 841.74,
+      "learning_rate": 8.792114695340502e-06,
+      "loss": 0.184,
+      "step": 130470
+    },
+    {
+      "epoch": 841.8,
+      "learning_rate": 8.78853046594982e-06,
+      "loss": 0.2252,
+      "step": 130480
+    },
+    {
+      "epoch": 841.87,
+      "learning_rate": 8.78494623655914e-06,
+      "loss": 0.1912,
+      "step": 130490
+    },
+    {
+      "epoch": 841.93,
+      "learning_rate": 8.781362007168458e-06,
+      "loss": 0.1755,
+      "step": 130500
+    },
+    {
+      "epoch": 842.0,
+      "learning_rate": 8.777777777777778e-06,
+      "loss": 0.1672,
+      "step": 130510
+    },
+    {
+      "epoch": 842.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.908940315246582,
+      "eval_runtime": 0.8897,
+      "eval_samples_per_second": 225.926,
+      "eval_steps_per_second": 7.868,
+      "step": 130510
+    },
+    {
+      "epoch": 842.06,
+      "learning_rate": 8.774193548387096e-06,
+      "loss": 0.2039,
+      "step": 130520
+    },
+    {
+      "epoch": 842.13,
+      "learning_rate": 8.770609318996416e-06,
+      "loss": 0.2225,
+      "step": 130530
+    },
+    {
+      "epoch": 842.19,
+      "learning_rate": 8.767025089605734e-06,
+      "loss": 0.2374,
+      "step": 130540
+    },
+    {
+      "epoch": 842.26,
+      "learning_rate": 8.763440860215054e-06,
+      "loss": 0.2121,
+      "step": 130550
+    },
+    {
+      "epoch": 842.32,
+      "learning_rate": 8.759856630824372e-06,
+      "loss": 0.1893,
+      "step": 130560
+    },
+    {
+      "epoch": 842.39,
+      "learning_rate": 8.756272401433692e-06,
+      "loss": 0.2254,
+      "step": 130570
+    },
+    {
+      "epoch": 842.45,
+      "learning_rate": 8.75268817204301e-06,
+      "loss": 0.1883,
+      "step": 130580
+    },
+    {
+      "epoch": 842.51,
+      "learning_rate": 8.74910394265233e-06,
+      "loss": 0.191,
+      "step": 130590
+    },
+    {
+      "epoch": 842.58,
+      "learning_rate": 8.745519713261649e-06,
+      "loss": 0.2075,
+      "step": 130600
+    },
+    {
+      "epoch": 842.64,
+      "learning_rate": 8.741935483870969e-06,
+      "loss": 0.1917,
+      "step": 130610
+    },
+    {
+      "epoch": 842.71,
+      "learning_rate": 8.738351254480287e-06,
+      "loss": 0.2361,
+      "step": 130620
+    },
+    {
+      "epoch": 842.77,
+      "learning_rate": 8.734767025089607e-06,
+      "loss": 0.2482,
+      "step": 130630
+    },
+    {
+      "epoch": 842.84,
+      "learning_rate": 8.731182795698925e-06,
+      "loss": 0.218,
+      "step": 130640
+    },
+    {
+      "epoch": 842.9,
+      "learning_rate": 8.727598566308245e-06,
+      "loss": 0.1986,
+      "step": 130650
+    },
+    {
+      "epoch": 842.96,
+      "learning_rate": 8.724014336917563e-06,
+      "loss": 0.2021,
+      "step": 130660
+    },
+    {
+      "epoch": 843.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.92840576171875,
+      "eval_runtime": 0.8876,
+      "eval_samples_per_second": 226.444,
+      "eval_steps_per_second": 7.886,
+      "step": 130665
+    },
+    {
+      "epoch": 843.03,
+      "learning_rate": 8.720430107526883e-06,
+      "loss": 0.2399,
+      "step": 130670
+    },
+    {
+      "epoch": 843.1,
+      "learning_rate": 8.716845878136201e-06,
+      "loss": 0.1919,
+      "step": 130680
+    },
+    {
+      "epoch": 843.16,
+      "learning_rate": 8.71326164874552e-06,
+      "loss": 0.2032,
+      "step": 130690
+    },
+    {
+      "epoch": 843.23,
+      "learning_rate": 8.70967741935484e-06,
+      "loss": 0.1818,
+      "step": 130700
+    },
+    {
+      "epoch": 843.29,
+      "learning_rate": 8.706093189964157e-06,
+      "loss": 0.2074,
+      "step": 130710
+    },
+    {
+      "epoch": 843.35,
+      "learning_rate": 8.702508960573477e-06,
+      "loss": 0.2501,
+      "step": 130720
+    },
+    {
+      "epoch": 843.42,
+      "learning_rate": 8.698924731182796e-06,
+      "loss": 0.1822,
+      "step": 130730
+    },
+    {
+      "epoch": 843.48,
+      "learning_rate": 8.695340501792115e-06,
+      "loss": 0.2014,
+      "step": 130740
+    },
+    {
+      "epoch": 843.55,
+      "learning_rate": 8.691756272401434e-06,
+      "loss": 0.2421,
+      "step": 130750
+    },
+    {
+      "epoch": 843.61,
+      "learning_rate": 8.688172043010754e-06,
+      "loss": 0.1942,
+      "step": 130760
+    },
+    {
+      "epoch": 843.68,
+      "learning_rate": 8.684587813620072e-06,
+      "loss": 0.272,
+      "step": 130770
+    },
+    {
+      "epoch": 843.74,
+      "learning_rate": 8.681003584229392e-06,
+      "loss": 0.2323,
+      "step": 130780
+    },
+    {
+      "epoch": 843.8,
+      "learning_rate": 8.67741935483871e-06,
+      "loss": 0.196,
+      "step": 130790
+    },
+    {
+      "epoch": 843.87,
+      "learning_rate": 8.67383512544803e-06,
+      "loss": 0.1804,
+      "step": 130800
+    },
+    {
+      "epoch": 843.93,
+      "learning_rate": 8.670250896057348e-06,
+      "loss": 0.2201,
+      "step": 130810
+    },
+    {
+      "epoch": 844.0,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 0.1984,
+      "step": 130820
+    },
+    {
+      "epoch": 844.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9314678311347961,
+      "eval_runtime": 0.868,
+      "eval_samples_per_second": 231.555,
+      "eval_steps_per_second": 8.064,
+      "step": 130820
+    },
+    {
+      "epoch": 844.06,
+      "learning_rate": 8.663082437275986e-06,
+      "loss": 0.1966,
+      "step": 130830
+    },
+    {
+      "epoch": 844.13,
+      "learning_rate": 8.659498207885306e-06,
+      "loss": 0.1973,
+      "step": 130840
+    },
+    {
+      "epoch": 844.19,
+      "learning_rate": 8.655913978494624e-06,
+      "loss": 0.1649,
+      "step": 130850
+    },
+    {
+      "epoch": 844.26,
+      "learning_rate": 8.652329749103944e-06,
+      "loss": 0.1684,
+      "step": 130860
+    },
+    {
+      "epoch": 844.32,
+      "learning_rate": 8.648745519713262e-06,
+      "loss": 0.1832,
+      "step": 130870
+    },
+    {
+      "epoch": 844.39,
+      "learning_rate": 8.645161290322582e-06,
+      "loss": 0.1855,
+      "step": 130880
+    },
+    {
+      "epoch": 844.45,
+      "learning_rate": 8.6415770609319e-06,
+      "loss": 0.1748,
+      "step": 130890
+    },
+    {
+      "epoch": 844.51,
+      "learning_rate": 8.63799283154122e-06,
+      "loss": 0.2252,
+      "step": 130900
+    },
+    {
+      "epoch": 844.58,
+      "learning_rate": 8.634408602150538e-06,
+      "loss": 0.2302,
+      "step": 130910
+    },
+    {
+      "epoch": 844.64,
+      "learning_rate": 8.630824372759858e-06,
+      "loss": 0.1999,
+      "step": 130920
+    },
+    {
+      "epoch": 844.71,
+      "learning_rate": 8.627240143369177e-06,
+      "loss": 0.2077,
+      "step": 130930
+    },
+    {
+      "epoch": 844.77,
+      "learning_rate": 8.623655913978495e-06,
+      "loss": 0.1896,
+      "step": 130940
+    },
+    {
+      "epoch": 844.84,
+      "learning_rate": 8.620071684587815e-06,
+      "loss": 0.1924,
+      "step": 130950
+    },
+    {
+      "epoch": 844.9,
+      "learning_rate": 8.616487455197133e-06,
+      "loss": 0.1941,
+      "step": 130960
+    },
+    {
+      "epoch": 844.96,
+      "learning_rate": 8.612903225806453e-06,
+      "loss": 0.2223,
+      "step": 130970
+    },
+    {
+      "epoch": 845.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9204798340797424,
+      "eval_runtime": 0.851,
+      "eval_samples_per_second": 236.205,
+      "eval_steps_per_second": 8.226,
+      "step": 130975
+    },
+    {
+      "epoch": 845.03,
+      "learning_rate": 8.609318996415771e-06,
+      "loss": 0.2681,
+      "step": 130980
+    },
+    {
+      "epoch": 845.1,
+      "learning_rate": 8.605734767025091e-06,
+      "loss": 0.2073,
+      "step": 130990
+    },
+    {
+      "epoch": 845.16,
+      "learning_rate": 8.602150537634409e-06,
+      "loss": 0.1868,
+      "step": 131000
+    },
+    {
+      "epoch": 845.23,
+      "learning_rate": 8.598566308243727e-06,
+      "loss": 0.2229,
+      "step": 131010
+    },
+    {
+      "epoch": 845.29,
+      "learning_rate": 8.594982078853047e-06,
+      "loss": 0.2088,
+      "step": 131020
+    },
+    {
+      "epoch": 845.35,
+      "learning_rate": 8.591397849462365e-06,
+      "loss": 0.2265,
+      "step": 131030
+    },
+    {
+      "epoch": 845.42,
+      "learning_rate": 8.587813620071684e-06,
+      "loss": 0.1612,
+      "step": 131040
+    },
+    {
+      "epoch": 845.48,
+      "learning_rate": 8.584229390681003e-06,
+      "loss": 0.2169,
+      "step": 131050
+    },
+    {
+      "epoch": 845.55,
+      "learning_rate": 8.580645161290322e-06,
+      "loss": 0.2101,
+      "step": 131060
+    },
+    {
+      "epoch": 845.61,
+      "learning_rate": 8.577060931899642e-06,
+      "loss": 0.224,
+      "step": 131070
+    },
+    {
+      "epoch": 845.68,
+      "learning_rate": 8.57347670250896e-06,
+      "loss": 0.1993,
+      "step": 131080
+    },
+    {
+      "epoch": 845.74,
+      "learning_rate": 8.56989247311828e-06,
+      "loss": 0.2412,
+      "step": 131090
+    },
+    {
+      "epoch": 845.8,
+      "learning_rate": 8.566308243727598e-06,
+      "loss": 0.1696,
+      "step": 131100
+    },
+    {
+      "epoch": 845.87,
+      "learning_rate": 8.562724014336918e-06,
+      "loss": 0.1742,
+      "step": 131110
+    },
+    {
+      "epoch": 845.93,
+      "learning_rate": 8.559139784946236e-06,
+      "loss": 0.1997,
+      "step": 131120
+    },
+    {
+      "epoch": 846.0,
+      "learning_rate": 8.555555555555556e-06,
+      "loss": 0.1963,
+      "step": 131130
+    },
+    {
+      "epoch": 846.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9149745106697083,
+      "eval_runtime": 0.9049,
+      "eval_samples_per_second": 222.119,
+      "eval_steps_per_second": 7.735,
+      "step": 131130
+    },
+    {
+      "epoch": 846.06,
+      "learning_rate": 8.551971326164874e-06,
+      "loss": 0.2428,
+      "step": 131140
+    },
+    {
+      "epoch": 846.13,
+      "learning_rate": 8.548387096774194e-06,
+      "loss": 0.1602,
+      "step": 131150
+    },
+    {
+      "epoch": 846.19,
+      "learning_rate": 8.544802867383512e-06,
+      "loss": 0.226,
+      "step": 131160
+    },
+    {
+      "epoch": 846.26,
+      "learning_rate": 8.541218637992832e-06,
+      "loss": 0.2176,
+      "step": 131170
+    },
+    {
+      "epoch": 846.32,
+      "learning_rate": 8.53763440860215e-06,
+      "loss": 0.2351,
+      "step": 131180
+    },
+    {
+      "epoch": 846.39,
+      "learning_rate": 8.53405017921147e-06,
+      "loss": 0.1936,
+      "step": 131190
+    },
+    {
+      "epoch": 846.45,
+      "learning_rate": 8.530465949820788e-06,
+      "loss": 0.2273,
+      "step": 131200
+    },
+    {
+      "epoch": 846.51,
+      "learning_rate": 8.526881720430108e-06,
+      "loss": 0.224,
+      "step": 131210
+    },
+    {
+      "epoch": 846.58,
+      "learning_rate": 8.523297491039427e-06,
+      "loss": 0.2389,
+      "step": 131220
+    },
+    {
+      "epoch": 846.64,
+      "learning_rate": 8.519713261648746e-06,
+      "loss": 0.1815,
+      "step": 131230
+    },
+    {
+      "epoch": 846.71,
+      "learning_rate": 8.516129032258065e-06,
+      "loss": 0.2147,
+      "step": 131240
+    },
+    {
+      "epoch": 846.77,
+      "learning_rate": 8.512544802867385e-06,
+      "loss": 0.2107,
+      "step": 131250
+    },
+    {
+      "epoch": 846.84,
+      "learning_rate": 8.508960573476703e-06,
+      "loss": 0.2203,
+      "step": 131260
+    },
+    {
+      "epoch": 846.9,
+      "learning_rate": 8.505376344086023e-06,
+      "loss": 0.1992,
+      "step": 131270
+    },
+    {
+      "epoch": 846.96,
+      "learning_rate": 8.50179211469534e-06,
+      "loss": 0.1974,
+      "step": 131280
+    },
+    {
+      "epoch": 847.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9006843566894531,
+      "eval_runtime": 0.8639,
+      "eval_samples_per_second": 232.672,
+      "eval_steps_per_second": 8.103,
+      "step": 131285
+    },
+    {
+      "epoch": 847.03,
+      "learning_rate": 8.498207885304659e-06,
+      "loss": 0.276,
+      "step": 131290
+    },
+    {
+      "epoch": 847.1,
+      "learning_rate": 8.494623655913979e-06,
+      "loss": 0.1902,
+      "step": 131300
+    },
+    {
+      "epoch": 847.16,
+      "learning_rate": 8.491039426523297e-06,
+      "loss": 0.1691,
+      "step": 131310
+    },
+    {
+      "epoch": 847.23,
+      "learning_rate": 8.487455197132617e-06,
+      "loss": 0.2045,
+      "step": 131320
+    },
+    {
+      "epoch": 847.29,
+      "learning_rate": 8.483870967741935e-06,
+      "loss": 0.1641,
+      "step": 131330
+    },
+    {
+      "epoch": 847.35,
+      "learning_rate": 8.480286738351255e-06,
+      "loss": 0.2348,
+      "step": 131340
+    },
+    {
+      "epoch": 847.42,
+      "learning_rate": 8.476702508960573e-06,
+      "loss": 0.186,
+      "step": 131350
+    },
+    {
+      "epoch": 847.48,
+      "learning_rate": 8.473118279569893e-06,
+      "loss": 0.2378,
+      "step": 131360
+    },
+    {
+      "epoch": 847.55,
+      "learning_rate": 8.469534050179211e-06,
+      "loss": 0.1718,
+      "step": 131370
+    },
+    {
+      "epoch": 847.61,
+      "learning_rate": 8.465949820788531e-06,
+      "loss": 0.2084,
+      "step": 131380
+    },
+    {
+      "epoch": 847.68,
+      "learning_rate": 8.46236559139785e-06,
+      "loss": 0.2405,
+      "step": 131390
+    },
+    {
+      "epoch": 847.74,
+      "learning_rate": 8.45878136200717e-06,
+      "loss": 0.2266,
+      "step": 131400
+    },
+    {
+      "epoch": 847.8,
+      "learning_rate": 8.455197132616488e-06,
+      "loss": 0.2148,
+      "step": 131410
+    },
+    {
+      "epoch": 847.87,
+      "learning_rate": 8.451612903225808e-06,
+      "loss": 0.2804,
+      "step": 131420
+    },
+    {
+      "epoch": 847.93,
+      "learning_rate": 8.448028673835126e-06,
+      "loss": 0.1826,
+      "step": 131430
+    },
+    {
+      "epoch": 848.0,
+      "learning_rate": 8.444444444444446e-06,
+      "loss": 0.1714,
+      "step": 131440
+    },
+    {
+      "epoch": 848.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.8926525115966797,
+      "eval_runtime": 0.8832,
+      "eval_samples_per_second": 227.59,
+      "eval_steps_per_second": 7.926,
+      "step": 131440
+    },
+    {
+      "epoch": 848.06,
+      "learning_rate": 8.440860215053764e-06,
+      "loss": 0.244,
+      "step": 131450
+    },
+    {
+      "epoch": 848.13,
+      "learning_rate": 8.437275985663084e-06,
+      "loss": 0.1586,
+      "step": 131460
+    },
+    {
+      "epoch": 848.19,
+      "learning_rate": 8.433691756272402e-06,
+      "loss": 0.2006,
+      "step": 131470
+    },
+    {
+      "epoch": 848.26,
+      "learning_rate": 8.430107526881722e-06,
+      "loss": 0.2086,
+      "step": 131480
+    },
+    {
+      "epoch": 848.32,
+      "learning_rate": 8.42652329749104e-06,
+      "loss": 0.1854,
+      "step": 131490
+    },
+    {
+      "epoch": 848.39,
+      "learning_rate": 8.42293906810036e-06,
+      "loss": 0.2231,
+      "step": 131500
+    },
+    {
+      "epoch": 848.45,
+      "learning_rate": 8.419354838709678e-06,
+      "loss": 0.2368,
+      "step": 131510
+    },
+    {
+      "epoch": 848.51,
+      "learning_rate": 8.415770609318998e-06,
+      "loss": 0.2156,
+      "step": 131520
+    },
+    {
+      "epoch": 848.58,
+      "learning_rate": 8.412186379928316e-06,
+      "loss": 0.2104,
+      "step": 131530
+    },
+    {
+      "epoch": 848.64,
+      "learning_rate": 8.408602150537634e-06,
+      "loss": 0.1911,
+      "step": 131540
+    },
+    {
+      "epoch": 848.71,
+      "learning_rate": 8.405017921146954e-06,
+      "loss": 0.1936,
+      "step": 131550
+    },
+    {
+      "epoch": 848.77,
+      "learning_rate": 8.401433691756273e-06,
+      "loss": 0.2088,
+      "step": 131560
+    },
+    {
+      "epoch": 848.84,
+      "learning_rate": 8.397849462365592e-06,
+      "loss": 0.2018,
+      "step": 131570
+    },
+    {
+      "epoch": 848.9,
+      "learning_rate": 8.39426523297491e-06,
+      "loss": 0.1603,
+      "step": 131580
+    },
+    {
+      "epoch": 848.96,
+      "learning_rate": 8.39068100358423e-06,
+      "loss": 0.2223,
+      "step": 131590
+    },
+    {
+      "epoch": 849.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9520587921142578,
+      "eval_runtime": 0.9093,
+      "eval_samples_per_second": 221.043,
+      "eval_steps_per_second": 7.698,
+      "step": 131595
+    },
+    {
+      "epoch": 849.03,
+      "learning_rate": 8.387096774193549e-06,
+      "loss": 0.2449,
+      "step": 131600
+    },
+    {
+      "epoch": 849.1,
+      "learning_rate": 8.383512544802869e-06,
+      "loss": 0.2125,
+      "step": 131610
+    },
+    {
+      "epoch": 849.16,
+      "learning_rate": 8.379928315412187e-06,
+      "loss": 0.2026,
+      "step": 131620
+    },
+    {
+      "epoch": 849.23,
+      "learning_rate": 8.376344086021507e-06,
+      "loss": 0.1946,
+      "step": 131630
+    },
+    {
+      "epoch": 849.29,
+      "learning_rate": 8.372759856630825e-06,
+      "loss": 0.2036,
+      "step": 131640
+    },
+    {
+      "epoch": 849.35,
+      "learning_rate": 8.369175627240143e-06,
+      "loss": 0.2396,
+      "step": 131650
+    },
+    {
+      "epoch": 849.42,
+      "learning_rate": 8.365591397849461e-06,
+      "loss": 0.2184,
+      "step": 131660
+    },
+    {
+      "epoch": 849.48,
+      "learning_rate": 8.362007168458781e-06,
+      "loss": 0.1908,
+      "step": 131670
+    },
+    {
+      "epoch": 849.55,
+      "learning_rate": 8.3584229390681e-06,
+      "loss": 0.2112,
+      "step": 131680
+    },
+    {
+      "epoch": 849.61,
+      "learning_rate": 8.35483870967742e-06,
+      "loss": 0.2349,
+      "step": 131690
+    },
+    {
+      "epoch": 849.68,
+      "learning_rate": 8.351254480286738e-06,
+      "loss": 0.2297,
+      "step": 131700
+    },
+    {
+      "epoch": 849.74,
+      "learning_rate": 8.347670250896058e-06,
+      "loss": 0.2087,
+      "step": 131710
+    },
+    {
+      "epoch": 849.8,
+      "learning_rate": 8.344086021505376e-06,
+      "loss": 0.2167,
+      "step": 131720
+    },
+    {
+      "epoch": 849.87,
+      "learning_rate": 8.340501792114696e-06,
+      "loss": 0.1895,
+      "step": 131730
+    },
+    {
+      "epoch": 849.93,
+      "learning_rate": 8.336917562724014e-06,
+      "loss": 0.1933,
+      "step": 131740
+    },
+    {
+      "epoch": 850.0,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.2352,
+      "step": 131750
+    },
+    {
+      "epoch": 850.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9149027466773987,
+      "eval_runtime": 0.8653,
+      "eval_samples_per_second": 232.299,
+      "eval_steps_per_second": 8.09,
+      "step": 131750
+    },
+    {
+      "epoch": 850.06,
+      "learning_rate": 8.329749103942652e-06,
+      "loss": 0.2565,
+      "step": 131760
+    },
+    {
+      "epoch": 850.13,
+      "learning_rate": 8.326164874551972e-06,
+      "loss": 0.2184,
+      "step": 131770
+    },
+    {
+      "epoch": 850.19,
+      "learning_rate": 8.32258064516129e-06,
+      "loss": 0.2007,
+      "step": 131780
+    },
+    {
+      "epoch": 850.26,
+      "learning_rate": 8.31899641577061e-06,
+      "loss": 0.1999,
+      "step": 131790
+    },
+    {
+      "epoch": 850.32,
+      "learning_rate": 8.315412186379928e-06,
+      "loss": 0.2244,
+      "step": 131800
+    },
+    {
+      "epoch": 850.39,
+      "learning_rate": 8.311827956989248e-06,
+      "loss": 0.1982,
+      "step": 131810
+    },
+    {
+      "epoch": 850.45,
+      "learning_rate": 8.308243727598566e-06,
+      "loss": 0.187,
+      "step": 131820
+    },
+    {
+      "epoch": 850.51,
+      "learning_rate": 8.304659498207886e-06,
+      "loss": 0.2109,
+      "step": 131830
+    },
+    {
+      "epoch": 850.58,
+      "learning_rate": 8.301075268817204e-06,
+      "loss": 0.177,
+      "step": 131840
+    },
+    {
+      "epoch": 850.64,
+      "learning_rate": 8.297491039426524e-06,
+      "loss": 0.1907,
+      "step": 131850
+    },
+    {
+      "epoch": 850.71,
+      "learning_rate": 8.293906810035842e-06,
+      "loss": 0.1516,
+      "step": 131860
+    },
+    {
+      "epoch": 850.77,
+      "learning_rate": 8.29032258064516e-06,
+      "loss": 0.2096,
+      "step": 131870
+    },
+    {
+      "epoch": 850.84,
+      "learning_rate": 8.28673835125448e-06,
+      "loss": 0.2226,
+      "step": 131880
+    },
+    {
+      "epoch": 850.9,
+      "learning_rate": 8.283154121863799e-06,
+      "loss": 0.2256,
+      "step": 131890
+    },
+    {
+      "epoch": 850.96,
+      "learning_rate": 8.279569892473119e-06,
+      "loss": 0.2049,
+      "step": 131900
+    },
+    {
+      "epoch": 851.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9140921235084534,
+      "eval_runtime": 0.8721,
+      "eval_samples_per_second": 230.482,
+      "eval_steps_per_second": 8.027,
+      "step": 131905
+    },
+    {
+      "epoch": 851.03,
+      "learning_rate": 8.275985663082437e-06,
+      "loss": 0.2095,
+      "step": 131910
+    },
+    {
+      "epoch": 851.1,
+      "learning_rate": 8.272401433691757e-06,
+      "loss": 0.2166,
+      "step": 131920
+    },
+    {
+      "epoch": 851.16,
+      "learning_rate": 8.268817204301075e-06,
+      "loss": 0.19,
+      "step": 131930
+    },
+    {
+      "epoch": 851.23,
+      "learning_rate": 8.265232974910395e-06,
+      "loss": 0.1682,
+      "step": 131940
+    },
+    {
+      "epoch": 851.29,
+      "learning_rate": 8.261648745519713e-06,
+      "loss": 0.2353,
+      "step": 131950
+    },
+    {
+      "epoch": 851.35,
+      "learning_rate": 8.258064516129033e-06,
+      "loss": 0.2286,
+      "step": 131960
+    },
+    {
+      "epoch": 851.42,
+      "learning_rate": 8.254480286738351e-06,
+      "loss": 0.2059,
+      "step": 131970
+    },
+    {
+      "epoch": 851.48,
+      "learning_rate": 8.250896057347671e-06,
+      "loss": 0.1667,
+      "step": 131980
+    },
+    {
+      "epoch": 851.55,
+      "learning_rate": 8.24731182795699e-06,
+      "loss": 0.2198,
+      "step": 131990
+    },
+    {
+      "epoch": 851.61,
+      "learning_rate": 8.24372759856631e-06,
+      "loss": 0.229,
+      "step": 132000
+    },
+    {
+      "epoch": 851.68,
+      "learning_rate": 8.240143369175627e-06,
+      "loss": 0.211,
+      "step": 132010
+    },
+    {
+      "epoch": 851.74,
+      "learning_rate": 8.236559139784947e-06,
+      "loss": 0.2011,
+      "step": 132020
+    },
+    {
+      "epoch": 851.8,
+      "learning_rate": 8.232974910394265e-06,
+      "loss": 0.2005,
+      "step": 132030
+    },
+    {
+      "epoch": 851.87,
+      "learning_rate": 8.229390681003585e-06,
+      "loss": 0.2051,
+      "step": 132040
+    },
+    {
+      "epoch": 851.93,
+      "learning_rate": 8.225806451612904e-06,
+      "loss": 0.221,
+      "step": 132050
+    },
+    {
+      "epoch": 852.0,
+      "learning_rate": 8.222222222222223e-06,
+      "loss": 0.1875,
+      "step": 132060
+    },
+    {
+      "epoch": 852.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9273878335952759,
+      "eval_runtime": 0.871,
+      "eval_samples_per_second": 230.766,
+      "eval_steps_per_second": 8.037,
+      "step": 132060
+    },
+    {
+      "epoch": 852.06,
+      "learning_rate": 8.218637992831542e-06,
+      "loss": 0.2231,
+      "step": 132070
+    },
+    {
+      "epoch": 852.13,
+      "learning_rate": 8.215053763440862e-06,
+      "loss": 0.2266,
+      "step": 132080
+    },
+    {
+      "epoch": 852.19,
+      "learning_rate": 8.21146953405018e-06,
+      "loss": 0.1932,
+      "step": 132090
+    },
+    {
+      "epoch": 852.26,
+      "learning_rate": 8.2078853046595e-06,
+      "loss": 0.1842,
+      "step": 132100
+    },
+    {
+      "epoch": 852.32,
+      "learning_rate": 8.204301075268818e-06,
+      "loss": 0.194,
+      "step": 132110
+    },
+    {
+      "epoch": 852.39,
+      "learning_rate": 8.200716845878136e-06,
+      "loss": 0.2128,
+      "step": 132120
+    },
+    {
+      "epoch": 852.45,
+      "learning_rate": 8.197132616487456e-06,
+      "loss": 0.1844,
+      "step": 132130
+    },
+    {
+      "epoch": 852.51,
+      "learning_rate": 8.193548387096774e-06,
+      "loss": 0.1758,
+      "step": 132140
+    },
+    {
+      "epoch": 852.58,
+      "learning_rate": 8.189964157706094e-06,
+      "loss": 0.2047,
+      "step": 132150
+    },
+    {
+      "epoch": 852.64,
+      "learning_rate": 8.186379928315412e-06,
+      "loss": 0.2126,
+      "step": 132160
+    },
+    {
+      "epoch": 852.71,
+      "learning_rate": 8.182795698924732e-06,
+      "loss": 0.211,
+      "step": 132170
+    },
+    {
+      "epoch": 852.77,
+      "learning_rate": 8.17921146953405e-06,
+      "loss": 0.1755,
+      "step": 132180
+    },
+    {
+      "epoch": 852.84,
+      "learning_rate": 8.17562724014337e-06,
+      "loss": 0.2053,
+      "step": 132190
+    },
+    {
+      "epoch": 852.9,
+      "learning_rate": 8.172043010752689e-06,
+      "loss": 0.1836,
+      "step": 132200
+    },
+    {
+      "epoch": 852.96,
+      "learning_rate": 8.168458781362008e-06,
+      "loss": 0.2248,
+      "step": 132210
+    },
+    {
+      "epoch": 853.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9296290278434753,
+      "eval_runtime": 0.8805,
+      "eval_samples_per_second": 228.276,
+      "eval_steps_per_second": 7.95,
+      "step": 132215
+    },
+    {
+      "epoch": 853.03,
+      "learning_rate": 8.164874551971327e-06,
+      "loss": 0.2536,
+      "step": 132220
+    },
+    {
+      "epoch": 853.1,
+      "learning_rate": 8.161290322580647e-06,
+      "loss": 0.1937,
+      "step": 132230
+    },
+    {
+      "epoch": 853.16,
+      "learning_rate": 8.157706093189965e-06,
+      "loss": 0.2044,
+      "step": 132240
+    },
+    {
+      "epoch": 853.23,
+      "learning_rate": 8.154121863799285e-06,
+      "loss": 0.1691,
+      "step": 132250
+    },
+    {
+      "epoch": 853.29,
+      "learning_rate": 8.150537634408603e-06,
+      "loss": 0.2373,
+      "step": 132260
+    },
+    {
+      "epoch": 853.35,
+      "learning_rate": 8.146953405017923e-06,
+      "loss": 0.2282,
+      "step": 132270
+    },
+    {
+      "epoch": 853.42,
+      "learning_rate": 8.143369175627241e-06,
+      "loss": 0.1634,
+      "step": 132280
+    },
+    {
+      "epoch": 853.48,
+      "learning_rate": 8.139784946236559e-06,
+      "loss": 0.1942,
+      "step": 132290
+    },
+    {
+      "epoch": 853.55,
+      "learning_rate": 8.136200716845877e-06,
+      "loss": 0.2322,
+      "step": 132300
+    },
+    {
+      "epoch": 853.61,
+      "learning_rate": 8.132616487455197e-06,
+      "loss": 0.222,
+      "step": 132310
+    },
+    {
+      "epoch": 853.68,
+      "learning_rate": 8.129032258064515e-06,
+      "loss": 0.239,
+      "step": 132320
+    },
+    {
+      "epoch": 853.74,
+      "learning_rate": 8.125448028673835e-06,
+      "loss": 0.1704,
+      "step": 132330
+    },
+    {
+      "epoch": 853.8,
+      "learning_rate": 8.121863799283154e-06,
+      "loss": 0.1936,
+      "step": 132340
+    },
+    {
+      "epoch": 853.87,
+      "learning_rate": 8.118279569892473e-06,
+      "loss": 0.1963,
+      "step": 132350
+    },
+    {
+      "epoch": 853.93,
+      "learning_rate": 8.114695340501792e-06,
+      "loss": 0.2161,
+      "step": 132360
+    },
+    {
+      "epoch": 854.0,
+      "learning_rate": 8.111111111111112e-06,
+      "loss": 0.1615,
+      "step": 132370
+    },
+    {
+      "epoch": 854.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9393577575683594,
+      "eval_runtime": 0.8588,
+      "eval_samples_per_second": 234.04,
+      "eval_steps_per_second": 8.151,
+      "step": 132370
+    },
+    {
+      "epoch": 854.06,
+      "learning_rate": 8.10752688172043e-06,
+      "loss": 0.2397,
+      "step": 132380
+    },
+    {
+      "epoch": 854.13,
+      "learning_rate": 8.10394265232975e-06,
+      "loss": 0.2202,
+      "step": 132390
+    },
+    {
+      "epoch": 854.19,
+      "learning_rate": 8.100358422939068e-06,
+      "loss": 0.2038,
+      "step": 132400
+    },
+    {
+      "epoch": 854.26,
+      "learning_rate": 8.096774193548388e-06,
+      "loss": 0.1722,
+      "step": 132410
+    },
+    {
+      "epoch": 854.32,
+      "learning_rate": 8.093189964157706e-06,
+      "loss": 0.2235,
+      "step": 132420
+    },
+    {
+      "epoch": 854.39,
+      "learning_rate": 8.089605734767026e-06,
+      "loss": 0.1652,
+      "step": 132430
+    },
+    {
+      "epoch": 854.45,
+      "learning_rate": 8.086021505376344e-06,
+      "loss": 0.1389,
+      "step": 132440
+    },
+    {
+      "epoch": 854.51,
+      "learning_rate": 8.082437275985664e-06,
+      "loss": 0.1628,
+      "step": 132450
+    },
+    {
+      "epoch": 854.58,
+      "learning_rate": 8.078853046594982e-06,
+      "loss": 0.1825,
+      "step": 132460
+    },
+    {
+      "epoch": 854.64,
+      "learning_rate": 8.0752688172043e-06,
+      "loss": 0.2177,
+      "step": 132470
+    },
+    {
+      "epoch": 854.71,
+      "learning_rate": 8.07168458781362e-06,
+      "loss": 0.1789,
+      "step": 132480
+    },
+    {
+      "epoch": 854.77,
+      "learning_rate": 8.068100358422938e-06,
+      "loss": 0.2057,
+      "step": 132490
+    },
+    {
+      "epoch": 854.84,
+      "learning_rate": 8.064516129032258e-06,
+      "loss": 0.2219,
+      "step": 132500
+    },
+    {
+      "epoch": 854.9,
+      "learning_rate": 8.060931899641577e-06,
+      "loss": 0.2164,
+      "step": 132510
+    },
+    {
+      "epoch": 854.96,
+      "learning_rate": 8.057347670250897e-06,
+      "loss": 0.2181,
+      "step": 132520
+    },
+    {
+      "epoch": 855.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9321929812431335,
+      "eval_runtime": 0.8888,
+      "eval_samples_per_second": 226.139,
+      "eval_steps_per_second": 7.875,
+      "step": 132525
+    },
+    {
+      "epoch": 855.03,
+      "learning_rate": 8.053763440860215e-06,
+      "loss": 0.2078,
+      "step": 132530
+    },
+    {
+      "epoch": 855.1,
+      "learning_rate": 8.050179211469535e-06,
+      "loss": 0.2565,
+      "step": 132540
+    },
+    {
+      "epoch": 855.16,
+      "learning_rate": 8.046594982078853e-06,
+      "loss": 0.2182,
+      "step": 132550
+    },
+    {
+      "epoch": 855.23,
+      "learning_rate": 8.043010752688173e-06,
+      "loss": 0.1951,
+      "step": 132560
+    },
+    {
+      "epoch": 855.29,
+      "learning_rate": 8.039426523297491e-06,
+      "loss": 0.2285,
+      "step": 132570
+    },
+    {
+      "epoch": 855.35,
+      "learning_rate": 8.03584229390681e-06,
+      "loss": 0.1828,
+      "step": 132580
+    },
+    {
+      "epoch": 855.42,
+      "learning_rate": 8.032258064516129e-06,
+      "loss": 0.2098,
+      "step": 132590
+    },
+    {
+      "epoch": 855.48,
+      "learning_rate": 8.028673835125449e-06,
+      "loss": 0.1634,
+      "step": 132600
+    },
+    {
+      "epoch": 855.55,
+      "learning_rate": 8.025089605734767e-06,
+      "loss": 0.2524,
+      "step": 132610
+    },
+    {
+      "epoch": 855.61,
+      "learning_rate": 8.021505376344087e-06,
+      "loss": 0.2106,
+      "step": 132620
+    },
+    {
+      "epoch": 855.68,
+      "learning_rate": 8.017921146953405e-06,
+      "loss": 0.2032,
+      "step": 132630
+    },
+    {
+      "epoch": 855.74,
+      "learning_rate": 8.014336917562725e-06,
+      "loss": 0.1515,
+      "step": 132640
+    },
+    {
+      "epoch": 855.8,
+      "learning_rate": 8.010752688172043e-06,
+      "loss": 0.1827,
+      "step": 132650
+    },
+    {
+      "epoch": 855.87,
+      "learning_rate": 8.007168458781363e-06,
+      "loss": 0.2109,
+      "step": 132660
+    },
+    {
+      "epoch": 855.93,
+      "learning_rate": 8.003584229390681e-06,
+      "loss": 0.1936,
+      "step": 132670
+    },
+    {
+      "epoch": 856.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1831,
+      "step": 132680
+    },
+    {
+      "epoch": 856.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9493621587753296,
+      "eval_runtime": 0.8637,
+      "eval_samples_per_second": 232.728,
+      "eval_steps_per_second": 8.105,
+      "step": 132680
+    },
+    {
+      "epoch": 856.06,
+      "learning_rate": 7.99641577060932e-06,
+      "loss": 0.1833,
+      "step": 132690
+    },
+    {
+      "epoch": 856.13,
+      "learning_rate": 7.992831541218638e-06,
+      "loss": 0.2231,
+      "step": 132700
+    },
+    {
+      "epoch": 856.19,
+      "learning_rate": 7.989247311827958e-06,
+      "loss": 0.178,
+      "step": 132710
+    },
+    {
+      "epoch": 856.26,
+      "learning_rate": 7.985663082437276e-06,
+      "loss": 0.2154,
+      "step": 132720
+    },
+    {
+      "epoch": 856.32,
+      "learning_rate": 7.982078853046596e-06,
+      "loss": 0.1776,
+      "step": 132730
+    },
+    {
+      "epoch": 856.39,
+      "learning_rate": 7.978494623655914e-06,
+      "loss": 0.2143,
+      "step": 132740
+    },
+    {
+      "epoch": 856.45,
+      "learning_rate": 7.974910394265234e-06,
+      "loss": 0.1636,
+      "step": 132750
+    },
+    {
+      "epoch": 856.51,
+      "learning_rate": 7.971326164874552e-06,
+      "loss": 0.2209,
+      "step": 132760
+    },
+    {
+      "epoch": 856.58,
+      "learning_rate": 7.967741935483872e-06,
+      "loss": 0.2184,
+      "step": 132770
+    },
+    {
+      "epoch": 856.64,
+      "learning_rate": 7.96415770609319e-06,
+      "loss": 0.2341,
+      "step": 132780
+    },
+    {
+      "epoch": 856.71,
+      "learning_rate": 7.96057347670251e-06,
+      "loss": 0.1899,
+      "step": 132790
+    },
+    {
+      "epoch": 856.77,
+      "learning_rate": 7.956989247311828e-06,
+      "loss": 0.1834,
+      "step": 132800
+    },
+    {
+      "epoch": 856.84,
+      "learning_rate": 7.953405017921148e-06,
+      "loss": 0.1992,
+      "step": 132810
+    },
+    {
+      "epoch": 856.9,
+      "learning_rate": 7.949820788530466e-06,
+      "loss": 0.2669,
+      "step": 132820
+    },
+    {
+      "epoch": 856.96,
+      "learning_rate": 7.946236559139786e-06,
+      "loss": 0.1739,
+      "step": 132830
+    },
+    {
+      "epoch": 857.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9396321177482605,
+      "eval_runtime": 0.8783,
+      "eval_samples_per_second": 228.843,
+      "eval_steps_per_second": 7.97,
+      "step": 132835
+    },
+    {
+      "epoch": 857.03,
+      "learning_rate": 7.942652329749104e-06,
+      "loss": 0.2506,
+      "step": 132840
+    },
+    {
+      "epoch": 857.1,
+      "learning_rate": 7.939068100358424e-06,
+      "loss": 0.182,
+      "step": 132850
+    },
+    {
+      "epoch": 857.16,
+      "learning_rate": 7.935483870967743e-06,
+      "loss": 0.2062,
+      "step": 132860
+    },
+    {
+      "epoch": 857.23,
+      "learning_rate": 7.931899641577062e-06,
+      "loss": 0.1865,
+      "step": 132870
+    },
+    {
+      "epoch": 857.29,
+      "learning_rate": 7.92831541218638e-06,
+      "loss": 0.201,
+      "step": 132880
+    },
+    {
+      "epoch": 857.35,
+      "learning_rate": 7.9247311827957e-06,
+      "loss": 0.1536,
+      "step": 132890
+    },
+    {
+      "epoch": 857.42,
+      "learning_rate": 7.921146953405019e-06,
+      "loss": 0.2051,
+      "step": 132900
+    },
+    {
+      "epoch": 857.48,
+      "learning_rate": 7.917562724014339e-06,
+      "loss": 0.2385,
+      "step": 132910
+    },
+    {
+      "epoch": 857.55,
+      "learning_rate": 7.913978494623657e-06,
+      "loss": 0.1839,
+      "step": 132920
+    },
+    {
+      "epoch": 857.61,
+      "learning_rate": 7.910394265232975e-06,
+      "loss": 0.1866,
+      "step": 132930
+    },
+    {
+      "epoch": 857.68,
+      "learning_rate": 7.906810035842293e-06,
+      "loss": 0.2248,
+      "step": 132940
+    },
+    {
+      "epoch": 857.74,
+      "learning_rate": 7.903225806451613e-06,
+      "loss": 0.1836,
+      "step": 132950
+    },
+    {
+      "epoch": 857.8,
+      "learning_rate": 7.899641577060931e-06,
+      "loss": 0.1993,
+      "step": 132960
+    },
+    {
+      "epoch": 857.87,
+      "learning_rate": 7.896057347670251e-06,
+      "loss": 0.1831,
+      "step": 132970
+    },
+    {
+      "epoch": 857.93,
+      "learning_rate": 7.89247311827957e-06,
+      "loss": 0.2288,
+      "step": 132980
+    },
+    {
+      "epoch": 858.0,
+      "learning_rate": 7.88888888888889e-06,
+      "loss": 0.1696,
+      "step": 132990
+    },
+    {
+      "epoch": 858.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.8989540338516235,
+      "eval_runtime": 0.8963,
+      "eval_samples_per_second": 224.244,
+      "eval_steps_per_second": 7.809,
+      "step": 132990
+    },
+    {
+      "epoch": 858.06,
+      "learning_rate": 7.885304659498208e-06,
+      "loss": 0.2585,
+      "step": 133000
+    },
+    {
+      "epoch": 858.13,
+      "learning_rate": 7.881720430107528e-06,
+      "loss": 0.2229,
+      "step": 133010
+    },
+    {
+      "epoch": 858.19,
+      "learning_rate": 7.878136200716846e-06,
+      "loss": 0.1931,
+      "step": 133020
+    },
+    {
+      "epoch": 858.26,
+      "learning_rate": 7.874551971326166e-06,
+      "loss": 0.1969,
+      "step": 133030
+    },
+    {
+      "epoch": 858.32,
+      "learning_rate": 7.870967741935484e-06,
+      "loss": 0.2126,
+      "step": 133040
+    },
+    {
+      "epoch": 858.39,
+      "learning_rate": 7.867383512544802e-06,
+      "loss": 0.1971,
+      "step": 133050
+    },
+    {
+      "epoch": 858.45,
+      "learning_rate": 7.863799283154122e-06,
+      "loss": 0.2055,
+      "step": 133060
+    },
+    {
+      "epoch": 858.51,
+      "learning_rate": 7.86021505376344e-06,
+      "loss": 0.292,
+      "step": 133070
+    },
+    {
+      "epoch": 858.58,
+      "learning_rate": 7.85663082437276e-06,
+      "loss": 0.1779,
+      "step": 133080
+    },
+    {
+      "epoch": 858.64,
+      "learning_rate": 7.853046594982078e-06,
+      "loss": 0.2104,
+      "step": 133090
+    },
+    {
+      "epoch": 858.71,
+      "learning_rate": 7.849462365591398e-06,
+      "loss": 0.2066,
+      "step": 133100
+    },
+    {
+      "epoch": 858.77,
+      "learning_rate": 7.845878136200716e-06,
+      "loss": 0.1687,
+      "step": 133110
+    },
+    {
+      "epoch": 858.84,
+      "learning_rate": 7.842293906810036e-06,
+      "loss": 0.2492,
+      "step": 133120
+    },
+    {
+      "epoch": 858.9,
+      "learning_rate": 7.838709677419354e-06,
+      "loss": 0.176,
+      "step": 133130
+    },
+    {
+      "epoch": 858.96,
+      "learning_rate": 7.835125448028674e-06,
+      "loss": 0.2108,
+      "step": 133140
+    },
+    {
+      "epoch": 859.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9563379287719727,
+      "eval_runtime": 0.8603,
+      "eval_samples_per_second": 233.647,
+      "eval_steps_per_second": 8.137,
+      "step": 133145
+    },
+    {
+      "epoch": 859.03,
+      "learning_rate": 7.831541218637993e-06,
+      "loss": 0.2273,
+      "step": 133150
+    },
+    {
+      "epoch": 859.1,
+      "learning_rate": 7.827956989247312e-06,
+      "loss": 0.1794,
+      "step": 133160
+    },
+    {
+      "epoch": 859.16,
+      "learning_rate": 7.82437275985663e-06,
+      "loss": 0.2142,
+      "step": 133170
+    },
+    {
+      "epoch": 859.23,
+      "learning_rate": 7.82078853046595e-06,
+      "loss": 0.2464,
+      "step": 133180
+    },
+    {
+      "epoch": 859.29,
+      "learning_rate": 7.817204301075269e-06,
+      "loss": 0.1987,
+      "step": 133190
+    },
+    {
+      "epoch": 859.35,
+      "learning_rate": 7.813620071684589e-06,
+      "loss": 0.2299,
+      "step": 133200
+    },
+    {
+      "epoch": 859.42,
+      "learning_rate": 7.810035842293907e-06,
+      "loss": 0.1951,
+      "step": 133210
+    },
+    {
+      "epoch": 859.48,
+      "learning_rate": 7.806451612903227e-06,
+      "loss": 0.2004,
+      "step": 133220
+    },
+    {
+      "epoch": 859.55,
+      "learning_rate": 7.802867383512545e-06,
+      "loss": 0.1884,
+      "step": 133230
+    },
+    {
+      "epoch": 859.61,
+      "learning_rate": 7.799283154121865e-06,
+      "loss": 0.2204,
+      "step": 133240
+    },
+    {
+      "epoch": 859.68,
+      "learning_rate": 7.795698924731183e-06,
+      "loss": 0.2427,
+      "step": 133250
+    },
+    {
+      "epoch": 859.74,
+      "learning_rate": 7.792114695340503e-06,
+      "loss": 0.2268,
+      "step": 133260
+    },
+    {
+      "epoch": 859.8,
+      "learning_rate": 7.788530465949821e-06,
+      "loss": 0.2123,
+      "step": 133270
+    },
+    {
+      "epoch": 859.87,
+      "learning_rate": 7.784946236559141e-06,
+      "loss": 0.2644,
+      "step": 133280
+    },
+    {
+      "epoch": 859.93,
+      "learning_rate": 7.78136200716846e-06,
+      "loss": 0.2089,
+      "step": 133290
+    },
+    {
+      "epoch": 860.0,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 0.1937,
+      "step": 133300
+    },
+    {
+      "epoch": 860.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.880933940410614,
+      "eval_runtime": 0.9087,
+      "eval_samples_per_second": 221.199,
+      "eval_steps_per_second": 7.703,
+      "step": 133300
+    },
+    {
+      "epoch": 860.06,
+      "learning_rate": 7.774193548387097e-06,
+      "loss": 0.2587,
+      "step": 133310
+    },
+    {
+      "epoch": 860.13,
+      "learning_rate": 7.770609318996416e-06,
+      "loss": 0.1776,
+      "step": 133320
+    },
+    {
+      "epoch": 860.19,
+      "learning_rate": 7.767025089605735e-06,
+      "loss": 0.1803,
+      "step": 133330
+    },
+    {
+      "epoch": 860.26,
+      "learning_rate": 7.763440860215054e-06,
+      "loss": 0.1822,
+      "step": 133340
+    },
+    {
+      "epoch": 860.32,
+      "learning_rate": 7.759856630824374e-06,
+      "loss": 0.22,
+      "step": 133350
+    },
+    {
+      "epoch": 860.39,
+      "learning_rate": 7.756272401433692e-06,
+      "loss": 0.2304,
+      "step": 133360
+    },
+    {
+      "epoch": 860.45,
+      "learning_rate": 7.752688172043012e-06,
+      "loss": 0.1738,
+      "step": 133370
+    },
+    {
+      "epoch": 860.51,
+      "learning_rate": 7.74910394265233e-06,
+      "loss": 0.2112,
+      "step": 133380
+    },
+    {
+      "epoch": 860.58,
+      "learning_rate": 7.74551971326165e-06,
+      "loss": 0.216,
+      "step": 133390
+    },
+    {
+      "epoch": 860.64,
+      "learning_rate": 7.741935483870968e-06,
+      "loss": 0.2551,
+      "step": 133400
+    },
+    {
+      "epoch": 860.71,
+      "learning_rate": 7.738351254480288e-06,
+      "loss": 0.1765,
+      "step": 133410
+    },
+    {
+      "epoch": 860.77,
+      "learning_rate": 7.734767025089606e-06,
+      "loss": 0.1677,
+      "step": 133420
+    },
+    {
+      "epoch": 860.84,
+      "learning_rate": 7.731182795698926e-06,
+      "loss": 0.1867,
+      "step": 133430
+    },
+    {
+      "epoch": 860.9,
+      "learning_rate": 7.727598566308244e-06,
+      "loss": 0.1942,
+      "step": 133440
+    },
+    {
+      "epoch": 860.96,
+      "learning_rate": 7.724014336917564e-06,
+      "loss": 0.2123,
+      "step": 133450
+    },
+    {
+      "epoch": 861.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9196632504463196,
+      "eval_runtime": 0.8892,
+      "eval_samples_per_second": 226.041,
+      "eval_steps_per_second": 7.872,
+      "step": 133455
+    },
+    {
+      "epoch": 861.03,
+      "learning_rate": 7.720430107526882e-06,
+      "loss": 0.2032,
+      "step": 133460
+    },
+    {
+      "epoch": 861.1,
+      "learning_rate": 7.716845878136202e-06,
+      "loss": 0.1766,
+      "step": 133470
+    },
+    {
+      "epoch": 861.16,
+      "learning_rate": 7.71326164874552e-06,
+      "loss": 0.2169,
+      "step": 133480
+    },
+    {
+      "epoch": 861.23,
+      "learning_rate": 7.70967741935484e-06,
+      "loss": 0.1841,
+      "step": 133490
+    },
+    {
+      "epoch": 861.29,
+      "learning_rate": 7.706093189964159e-06,
+      "loss": 0.213,
+      "step": 133500
+    },
+    {
+      "epoch": 861.35,
+      "learning_rate": 7.702508960573478e-06,
+      "loss": 0.1788,
+      "step": 133510
+    },
+    {
+      "epoch": 861.42,
+      "learning_rate": 7.698924731182797e-06,
+      "loss": 0.2122,
+      "step": 133520
+    },
+    {
+      "epoch": 861.48,
+      "learning_rate": 7.695340501792117e-06,
+      "loss": 0.1266,
+      "step": 133530
+    },
+    {
+      "epoch": 861.55,
+      "learning_rate": 7.691756272401435e-06,
+      "loss": 0.2094,
+      "step": 133540
+    },
+    {
+      "epoch": 861.61,
+      "learning_rate": 7.688172043010753e-06,
+      "loss": 0.1964,
+      "step": 133550
+    },
+    {
+      "epoch": 861.68,
+      "learning_rate": 7.684587813620073e-06,
+      "loss": 0.2394,
+      "step": 133560
+    },
+    {
+      "epoch": 861.74,
+      "learning_rate": 7.681003584229391e-06,
+      "loss": 0.2082,
+      "step": 133570
+    },
+    {
+      "epoch": 861.8,
+      "learning_rate": 7.67741935483871e-06,
+      "loss": 0.2296,
+      "step": 133580
+    },
+    {
+      "epoch": 861.87,
+      "learning_rate": 7.673835125448029e-06,
+      "loss": 0.1815,
+      "step": 133590
+    },
+    {
+      "epoch": 861.93,
+      "learning_rate": 7.670250896057347e-06,
+      "loss": 0.1968,
+      "step": 133600
+    },
+    {
+      "epoch": 862.0,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 0.2056,
+      "step": 133610
+    },
+    {
+      "epoch": 862.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9398884177207947,
+      "eval_runtime": 0.8854,
+      "eval_samples_per_second": 227.011,
+      "eval_steps_per_second": 7.906,
+      "step": 133610
+    },
+    {
+      "epoch": 862.06,
+      "learning_rate": 7.663082437275985e-06,
+      "loss": 0.2328,
+      "step": 133620
+    },
+    {
+      "epoch": 862.13,
+      "learning_rate": 7.659498207885304e-06,
+      "loss": 0.2192,
+      "step": 133630
+    },
+    {
+      "epoch": 862.19,
+      "learning_rate": 7.655913978494624e-06,
+      "loss": 0.217,
+      "step": 133640
+    },
+    {
+      "epoch": 862.26,
+      "learning_rate": 7.652329749103942e-06,
+      "loss": 0.2257,
+      "step": 133650
+    },
+    {
+      "epoch": 862.32,
+      "learning_rate": 7.648745519713262e-06,
+      "loss": 0.1951,
+      "step": 133660
+    },
+    {
+      "epoch": 862.39,
+      "learning_rate": 7.64516129032258e-06,
+      "loss": 0.185,
+      "step": 133670
+    },
+    {
+      "epoch": 862.45,
+      "learning_rate": 7.6415770609319e-06,
+      "loss": 0.2192,
+      "step": 133680
+    },
+    {
+      "epoch": 862.51,
+      "learning_rate": 7.637992831541218e-06,
+      "loss": 0.1816,
+      "step": 133690
+    },
+    {
+      "epoch": 862.58,
+      "learning_rate": 7.634408602150538e-06,
+      "loss": 0.1861,
+      "step": 133700
+    },
+    {
+      "epoch": 862.64,
+      "learning_rate": 7.630824372759856e-06,
+      "loss": 0.2244,
+      "step": 133710
+    },
+    {
+      "epoch": 862.71,
+      "learning_rate": 7.627240143369176e-06,
+      "loss": 0.267,
+      "step": 133720
+    },
+    {
+      "epoch": 862.77,
+      "learning_rate": 7.623655913978495e-06,
+      "loss": 0.1947,
+      "step": 133730
+    },
+    {
+      "epoch": 862.84,
+      "learning_rate": 7.620071684587814e-06,
+      "loss": 0.2087,
+      "step": 133740
+    },
+    {
+      "epoch": 862.9,
+      "learning_rate": 7.616487455197132e-06,
+      "loss": 0.1954,
+      "step": 133750
+    },
+    {
+      "epoch": 862.96,
+      "learning_rate": 7.612903225806451e-06,
+      "loss": 0.1919,
+      "step": 133760
+    },
+    {
+      "epoch": 863.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9599234461784363,
+      "eval_runtime": 0.897,
+      "eval_samples_per_second": 224.088,
+      "eval_steps_per_second": 7.804,
+      "step": 133765
+    },
+    {
+      "epoch": 863.03,
+      "learning_rate": 7.60931899641577e-06,
+      "loss": 0.2069,
+      "step": 133770
+    },
+    {
+      "epoch": 863.1,
+      "learning_rate": 7.6057347670250894e-06,
+      "loss": 0.1968,
+      "step": 133780
+    },
+    {
+      "epoch": 863.16,
+      "learning_rate": 7.6021505376344085e-06,
+      "loss": 0.244,
+      "step": 133790
+    },
+    {
+      "epoch": 863.23,
+      "learning_rate": 7.5985663082437275e-06,
+      "loss": 0.1966,
+      "step": 133800
+    },
+    {
+      "epoch": 863.29,
+      "learning_rate": 7.594982078853047e-06,
+      "loss": 0.2035,
+      "step": 133810
+    },
+    {
+      "epoch": 863.35,
+      "learning_rate": 7.591397849462366e-06,
+      "loss": 0.2211,
+      "step": 133820
+    },
+    {
+      "epoch": 863.42,
+      "learning_rate": 7.587813620071685e-06,
+      "loss": 0.1999,
+      "step": 133830
+    },
+    {
+      "epoch": 863.48,
+      "learning_rate": 7.584229390681004e-06,
+      "loss": 0.1874,
+      "step": 133840
+    },
+    {
+      "epoch": 863.55,
+      "learning_rate": 7.580645161290323e-06,
+      "loss": 0.241,
+      "step": 133850
+    },
+    {
+      "epoch": 863.61,
+      "learning_rate": 7.577060931899642e-06,
+      "loss": 0.2215,
+      "step": 133860
+    },
+    {
+      "epoch": 863.68,
+      "learning_rate": 7.573476702508961e-06,
+      "loss": 0.1624,
+      "step": 133870
+    },
+    {
+      "epoch": 863.74,
+      "learning_rate": 7.56989247311828e-06,
+      "loss": 0.2367,
+      "step": 133880
+    },
+    {
+      "epoch": 863.8,
+      "learning_rate": 7.566308243727599e-06,
+      "loss": 0.196,
+      "step": 133890
+    },
+    {
+      "epoch": 863.87,
+      "learning_rate": 7.562724014336918e-06,
+      "loss": 0.221,
+      "step": 133900
+    },
+    {
+      "epoch": 863.93,
+      "learning_rate": 7.559139784946237e-06,
+      "loss": 0.2313,
+      "step": 133910
+    },
+    {
+      "epoch": 864.0,
+      "learning_rate": 7.555555555555556e-06,
+      "loss": 0.1786,
+      "step": 133920
+    },
+    {
+      "epoch": 864.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.921306848526001,
+      "eval_runtime": 0.8768,
+      "eval_samples_per_second": 229.251,
+      "eval_steps_per_second": 7.984,
+      "step": 133920
+    },
+    {
+      "epoch": 864.06,
+      "learning_rate": 7.551971326164875e-06,
+      "loss": 0.1798,
+      "step": 133930
+    },
+    {
+      "epoch": 864.13,
+      "learning_rate": 7.548387096774194e-06,
+      "loss": 0.2205,
+      "step": 133940
+    },
+    {
+      "epoch": 864.19,
+      "learning_rate": 7.544802867383513e-06,
+      "loss": 0.2084,
+      "step": 133950
+    },
+    {
+      "epoch": 864.26,
+      "learning_rate": 7.541218637992832e-06,
+      "loss": 0.2378,
+      "step": 133960
+    },
+    {
+      "epoch": 864.32,
+      "learning_rate": 7.537634408602151e-06,
+      "loss": 0.2167,
+      "step": 133970
+    },
+    {
+      "epoch": 864.39,
+      "learning_rate": 7.5340501792114705e-06,
+      "loss": 0.2051,
+      "step": 133980
+    },
+    {
+      "epoch": 864.45,
+      "learning_rate": 7.5304659498207895e-06,
+      "loss": 0.2099,
+      "step": 133990
+    },
+    {
+      "epoch": 864.51,
+      "learning_rate": 7.526881720430108e-06,
+      "loss": 0.1858,
+      "step": 134000
+    },
+    {
+      "epoch": 864.58,
+      "learning_rate": 7.523297491039427e-06,
+      "loss": 0.1794,
+      "step": 134010
+    },
+    {
+      "epoch": 864.64,
+      "learning_rate": 7.519713261648746e-06,
+      "loss": 0.2235,
+      "step": 134020
+    },
+    {
+      "epoch": 864.71,
+      "learning_rate": 7.516129032258065e-06,
+      "loss": 0.2011,
+      "step": 134030
+    },
+    {
+      "epoch": 864.77,
+      "learning_rate": 7.512544802867384e-06,
+      "loss": 0.1929,
+      "step": 134040
+    },
+    {
+      "epoch": 864.84,
+      "learning_rate": 7.508960573476703e-06,
+      "loss": 0.1931,
+      "step": 134050
+    },
+    {
+      "epoch": 864.9,
+      "learning_rate": 7.505376344086022e-06,
+      "loss": 0.2047,
+      "step": 134060
+    },
+    {
+      "epoch": 864.96,
+      "learning_rate": 7.501792114695341e-06,
+      "loss": 0.2029,
+      "step": 134070
+    },
+    {
+      "epoch": 865.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9407548904418945,
+      "eval_runtime": 0.8675,
+      "eval_samples_per_second": 231.707,
+      "eval_steps_per_second": 8.069,
+      "step": 134075
+    },
+    {
+      "epoch": 865.03,
+      "learning_rate": 7.49820788530466e-06,
+      "loss": 0.2277,
+      "step": 134080
+    },
+    {
+      "epoch": 865.1,
+      "learning_rate": 7.494623655913979e-06,
+      "loss": 0.2473,
+      "step": 134090
+    },
+    {
+      "epoch": 865.16,
+      "learning_rate": 7.491039426523298e-06,
+      "loss": 0.2386,
+      "step": 134100
+    },
+    {
+      "epoch": 865.23,
+      "learning_rate": 7.487455197132617e-06,
+      "loss": 0.189,
+      "step": 134110
+    },
+    {
+      "epoch": 865.29,
+      "learning_rate": 7.483870967741936e-06,
+      "loss": 0.1939,
+      "step": 134120
+    },
+    {
+      "epoch": 865.35,
+      "learning_rate": 7.480286738351255e-06,
+      "loss": 0.2083,
+      "step": 134130
+    },
+    {
+      "epoch": 865.42,
+      "learning_rate": 7.4767025089605745e-06,
+      "loss": 0.2172,
+      "step": 134140
+    },
+    {
+      "epoch": 865.48,
+      "learning_rate": 7.4731182795698935e-06,
+      "loss": 0.1764,
+      "step": 134150
+    },
+    {
+      "epoch": 865.55,
+      "learning_rate": 7.4695340501792126e-06,
+      "loss": 0.2373,
+      "step": 134160
+    },
+    {
+      "epoch": 865.61,
+      "learning_rate": 7.465949820788532e-06,
+      "loss": 0.1886,
+      "step": 134170
+    },
+    {
+      "epoch": 865.68,
+      "learning_rate": 7.462365591397851e-06,
+      "loss": 0.2008,
+      "step": 134180
+    },
+    {
+      "epoch": 865.74,
+      "learning_rate": 7.45878136200717e-06,
+      "loss": 0.1705,
+      "step": 134190
+    },
+    {
+      "epoch": 865.8,
+      "learning_rate": 7.455197132616489e-06,
+      "loss": 0.1869,
+      "step": 134200
+    },
+    {
+      "epoch": 865.87,
+      "learning_rate": 7.451612903225806e-06,
+      "loss": 0.1922,
+      "step": 134210
+    },
+    {
+      "epoch": 865.93,
+      "learning_rate": 7.448028673835125e-06,
+      "loss": 0.2338,
+      "step": 134220
+    },
+    {
+      "epoch": 866.0,
+      "learning_rate": 7.444444444444444e-06,
+      "loss": 0.1666,
+      "step": 134230
+    },
+    {
+      "epoch": 866.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9596779346466064,
+      "eval_runtime": 0.8648,
+      "eval_samples_per_second": 232.416,
+      "eval_steps_per_second": 8.094,
+      "step": 134230
+    },
+    {
+      "epoch": 866.06,
+      "learning_rate": 7.440860215053763e-06,
+      "loss": 0.1699,
+      "step": 134240
+    },
+    {
+      "epoch": 866.13,
+      "learning_rate": 7.437275985663082e-06,
+      "loss": 0.2281,
+      "step": 134250
+    },
+    {
+      "epoch": 866.19,
+      "learning_rate": 7.433691756272401e-06,
+      "loss": 0.1801,
+      "step": 134260
+    },
+    {
+      "epoch": 866.26,
+      "learning_rate": 7.4301075268817204e-06,
+      "loss": 0.2253,
+      "step": 134270
+    },
+    {
+      "epoch": 866.32,
+      "learning_rate": 7.4265232974910395e-06,
+      "loss": 0.198,
+      "step": 134280
+    },
+    {
+      "epoch": 866.39,
+      "learning_rate": 7.4229390681003585e-06,
+      "loss": 0.2083,
+      "step": 134290
+    },
+    {
+      "epoch": 866.45,
+      "learning_rate": 7.419354838709678e-06,
+      "loss": 0.2095,
+      "step": 134300
+    },
+    {
+      "epoch": 866.51,
+      "learning_rate": 7.415770609318997e-06,
+      "loss": 0.1605,
+      "step": 134310
+    },
+    {
+      "epoch": 866.58,
+      "learning_rate": 7.412186379928316e-06,
+      "loss": 0.203,
+      "step": 134320
+    },
+    {
+      "epoch": 866.64,
+      "learning_rate": 7.408602150537635e-06,
+      "loss": 0.1903,
+      "step": 134330
+    },
+    {
+      "epoch": 866.71,
+      "learning_rate": 7.405017921146953e-06,
+      "loss": 0.2114,
+      "step": 134340
+    },
+    {
+      "epoch": 866.77,
+      "learning_rate": 7.401433691756272e-06,
+      "loss": 0.1653,
+      "step": 134350
+    },
+    {
+      "epoch": 866.84,
+      "learning_rate": 7.397849462365591e-06,
+      "loss": 0.1491,
+      "step": 134360
+    },
+    {
+      "epoch": 866.9,
+      "learning_rate": 7.39426523297491e-06,
+      "loss": 0.1947,
+      "step": 134370
+    },
+    {
+      "epoch": 866.96,
+      "learning_rate": 7.390681003584229e-06,
+      "loss": 0.2306,
+      "step": 134380
+    },
+    {
+      "epoch": 867.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9477769136428833,
+      "eval_runtime": 0.8753,
+      "eval_samples_per_second": 229.638,
+      "eval_steps_per_second": 7.997,
+      "step": 134385
+    },
+    {
+      "epoch": 867.03,
+      "learning_rate": 7.387096774193548e-06,
+      "loss": 0.2226,
+      "step": 134390
+    },
+    {
+      "epoch": 867.1,
+      "learning_rate": 7.383512544802867e-06,
+      "loss": 0.1913,
+      "step": 134400
+    },
+    {
+      "epoch": 867.16,
+      "learning_rate": 7.379928315412186e-06,
+      "loss": 0.1956,
+      "step": 134410
+    },
+    {
+      "epoch": 867.23,
+      "learning_rate": 7.376344086021505e-06,
+      "loss": 0.2561,
+      "step": 134420
+    },
+    {
+      "epoch": 867.29,
+      "learning_rate": 7.372759856630824e-06,
+      "loss": 0.1887,
+      "step": 134430
+    },
+    {
+      "epoch": 867.35,
+      "learning_rate": 7.3691756272401435e-06,
+      "loss": 0.2009,
+      "step": 134440
+    },
+    {
+      "epoch": 867.42,
+      "learning_rate": 7.3655913978494625e-06,
+      "loss": 0.2319,
+      "step": 134450
+    },
+    {
+      "epoch": 867.48,
+      "learning_rate": 7.3620071684587816e-06,
+      "loss": 0.231,
+      "step": 134460
+    },
+    {
+      "epoch": 867.55,
+      "learning_rate": 7.358422939068101e-06,
+      "loss": 0.1976,
+      "step": 134470
+    },
+    {
+      "epoch": 867.61,
+      "learning_rate": 7.35483870967742e-06,
+      "loss": 0.1876,
+      "step": 134480
+    },
+    {
+      "epoch": 867.68,
+      "learning_rate": 7.351254480286739e-06,
+      "loss": 0.1641,
+      "step": 134490
+    },
+    {
+      "epoch": 867.74,
+      "learning_rate": 7.347670250896058e-06,
+      "loss": 0.2132,
+      "step": 134500
+    },
+    {
+      "epoch": 867.8,
+      "learning_rate": 7.344086021505377e-06,
+      "loss": 0.2321,
+      "step": 134510
+    },
+    {
+      "epoch": 867.87,
+      "learning_rate": 7.340501792114696e-06,
+      "loss": 0.2038,
+      "step": 134520
+    },
+    {
+      "epoch": 867.93,
+      "learning_rate": 7.336917562724015e-06,
+      "loss": 0.1782,
+      "step": 134530
+    },
+    {
+      "epoch": 868.0,
+      "learning_rate": 7.333333333333334e-06,
+      "loss": 0.2263,
+      "step": 134540
+    },
+    {
+      "epoch": 868.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.939186155796051,
+      "eval_runtime": 0.8688,
+      "eval_samples_per_second": 231.343,
+      "eval_steps_per_second": 8.057,
+      "step": 134540
+    },
+    {
+      "epoch": 868.06,
+      "learning_rate": 7.329749103942653e-06,
+      "loss": 0.2865,
+      "step": 134550
+    },
+    {
+      "epoch": 868.13,
+      "learning_rate": 7.326164874551972e-06,
+      "loss": 0.176,
+      "step": 134560
+    },
+    {
+      "epoch": 868.19,
+      "learning_rate": 7.322580645161291e-06,
+      "loss": 0.1913,
+      "step": 134570
+    },
+    {
+      "epoch": 868.26,
+      "learning_rate": 7.31899641577061e-06,
+      "loss": 0.225,
+      "step": 134580
+    },
+    {
+      "epoch": 868.32,
+      "learning_rate": 7.315412186379928e-06,
+      "loss": 0.1701,
+      "step": 134590
+    },
+    {
+      "epoch": 868.39,
+      "learning_rate": 7.3118279569892475e-06,
+      "loss": 0.2153,
+      "step": 134600
+    },
+    {
+      "epoch": 868.45,
+      "learning_rate": 7.3082437275985665e-06,
+      "loss": 0.2131,
+      "step": 134610
+    },
+    {
+      "epoch": 868.51,
+      "learning_rate": 7.3046594982078856e-06,
+      "loss": 0.2472,
+      "step": 134620
+    },
+    {
+      "epoch": 868.58,
+      "learning_rate": 7.301075268817205e-06,
+      "loss": 0.171,
+      "step": 134630
+    },
+    {
+      "epoch": 868.64,
+      "learning_rate": 7.297491039426524e-06,
+      "loss": 0.2091,
+      "step": 134640
+    },
+    {
+      "epoch": 868.71,
+      "learning_rate": 7.293906810035843e-06,
+      "loss": 0.1946,
+      "step": 134650
+    },
+    {
+      "epoch": 868.77,
+      "learning_rate": 7.290322580645162e-06,
+      "loss": 0.1757,
+      "step": 134660
+    },
+    {
+      "epoch": 868.84,
+      "learning_rate": 7.286738351254481e-06,
+      "loss": 0.2145,
+      "step": 134670
+    },
+    {
+      "epoch": 868.9,
+      "learning_rate": 7.2831541218638e-06,
+      "loss": 0.1507,
+      "step": 134680
+    },
+    {
+      "epoch": 868.96,
+      "learning_rate": 7.279569892473119e-06,
+      "loss": 0.1858,
+      "step": 134690
+    },
+    {
+      "epoch": 869.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9529990553855896,
+      "eval_runtime": 0.9432,
+      "eval_samples_per_second": 213.107,
+      "eval_steps_per_second": 7.422,
+      "step": 134695
+    },
+    {
+      "epoch": 869.03,
+      "learning_rate": 7.275985663082438e-06,
+      "loss": 0.2456,
+      "step": 134700
+    },
+    {
+      "epoch": 869.1,
+      "learning_rate": 7.272401433691757e-06,
+      "loss": 0.2158,
+      "step": 134710
+    },
+    {
+      "epoch": 869.16,
+      "learning_rate": 7.268817204301076e-06,
+      "loss": 0.1526,
+      "step": 134720
+    },
+    {
+      "epoch": 869.23,
+      "learning_rate": 7.265232974910395e-06,
+      "loss": 0.2172,
+      "step": 134730
+    },
+    {
+      "epoch": 869.29,
+      "learning_rate": 7.261648745519714e-06,
+      "loss": 0.2234,
+      "step": 134740
+    },
+    {
+      "epoch": 869.35,
+      "learning_rate": 7.258064516129033e-06,
+      "loss": 0.2756,
+      "step": 134750
+    },
+    {
+      "epoch": 869.42,
+      "learning_rate": 7.254480286738352e-06,
+      "loss": 0.2497,
+      "step": 134760
+    },
+    {
+      "epoch": 869.48,
+      "learning_rate": 7.250896057347671e-06,
+      "loss": 0.1538,
+      "step": 134770
+    },
+    {
+      "epoch": 869.55,
+      "learning_rate": 7.24731182795699e-06,
+      "loss": 0.2034,
+      "step": 134780
+    },
+    {
+      "epoch": 869.61,
+      "learning_rate": 7.2437275985663094e-06,
+      "loss": 0.1978,
+      "step": 134790
+    },
+    {
+      "epoch": 869.68,
+      "learning_rate": 7.2401433691756285e-06,
+      "loss": 0.2063,
+      "step": 134800
+    },
+    {
+      "epoch": 869.74,
+      "learning_rate": 7.2365591397849475e-06,
+      "loss": 0.219,
+      "step": 134810
+    },
+    {
+      "epoch": 869.8,
+      "learning_rate": 7.232974910394267e-06,
+      "loss": 0.2133,
+      "step": 134820
+    },
+    {
+      "epoch": 869.87,
+      "learning_rate": 7.229390681003586e-06,
+      "loss": 0.2189,
+      "step": 134830
+    },
+    {
+      "epoch": 869.93,
+      "learning_rate": 7.225806451612904e-06,
+      "loss": 0.2096,
+      "step": 134840
+    },
+    {
+      "epoch": 870.0,
+      "learning_rate": 7.222222222222222e-06,
+      "loss": 0.189,
+      "step": 134850
+    },
+    {
+      "epoch": 870.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9599297642707825,
+      "eval_runtime": 0.8679,
+      "eval_samples_per_second": 231.591,
+      "eval_steps_per_second": 8.065,
+      "step": 134850
+    },
+    {
+      "epoch": 870.06,
+      "learning_rate": 7.218637992831541e-06,
+      "loss": 0.2257,
+      "step": 134860
+    },
+    {
+      "epoch": 870.13,
+      "learning_rate": 7.21505376344086e-06,
+      "loss": 0.1951,
+      "step": 134870
+    },
+    {
+      "epoch": 870.19,
+      "learning_rate": 7.211469534050179e-06,
+      "loss": 0.2171,
+      "step": 134880
+    },
+    {
+      "epoch": 870.26,
+      "learning_rate": 7.207885304659498e-06,
+      "loss": 0.1782,
+      "step": 134890
+    },
+    {
+      "epoch": 870.32,
+      "learning_rate": 7.204301075268817e-06,
+      "loss": 0.2033,
+      "step": 134900
+    },
+    {
+      "epoch": 870.39,
+      "learning_rate": 7.200716845878136e-06,
+      "loss": 0.1914,
+      "step": 134910
+    },
+    {
+      "epoch": 870.45,
+      "learning_rate": 7.1971326164874554e-06,
+      "loss": 0.2097,
+      "step": 134920
+    },
+    {
+      "epoch": 870.51,
+      "learning_rate": 7.193548387096774e-06,
+      "loss": 0.1846,
+      "step": 134930
+    },
+    {
+      "epoch": 870.58,
+      "learning_rate": 7.189964157706093e-06,
+      "loss": 0.1854,
+      "step": 134940
+    },
+    {
+      "epoch": 870.64,
+      "learning_rate": 7.186379928315412e-06,
+      "loss": 0.1999,
+      "step": 134950
+    },
+    {
+      "epoch": 870.71,
+      "learning_rate": 7.182795698924731e-06,
+      "loss": 0.2288,
+      "step": 134960
+    },
+    {
+      "epoch": 870.77,
+      "learning_rate": 7.17921146953405e-06,
+      "loss": 0.1904,
+      "step": 134970
+    },
+    {
+      "epoch": 870.84,
+      "learning_rate": 7.175627240143369e-06,
+      "loss": 0.2299,
+      "step": 134980
+    },
+    {
+      "epoch": 870.9,
+      "learning_rate": 7.172043010752688e-06,
+      "loss": 0.1995,
+      "step": 134990
+    },
+    {
+      "epoch": 870.96,
+      "learning_rate": 7.168458781362007e-06,
+      "loss": 0.2055,
+      "step": 135000
+    },
+    {
+      "epoch": 871.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9545081853866577,
+      "eval_runtime": 0.8713,
+      "eval_samples_per_second": 230.696,
+      "eval_steps_per_second": 8.034,
+      "step": 135005
+    },
+    {
+      "epoch": 871.03,
+      "learning_rate": 7.164874551971326e-06,
+      "loss": 0.232,
+      "step": 135010
+    },
+    {
+      "epoch": 871.1,
+      "learning_rate": 7.161290322580645e-06,
+      "loss": 0.1719,
+      "step": 135020
+    },
+    {
+      "epoch": 871.16,
+      "learning_rate": 7.157706093189964e-06,
+      "loss": 0.2319,
+      "step": 135030
+    },
+    {
+      "epoch": 871.23,
+      "learning_rate": 7.154121863799283e-06,
+      "loss": 0.2272,
+      "step": 135040
+    },
+    {
+      "epoch": 871.29,
+      "learning_rate": 7.150537634408602e-06,
+      "loss": 0.2162,
+      "step": 135050
+    },
+    {
+      "epoch": 871.35,
+      "learning_rate": 7.146953405017921e-06,
+      "loss": 0.2053,
+      "step": 135060
+    },
+    {
+      "epoch": 871.42,
+      "learning_rate": 7.14336917562724e-06,
+      "loss": 0.207,
+      "step": 135070
+    },
+    {
+      "epoch": 871.48,
+      "learning_rate": 7.139784946236559e-06,
+      "loss": 0.2277,
+      "step": 135080
+    },
+    {
+      "epoch": 871.55,
+      "learning_rate": 7.1362007168458785e-06,
+      "loss": 0.2117,
+      "step": 135090
+    },
+    {
+      "epoch": 871.61,
+      "learning_rate": 7.1326164874551975e-06,
+      "loss": 0.2078,
+      "step": 135100
+    },
+    {
+      "epoch": 871.68,
+      "learning_rate": 7.1290322580645166e-06,
+      "loss": 0.2082,
+      "step": 135110
+    },
+    {
+      "epoch": 871.74,
+      "learning_rate": 7.125448028673836e-06,
+      "loss": 0.1675,
+      "step": 135120
+    },
+    {
+      "epoch": 871.8,
+      "learning_rate": 7.121863799283155e-06,
+      "loss": 0.1911,
+      "step": 135130
+    },
+    {
+      "epoch": 871.87,
+      "learning_rate": 7.118279569892474e-06,
+      "loss": 0.2079,
+      "step": 135140
+    },
+    {
+      "epoch": 871.93,
+      "learning_rate": 7.114695340501793e-06,
+      "loss": 0.226,
+      "step": 135150
+    },
+    {
+      "epoch": 872.0,
+      "learning_rate": 7.111111111111112e-06,
+      "loss": 0.1912,
+      "step": 135160
+    },
+    {
+      "epoch": 872.0,
+      "eval_accuracy": 0.7412935323383084,
+      "eval_loss": 0.973096489906311,
+      "eval_runtime": 0.9047,
+      "eval_samples_per_second": 222.185,
+      "eval_steps_per_second": 7.738,
+      "step": 135160
+    },
+    {
+      "epoch": 872.06,
+      "learning_rate": 7.107526881720431e-06,
+      "loss": 0.1971,
+      "step": 135170
+    },
+    {
+      "epoch": 872.13,
+      "learning_rate": 7.103942652329749e-06,
+      "loss": 0.1987,
+      "step": 135180
+    },
+    {
+      "epoch": 872.19,
+      "learning_rate": 7.100358422939068e-06,
+      "loss": 0.2197,
+      "step": 135190
+    },
+    {
+      "epoch": 872.26,
+      "learning_rate": 7.096774193548387e-06,
+      "loss": 0.2175,
+      "step": 135200
+    },
+    {
+      "epoch": 872.32,
+      "learning_rate": 7.093189964157706e-06,
+      "loss": 0.2147,
+      "step": 135210
+    },
+    {
+      "epoch": 872.39,
+      "learning_rate": 7.089605734767025e-06,
+      "loss": 0.2182,
+      "step": 135220
+    },
+    {
+      "epoch": 872.45,
+      "learning_rate": 7.086021505376344e-06,
+      "loss": 0.2034,
+      "step": 135230
+    },
+    {
+      "epoch": 872.51,
+      "learning_rate": 7.082437275985663e-06,
+      "loss": 0.2191,
+      "step": 135240
+    },
+    {
+      "epoch": 872.58,
+      "learning_rate": 7.0788530465949824e-06,
+      "loss": 0.2009,
+      "step": 135250
+    },
+    {
+      "epoch": 872.64,
+      "learning_rate": 7.0752688172043015e-06,
+      "loss": 0.1904,
+      "step": 135260
+    },
+    {
+      "epoch": 872.71,
+      "learning_rate": 7.0716845878136205e-06,
+      "loss": 0.1631,
+      "step": 135270
+    },
+    {
+      "epoch": 872.77,
+      "learning_rate": 7.06810035842294e-06,
+      "loss": 0.2104,
+      "step": 135280
+    },
+    {
+      "epoch": 872.84,
+      "learning_rate": 7.064516129032259e-06,
+      "loss": 0.2121,
+      "step": 135290
+    },
+    {
+      "epoch": 872.9,
+      "learning_rate": 7.060931899641578e-06,
+      "loss": 0.1822,
+      "step": 135300
+    },
+    {
+      "epoch": 872.96,
+      "learning_rate": 7.057347670250897e-06,
+      "loss": 0.224,
+      "step": 135310
+    },
+    {
+      "epoch": 873.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.946251392364502,
+      "eval_runtime": 0.885,
+      "eval_samples_per_second": 227.115,
+      "eval_steps_per_second": 7.909,
+      "step": 135315
+    },
+    {
+      "epoch": 873.03,
+      "learning_rate": 7.053763440860216e-06,
+      "loss": 0.2034,
+      "step": 135320
+    },
+    {
+      "epoch": 873.1,
+      "learning_rate": 7.050179211469535e-06,
+      "loss": 0.2187,
+      "step": 135330
+    },
+    {
+      "epoch": 873.16,
+      "learning_rate": 7.046594982078854e-06,
+      "loss": 0.1841,
+      "step": 135340
+    },
+    {
+      "epoch": 873.23,
+      "learning_rate": 7.043010752688173e-06,
+      "loss": 0.2326,
+      "step": 135350
+    },
+    {
+      "epoch": 873.29,
+      "learning_rate": 7.039426523297492e-06,
+      "loss": 0.1926,
+      "step": 135360
+    },
+    {
+      "epoch": 873.35,
+      "learning_rate": 7.035842293906811e-06,
+      "loss": 0.23,
+      "step": 135370
+    },
+    {
+      "epoch": 873.42,
+      "learning_rate": 7.03225806451613e-06,
+      "loss": 0.1878,
+      "step": 135380
+    },
+    {
+      "epoch": 873.48,
+      "learning_rate": 7.028673835125449e-06,
+      "loss": 0.2024,
+      "step": 135390
+    },
+    {
+      "epoch": 873.55,
+      "learning_rate": 7.025089605734768e-06,
+      "loss": 0.2273,
+      "step": 135400
+    },
+    {
+      "epoch": 873.61,
+      "learning_rate": 7.021505376344087e-06,
+      "loss": 0.2386,
+      "step": 135410
+    },
+    {
+      "epoch": 873.68,
+      "learning_rate": 7.017921146953406e-06,
+      "loss": 0.1953,
+      "step": 135420
+    },
+    {
+      "epoch": 873.74,
+      "learning_rate": 7.0143369175627245e-06,
+      "loss": 0.2141,
+      "step": 135430
+    },
+    {
+      "epoch": 873.8,
+      "learning_rate": 7.010752688172044e-06,
+      "loss": 0.1663,
+      "step": 135440
+    },
+    {
+      "epoch": 873.87,
+      "learning_rate": 7.007168458781363e-06,
+      "loss": 0.2317,
+      "step": 135450
+    },
+    {
+      "epoch": 873.93,
+      "learning_rate": 7.003584229390682e-06,
+      "loss": 0.1598,
+      "step": 135460
+    },
+    {
+      "epoch": 874.0,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 0.187,
+      "step": 135470
+    },
+    {
+      "epoch": 874.0,
+      "eval_accuracy": 0.746268656716418,
+      "eval_loss": 0.9495944976806641,
+      "eval_runtime": 0.8718,
+      "eval_samples_per_second": 230.562,
+      "eval_steps_per_second": 8.03,
+      "step": 135470
+    },
+    {
+      "epoch": 874.06,
+      "learning_rate": 6.99641577060932e-06,
+      "loss": 0.2101,
+      "step": 135480
+    },
+    {
+      "epoch": 874.13,
+      "learning_rate": 6.992831541218638e-06,
+      "loss": 0.2328,
+      "step": 135490
+    },
+    {
+      "epoch": 874.19,
+      "learning_rate": 6.989247311827957e-06,
+      "loss": 0.2357,
+      "step": 135500
+    },
+    {
+      "epoch": 874.26,
+      "learning_rate": 6.985663082437276e-06,
+      "loss": 0.2294,
+      "step": 135510
+    },
+    {
+      "epoch": 874.32,
+      "learning_rate": 6.982078853046594e-06,
+      "loss": 0.2065,
+      "step": 135520
+    },
+    {
+      "epoch": 874.39,
+      "learning_rate": 6.978494623655913e-06,
+      "loss": 0.1897,
+      "step": 135530
+    },
+    {
+      "epoch": 874.45,
+      "learning_rate": 6.974910394265232e-06,
+      "loss": 0.2443,
+      "step": 135540
+    },
+    {
+      "epoch": 874.51,
+      "learning_rate": 6.9713261648745515e-06,
+      "loss": 0.1914,
+      "step": 135550
+    },
+    {
+      "epoch": 874.58,
+      "learning_rate": 6.9677419354838705e-06,
+      "loss": 0.1923,
+      "step": 135560
+    },
+    {
+      "epoch": 874.64,
+      "learning_rate": 6.9641577060931896e-06,
+      "loss": 0.2154,
+      "step": 135570
+    },
+    {
+      "epoch": 874.71,
+      "learning_rate": 6.960573476702509e-06,
+      "loss": 0.2301,
+      "step": 135580
+    },
+    {
+      "epoch": 874.77,
+      "learning_rate": 6.956989247311828e-06,
+      "loss": 0.224,
+      "step": 135590
+    },
+    {
+      "epoch": 874.84,
+      "learning_rate": 6.953405017921147e-06,
+      "loss": 0.1609,
+      "step": 135600
+    },
+    {
+      "epoch": 874.9,
+      "learning_rate": 6.949820788530466e-06,
+      "loss": 0.1751,
+      "step": 135610
+    },
+    {
+      "epoch": 874.96,
+      "learning_rate": 6.946236559139785e-06,
+      "loss": 0.218,
+      "step": 135620
+    },
+    {
+      "epoch": 875.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.960318922996521,
+      "eval_runtime": 0.8705,
+      "eval_samples_per_second": 230.899,
+      "eval_steps_per_second": 8.041,
+      "step": 135625
+    },
+    {
+      "epoch": 875.03,
+      "learning_rate": 6.942652329749104e-06,
+      "loss": 0.2132,
+      "step": 135630
+    },
+    {
+      "epoch": 875.1,
+      "learning_rate": 6.939068100358423e-06,
+      "loss": 0.1767,
+      "step": 135640
+    },
+    {
+      "epoch": 875.16,
+      "learning_rate": 6.935483870967742e-06,
+      "loss": 0.218,
+      "step": 135650
+    },
+    {
+      "epoch": 875.23,
+      "learning_rate": 6.931899641577061e-06,
+      "loss": 0.1973,
+      "step": 135660
+    },
+    {
+      "epoch": 875.29,
+      "learning_rate": 6.92831541218638e-06,
+      "loss": 0.2016,
+      "step": 135670
+    },
+    {
+      "epoch": 875.35,
+      "learning_rate": 6.924731182795699e-06,
+      "loss": 0.1988,
+      "step": 135680
+    },
+    {
+      "epoch": 875.42,
+      "learning_rate": 6.921146953405018e-06,
+      "loss": 0.1839,
+      "step": 135690
+    },
+    {
+      "epoch": 875.48,
+      "learning_rate": 6.917562724014337e-06,
+      "loss": 0.2037,
+      "step": 135700
+    },
+    {
+      "epoch": 875.55,
+      "learning_rate": 6.913978494623656e-06,
+      "loss": 0.2289,
+      "step": 135710
+    },
+    {
+      "epoch": 875.61,
+      "learning_rate": 6.910394265232975e-06,
+      "loss": 0.233,
+      "step": 135720
+    },
+    {
+      "epoch": 875.68,
+      "learning_rate": 6.906810035842294e-06,
+      "loss": 0.2088,
+      "step": 135730
+    },
+    {
+      "epoch": 875.74,
+      "learning_rate": 6.9032258064516135e-06,
+      "loss": 0.2054,
+      "step": 135740
+    },
+    {
+      "epoch": 875.8,
+      "learning_rate": 6.8996415770609325e-06,
+      "loss": 0.1872,
+      "step": 135750
+    },
+    {
+      "epoch": 875.87,
+      "learning_rate": 6.8960573476702516e-06,
+      "loss": 0.1946,
+      "step": 135760
+    },
+    {
+      "epoch": 875.93,
+      "learning_rate": 6.89247311827957e-06,
+      "loss": 0.195,
+      "step": 135770
+    },
+    {
+      "epoch": 876.0,
+      "learning_rate": 6.888888888888889e-06,
+      "loss": 0.1882,
+      "step": 135780
+    },
+    {
+      "epoch": 876.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9332004189491272,
+      "eval_runtime": 0.8894,
+      "eval_samples_per_second": 226.007,
+      "eval_steps_per_second": 7.871,
+      "step": 135780
+    },
+    {
+      "epoch": 876.06,
+      "learning_rate": 6.885304659498208e-06,
+      "loss": 0.2175,
+      "step": 135790
+    },
+    {
+      "epoch": 876.13,
+      "learning_rate": 6.881720430107527e-06,
+      "loss": 0.1996,
+      "step": 135800
+    },
+    {
+      "epoch": 876.19,
+      "learning_rate": 6.878136200716846e-06,
+      "loss": 0.1642,
+      "step": 135810
+    },
+    {
+      "epoch": 876.26,
+      "learning_rate": 6.874551971326165e-06,
+      "loss": 0.25,
+      "step": 135820
+    },
+    {
+      "epoch": 876.32,
+      "learning_rate": 6.870967741935484e-06,
+      "loss": 0.1984,
+      "step": 135830
+    },
+    {
+      "epoch": 876.39,
+      "learning_rate": 6.867383512544803e-06,
+      "loss": 0.2267,
+      "step": 135840
+    },
+    {
+      "epoch": 876.45,
+      "learning_rate": 6.863799283154122e-06,
+      "loss": 0.1644,
+      "step": 135850
+    },
+    {
+      "epoch": 876.51,
+      "learning_rate": 6.860215053763441e-06,
+      "loss": 0.2079,
+      "step": 135860
+    },
+    {
+      "epoch": 876.58,
+      "learning_rate": 6.85663082437276e-06,
+      "loss": 0.2246,
+      "step": 135870
+    },
+    {
+      "epoch": 876.64,
+      "learning_rate": 6.853046594982079e-06,
+      "loss": 0.2436,
+      "step": 135880
+    },
+    {
+      "epoch": 876.71,
+      "learning_rate": 6.849462365591398e-06,
+      "loss": 0.2324,
+      "step": 135890
+    },
+    {
+      "epoch": 876.77,
+      "learning_rate": 6.8458781362007174e-06,
+      "loss": 0.2666,
+      "step": 135900
+    },
+    {
+      "epoch": 876.84,
+      "learning_rate": 6.8422939068100365e-06,
+      "loss": 0.2189,
+      "step": 135910
+    },
+    {
+      "epoch": 876.9,
+      "learning_rate": 6.8387096774193555e-06,
+      "loss": 0.2115,
+      "step": 135920
+    },
+    {
+      "epoch": 876.96,
+      "learning_rate": 6.835125448028675e-06,
+      "loss": 0.1605,
+      "step": 135930
+    },
+    {
+      "epoch": 877.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9545229077339172,
+      "eval_runtime": 0.8563,
+      "eval_samples_per_second": 234.719,
+      "eval_steps_per_second": 8.174,
+      "step": 135935
+    },
+    {
+      "epoch": 877.03,
+      "learning_rate": 6.831541218637994e-06,
+      "loss": 0.1961,
+      "step": 135940
+    },
+    {
+      "epoch": 877.1,
+      "learning_rate": 6.827956989247313e-06,
+      "loss": 0.1811,
+      "step": 135950
+    },
+    {
+      "epoch": 877.16,
+      "learning_rate": 6.824372759856632e-06,
+      "loss": 0.2131,
+      "step": 135960
+    },
+    {
+      "epoch": 877.23,
+      "learning_rate": 6.820788530465951e-06,
+      "loss": 0.1989,
+      "step": 135970
+    },
+    {
+      "epoch": 877.29,
+      "learning_rate": 6.81720430107527e-06,
+      "loss": 0.1914,
+      "step": 135980
+    },
+    {
+      "epoch": 877.35,
+      "learning_rate": 6.813620071684589e-06,
+      "loss": 0.2021,
+      "step": 135990
+    },
+    {
+      "epoch": 877.42,
+      "learning_rate": 6.810035842293908e-06,
+      "loss": 0.1916,
+      "step": 136000
+    },
+    {
+      "epoch": 877.48,
+      "learning_rate": 6.806451612903226e-06,
+      "loss": 0.2379,
+      "step": 136010
+    },
+    {
+      "epoch": 877.55,
+      "learning_rate": 6.802867383512545e-06,
+      "loss": 0.1828,
+      "step": 136020
+    },
+    {
+      "epoch": 877.61,
+      "learning_rate": 6.799283154121864e-06,
+      "loss": 0.2153,
+      "step": 136030
+    },
+    {
+      "epoch": 877.68,
+      "learning_rate": 6.795698924731183e-06,
+      "loss": 0.2632,
+      "step": 136040
+    },
+    {
+      "epoch": 877.74,
+      "learning_rate": 6.792114695340502e-06,
+      "loss": 0.1764,
+      "step": 136050
+    },
+    {
+      "epoch": 877.8,
+      "learning_rate": 6.788530465949821e-06,
+      "loss": 0.2174,
+      "step": 136060
+    },
+    {
+      "epoch": 877.87,
+      "learning_rate": 6.7849462365591405e-06,
+      "loss": 0.1829,
+      "step": 136070
+    },
+    {
+      "epoch": 877.93,
+      "learning_rate": 6.7813620071684595e-06,
+      "loss": 0.2275,
+      "step": 136080
+    },
+    {
+      "epoch": 878.0,
+      "learning_rate": 6.777777777777779e-06,
+      "loss": 0.2007,
+      "step": 136090
+    },
+    {
+      "epoch": 878.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9756190180778503,
+      "eval_runtime": 0.8673,
+      "eval_samples_per_second": 231.755,
+      "eval_steps_per_second": 8.071,
+      "step": 136090
+    },
+    {
+      "epoch": 878.06,
+      "learning_rate": 6.774193548387098e-06,
+      "loss": 0.2464,
+      "step": 136100
+    },
+    {
+      "epoch": 878.13,
+      "learning_rate": 6.770609318996417e-06,
+      "loss": 0.2258,
+      "step": 136110
+    },
+    {
+      "epoch": 878.19,
+      "learning_rate": 6.767025089605736e-06,
+      "loss": 0.2289,
+      "step": 136120
+    },
+    {
+      "epoch": 878.26,
+      "learning_rate": 6.763440860215053e-06,
+      "loss": 0.2144,
+      "step": 136130
+    },
+    {
+      "epoch": 878.32,
+      "learning_rate": 6.759856630824372e-06,
+      "loss": 0.1929,
+      "step": 136140
+    },
+    {
+      "epoch": 878.39,
+      "learning_rate": 6.756272401433691e-06,
+      "loss": 0.1833,
+      "step": 136150
+    },
+    {
+      "epoch": 878.45,
+      "learning_rate": 6.75268817204301e-06,
+      "loss": 0.2524,
+      "step": 136160
+    },
+    {
+      "epoch": 878.51,
+      "learning_rate": 6.749103942652329e-06,
+      "loss": 0.1847,
+      "step": 136170
+    },
+    {
+      "epoch": 878.58,
+      "learning_rate": 6.745519713261648e-06,
+      "loss": 0.219,
+      "step": 136180
+    },
+    {
+      "epoch": 878.64,
+      "learning_rate": 6.741935483870967e-06,
+      "loss": 0.2042,
+      "step": 136190
+    },
+    {
+      "epoch": 878.71,
+      "learning_rate": 6.7383512544802865e-06,
+      "loss": 0.2236,
+      "step": 136200
+    },
+    {
+      "epoch": 878.77,
+      "learning_rate": 6.7347670250896055e-06,
+      "loss": 0.2029,
+      "step": 136210
+    },
+    {
+      "epoch": 878.84,
+      "learning_rate": 6.7311827956989246e-06,
+      "loss": 0.1467,
+      "step": 136220
+    },
+    {
+      "epoch": 878.9,
+      "learning_rate": 6.727598566308244e-06,
+      "loss": 0.2098,
+      "step": 136230
+    },
+    {
+      "epoch": 878.96,
+      "learning_rate": 6.724014336917563e-06,
+      "loss": 0.2177,
+      "step": 136240
+    },
+    {
+      "epoch": 879.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9881079196929932,
+      "eval_runtime": 0.871,
+      "eval_samples_per_second": 230.771,
+      "eval_steps_per_second": 8.037,
+      "step": 136245
+    },
+    {
+      "epoch": 879.03,
+      "learning_rate": 6.720430107526882e-06,
+      "loss": 0.239,
+      "step": 136250
+    },
+    {
+      "epoch": 879.1,
+      "learning_rate": 6.716845878136201e-06,
+      "loss": 0.2369,
+      "step": 136260
+    },
+    {
+      "epoch": 879.16,
+      "learning_rate": 6.71326164874552e-06,
+      "loss": 0.2147,
+      "step": 136270
+    },
+    {
+      "epoch": 879.23,
+      "learning_rate": 6.709677419354839e-06,
+      "loss": 0.2149,
+      "step": 136280
+    },
+    {
+      "epoch": 879.29,
+      "learning_rate": 6.706093189964158e-06,
+      "loss": 0.2052,
+      "step": 136290
+    },
+    {
+      "epoch": 879.35,
+      "learning_rate": 6.702508960573477e-06,
+      "loss": 0.2209,
+      "step": 136300
+    },
+    {
+      "epoch": 879.42,
+      "learning_rate": 6.698924731182796e-06,
+      "loss": 0.218,
+      "step": 136310
+    },
+    {
+      "epoch": 879.48,
+      "learning_rate": 6.695340501792115e-06,
+      "loss": 0.2354,
+      "step": 136320
+    },
+    {
+      "epoch": 879.55,
+      "learning_rate": 6.691756272401434e-06,
+      "loss": 0.1811,
+      "step": 136330
+    },
+    {
+      "epoch": 879.61,
+      "learning_rate": 6.688172043010753e-06,
+      "loss": 0.1839,
+      "step": 136340
+    },
+    {
+      "epoch": 879.68,
+      "learning_rate": 6.684587813620071e-06,
+      "loss": 0.23,
+      "step": 136350
+    },
+    {
+      "epoch": 879.74,
+      "learning_rate": 6.6810035842293904e-06,
+      "loss": 0.2198,
+      "step": 136360
+    },
+    {
+      "epoch": 879.8,
+      "learning_rate": 6.6774193548387095e-06,
+      "loss": 0.2045,
+      "step": 136370
+    },
+    {
+      "epoch": 879.87,
+      "learning_rate": 6.6738351254480285e-06,
+      "loss": 0.1851,
+      "step": 136380
+    },
+    {
+      "epoch": 879.93,
+      "learning_rate": 6.670250896057348e-06,
+      "loss": 0.2258,
+      "step": 136390
+    },
+    {
+      "epoch": 880.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.2456,
+      "step": 136400
+    },
+    {
+      "epoch": 880.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.972332775592804,
+      "eval_runtime": 0.8758,
+      "eval_samples_per_second": 229.502,
+      "eval_steps_per_second": 7.993,
+      "step": 136400
+    },
+    {
+      "epoch": 880.06,
+      "learning_rate": 6.663082437275986e-06,
+      "loss": 0.1921,
+      "step": 136410
+    },
+    {
+      "epoch": 880.13,
+      "learning_rate": 6.659498207885305e-06,
+      "loss": 0.243,
+      "step": 136420
+    },
+    {
+      "epoch": 880.19,
+      "learning_rate": 6.655913978494624e-06,
+      "loss": 0.2063,
+      "step": 136430
+    },
+    {
+      "epoch": 880.26,
+      "learning_rate": 6.652329749103943e-06,
+      "loss": 0.1766,
+      "step": 136440
+    },
+    {
+      "epoch": 880.32,
+      "learning_rate": 6.648745519713262e-06,
+      "loss": 0.1734,
+      "step": 136450
+    },
+    {
+      "epoch": 880.39,
+      "learning_rate": 6.645161290322581e-06,
+      "loss": 0.1712,
+      "step": 136460
+    },
+    {
+      "epoch": 880.45,
+      "learning_rate": 6.6415770609319e-06,
+      "loss": 0.1972,
+      "step": 136470
+    },
+    {
+      "epoch": 880.51,
+      "learning_rate": 6.637992831541219e-06,
+      "loss": 0.27,
+      "step": 136480
+    },
+    {
+      "epoch": 880.58,
+      "learning_rate": 6.634408602150538e-06,
+      "loss": 0.2107,
+      "step": 136490
+    },
+    {
+      "epoch": 880.64,
+      "learning_rate": 6.630824372759857e-06,
+      "loss": 0.2129,
+      "step": 136500
+    },
+    {
+      "epoch": 880.71,
+      "learning_rate": 6.627240143369176e-06,
+      "loss": 0.1824,
+      "step": 136510
+    },
+    {
+      "epoch": 880.77,
+      "learning_rate": 6.623655913978495e-06,
+      "loss": 0.2021,
+      "step": 136520
+    },
+    {
+      "epoch": 880.84,
+      "learning_rate": 6.620071684587814e-06,
+      "loss": 0.2307,
+      "step": 136530
+    },
+    {
+      "epoch": 880.9,
+      "learning_rate": 6.616487455197133e-06,
+      "loss": 0.2083,
+      "step": 136540
+    },
+    {
+      "epoch": 880.96,
+      "learning_rate": 6.6129032258064524e-06,
+      "loss": 0.2423,
+      "step": 136550
+    },
+    {
+      "epoch": 881.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.9733842611312866,
+      "eval_runtime": 0.8756,
+      "eval_samples_per_second": 229.558,
+      "eval_steps_per_second": 7.995,
+      "step": 136555
+    },
+    {
+      "epoch": 881.03,
+      "learning_rate": 6.6093189964157715e-06,
+      "loss": 0.1734,
+      "step": 136560
+    },
+    {
+      "epoch": 881.1,
+      "learning_rate": 6.6057347670250905e-06,
+      "loss": 0.1977,
+      "step": 136570
+    },
+    {
+      "epoch": 881.16,
+      "learning_rate": 6.60215053763441e-06,
+      "loss": 0.224,
+      "step": 136580
+    },
+    {
+      "epoch": 881.23,
+      "learning_rate": 6.598566308243729e-06,
+      "loss": 0.2238,
+      "step": 136590
+    },
+    {
+      "epoch": 881.29,
+      "learning_rate": 6.594982078853047e-06,
+      "loss": 0.2654,
+      "step": 136600
+    },
+    {
+      "epoch": 881.35,
+      "learning_rate": 6.591397849462366e-06,
+      "loss": 0.1705,
+      "step": 136610
+    },
+    {
+      "epoch": 881.42,
+      "learning_rate": 6.587813620071685e-06,
+      "loss": 0.2519,
+      "step": 136620
+    },
+    {
+      "epoch": 881.48,
+      "learning_rate": 6.584229390681004e-06,
+      "loss": 0.2494,
+      "step": 136630
+    },
+    {
+      "epoch": 881.55,
+      "learning_rate": 6.580645161290323e-06,
+      "loss": 0.2412,
+      "step": 136640
+    },
+    {
+      "epoch": 881.61,
+      "learning_rate": 6.577060931899642e-06,
+      "loss": 0.2013,
+      "step": 136650
+    },
+    {
+      "epoch": 881.68,
+      "learning_rate": 6.573476702508961e-06,
+      "loss": 0.209,
+      "step": 136660
+    },
+    {
+      "epoch": 881.74,
+      "learning_rate": 6.56989247311828e-06,
+      "loss": 0.1895,
+      "step": 136670
+    },
+    {
+      "epoch": 881.8,
+      "learning_rate": 6.566308243727599e-06,
+      "loss": 0.1698,
+      "step": 136680
+    },
+    {
+      "epoch": 881.87,
+      "learning_rate": 6.562724014336918e-06,
+      "loss": 0.1949,
+      "step": 136690
+    },
+    {
+      "epoch": 881.93,
+      "learning_rate": 6.559139784946237e-06,
+      "loss": 0.2083,
+      "step": 136700
+    },
+    {
+      "epoch": 882.0,
+      "learning_rate": 6.555555555555556e-06,
+      "loss": 0.2249,
+      "step": 136710
+    },
+    {
+      "epoch": 882.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9833352565765381,
+      "eval_runtime": 0.8845,
+      "eval_samples_per_second": 227.25,
+      "eval_steps_per_second": 7.914,
+      "step": 136710
+    },
+    {
+      "epoch": 882.06,
+      "learning_rate": 6.5519713261648755e-06,
+      "loss": 0.2342,
+      "step": 136720
+    },
+    {
+      "epoch": 882.13,
+      "learning_rate": 6.5483870967741945e-06,
+      "loss": 0.2381,
+      "step": 136730
+    },
+    {
+      "epoch": 882.19,
+      "learning_rate": 6.5448028673835136e-06,
+      "loss": 0.1752,
+      "step": 136740
+    },
+    {
+      "epoch": 882.26,
+      "learning_rate": 6.541218637992833e-06,
+      "loss": 0.2106,
+      "step": 136750
+    },
+    {
+      "epoch": 882.32,
+      "learning_rate": 6.537634408602152e-06,
+      "loss": 0.2696,
+      "step": 136760
+    },
+    {
+      "epoch": 882.39,
+      "learning_rate": 6.534050179211469e-06,
+      "loss": 0.2518,
+      "step": 136770
+    },
+    {
+      "epoch": 882.45,
+      "learning_rate": 6.530465949820788e-06,
+      "loss": 0.2508,
+      "step": 136780
+    },
+    {
+      "epoch": 882.51,
+      "learning_rate": 6.526881720430107e-06,
+      "loss": 0.2055,
+      "step": 136790
+    },
+    {
+      "epoch": 882.58,
+      "learning_rate": 6.523297491039426e-06,
+      "loss": 0.2365,
+      "step": 136800
+    },
+    {
+      "epoch": 882.64,
+      "learning_rate": 6.519713261648745e-06,
+      "loss": 0.2067,
+      "step": 136810
+    },
+    {
+      "epoch": 882.71,
+      "learning_rate": 6.516129032258064e-06,
+      "loss": 0.1612,
+      "step": 136820
+    },
+    {
+      "epoch": 882.77,
+      "learning_rate": 6.512544802867383e-06,
+      "loss": 0.1953,
+      "step": 136830
+    },
+    {
+      "epoch": 882.84,
+      "learning_rate": 6.508960573476702e-06,
+      "loss": 0.1921,
+      "step": 136840
+    },
+    {
+      "epoch": 882.9,
+      "learning_rate": 6.5053763440860214e-06,
+      "loss": 0.2167,
+      "step": 136850
+    },
+    {
+      "epoch": 882.96,
+      "learning_rate": 6.5017921146953405e-06,
+      "loss": 0.2417,
+      "step": 136860
+    },
+    {
+      "epoch": 883.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9601426720619202,
+      "eval_runtime": 0.8763,
+      "eval_samples_per_second": 229.369,
+      "eval_steps_per_second": 7.988,
+      "step": 136865
+    },
+    {
+      "epoch": 883.03,
+      "learning_rate": 6.4982078853046595e-06,
+      "loss": 0.1761,
+      "step": 136870
+    },
+    {
+      "epoch": 883.1,
+      "learning_rate": 6.494623655913979e-06,
+      "loss": 0.2248,
+      "step": 136880
+    },
+    {
+      "epoch": 883.16,
+      "learning_rate": 6.491039426523298e-06,
+      "loss": 0.2094,
+      "step": 136890
+    },
+    {
+      "epoch": 883.23,
+      "learning_rate": 6.487455197132617e-06,
+      "loss": 0.2212,
+      "step": 136900
+    },
+    {
+      "epoch": 883.29,
+      "learning_rate": 6.483870967741936e-06,
+      "loss": 0.1937,
+      "step": 136910
+    },
+    {
+      "epoch": 883.35,
+      "learning_rate": 6.480286738351255e-06,
+      "loss": 0.1678,
+      "step": 136920
+    },
+    {
+      "epoch": 883.42,
+      "learning_rate": 6.476702508960574e-06,
+      "loss": 0.1933,
+      "step": 136930
+    },
+    {
+      "epoch": 883.48,
+      "learning_rate": 6.473118279569892e-06,
+      "loss": 0.1475,
+      "step": 136940
+    },
+    {
+      "epoch": 883.55,
+      "learning_rate": 6.469534050179211e-06,
+      "loss": 0.1942,
+      "step": 136950
+    },
+    {
+      "epoch": 883.61,
+      "learning_rate": 6.46594982078853e-06,
+      "loss": 0.2108,
+      "step": 136960
+    },
+    {
+      "epoch": 883.68,
+      "learning_rate": 6.462365591397849e-06,
+      "loss": 0.2229,
+      "step": 136970
+    },
+    {
+      "epoch": 883.74,
+      "learning_rate": 6.458781362007168e-06,
+      "loss": 0.2309,
+      "step": 136980
+    },
+    {
+      "epoch": 883.8,
+      "learning_rate": 6.455197132616487e-06,
+      "loss": 0.1965,
+      "step": 136990
+    },
+    {
+      "epoch": 883.87,
+      "learning_rate": 6.451612903225806e-06,
+      "loss": 0.2148,
+      "step": 137000
+    },
+    {
+      "epoch": 883.93,
+      "learning_rate": 6.4480286738351254e-06,
+      "loss": 0.1761,
+      "step": 137010
+    },
+    {
+      "epoch": 884.0,
+      "learning_rate": 6.4444444444444445e-06,
+      "loss": 0.1801,
+      "step": 137020
+    },
+    {
+      "epoch": 884.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9594831466674805,
+      "eval_runtime": 0.9421,
+      "eval_samples_per_second": 213.355,
+      "eval_steps_per_second": 7.43,
+      "step": 137020
+    },
+    {
+      "epoch": 884.06,
+      "learning_rate": 6.4408602150537635e-06,
+      "loss": 0.2274,
+      "step": 137030
+    },
+    {
+      "epoch": 884.13,
+      "learning_rate": 6.437275985663083e-06,
+      "loss": 0.1928,
+      "step": 137040
+    },
+    {
+      "epoch": 884.19,
+      "learning_rate": 6.433691756272402e-06,
+      "loss": 0.1851,
+      "step": 137050
+    },
+    {
+      "epoch": 884.26,
+      "learning_rate": 6.430107526881721e-06,
+      "loss": 0.2164,
+      "step": 137060
+    },
+    {
+      "epoch": 884.32,
+      "learning_rate": 6.42652329749104e-06,
+      "loss": 0.2301,
+      "step": 137070
+    },
+    {
+      "epoch": 884.39,
+      "learning_rate": 6.422939068100359e-06,
+      "loss": 0.2203,
+      "step": 137080
+    },
+    {
+      "epoch": 884.45,
+      "learning_rate": 6.419354838709678e-06,
+      "loss": 0.2083,
+      "step": 137090
+    },
+    {
+      "epoch": 884.51,
+      "learning_rate": 6.415770609318997e-06,
+      "loss": 0.1856,
+      "step": 137100
+    },
+    {
+      "epoch": 884.58,
+      "learning_rate": 6.412186379928316e-06,
+      "loss": 0.1282,
+      "step": 137110
+    },
+    {
+      "epoch": 884.64,
+      "learning_rate": 6.408602150537635e-06,
+      "loss": 0.239,
+      "step": 137120
+    },
+    {
+      "epoch": 884.71,
+      "learning_rate": 6.405017921146954e-06,
+      "loss": 0.2705,
+      "step": 137130
+    },
+    {
+      "epoch": 884.77,
+      "learning_rate": 6.401433691756273e-06,
+      "loss": 0.1868,
+      "step": 137140
+    },
+    {
+      "epoch": 884.84,
+      "learning_rate": 6.397849462365592e-06,
+      "loss": 0.2236,
+      "step": 137150
+    },
+    {
+      "epoch": 884.9,
+      "learning_rate": 6.394265232974911e-06,
+      "loss": 0.2337,
+      "step": 137160
+    },
+    {
+      "epoch": 884.96,
+      "learning_rate": 6.39068100358423e-06,
+      "loss": 0.223,
+      "step": 137170
+    },
+    {
+      "epoch": 885.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9719402194023132,
+      "eval_runtime": 0.8779,
+      "eval_samples_per_second": 228.963,
+      "eval_steps_per_second": 7.974,
+      "step": 137175
+    },
+    {
+      "epoch": 885.03,
+      "learning_rate": 6.387096774193549e-06,
+      "loss": 0.2672,
+      "step": 137180
+    },
+    {
+      "epoch": 885.1,
+      "learning_rate": 6.3835125448028675e-06,
+      "loss": 0.2581,
+      "step": 137190
+    },
+    {
+      "epoch": 885.16,
+      "learning_rate": 6.3799283154121866e-06,
+      "loss": 0.1899,
+      "step": 137200
+    },
+    {
+      "epoch": 885.23,
+      "learning_rate": 6.376344086021506e-06,
+      "loss": 0.214,
+      "step": 137210
+    },
+    {
+      "epoch": 885.29,
+      "learning_rate": 6.372759856630825e-06,
+      "loss": 0.2274,
+      "step": 137220
+    },
+    {
+      "epoch": 885.35,
+      "learning_rate": 6.369175627240144e-06,
+      "loss": 0.1919,
+      "step": 137230
+    },
+    {
+      "epoch": 885.42,
+      "learning_rate": 6.365591397849463e-06,
+      "loss": 0.22,
+      "step": 137240
+    },
+    {
+      "epoch": 885.48,
+      "learning_rate": 6.362007168458782e-06,
+      "loss": 0.1748,
+      "step": 137250
+    },
+    {
+      "epoch": 885.55,
+      "learning_rate": 6.358422939068101e-06,
+      "loss": 0.1956,
+      "step": 137260
+    },
+    {
+      "epoch": 885.61,
+      "learning_rate": 6.35483870967742e-06,
+      "loss": 0.225,
+      "step": 137270
+    },
+    {
+      "epoch": 885.68,
+      "learning_rate": 6.351254480286739e-06,
+      "loss": 0.2059,
+      "step": 137280
+    },
+    {
+      "epoch": 885.74,
+      "learning_rate": 6.347670250896058e-06,
+      "loss": 0.1614,
+      "step": 137290
+    },
+    {
+      "epoch": 885.8,
+      "learning_rate": 6.344086021505377e-06,
+      "loss": 0.2134,
+      "step": 137300
+    },
+    {
+      "epoch": 885.87,
+      "learning_rate": 6.340501792114696e-06,
+      "loss": 0.2336,
+      "step": 137310
+    },
+    {
+      "epoch": 885.93,
+      "learning_rate": 6.336917562724015e-06,
+      "loss": 0.1961,
+      "step": 137320
+    },
+    {
+      "epoch": 886.0,
+      "learning_rate": 6.333333333333334e-06,
+      "loss": 0.2521,
+      "step": 137330
+    },
+    {
+      "epoch": 886.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9420933723449707,
+      "eval_runtime": 0.8771,
+      "eval_samples_per_second": 229.174,
+      "eval_steps_per_second": 7.981,
+      "step": 137330
+    },
+    {
+      "epoch": 886.06,
+      "learning_rate": 6.329749103942653e-06,
+      "loss": 0.1782,
+      "step": 137340
+    },
+    {
+      "epoch": 886.13,
+      "learning_rate": 6.326164874551972e-06,
+      "loss": 0.2568,
+      "step": 137350
+    },
+    {
+      "epoch": 886.19,
+      "learning_rate": 6.322580645161291e-06,
+      "loss": 0.2145,
+      "step": 137360
+    },
+    {
+      "epoch": 886.26,
+      "learning_rate": 6.3189964157706105e-06,
+      "loss": 0.2168,
+      "step": 137370
+    },
+    {
+      "epoch": 886.32,
+      "learning_rate": 6.3154121863799295e-06,
+      "loss": 0.2442,
+      "step": 137380
+    },
+    {
+      "epoch": 886.39,
+      "learning_rate": 6.3118279569892486e-06,
+      "loss": 0.2839,
+      "step": 137390
+    },
+    {
+      "epoch": 886.45,
+      "learning_rate": 6.308243727598568e-06,
+      "loss": 0.2219,
+      "step": 137400
+    },
+    {
+      "epoch": 886.51,
+      "learning_rate": 6.304659498207885e-06,
+      "loss": 0.1704,
+      "step": 137410
+    },
+    {
+      "epoch": 886.58,
+      "learning_rate": 6.301075268817204e-06,
+      "loss": 0.2146,
+      "step": 137420
+    },
+    {
+      "epoch": 886.64,
+      "learning_rate": 6.297491039426523e-06,
+      "loss": 0.2463,
+      "step": 137430
+    },
+    {
+      "epoch": 886.71,
+      "learning_rate": 6.293906810035842e-06,
+      "loss": 0.2032,
+      "step": 137440
+    },
+    {
+      "epoch": 886.77,
+      "learning_rate": 6.290322580645161e-06,
+      "loss": 0.2074,
+      "step": 137450
+    },
+    {
+      "epoch": 886.84,
+      "learning_rate": 6.28673835125448e-06,
+      "loss": 0.1881,
+      "step": 137460
+    },
+    {
+      "epoch": 886.9,
+      "learning_rate": 6.283154121863799e-06,
+      "loss": 0.2275,
+      "step": 137470
+    },
+    {
+      "epoch": 886.96,
+      "learning_rate": 6.279569892473118e-06,
+      "loss": 0.2106,
+      "step": 137480
+    },
+    {
+      "epoch": 887.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9469612240791321,
+      "eval_runtime": 0.8796,
+      "eval_samples_per_second": 228.501,
+      "eval_steps_per_second": 7.958,
+      "step": 137485
+    },
+    {
+      "epoch": 887.03,
+      "learning_rate": 6.275985663082437e-06,
+      "loss": 0.2222,
+      "step": 137490
+    },
+    {
+      "epoch": 887.1,
+      "learning_rate": 6.2724014336917564e-06,
+      "loss": 0.1769,
+      "step": 137500
+    },
+    {
+      "epoch": 887.16,
+      "learning_rate": 6.2688172043010755e-06,
+      "loss": 0.1838,
+      "step": 137510
+    },
+    {
+      "epoch": 887.23,
+      "learning_rate": 6.2652329749103945e-06,
+      "loss": 0.1923,
+      "step": 137520
+    },
+    {
+      "epoch": 887.29,
+      "learning_rate": 6.261648745519713e-06,
+      "loss": 0.1895,
+      "step": 137530
+    },
+    {
+      "epoch": 887.35,
+      "learning_rate": 6.258064516129032e-06,
+      "loss": 0.2123,
+      "step": 137540
+    },
+    {
+      "epoch": 887.42,
+      "learning_rate": 6.254480286738351e-06,
+      "loss": 0.2121,
+      "step": 137550
+    },
+    {
+      "epoch": 887.48,
+      "learning_rate": 6.25089605734767e-06,
+      "loss": 0.173,
+      "step": 137560
+    },
+    {
+      "epoch": 887.55,
+      "learning_rate": 6.247311827956989e-06,
+      "loss": 0.2272,
+      "step": 137570
+    },
+    {
+      "epoch": 887.61,
+      "learning_rate": 6.243727598566308e-06,
+      "loss": 0.1872,
+      "step": 137580
+    },
+    {
+      "epoch": 887.68,
+      "learning_rate": 6.240143369175627e-06,
+      "loss": 0.183,
+      "step": 137590
+    },
+    {
+      "epoch": 887.74,
+      "learning_rate": 6.236559139784946e-06,
+      "loss": 0.2108,
+      "step": 137600
+    },
+    {
+      "epoch": 887.8,
+      "learning_rate": 6.232974910394265e-06,
+      "loss": 0.2024,
+      "step": 137610
+    },
+    {
+      "epoch": 887.87,
+      "learning_rate": 6.229390681003584e-06,
+      "loss": 0.205,
+      "step": 137620
+    },
+    {
+      "epoch": 887.93,
+      "learning_rate": 6.225806451612903e-06,
+      "loss": 0.1814,
+      "step": 137630
+    },
+    {
+      "epoch": 888.0,
+      "learning_rate": 6.222222222222222e-06,
+      "loss": 0.2272,
+      "step": 137640
+    },
+    {
+      "epoch": 888.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9220461249351501,
+      "eval_runtime": 0.8682,
+      "eval_samples_per_second": 231.503,
+      "eval_steps_per_second": 8.062,
+      "step": 137640
+    },
+    {
+      "epoch": 888.06,
+      "learning_rate": 6.218637992831541e-06,
+      "loss": 0.1807,
+      "step": 137650
+    },
+    {
+      "epoch": 888.13,
+      "learning_rate": 6.21505376344086e-06,
+      "loss": 0.1979,
+      "step": 137660
+    },
+    {
+      "epoch": 888.19,
+      "learning_rate": 6.2114695340501795e-06,
+      "loss": 0.27,
+      "step": 137670
+    },
+    {
+      "epoch": 888.26,
+      "learning_rate": 6.2078853046594985e-06,
+      "loss": 0.1762,
+      "step": 137680
+    },
+    {
+      "epoch": 888.32,
+      "learning_rate": 6.204301075268818e-06,
+      "loss": 0.1947,
+      "step": 137690
+    },
+    {
+      "epoch": 888.39,
+      "learning_rate": 6.200716845878137e-06,
+      "loss": 0.177,
+      "step": 137700
+    },
+    {
+      "epoch": 888.45,
+      "learning_rate": 6.197132616487456e-06,
+      "loss": 0.2265,
+      "step": 137710
+    },
+    {
+      "epoch": 888.51,
+      "learning_rate": 6.193548387096775e-06,
+      "loss": 0.2003,
+      "step": 137720
+    },
+    {
+      "epoch": 888.58,
+      "learning_rate": 6.189964157706094e-06,
+      "loss": 0.2164,
+      "step": 137730
+    },
+    {
+      "epoch": 888.64,
+      "learning_rate": 6.186379928315413e-06,
+      "loss": 0.2228,
+      "step": 137740
+    },
+    {
+      "epoch": 888.71,
+      "learning_rate": 6.182795698924732e-06,
+      "loss": 0.1987,
+      "step": 137750
+    },
+    {
+      "epoch": 888.77,
+      "learning_rate": 6.179211469534051e-06,
+      "loss": 0.2578,
+      "step": 137760
+    },
+    {
+      "epoch": 888.84,
+      "learning_rate": 6.17562724014337e-06,
+      "loss": 0.1896,
+      "step": 137770
+    },
+    {
+      "epoch": 888.9,
+      "learning_rate": 6.172043010752688e-06,
+      "loss": 0.2135,
+      "step": 137780
+    },
+    {
+      "epoch": 888.96,
+      "learning_rate": 6.168458781362007e-06,
+      "loss": 0.1732,
+      "step": 137790
+    },
+    {
+      "epoch": 889.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9395360946655273,
+      "eval_runtime": 0.872,
+      "eval_samples_per_second": 230.504,
+      "eval_steps_per_second": 8.028,
+      "step": 137795
+    },
+    {
+      "epoch": 889.03,
+      "learning_rate": 6.164874551971326e-06,
+      "loss": 0.2515,
+      "step": 137800
+    },
+    {
+      "epoch": 889.1,
+      "learning_rate": 6.161290322580645e-06,
+      "loss": 0.2422,
+      "step": 137810
+    },
+    {
+      "epoch": 889.16,
+      "learning_rate": 6.157706093189964e-06,
+      "loss": 0.2234,
+      "step": 137820
+    },
+    {
+      "epoch": 889.23,
+      "learning_rate": 6.1541218637992835e-06,
+      "loss": 0.2276,
+      "step": 137830
+    },
+    {
+      "epoch": 889.29,
+      "learning_rate": 6.1505376344086025e-06,
+      "loss": 0.1759,
+      "step": 137840
+    },
+    {
+      "epoch": 889.35,
+      "learning_rate": 6.1469534050179216e-06,
+      "loss": 0.1814,
+      "step": 137850
+    },
+    {
+      "epoch": 889.42,
+      "learning_rate": 6.143369175627241e-06,
+      "loss": 0.2271,
+      "step": 137860
+    },
+    {
+      "epoch": 889.48,
+      "learning_rate": 6.13978494623656e-06,
+      "loss": 0.2057,
+      "step": 137870
+    },
+    {
+      "epoch": 889.55,
+      "learning_rate": 6.136200716845879e-06,
+      "loss": 0.204,
+      "step": 137880
+    },
+    {
+      "epoch": 889.61,
+      "learning_rate": 6.132616487455197e-06,
+      "loss": 0.1993,
+      "step": 137890
+    },
+    {
+      "epoch": 889.68,
+      "learning_rate": 6.129032258064516e-06,
+      "loss": 0.1658,
+      "step": 137900
+    },
+    {
+      "epoch": 889.74,
+      "learning_rate": 6.125448028673835e-06,
+      "loss": 0.1799,
+      "step": 137910
+    },
+    {
+      "epoch": 889.8,
+      "learning_rate": 6.121863799283154e-06,
+      "loss": 0.1649,
+      "step": 137920
+    },
+    {
+      "epoch": 889.87,
+      "learning_rate": 6.118279569892473e-06,
+      "loss": 0.2077,
+      "step": 137930
+    },
+    {
+      "epoch": 889.93,
+      "learning_rate": 6.114695340501792e-06,
+      "loss": 0.1629,
+      "step": 137940
+    },
+    {
+      "epoch": 890.0,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 0.215,
+      "step": 137950
+    },
+    {
+      "epoch": 890.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.932560384273529,
+      "eval_runtime": 0.8621,
+      "eval_samples_per_second": 233.154,
+      "eval_steps_per_second": 8.12,
+      "step": 137950
+    },
+    {
+      "epoch": 890.06,
+      "learning_rate": 6.10752688172043e-06,
+      "loss": 0.1616,
+      "step": 137960
+    },
+    {
+      "epoch": 890.13,
+      "learning_rate": 6.103942652329749e-06,
+      "loss": 0.1695,
+      "step": 137970
+    },
+    {
+      "epoch": 890.19,
+      "learning_rate": 6.100358422939068e-06,
+      "loss": 0.2092,
+      "step": 137980
+    },
+    {
+      "epoch": 890.26,
+      "learning_rate": 6.0967741935483874e-06,
+      "loss": 0.1796,
+      "step": 137990
+    },
+    {
+      "epoch": 890.32,
+      "learning_rate": 6.0931899641577065e-06,
+      "loss": 0.2089,
+      "step": 138000
+    },
+    {
+      "epoch": 890.39,
+      "learning_rate": 6.0896057347670255e-06,
+      "loss": 0.2376,
+      "step": 138010
+    },
+    {
+      "epoch": 890.45,
+      "learning_rate": 6.086021505376345e-06,
+      "loss": 0.2006,
+      "step": 138020
+    },
+    {
+      "epoch": 890.51,
+      "learning_rate": 6.082437275985664e-06,
+      "loss": 0.2143,
+      "step": 138030
+    },
+    {
+      "epoch": 890.58,
+      "learning_rate": 6.078853046594983e-06,
+      "loss": 0.1997,
+      "step": 138040
+    },
+    {
+      "epoch": 890.64,
+      "learning_rate": 6.075268817204302e-06,
+      "loss": 0.1843,
+      "step": 138050
+    },
+    {
+      "epoch": 890.71,
+      "learning_rate": 6.071684587813621e-06,
+      "loss": 0.2197,
+      "step": 138060
+    },
+    {
+      "epoch": 890.77,
+      "learning_rate": 6.068100358422939e-06,
+      "loss": 0.2142,
+      "step": 138070
+    },
+    {
+      "epoch": 890.84,
+      "learning_rate": 6.064516129032258e-06,
+      "loss": 0.2071,
+      "step": 138080
+    },
+    {
+      "epoch": 890.9,
+      "learning_rate": 6.060931899641577e-06,
+      "loss": 0.2046,
+      "step": 138090
+    },
+    {
+      "epoch": 890.96,
+      "learning_rate": 6.057347670250896e-06,
+      "loss": 0.2064,
+      "step": 138100
+    },
+    {
+      "epoch": 891.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9463999271392822,
+      "eval_runtime": 0.8713,
+      "eval_samples_per_second": 230.696,
+      "eval_steps_per_second": 8.034,
+      "step": 138105
+    },
+    {
+      "epoch": 891.03,
+      "learning_rate": 6.053763440860215e-06,
+      "loss": 0.1946,
+      "step": 138110
+    },
+    {
+      "epoch": 891.1,
+      "learning_rate": 6.050179211469534e-06,
+      "loss": 0.2071,
+      "step": 138120
+    },
+    {
+      "epoch": 891.16,
+      "learning_rate": 6.046594982078853e-06,
+      "loss": 0.2211,
+      "step": 138130
+    },
+    {
+      "epoch": 891.23,
+      "learning_rate": 6.043010752688172e-06,
+      "loss": 0.2339,
+      "step": 138140
+    },
+    {
+      "epoch": 891.29,
+      "learning_rate": 6.0394265232974914e-06,
+      "loss": 0.1918,
+      "step": 138150
+    },
+    {
+      "epoch": 891.35,
+      "learning_rate": 6.0358422939068105e-06,
+      "loss": 0.162,
+      "step": 138160
+    },
+    {
+      "epoch": 891.42,
+      "learning_rate": 6.0322580645161295e-06,
+      "loss": 0.2358,
+      "step": 138170
+    },
+    {
+      "epoch": 891.48,
+      "learning_rate": 6.028673835125449e-06,
+      "loss": 0.175,
+      "step": 138180
+    },
+    {
+      "epoch": 891.55,
+      "learning_rate": 6.025089605734768e-06,
+      "loss": 0.2325,
+      "step": 138190
+    },
+    {
+      "epoch": 891.61,
+      "learning_rate": 6.021505376344087e-06,
+      "loss": 0.2489,
+      "step": 138200
+    },
+    {
+      "epoch": 891.68,
+      "learning_rate": 6.017921146953405e-06,
+      "loss": 0.2147,
+      "step": 138210
+    },
+    {
+      "epoch": 891.74,
+      "learning_rate": 6.014336917562724e-06,
+      "loss": 0.161,
+      "step": 138220
+    },
+    {
+      "epoch": 891.8,
+      "learning_rate": 6.010752688172043e-06,
+      "loss": 0.1935,
+      "step": 138230
+    },
+    {
+      "epoch": 891.87,
+      "learning_rate": 6.007168458781362e-06,
+      "loss": 0.183,
+      "step": 138240
+    },
+    {
+      "epoch": 891.93,
+      "learning_rate": 6.003584229390681e-06,
+      "loss": 0.2007,
+      "step": 138250
+    },
+    {
+      "epoch": 892.0,
+      "learning_rate": 6e-06,
+      "loss": 0.1928,
+      "step": 138260
+    },
+    {
+      "epoch": 892.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.925355076789856,
+      "eval_runtime": 0.8756,
+      "eval_samples_per_second": 229.558,
+      "eval_steps_per_second": 7.995,
+      "step": 138260
+    },
+    {
+      "epoch": 892.06,
+      "learning_rate": 5.996415770609319e-06,
+      "loss": 0.257,
+      "step": 138270
+    },
+    {
+      "epoch": 892.13,
+      "learning_rate": 5.992831541218638e-06,
+      "loss": 0.1647,
+      "step": 138280
+    },
+    {
+      "epoch": 892.19,
+      "learning_rate": 5.989247311827957e-06,
+      "loss": 0.2436,
+      "step": 138290
+    },
+    {
+      "epoch": 892.26,
+      "learning_rate": 5.985663082437276e-06,
+      "loss": 0.2051,
+      "step": 138300
+    },
+    {
+      "epoch": 892.32,
+      "learning_rate": 5.982078853046595e-06,
+      "loss": 0.166,
+      "step": 138310
+    },
+    {
+      "epoch": 892.39,
+      "learning_rate": 5.9784946236559145e-06,
+      "loss": 0.211,
+      "step": 138320
+    },
+    {
+      "epoch": 892.45,
+      "learning_rate": 5.9749103942652335e-06,
+      "loss": 0.1674,
+      "step": 138330
+    },
+    {
+      "epoch": 892.51,
+      "learning_rate": 5.9713261648745526e-06,
+      "loss": 0.2246,
+      "step": 138340
+    },
+    {
+      "epoch": 892.58,
+      "learning_rate": 5.967741935483872e-06,
+      "loss": 0.2028,
+      "step": 138350
+    },
+    {
+      "epoch": 892.64,
+      "learning_rate": 5.96415770609319e-06,
+      "loss": 0.2011,
+      "step": 138360
+    },
+    {
+      "epoch": 892.71,
+      "learning_rate": 5.960573476702509e-06,
+      "loss": 0.2073,
+      "step": 138370
+    },
+    {
+      "epoch": 892.77,
+      "learning_rate": 5.956989247311828e-06,
+      "loss": 0.1778,
+      "step": 138380
+    },
+    {
+      "epoch": 892.84,
+      "learning_rate": 5.953405017921147e-06,
+      "loss": 0.1791,
+      "step": 138390
+    },
+    {
+      "epoch": 892.9,
+      "learning_rate": 5.949820788530466e-06,
+      "loss": 0.2188,
+      "step": 138400
+    },
+    {
+      "epoch": 892.96,
+      "learning_rate": 5.946236559139785e-06,
+      "loss": 0.2203,
+      "step": 138410
+    },
+    {
+      "epoch": 893.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9162575006484985,
+      "eval_runtime": 0.8782,
+      "eval_samples_per_second": 228.884,
+      "eval_steps_per_second": 7.971,
+      "step": 138415
+    },
+    {
+      "epoch": 893.03,
+      "learning_rate": 5.942652329749104e-06,
+      "loss": 0.2069,
+      "step": 138420
+    },
+    {
+      "epoch": 893.1,
+      "learning_rate": 5.939068100358423e-06,
+      "loss": 0.2152,
+      "step": 138430
+    },
+    {
+      "epoch": 893.16,
+      "learning_rate": 5.935483870967742e-06,
+      "loss": 0.2485,
+      "step": 138440
+    },
+    {
+      "epoch": 893.23,
+      "learning_rate": 5.931899641577061e-06,
+      "loss": 0.2181,
+      "step": 138450
+    },
+    {
+      "epoch": 893.29,
+      "learning_rate": 5.92831541218638e-06,
+      "loss": 0.2072,
+      "step": 138460
+    },
+    {
+      "epoch": 893.35,
+      "learning_rate": 5.924731182795699e-06,
+      "loss": 0.2018,
+      "step": 138470
+    },
+    {
+      "epoch": 893.42,
+      "learning_rate": 5.9211469534050184e-06,
+      "loss": 0.2156,
+      "step": 138480
+    },
+    {
+      "epoch": 893.48,
+      "learning_rate": 5.9175627240143375e-06,
+      "loss": 0.1973,
+      "step": 138490
+    },
+    {
+      "epoch": 893.55,
+      "learning_rate": 5.9139784946236566e-06,
+      "loss": 0.2067,
+      "step": 138500
+    },
+    {
+      "epoch": 893.61,
+      "learning_rate": 5.910394265232976e-06,
+      "loss": 0.2203,
+      "step": 138510
+    },
+    {
+      "epoch": 893.68,
+      "learning_rate": 5.906810035842295e-06,
+      "loss": 0.1789,
+      "step": 138520
+    },
+    {
+      "epoch": 893.74,
+      "learning_rate": 5.903225806451613e-06,
+      "loss": 0.2012,
+      "step": 138530
+    },
+    {
+      "epoch": 893.8,
+      "learning_rate": 5.899641577060932e-06,
+      "loss": 0.1808,
+      "step": 138540
+    },
+    {
+      "epoch": 893.87,
+      "learning_rate": 5.896057347670251e-06,
+      "loss": 0.215,
+      "step": 138550
+    },
+    {
+      "epoch": 893.93,
+      "learning_rate": 5.89247311827957e-06,
+      "loss": 0.2077,
+      "step": 138560
+    },
+    {
+      "epoch": 894.0,
+      "learning_rate": 5.888888888888889e-06,
+      "loss": 0.2014,
+      "step": 138570
+    },
+    {
+      "epoch": 894.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9064329266548157,
+      "eval_runtime": 0.8878,
+      "eval_samples_per_second": 226.399,
+      "eval_steps_per_second": 7.885,
+      "step": 138570
+    },
+    {
+      "epoch": 894.06,
+      "learning_rate": 5.885304659498208e-06,
+      "loss": 0.2171,
+      "step": 138580
+    },
+    {
+      "epoch": 894.13,
+      "learning_rate": 5.881720430107527e-06,
+      "loss": 0.2212,
+      "step": 138590
+    },
+    {
+      "epoch": 894.19,
+      "learning_rate": 5.878136200716846e-06,
+      "loss": 0.2592,
+      "step": 138600
+    },
+    {
+      "epoch": 894.26,
+      "learning_rate": 5.874551971326165e-06,
+      "loss": 0.2508,
+      "step": 138610
+    },
+    {
+      "epoch": 894.32,
+      "learning_rate": 5.870967741935484e-06,
+      "loss": 0.224,
+      "step": 138620
+    },
+    {
+      "epoch": 894.39,
+      "learning_rate": 5.867383512544803e-06,
+      "loss": 0.2011,
+      "step": 138630
+    },
+    {
+      "epoch": 894.45,
+      "learning_rate": 5.8637992831541224e-06,
+      "loss": 0.2426,
+      "step": 138640
+    },
+    {
+      "epoch": 894.51,
+      "learning_rate": 5.8602150537634415e-06,
+      "loss": 0.1956,
+      "step": 138650
+    },
+    {
+      "epoch": 894.58,
+      "learning_rate": 5.85663082437276e-06,
+      "loss": 0.2035,
+      "step": 138660
+    },
+    {
+      "epoch": 894.64,
+      "learning_rate": 5.853046594982079e-06,
+      "loss": 0.1946,
+      "step": 138670
+    },
+    {
+      "epoch": 894.71,
+      "learning_rate": 5.849462365591398e-06,
+      "loss": 0.1885,
+      "step": 138680
+    },
+    {
+      "epoch": 894.77,
+      "learning_rate": 5.845878136200717e-06,
+      "loss": 0.2127,
+      "step": 138690
+    },
+    {
+      "epoch": 894.84,
+      "learning_rate": 5.842293906810036e-06,
+      "loss": 0.237,
+      "step": 138700
+    },
+    {
+      "epoch": 894.9,
+      "learning_rate": 5.838709677419355e-06,
+      "loss": 0.22,
+      "step": 138710
+    },
+    {
+      "epoch": 894.96,
+      "learning_rate": 5.835125448028674e-06,
+      "loss": 0.2014,
+      "step": 138720
+    },
+    {
+      "epoch": 895.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9073975682258606,
+      "eval_runtime": 0.8623,
+      "eval_samples_per_second": 233.084,
+      "eval_steps_per_second": 8.117,
+      "step": 138725
+    },
+    {
+      "epoch": 895.03,
+      "learning_rate": 5.831541218637993e-06,
+      "loss": 0.2467,
+      "step": 138730
+    },
+    {
+      "epoch": 895.1,
+      "learning_rate": 5.827956989247312e-06,
+      "loss": 0.2108,
+      "step": 138740
+    },
+    {
+      "epoch": 895.16,
+      "learning_rate": 5.824372759856631e-06,
+      "loss": 0.1661,
+      "step": 138750
+    },
+    {
+      "epoch": 895.23,
+      "learning_rate": 5.82078853046595e-06,
+      "loss": 0.2176,
+      "step": 138760
+    },
+    {
+      "epoch": 895.29,
+      "learning_rate": 5.817204301075269e-06,
+      "loss": 0.1924,
+      "step": 138770
+    },
+    {
+      "epoch": 895.35,
+      "learning_rate": 5.813620071684588e-06,
+      "loss": 0.2198,
+      "step": 138780
+    },
+    {
+      "epoch": 895.42,
+      "learning_rate": 5.810035842293907e-06,
+      "loss": 0.2514,
+      "step": 138790
+    },
+    {
+      "epoch": 895.48,
+      "learning_rate": 5.806451612903226e-06,
+      "loss": 0.1912,
+      "step": 138800
+    },
+    {
+      "epoch": 895.55,
+      "learning_rate": 5.8028673835125455e-06,
+      "loss": 0.1469,
+      "step": 138810
+    },
+    {
+      "epoch": 895.61,
+      "learning_rate": 5.7992831541218645e-06,
+      "loss": 0.184,
+      "step": 138820
+    },
+    {
+      "epoch": 895.68,
+      "learning_rate": 5.7956989247311836e-06,
+      "loss": 0.1814,
+      "step": 138830
+    },
+    {
+      "epoch": 895.74,
+      "learning_rate": 5.792114695340503e-06,
+      "loss": 0.2032,
+      "step": 138840
+    },
+    {
+      "epoch": 895.8,
+      "learning_rate": 5.788530465949821e-06,
+      "loss": 0.1672,
+      "step": 138850
+    },
+    {
+      "epoch": 895.87,
+      "learning_rate": 5.78494623655914e-06,
+      "loss": 0.2264,
+      "step": 138860
+    },
+    {
+      "epoch": 895.93,
+      "learning_rate": 5.781362007168459e-06,
+      "loss": 0.217,
+      "step": 138870
+    },
+    {
+      "epoch": 896.0,
+      "learning_rate": 5.777777777777778e-06,
+      "loss": 0.2448,
+      "step": 138880
+    },
+    {
+      "epoch": 896.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9198402762413025,
+      "eval_runtime": 0.8932,
+      "eval_samples_per_second": 225.046,
+      "eval_steps_per_second": 7.837,
+      "step": 138880
+    },
+    {
+      "epoch": 896.06,
+      "learning_rate": 5.774193548387097e-06,
+      "loss": 0.2438,
+      "step": 138890
+    },
+    {
+      "epoch": 896.13,
+      "learning_rate": 5.770609318996416e-06,
+      "loss": 0.229,
+      "step": 138900
+    },
+    {
+      "epoch": 896.19,
+      "learning_rate": 5.767025089605735e-06,
+      "loss": 0.1753,
+      "step": 138910
+    },
+    {
+      "epoch": 896.26,
+      "learning_rate": 5.763440860215054e-06,
+      "loss": 0.1879,
+      "step": 138920
+    },
+    {
+      "epoch": 896.32,
+      "learning_rate": 5.759856630824373e-06,
+      "loss": 0.196,
+      "step": 138930
+    },
+    {
+      "epoch": 896.39,
+      "learning_rate": 5.756272401433692e-06,
+      "loss": 0.2137,
+      "step": 138940
+    },
+    {
+      "epoch": 896.45,
+      "learning_rate": 5.7526881720430105e-06,
+      "loss": 0.2031,
+      "step": 138950
+    },
+    {
+      "epoch": 896.51,
+      "learning_rate": 5.7491039426523296e-06,
+      "loss": 0.1762,
+      "step": 138960
+    },
+    {
+      "epoch": 896.58,
+      "learning_rate": 5.745519713261649e-06,
+      "loss": 0.2068,
+      "step": 138970
+    },
+    {
+      "epoch": 896.64,
+      "learning_rate": 5.741935483870968e-06,
+      "loss": 0.1874,
+      "step": 138980
+    },
+    {
+      "epoch": 896.71,
+      "learning_rate": 5.738351254480287e-06,
+      "loss": 0.1779,
+      "step": 138990
+    },
+    {
+      "epoch": 896.77,
+      "learning_rate": 5.734767025089606e-06,
+      "loss": 0.1835,
+      "step": 139000
+    },
+    {
+      "epoch": 896.84,
+      "learning_rate": 5.731182795698925e-06,
+      "loss": 0.2201,
+      "step": 139010
+    },
+    {
+      "epoch": 896.9,
+      "learning_rate": 5.727598566308244e-06,
+      "loss": 0.2064,
+      "step": 139020
+    },
+    {
+      "epoch": 896.96,
+      "learning_rate": 5.724014336917563e-06,
+      "loss": 0.1646,
+      "step": 139030
+    },
+    {
+      "epoch": 897.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9195365905761719,
+      "eval_runtime": 0.8945,
+      "eval_samples_per_second": 224.698,
+      "eval_steps_per_second": 7.825,
+      "step": 139035
+    },
+    {
+      "epoch": 897.03,
+      "learning_rate": 5.720430107526882e-06,
+      "loss": 0.1827,
+      "step": 139040
+    },
+    {
+      "epoch": 897.1,
+      "learning_rate": 5.716845878136201e-06,
+      "loss": 0.2234,
+      "step": 139050
+    },
+    {
+      "epoch": 897.16,
+      "learning_rate": 5.71326164874552e-06,
+      "loss": 0.2096,
+      "step": 139060
+    },
+    {
+      "epoch": 897.23,
+      "learning_rate": 5.709677419354839e-06,
+      "loss": 0.1914,
+      "step": 139070
+    },
+    {
+      "epoch": 897.29,
+      "learning_rate": 5.706093189964158e-06,
+      "loss": 0.2164,
+      "step": 139080
+    },
+    {
+      "epoch": 897.35,
+      "learning_rate": 5.702508960573477e-06,
+      "loss": 0.1814,
+      "step": 139090
+    },
+    {
+      "epoch": 897.42,
+      "learning_rate": 5.698924731182796e-06,
+      "loss": 0.2002,
+      "step": 139100
+    },
+    {
+      "epoch": 897.48,
+      "learning_rate": 5.695340501792115e-06,
+      "loss": 0.197,
+      "step": 139110
+    },
+    {
+      "epoch": 897.55,
+      "learning_rate": 5.691756272401434e-06,
+      "loss": 0.1699,
+      "step": 139120
+    },
+    {
+      "epoch": 897.61,
+      "learning_rate": 5.6881720430107534e-06,
+      "loss": 0.2391,
+      "step": 139130
+    },
+    {
+      "epoch": 897.68,
+      "learning_rate": 5.6845878136200725e-06,
+      "loss": 0.2345,
+      "step": 139140
+    },
+    {
+      "epoch": 897.74,
+      "learning_rate": 5.6810035842293915e-06,
+      "loss": 0.2503,
+      "step": 139150
+    },
+    {
+      "epoch": 897.8,
+      "learning_rate": 5.677419354838711e-06,
+      "loss": 0.1825,
+      "step": 139160
+    },
+    {
+      "epoch": 897.87,
+      "learning_rate": 5.673835125448029e-06,
+      "loss": 0.1935,
+      "step": 139170
+    },
+    {
+      "epoch": 897.93,
+      "learning_rate": 5.670250896057348e-06,
+      "loss": 0.1808,
+      "step": 139180
+    },
+    {
+      "epoch": 898.0,
+      "learning_rate": 5.666666666666667e-06,
+      "loss": 0.2141,
+      "step": 139190
+    },
+    {
+      "epoch": 898.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9350091218948364,
+      "eval_runtime": 0.8734,
+      "eval_samples_per_second": 230.139,
+      "eval_steps_per_second": 8.015,
+      "step": 139190
+    },
+    {
+      "epoch": 898.06,
+      "learning_rate": 5.663082437275986e-06,
+      "loss": 0.2055,
+      "step": 139200
+    },
+    {
+      "epoch": 898.13,
+      "learning_rate": 5.659498207885305e-06,
+      "loss": 0.2016,
+      "step": 139210
+    },
+    {
+      "epoch": 898.19,
+      "learning_rate": 5.655913978494624e-06,
+      "loss": 0.1705,
+      "step": 139220
+    },
+    {
+      "epoch": 898.26,
+      "learning_rate": 5.652329749103943e-06,
+      "loss": 0.1943,
+      "step": 139230
+    },
+    {
+      "epoch": 898.32,
+      "learning_rate": 5.648745519713262e-06,
+      "loss": 0.2192,
+      "step": 139240
+    },
+    {
+      "epoch": 898.39,
+      "learning_rate": 5.64516129032258e-06,
+      "loss": 0.2081,
+      "step": 139250
+    },
+    {
+      "epoch": 898.45,
+      "learning_rate": 5.641577060931899e-06,
+      "loss": 0.1942,
+      "step": 139260
+    },
+    {
+      "epoch": 898.51,
+      "learning_rate": 5.6379928315412185e-06,
+      "loss": 0.1641,
+      "step": 139270
+    },
+    {
+      "epoch": 898.58,
+      "learning_rate": 5.6344086021505375e-06,
+      "loss": 0.2077,
+      "step": 139280
+    },
+    {
+      "epoch": 898.64,
+      "learning_rate": 5.630824372759857e-06,
+      "loss": 0.1811,
+      "step": 139290
+    },
+    {
+      "epoch": 898.71,
+      "learning_rate": 5.627240143369176e-06,
+      "loss": 0.1866,
+      "step": 139300
+    },
+    {
+      "epoch": 898.77,
+      "learning_rate": 5.623655913978495e-06,
+      "loss": 0.2235,
+      "step": 139310
+    },
+    {
+      "epoch": 898.84,
+      "learning_rate": 5.620071684587814e-06,
+      "loss": 0.2038,
+      "step": 139320
+    },
+    {
+      "epoch": 898.9,
+      "learning_rate": 5.616487455197133e-06,
+      "loss": 0.1849,
+      "step": 139330
+    },
+    {
+      "epoch": 898.96,
+      "learning_rate": 5.612903225806452e-06,
+      "loss": 0.2238,
+      "step": 139340
+    },
+    {
+      "epoch": 899.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9182186722755432,
+      "eval_runtime": 0.8705,
+      "eval_samples_per_second": 230.899,
+      "eval_steps_per_second": 8.041,
+      "step": 139345
+    },
+    {
+      "epoch": 899.03,
+      "learning_rate": 5.609318996415771e-06,
+      "loss": 0.1616,
+      "step": 139350
+    },
+    {
+      "epoch": 899.1,
+      "learning_rate": 5.60573476702509e-06,
+      "loss": 0.1955,
+      "step": 139360
+    },
+    {
+      "epoch": 899.16,
+      "learning_rate": 5.602150537634409e-06,
+      "loss": 0.203,
+      "step": 139370
+    },
+    {
+      "epoch": 899.23,
+      "learning_rate": 5.598566308243728e-06,
+      "loss": 0.1979,
+      "step": 139380
+    },
+    {
+      "epoch": 899.29,
+      "learning_rate": 5.594982078853047e-06,
+      "loss": 0.1675,
+      "step": 139390
+    },
+    {
+      "epoch": 899.35,
+      "learning_rate": 5.591397849462366e-06,
+      "loss": 0.2187,
+      "step": 139400
+    },
+    {
+      "epoch": 899.42,
+      "learning_rate": 5.587813620071685e-06,
+      "loss": 0.2096,
+      "step": 139410
+    },
+    {
+      "epoch": 899.48,
+      "learning_rate": 5.584229390681004e-06,
+      "loss": 0.2109,
+      "step": 139420
+    },
+    {
+      "epoch": 899.55,
+      "learning_rate": 5.580645161290323e-06,
+      "loss": 0.225,
+      "step": 139430
+    },
+    {
+      "epoch": 899.61,
+      "learning_rate": 5.577060931899642e-06,
+      "loss": 0.2299,
+      "step": 139440
+    },
+    {
+      "epoch": 899.68,
+      "learning_rate": 5.573476702508961e-06,
+      "loss": 0.2775,
+      "step": 139450
+    },
+    {
+      "epoch": 899.74,
+      "learning_rate": 5.5698924731182805e-06,
+      "loss": 0.2038,
+      "step": 139460
+    },
+    {
+      "epoch": 899.8,
+      "learning_rate": 5.5663082437275995e-06,
+      "loss": 0.1781,
+      "step": 139470
+    },
+    {
+      "epoch": 899.87,
+      "learning_rate": 5.5627240143369186e-06,
+      "loss": 0.1957,
+      "step": 139480
+    },
+    {
+      "epoch": 899.93,
+      "learning_rate": 5.559139784946237e-06,
+      "loss": 0.1697,
+      "step": 139490
+    },
+    {
+      "epoch": 900.0,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 0.1499,
+      "step": 139500
+    },
+    {
+      "epoch": 900.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9444724321365356,
+      "eval_runtime": 0.8671,
+      "eval_samples_per_second": 231.806,
+      "eval_steps_per_second": 8.073,
+      "step": 139500
+    },
+    {
+      "epoch": 900.06,
+      "learning_rate": 5.551971326164875e-06,
+      "loss": 0.1859,
+      "step": 139510
+    },
+    {
+      "epoch": 900.13,
+      "learning_rate": 5.548387096774194e-06,
+      "loss": 0.2202,
+      "step": 139520
+    },
+    {
+      "epoch": 900.19,
+      "learning_rate": 5.544802867383513e-06,
+      "loss": 0.1749,
+      "step": 139530
+    },
+    {
+      "epoch": 900.26,
+      "learning_rate": 5.541218637992831e-06,
+      "loss": 0.1804,
+      "step": 139540
+    },
+    {
+      "epoch": 900.32,
+      "learning_rate": 5.53763440860215e-06,
+      "loss": 0.1827,
+      "step": 139550
+    },
+    {
+      "epoch": 900.39,
+      "learning_rate": 5.534050179211469e-06,
+      "loss": 0.1905,
+      "step": 139560
+    },
+    {
+      "epoch": 900.45,
+      "learning_rate": 5.530465949820788e-06,
+      "loss": 0.1889,
+      "step": 139570
+    },
+    {
+      "epoch": 900.51,
+      "learning_rate": 5.526881720430107e-06,
+      "loss": 0.1564,
+      "step": 139580
+    },
+    {
+      "epoch": 900.58,
+      "learning_rate": 5.5232974910394264e-06,
+      "loss": 0.227,
+      "step": 139590
+    },
+    {
+      "epoch": 900.64,
+      "learning_rate": 5.5197132616487455e-06,
+      "loss": 0.1511,
+      "step": 139600
+    },
+    {
+      "epoch": 900.71,
+      "learning_rate": 5.5161290322580645e-06,
+      "loss": 0.1563,
+      "step": 139610
+    },
+    {
+      "epoch": 900.77,
+      "learning_rate": 5.512544802867384e-06,
+      "loss": 0.1887,
+      "step": 139620
+    },
+    {
+      "epoch": 900.84,
+      "learning_rate": 5.508960573476703e-06,
+      "loss": 0.1658,
+      "step": 139630
+    },
+    {
+      "epoch": 900.9,
+      "learning_rate": 5.505376344086022e-06,
+      "loss": 0.2112,
+      "step": 139640
+    },
+    {
+      "epoch": 900.96,
+      "learning_rate": 5.501792114695341e-06,
+      "loss": 0.1832,
+      "step": 139650
+    },
+    {
+      "epoch": 901.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9245242476463318,
+      "eval_runtime": 0.859,
+      "eval_samples_per_second": 234.003,
+      "eval_steps_per_second": 8.149,
+      "step": 139655
+    },
+    {
+      "epoch": 901.03,
+      "learning_rate": 5.49820788530466e-06,
+      "loss": 0.2059,
+      "step": 139660
+    },
+    {
+      "epoch": 901.1,
+      "learning_rate": 5.494623655913979e-06,
+      "loss": 0.224,
+      "step": 139670
+    },
+    {
+      "epoch": 901.16,
+      "learning_rate": 5.491039426523298e-06,
+      "loss": 0.2008,
+      "step": 139680
+    },
+    {
+      "epoch": 901.23,
+      "learning_rate": 5.487455197132617e-06,
+      "loss": 0.2293,
+      "step": 139690
+    },
+    {
+      "epoch": 901.29,
+      "learning_rate": 5.483870967741936e-06,
+      "loss": 0.2003,
+      "step": 139700
+    },
+    {
+      "epoch": 901.35,
+      "learning_rate": 5.480286738351255e-06,
+      "loss": 0.2099,
+      "step": 139710
+    },
+    {
+      "epoch": 901.42,
+      "learning_rate": 5.476702508960574e-06,
+      "loss": 0.2208,
+      "step": 139720
+    },
+    {
+      "epoch": 901.48,
+      "learning_rate": 5.473118279569893e-06,
+      "loss": 0.2024,
+      "step": 139730
+    },
+    {
+      "epoch": 901.55,
+      "learning_rate": 5.469534050179212e-06,
+      "loss": 0.2194,
+      "step": 139740
+    },
+    {
+      "epoch": 901.61,
+      "learning_rate": 5.465949820788531e-06,
+      "loss": 0.2076,
+      "step": 139750
+    },
+    {
+      "epoch": 901.68,
+      "learning_rate": 5.46236559139785e-06,
+      "loss": 0.2081,
+      "step": 139760
+    },
+    {
+      "epoch": 901.74,
+      "learning_rate": 5.458781362007169e-06,
+      "loss": 0.1968,
+      "step": 139770
+    },
+    {
+      "epoch": 901.8,
+      "learning_rate": 5.4551971326164884e-06,
+      "loss": 0.1678,
+      "step": 139780
+    },
+    {
+      "epoch": 901.87,
+      "learning_rate": 5.451612903225807e-06,
+      "loss": 0.2188,
+      "step": 139790
+    },
+    {
+      "epoch": 901.93,
+      "learning_rate": 5.448028673835126e-06,
+      "loss": 0.2217,
+      "step": 139800
+    },
+    {
+      "epoch": 902.0,
+      "learning_rate": 5.444444444444445e-06,
+      "loss": 0.1971,
+      "step": 139810
+    },
+    {
+      "epoch": 902.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9505664110183716,
+      "eval_runtime": 0.8995,
+      "eval_samples_per_second": 223.461,
+      "eval_steps_per_second": 7.782,
+      "step": 139810
+    },
+    {
+      "epoch": 902.06,
+      "learning_rate": 5.440860215053764e-06,
+      "loss": 0.2048,
+      "step": 139820
+    },
+    {
+      "epoch": 902.13,
+      "learning_rate": 5.437275985663082e-06,
+      "loss": 0.1487,
+      "step": 139830
+    },
+    {
+      "epoch": 902.19,
+      "learning_rate": 5.433691756272401e-06,
+      "loss": 0.2264,
+      "step": 139840
+    },
+    {
+      "epoch": 902.26,
+      "learning_rate": 5.43010752688172e-06,
+      "loss": 0.234,
+      "step": 139850
+    },
+    {
+      "epoch": 902.32,
+      "learning_rate": 5.426523297491039e-06,
+      "loss": 0.2518,
+      "step": 139860
+    },
+    {
+      "epoch": 902.39,
+      "learning_rate": 5.422939068100358e-06,
+      "loss": 0.1936,
+      "step": 139870
+    },
+    {
+      "epoch": 902.45,
+      "learning_rate": 5.419354838709677e-06,
+      "loss": 0.1794,
+      "step": 139880
+    },
+    {
+      "epoch": 902.51,
+      "learning_rate": 5.415770609318996e-06,
+      "loss": 0.2291,
+      "step": 139890
+    },
+    {
+      "epoch": 902.58,
+      "learning_rate": 5.412186379928315e-06,
+      "loss": 0.2038,
+      "step": 139900
+    },
+    {
+      "epoch": 902.64,
+      "learning_rate": 5.408602150537634e-06,
+      "loss": 0.1742,
+      "step": 139910
+    },
+    {
+      "epoch": 902.71,
+      "learning_rate": 5.4050179211469535e-06,
+      "loss": 0.1568,
+      "step": 139920
+    },
+    {
+      "epoch": 902.77,
+      "learning_rate": 5.4014336917562725e-06,
+      "loss": 0.2429,
+      "step": 139930
+    },
+    {
+      "epoch": 902.84,
+      "learning_rate": 5.3978494623655916e-06,
+      "loss": 0.1855,
+      "step": 139940
+    },
+    {
+      "epoch": 902.9,
+      "learning_rate": 5.394265232974911e-06,
+      "loss": 0.2074,
+      "step": 139950
+    },
+    {
+      "epoch": 902.96,
+      "learning_rate": 5.39068100358423e-06,
+      "loss": 0.2148,
+      "step": 139960
+    },
+    {
+      "epoch": 903.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9224830865859985,
+      "eval_runtime": 0.867,
+      "eval_samples_per_second": 231.829,
+      "eval_steps_per_second": 8.074,
+      "step": 139965
+    },
+    {
+      "epoch": 903.03,
+      "learning_rate": 5.387096774193549e-06,
+      "loss": 0.2588,
+      "step": 139970
+    },
+    {
+      "epoch": 903.1,
+      "learning_rate": 5.383512544802868e-06,
+      "loss": 0.1911,
+      "step": 139980
+    },
+    {
+      "epoch": 903.16,
+      "learning_rate": 5.379928315412187e-06,
+      "loss": 0.1897,
+      "step": 139990
+    },
+    {
+      "epoch": 903.23,
+      "learning_rate": 5.376344086021506e-06,
+      "loss": 0.1885,
+      "step": 140000
+    },
+    {
+      "epoch": 903.29,
+      "learning_rate": 5.372759856630825e-06,
+      "loss": 0.2151,
+      "step": 140010
+    },
+    {
+      "epoch": 903.35,
+      "learning_rate": 5.369175627240144e-06,
+      "loss": 0.2292,
+      "step": 140020
+    },
+    {
+      "epoch": 903.42,
+      "learning_rate": 5.365591397849463e-06,
+      "loss": 0.2059,
+      "step": 140030
+    },
+    {
+      "epoch": 903.48,
+      "learning_rate": 5.362007168458782e-06,
+      "loss": 0.2065,
+      "step": 140040
+    },
+    {
+      "epoch": 903.55,
+      "learning_rate": 5.358422939068101e-06,
+      "loss": 0.1971,
+      "step": 140050
+    },
+    {
+      "epoch": 903.61,
+      "learning_rate": 5.35483870967742e-06,
+      "loss": 0.1942,
+      "step": 140060
+    },
+    {
+      "epoch": 903.68,
+      "learning_rate": 5.351254480286739e-06,
+      "loss": 0.2192,
+      "step": 140070
+    },
+    {
+      "epoch": 903.74,
+      "learning_rate": 5.3476702508960574e-06,
+      "loss": 0.1986,
+      "step": 140080
+    },
+    {
+      "epoch": 903.8,
+      "learning_rate": 5.3440860215053765e-06,
+      "loss": 0.231,
+      "step": 140090
+    },
+    {
+      "epoch": 903.87,
+      "learning_rate": 5.3405017921146956e-06,
+      "loss": 0.2336,
+      "step": 140100
+    },
+    {
+      "epoch": 903.93,
+      "learning_rate": 5.336917562724015e-06,
+      "loss": 0.2247,
+      "step": 140110
+    },
+    {
+      "epoch": 904.0,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 0.2266,
+      "step": 140120
+    },
+    {
+      "epoch": 904.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9292066693305969,
+      "eval_runtime": 0.8787,
+      "eval_samples_per_second": 228.741,
+      "eval_steps_per_second": 7.966,
+      "step": 140120
+    },
+    {
+      "epoch": 904.06,
+      "learning_rate": 5.329749103942652e-06,
+      "loss": 0.1922,
+      "step": 140130
+    },
+    {
+      "epoch": 904.13,
+      "learning_rate": 5.326164874551971e-06,
+      "loss": 0.1987,
+      "step": 140140
+    },
+    {
+      "epoch": 904.19,
+      "learning_rate": 5.32258064516129e-06,
+      "loss": 0.2352,
+      "step": 140150
+    },
+    {
+      "epoch": 904.26,
+      "learning_rate": 5.318996415770609e-06,
+      "loss": 0.2087,
+      "step": 140160
+    },
+    {
+      "epoch": 904.32,
+      "learning_rate": 5.315412186379928e-06,
+      "loss": 0.2609,
+      "step": 140170
+    },
+    {
+      "epoch": 904.39,
+      "learning_rate": 5.311827956989247e-06,
+      "loss": 0.1644,
+      "step": 140180
+    },
+    {
+      "epoch": 904.45,
+      "learning_rate": 5.308243727598566e-06,
+      "loss": 0.2266,
+      "step": 140190
+    },
+    {
+      "epoch": 904.51,
+      "learning_rate": 5.304659498207885e-06,
+      "loss": 0.2096,
+      "step": 140200
+    },
+    {
+      "epoch": 904.58,
+      "learning_rate": 5.301075268817204e-06,
+      "loss": 0.2149,
+      "step": 140210
+    },
+    {
+      "epoch": 904.64,
+      "learning_rate": 5.297491039426523e-06,
+      "loss": 0.1764,
+      "step": 140220
+    },
+    {
+      "epoch": 904.71,
+      "learning_rate": 5.293906810035842e-06,
+      "loss": 0.2014,
+      "step": 140230
+    },
+    {
+      "epoch": 904.77,
+      "learning_rate": 5.2903225806451614e-06,
+      "loss": 0.2116,
+      "step": 140240
+    },
+    {
+      "epoch": 904.84,
+      "learning_rate": 5.2867383512544805e-06,
+      "loss": 0.2483,
+      "step": 140250
+    },
+    {
+      "epoch": 904.9,
+      "learning_rate": 5.2831541218637995e-06,
+      "loss": 0.2128,
+      "step": 140260
+    },
+    {
+      "epoch": 904.96,
+      "learning_rate": 5.279569892473119e-06,
+      "loss": 0.2619,
+      "step": 140270
+    },
+    {
+      "epoch": 905.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.9084001779556274,
+      "eval_runtime": 0.8551,
+      "eval_samples_per_second": 235.069,
+      "eval_steps_per_second": 8.186,
+      "step": 140275
+    },
+    {
+      "epoch": 905.03,
+      "learning_rate": 5.275985663082438e-06,
+      "loss": 0.213,
+      "step": 140280
+    },
+    {
+      "epoch": 905.1,
+      "learning_rate": 5.272401433691757e-06,
+      "loss": 0.2316,
+      "step": 140290
+    },
+    {
+      "epoch": 905.16,
+      "learning_rate": 5.268817204301076e-06,
+      "loss": 0.2141,
+      "step": 140300
+    },
+    {
+      "epoch": 905.23,
+      "learning_rate": 5.265232974910395e-06,
+      "loss": 0.1896,
+      "step": 140310
+    },
+    {
+      "epoch": 905.29,
+      "learning_rate": 5.261648745519714e-06,
+      "loss": 0.1981,
+      "step": 140320
+    },
+    {
+      "epoch": 905.35,
+      "learning_rate": 5.258064516129033e-06,
+      "loss": 0.2779,
+      "step": 140330
+    },
+    {
+      "epoch": 905.42,
+      "learning_rate": 5.254480286738352e-06,
+      "loss": 0.2156,
+      "step": 140340
+    },
+    {
+      "epoch": 905.48,
+      "learning_rate": 5.250896057347671e-06,
+      "loss": 0.214,
+      "step": 140350
+    },
+    {
+      "epoch": 905.55,
+      "learning_rate": 5.24731182795699e-06,
+      "loss": 0.1941,
+      "step": 140360
+    },
+    {
+      "epoch": 905.61,
+      "learning_rate": 5.243727598566308e-06,
+      "loss": 0.204,
+      "step": 140370
+    },
+    {
+      "epoch": 905.68,
+      "learning_rate": 5.240143369175627e-06,
+      "loss": 0.1822,
+      "step": 140380
+    },
+    {
+      "epoch": 905.74,
+      "learning_rate": 5.236559139784946e-06,
+      "loss": 0.1711,
+      "step": 140390
+    },
+    {
+      "epoch": 905.8,
+      "learning_rate": 5.232974910394265e-06,
+      "loss": 0.2045,
+      "step": 140400
+    },
+    {
+      "epoch": 905.87,
+      "learning_rate": 5.2293906810035845e-06,
+      "loss": 0.2094,
+      "step": 140410
+    },
+    {
+      "epoch": 905.93,
+      "learning_rate": 5.2258064516129035e-06,
+      "loss": 0.1928,
+      "step": 140420
+    },
+    {
+      "epoch": 906.0,
+      "learning_rate": 5.2222222222222226e-06,
+      "loss": 0.2091,
+      "step": 140430
+    },
+    {
+      "epoch": 906.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.93620365858078,
+      "eval_runtime": 0.873,
+      "eval_samples_per_second": 230.254,
+      "eval_steps_per_second": 8.019,
+      "step": 140430
+    },
+    {
+      "epoch": 906.06,
+      "learning_rate": 5.218637992831542e-06,
+      "loss": 0.2485,
+      "step": 140440
+    },
+    {
+      "epoch": 906.13,
+      "learning_rate": 5.21505376344086e-06,
+      "loss": 0.1924,
+      "step": 140450
+    },
+    {
+      "epoch": 906.19,
+      "learning_rate": 5.211469534050179e-06,
+      "loss": 0.1998,
+      "step": 140460
+    },
+    {
+      "epoch": 906.26,
+      "learning_rate": 5.207885304659498e-06,
+      "loss": 0.2281,
+      "step": 140470
+    },
+    {
+      "epoch": 906.32,
+      "learning_rate": 5.204301075268817e-06,
+      "loss": 0.1808,
+      "step": 140480
+    },
+    {
+      "epoch": 906.39,
+      "learning_rate": 5.200716845878136e-06,
+      "loss": 0.1591,
+      "step": 140490
+    },
+    {
+      "epoch": 906.45,
+      "learning_rate": 5.197132616487455e-06,
+      "loss": 0.2038,
+      "step": 140500
+    },
+    {
+      "epoch": 906.51,
+      "learning_rate": 5.193548387096774e-06,
+      "loss": 0.1664,
+      "step": 140510
+    },
+    {
+      "epoch": 906.58,
+      "learning_rate": 5.189964157706093e-06,
+      "loss": 0.2089,
+      "step": 140520
+    },
+    {
+      "epoch": 906.64,
+      "learning_rate": 5.186379928315412e-06,
+      "loss": 0.1946,
+      "step": 140530
+    },
+    {
+      "epoch": 906.71,
+      "learning_rate": 5.182795698924731e-06,
+      "loss": 0.1839,
+      "step": 140540
+    },
+    {
+      "epoch": 906.77,
+      "learning_rate": 5.17921146953405e-06,
+      "loss": 0.2539,
+      "step": 140550
+    },
+    {
+      "epoch": 906.84,
+      "learning_rate": 5.175627240143369e-06,
+      "loss": 0.1908,
+      "step": 140560
+    },
+    {
+      "epoch": 906.9,
+      "learning_rate": 5.1720430107526885e-06,
+      "loss": 0.1732,
+      "step": 140570
+    },
+    {
+      "epoch": 906.96,
+      "learning_rate": 5.1684587813620075e-06,
+      "loss": 0.1623,
+      "step": 140580
+    },
+    {
+      "epoch": 907.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9269290566444397,
+      "eval_runtime": 0.8939,
+      "eval_samples_per_second": 224.859,
+      "eval_steps_per_second": 7.831,
+      "step": 140585
+    },
+    {
+      "epoch": 907.03,
+      "learning_rate": 5.1648745519713266e-06,
+      "loss": 0.181,
+      "step": 140590
+    },
+    {
+      "epoch": 907.1,
+      "learning_rate": 5.161290322580646e-06,
+      "loss": 0.1766,
+      "step": 140600
+    },
+    {
+      "epoch": 907.16,
+      "learning_rate": 5.157706093189965e-06,
+      "loss": 0.2097,
+      "step": 140610
+    },
+    {
+      "epoch": 907.23,
+      "learning_rate": 5.154121863799284e-06,
+      "loss": 0.1994,
+      "step": 140620
+    },
+    {
+      "epoch": 907.29,
+      "learning_rate": 5.150537634408603e-06,
+      "loss": 0.221,
+      "step": 140630
+    },
+    {
+      "epoch": 907.35,
+      "learning_rate": 5.146953405017922e-06,
+      "loss": 0.2014,
+      "step": 140640
+    },
+    {
+      "epoch": 907.42,
+      "learning_rate": 5.143369175627241e-06,
+      "loss": 0.1772,
+      "step": 140650
+    },
+    {
+      "epoch": 907.48,
+      "learning_rate": 5.13978494623656e-06,
+      "loss": 0.2094,
+      "step": 140660
+    },
+    {
+      "epoch": 907.55,
+      "learning_rate": 5.136200716845878e-06,
+      "loss": 0.1998,
+      "step": 140670
+    },
+    {
+      "epoch": 907.61,
+      "learning_rate": 5.132616487455197e-06,
+      "loss": 0.2204,
+      "step": 140680
+    },
+    {
+      "epoch": 907.68,
+      "learning_rate": 5.129032258064516e-06,
+      "loss": 0.1928,
+      "step": 140690
+    },
+    {
+      "epoch": 907.74,
+      "learning_rate": 5.125448028673835e-06,
+      "loss": 0.198,
+      "step": 140700
+    },
+    {
+      "epoch": 907.8,
+      "learning_rate": 5.121863799283154e-06,
+      "loss": 0.2349,
+      "step": 140710
+    },
+    {
+      "epoch": 907.87,
+      "learning_rate": 5.118279569892473e-06,
+      "loss": 0.2204,
+      "step": 140720
+    },
+    {
+      "epoch": 907.93,
+      "learning_rate": 5.1146953405017924e-06,
+      "loss": 0.2264,
+      "step": 140730
+    },
+    {
+      "epoch": 908.0,
+      "learning_rate": 5.1111111111111115e-06,
+      "loss": 0.1798,
+      "step": 140740
+    },
+    {
+      "epoch": 908.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.930547297000885,
+      "eval_runtime": 0.8699,
+      "eval_samples_per_second": 231.051,
+      "eval_steps_per_second": 8.047,
+      "step": 140740
+    },
+    {
+      "epoch": 908.06,
+      "learning_rate": 5.1075268817204305e-06,
+      "loss": 0.2231,
+      "step": 140750
+    },
+    {
+      "epoch": 908.13,
+      "learning_rate": 5.10394265232975e-06,
+      "loss": 0.2331,
+      "step": 140760
+    },
+    {
+      "epoch": 908.19,
+      "learning_rate": 5.100358422939068e-06,
+      "loss": 0.1922,
+      "step": 140770
+    },
+    {
+      "epoch": 908.26,
+      "learning_rate": 5.096774193548387e-06,
+      "loss": 0.1942,
+      "step": 140780
+    },
+    {
+      "epoch": 908.32,
+      "learning_rate": 5.093189964157706e-06,
+      "loss": 0.1824,
+      "step": 140790
+    },
+    {
+      "epoch": 908.39,
+      "learning_rate": 5.089605734767025e-06,
+      "loss": 0.1952,
+      "step": 140800
+    },
+    {
+      "epoch": 908.45,
+      "learning_rate": 5.086021505376344e-06,
+      "loss": 0.1923,
+      "step": 140810
+    },
+    {
+      "epoch": 908.51,
+      "learning_rate": 5.082437275985663e-06,
+      "loss": 0.2126,
+      "step": 140820
+    },
+    {
+      "epoch": 908.58,
+      "learning_rate": 5.078853046594982e-06,
+      "loss": 0.1705,
+      "step": 140830
+    },
+    {
+      "epoch": 908.64,
+      "learning_rate": 5.075268817204301e-06,
+      "loss": 0.2255,
+      "step": 140840
+    },
+    {
+      "epoch": 908.71,
+      "learning_rate": 5.07168458781362e-06,
+      "loss": 0.1656,
+      "step": 140850
+    },
+    {
+      "epoch": 908.77,
+      "learning_rate": 5.068100358422939e-06,
+      "loss": 0.1455,
+      "step": 140860
+    },
+    {
+      "epoch": 908.84,
+      "learning_rate": 5.064516129032258e-06,
+      "loss": 0.2065,
+      "step": 140870
+    },
+    {
+      "epoch": 908.9,
+      "learning_rate": 5.060931899641577e-06,
+      "loss": 0.2302,
+      "step": 140880
+    },
+    {
+      "epoch": 908.96,
+      "learning_rate": 5.0573476702508964e-06,
+      "loss": 0.2005,
+      "step": 140890
+    },
+    {
+      "epoch": 909.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9222986102104187,
+      "eval_runtime": 0.8822,
+      "eval_samples_per_second": 227.834,
+      "eval_steps_per_second": 7.935,
+      "step": 140895
+    },
+    {
+      "epoch": 909.03,
+      "learning_rate": 5.0537634408602155e-06,
+      "loss": 0.21,
+      "step": 140900
+    },
+    {
+      "epoch": 909.1,
+      "learning_rate": 5.0501792114695345e-06,
+      "loss": 0.2087,
+      "step": 140910
+    },
+    {
+      "epoch": 909.16,
+      "learning_rate": 5.046594982078854e-06,
+      "loss": 0.1887,
+      "step": 140920
+    },
+    {
+      "epoch": 909.23,
+      "learning_rate": 5.043010752688173e-06,
+      "loss": 0.2249,
+      "step": 140930
+    },
+    {
+      "epoch": 909.29,
+      "learning_rate": 5.039426523297492e-06,
+      "loss": 0.188,
+      "step": 140940
+    },
+    {
+      "epoch": 909.35,
+      "learning_rate": 5.035842293906811e-06,
+      "loss": 0.2329,
+      "step": 140950
+    },
+    {
+      "epoch": 909.42,
+      "learning_rate": 5.032258064516129e-06,
+      "loss": 0.2466,
+      "step": 140960
+    },
+    {
+      "epoch": 909.48,
+      "learning_rate": 5.028673835125448e-06,
+      "loss": 0.2045,
+      "step": 140970
+    },
+    {
+      "epoch": 909.55,
+      "learning_rate": 5.025089605734767e-06,
+      "loss": 0.2147,
+      "step": 140980
+    },
+    {
+      "epoch": 909.61,
+      "learning_rate": 5.021505376344086e-06,
+      "loss": 0.2084,
+      "step": 140990
+    },
+    {
+      "epoch": 909.68,
+      "learning_rate": 5.017921146953405e-06,
+      "loss": 0.1831,
+      "step": 141000
+    },
+    {
+      "epoch": 909.74,
+      "learning_rate": 5.014336917562724e-06,
+      "loss": 0.208,
+      "step": 141010
+    },
+    {
+      "epoch": 909.8,
+      "learning_rate": 5.010752688172043e-06,
+      "loss": 0.1713,
+      "step": 141020
+    },
+    {
+      "epoch": 909.87,
+      "learning_rate": 5.007168458781362e-06,
+      "loss": 0.2427,
+      "step": 141030
+    },
+    {
+      "epoch": 909.93,
+      "learning_rate": 5.003584229390681e-06,
+      "loss": 0.2032,
+      "step": 141040
+    },
+    {
+      "epoch": 910.0,
+      "learning_rate": 5e-06,
+      "loss": 0.2141,
+      "step": 141050
+    },
+    {
+      "epoch": 910.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9171804785728455,
+      "eval_runtime": 0.9067,
+      "eval_samples_per_second": 221.685,
+      "eval_steps_per_second": 7.72,
+      "step": 141050
+    },
+    {
+      "epoch": 910.06,
+      "learning_rate": 4.9964157706093195e-06,
+      "loss": 0.2025,
+      "step": 141060
+    },
+    {
+      "epoch": 910.13,
+      "learning_rate": 4.9928315412186385e-06,
+      "loss": 0.2064,
+      "step": 141070
+    },
+    {
+      "epoch": 910.19,
+      "learning_rate": 4.9892473118279576e-06,
+      "loss": 0.2156,
+      "step": 141080
+    },
+    {
+      "epoch": 910.26,
+      "learning_rate": 4.985663082437276e-06,
+      "loss": 0.2247,
+      "step": 141090
+    },
+    {
+      "epoch": 910.32,
+      "learning_rate": 4.982078853046595e-06,
+      "loss": 0.1988,
+      "step": 141100
+    },
+    {
+      "epoch": 910.39,
+      "learning_rate": 4.978494623655914e-06,
+      "loss": 0.1957,
+      "step": 141110
+    },
+    {
+      "epoch": 910.45,
+      "learning_rate": 4.974910394265233e-06,
+      "loss": 0.1983,
+      "step": 141120
+    },
+    {
+      "epoch": 910.51,
+      "learning_rate": 4.971326164874552e-06,
+      "loss": 0.2346,
+      "step": 141130
+    },
+    {
+      "epoch": 910.58,
+      "learning_rate": 4.967741935483871e-06,
+      "loss": 0.1601,
+      "step": 141140
+    },
+    {
+      "epoch": 910.64,
+      "learning_rate": 4.96415770609319e-06,
+      "loss": 0.2128,
+      "step": 141150
+    },
+    {
+      "epoch": 910.71,
+      "learning_rate": 4.960573476702509e-06,
+      "loss": 0.2266,
+      "step": 141160
+    },
+    {
+      "epoch": 910.77,
+      "learning_rate": 4.956989247311828e-06,
+      "loss": 0.2228,
+      "step": 141170
+    },
+    {
+      "epoch": 910.84,
+      "learning_rate": 4.953405017921147e-06,
+      "loss": 0.192,
+      "step": 141180
+    },
+    {
+      "epoch": 910.9,
+      "learning_rate": 4.949820788530466e-06,
+      "loss": 0.189,
+      "step": 141190
+    },
+    {
+      "epoch": 910.96,
+      "learning_rate": 4.946236559139785e-06,
+      "loss": 0.214,
+      "step": 141200
+    },
+    {
+      "epoch": 911.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9322085976600647,
+      "eval_runtime": 0.8787,
+      "eval_samples_per_second": 228.755,
+      "eval_steps_per_second": 7.967,
+      "step": 141205
+    },
+    {
+      "epoch": 911.03,
+      "learning_rate": 4.942652329749104e-06,
+      "loss": 0.1949,
+      "step": 141210
+    },
+    {
+      "epoch": 911.1,
+      "learning_rate": 4.9390681003584234e-06,
+      "loss": 0.1962,
+      "step": 141220
+    },
+    {
+      "epoch": 911.16,
+      "learning_rate": 4.9354838709677425e-06,
+      "loss": 0.2209,
+      "step": 141230
+    },
+    {
+      "epoch": 911.23,
+      "learning_rate": 4.9318996415770615e-06,
+      "loss": 0.1667,
+      "step": 141240
+    },
+    {
+      "epoch": 911.29,
+      "learning_rate": 4.928315412186381e-06,
+      "loss": 0.2065,
+      "step": 141250
+    },
+    {
+      "epoch": 911.35,
+      "learning_rate": 4.924731182795699e-06,
+      "loss": 0.2043,
+      "step": 141260
+    },
+    {
+      "epoch": 911.42,
+      "learning_rate": 4.921146953405018e-06,
+      "loss": 0.2274,
+      "step": 141270
+    },
+    {
+      "epoch": 911.48,
+      "learning_rate": 4.917562724014337e-06,
+      "loss": 0.1562,
+      "step": 141280
+    },
+    {
+      "epoch": 911.55,
+      "learning_rate": 4.913978494623656e-06,
+      "loss": 0.1864,
+      "step": 141290
+    },
+    {
+      "epoch": 911.61,
+      "learning_rate": 4.910394265232975e-06,
+      "loss": 0.2374,
+      "step": 141300
+    },
+    {
+      "epoch": 911.68,
+      "learning_rate": 4.906810035842294e-06,
+      "loss": 0.1658,
+      "step": 141310
+    },
+    {
+      "epoch": 911.74,
+      "learning_rate": 4.903225806451613e-06,
+      "loss": 0.1825,
+      "step": 141320
+    },
+    {
+      "epoch": 911.8,
+      "learning_rate": 4.899641577060932e-06,
+      "loss": 0.2422,
+      "step": 141330
+    },
+    {
+      "epoch": 911.87,
+      "learning_rate": 4.896057347670251e-06,
+      "loss": 0.2118,
+      "step": 141340
+    },
+    {
+      "epoch": 911.93,
+      "learning_rate": 4.89247311827957e-06,
+      "loss": 0.2176,
+      "step": 141350
+    },
+    {
+      "epoch": 912.0,
+      "learning_rate": 4.888888888888889e-06,
+      "loss": 0.1851,
+      "step": 141360
+    },
+    {
+      "epoch": 912.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9298938512802124,
+      "eval_runtime": 0.8874,
+      "eval_samples_per_second": 226.494,
+      "eval_steps_per_second": 7.888,
+      "step": 141360
+    },
+    {
+      "epoch": 912.06,
+      "learning_rate": 4.885304659498208e-06,
+      "loss": 0.1702,
+      "step": 141370
+    },
+    {
+      "epoch": 912.13,
+      "learning_rate": 4.8817204301075274e-06,
+      "loss": 0.1637,
+      "step": 141380
+    },
+    {
+      "epoch": 912.19,
+      "learning_rate": 4.8781362007168465e-06,
+      "loss": 0.248,
+      "step": 141390
+    },
+    {
+      "epoch": 912.26,
+      "learning_rate": 4.8745519713261655e-06,
+      "loss": 0.2054,
+      "step": 141400
+    },
+    {
+      "epoch": 912.32,
+      "learning_rate": 4.870967741935484e-06,
+      "loss": 0.2214,
+      "step": 141410
+    },
+    {
+      "epoch": 912.39,
+      "learning_rate": 4.867383512544803e-06,
+      "loss": 0.1881,
+      "step": 141420
+    },
+    {
+      "epoch": 912.45,
+      "learning_rate": 4.863799283154122e-06,
+      "loss": 0.1929,
+      "step": 141430
+    },
+    {
+      "epoch": 912.51,
+      "learning_rate": 4.860215053763441e-06,
+      "loss": 0.1885,
+      "step": 141440
+    },
+    {
+      "epoch": 912.58,
+      "learning_rate": 4.85663082437276e-06,
+      "loss": 0.2323,
+      "step": 141450
+    },
+    {
+      "epoch": 912.64,
+      "learning_rate": 4.853046594982079e-06,
+      "loss": 0.2209,
+      "step": 141460
+    },
+    {
+      "epoch": 912.71,
+      "learning_rate": 4.849462365591398e-06,
+      "loss": 0.1698,
+      "step": 141470
+    },
+    {
+      "epoch": 912.77,
+      "learning_rate": 4.845878136200717e-06,
+      "loss": 0.1669,
+      "step": 141480
+    },
+    {
+      "epoch": 912.84,
+      "learning_rate": 4.842293906810036e-06,
+      "loss": 0.2079,
+      "step": 141490
+    },
+    {
+      "epoch": 912.9,
+      "learning_rate": 4.838709677419355e-06,
+      "loss": 0.2039,
+      "step": 141500
+    },
+    {
+      "epoch": 912.96,
+      "learning_rate": 4.835125448028674e-06,
+      "loss": 0.1818,
+      "step": 141510
+    },
+    {
+      "epoch": 913.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9235844612121582,
+      "eval_runtime": 0.8816,
+      "eval_samples_per_second": 227.987,
+      "eval_steps_per_second": 7.94,
+      "step": 141515
+    },
+    {
+      "epoch": 913.03,
+      "learning_rate": 4.831541218637993e-06,
+      "loss": 0.2212,
+      "step": 141520
+    },
+    {
+      "epoch": 913.1,
+      "learning_rate": 4.827956989247312e-06,
+      "loss": 0.1573,
+      "step": 141530
+    },
+    {
+      "epoch": 913.16,
+      "learning_rate": 4.824372759856631e-06,
+      "loss": 0.2459,
+      "step": 141540
+    },
+    {
+      "epoch": 913.23,
+      "learning_rate": 4.82078853046595e-06,
+      "loss": 0.2164,
+      "step": 141550
+    },
+    {
+      "epoch": 913.29,
+      "learning_rate": 4.817204301075269e-06,
+      "loss": 0.1774,
+      "step": 141560
+    },
+    {
+      "epoch": 913.35,
+      "learning_rate": 4.813620071684588e-06,
+      "loss": 0.1881,
+      "step": 141570
+    },
+    {
+      "epoch": 913.42,
+      "learning_rate": 4.810035842293907e-06,
+      "loss": 0.2135,
+      "step": 141580
+    },
+    {
+      "epoch": 913.48,
+      "learning_rate": 4.806451612903226e-06,
+      "loss": 0.2115,
+      "step": 141590
+    },
+    {
+      "epoch": 913.55,
+      "learning_rate": 4.802867383512545e-06,
+      "loss": 0.2268,
+      "step": 141600
+    },
+    {
+      "epoch": 913.61,
+      "learning_rate": 4.799283154121864e-06,
+      "loss": 0.1901,
+      "step": 141610
+    },
+    {
+      "epoch": 913.68,
+      "learning_rate": 4.795698924731183e-06,
+      "loss": 0.1729,
+      "step": 141620
+    },
+    {
+      "epoch": 913.74,
+      "learning_rate": 4.792114695340502e-06,
+      "loss": 0.1885,
+      "step": 141630
+    },
+    {
+      "epoch": 913.8,
+      "learning_rate": 4.788530465949821e-06,
+      "loss": 0.2213,
+      "step": 141640
+    },
+    {
+      "epoch": 913.87,
+      "learning_rate": 4.78494623655914e-06,
+      "loss": 0.2198,
+      "step": 141650
+    },
+    {
+      "epoch": 913.93,
+      "learning_rate": 4.781362007168459e-06,
+      "loss": 0.1849,
+      "step": 141660
+    },
+    {
+      "epoch": 914.0,
+      "learning_rate": 4.777777777777778e-06,
+      "loss": 0.2197,
+      "step": 141670
+    },
+    {
+      "epoch": 914.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9278761744499207,
+      "eval_runtime": 0.8681,
+      "eval_samples_per_second": 231.528,
+      "eval_steps_per_second": 8.063,
+      "step": 141670
+    },
+    {
+      "epoch": 914.06,
+      "learning_rate": 4.774193548387097e-06,
+      "loss": 0.2261,
+      "step": 141680
+    },
+    {
+      "epoch": 914.13,
+      "learning_rate": 4.770609318996416e-06,
+      "loss": 0.2094,
+      "step": 141690
+    },
+    {
+      "epoch": 914.19,
+      "learning_rate": 4.767025089605735e-06,
+      "loss": 0.2418,
+      "step": 141700
+    },
+    {
+      "epoch": 914.26,
+      "learning_rate": 4.7634408602150545e-06,
+      "loss": 0.2219,
+      "step": 141710
+    },
+    {
+      "epoch": 914.32,
+      "learning_rate": 4.7598566308243735e-06,
+      "loss": 0.1785,
+      "step": 141720
+    },
+    {
+      "epoch": 914.39,
+      "learning_rate": 4.756272401433692e-06,
+      "loss": 0.1382,
+      "step": 141730
+    },
+    {
+      "epoch": 914.45,
+      "learning_rate": 4.752688172043011e-06,
+      "loss": 0.2106,
+      "step": 141740
+    },
+    {
+      "epoch": 914.51,
+      "learning_rate": 4.74910394265233e-06,
+      "loss": 0.189,
+      "step": 141750
+    },
+    {
+      "epoch": 914.58,
+      "learning_rate": 4.745519713261649e-06,
+      "loss": 0.2004,
+      "step": 141760
+    },
+    {
+      "epoch": 914.64,
+      "learning_rate": 4.741935483870968e-06,
+      "loss": 0.1783,
+      "step": 141770
+    },
+    {
+      "epoch": 914.71,
+      "learning_rate": 4.738351254480287e-06,
+      "loss": 0.2287,
+      "step": 141780
+    },
+    {
+      "epoch": 914.77,
+      "learning_rate": 4.734767025089606e-06,
+      "loss": 0.2781,
+      "step": 141790
+    },
+    {
+      "epoch": 914.84,
+      "learning_rate": 4.731182795698925e-06,
+      "loss": 0.2119,
+      "step": 141800
+    },
+    {
+      "epoch": 914.9,
+      "learning_rate": 4.727598566308244e-06,
+      "loss": 0.2422,
+      "step": 141810
+    },
+    {
+      "epoch": 914.96,
+      "learning_rate": 4.724014336917563e-06,
+      "loss": 0.1879,
+      "step": 141820
+    },
+    {
+      "epoch": 915.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.916326105594635,
+      "eval_runtime": 0.8744,
+      "eval_samples_per_second": 229.879,
+      "eval_steps_per_second": 8.006,
+      "step": 141825
+    },
+    {
+      "epoch": 915.03,
+      "learning_rate": 4.720430107526882e-06,
+      "loss": 0.2492,
+      "step": 141830
+    },
+    {
+      "epoch": 915.1,
+      "learning_rate": 4.7168458781362004e-06,
+      "loss": 0.1972,
+      "step": 141840
+    },
+    {
+      "epoch": 915.16,
+      "learning_rate": 4.7132616487455195e-06,
+      "loss": 0.161,
+      "step": 141850
+    },
+    {
+      "epoch": 915.23,
+      "learning_rate": 4.7096774193548385e-06,
+      "loss": 0.2055,
+      "step": 141860
+    },
+    {
+      "epoch": 915.29,
+      "learning_rate": 4.706093189964158e-06,
+      "loss": 0.2278,
+      "step": 141870
+    },
+    {
+      "epoch": 915.35,
+      "learning_rate": 4.702508960573477e-06,
+      "loss": 0.1718,
+      "step": 141880
+    },
+    {
+      "epoch": 915.42,
+      "learning_rate": 4.698924731182796e-06,
+      "loss": 0.195,
+      "step": 141890
+    },
+    {
+      "epoch": 915.48,
+      "learning_rate": 4.695340501792115e-06,
+      "loss": 0.1789,
+      "step": 141900
+    },
+    {
+      "epoch": 915.55,
+      "learning_rate": 4.691756272401434e-06,
+      "loss": 0.1737,
+      "step": 141910
+    },
+    {
+      "epoch": 915.61,
+      "learning_rate": 4.688172043010753e-06,
+      "loss": 0.1939,
+      "step": 141920
+    },
+    {
+      "epoch": 915.68,
+      "learning_rate": 4.684587813620072e-06,
+      "loss": 0.1956,
+      "step": 141930
+    },
+    {
+      "epoch": 915.74,
+      "learning_rate": 4.681003584229391e-06,
+      "loss": 0.2134,
+      "step": 141940
+    },
+    {
+      "epoch": 915.8,
+      "learning_rate": 4.67741935483871e-06,
+      "loss": 0.2055,
+      "step": 141950
+    },
+    {
+      "epoch": 915.87,
+      "learning_rate": 4.673835125448029e-06,
+      "loss": 0.208,
+      "step": 141960
+    },
+    {
+      "epoch": 915.93,
+      "learning_rate": 4.670250896057348e-06,
+      "loss": 0.1736,
+      "step": 141970
+    },
+    {
+      "epoch": 916.0,
+      "learning_rate": 4.666666666666667e-06,
+      "loss": 0.2135,
+      "step": 141980
+    },
+    {
+      "epoch": 916.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9091780185699463,
+      "eval_runtime": 0.8601,
+      "eval_samples_per_second": 233.697,
+      "eval_steps_per_second": 8.139,
+      "step": 141980
+    },
+    {
+      "epoch": 916.06,
+      "learning_rate": 4.663082437275986e-06,
+      "loss": 0.2235,
+      "step": 141990
+    },
+    {
+      "epoch": 916.13,
+      "learning_rate": 4.659498207885305e-06,
+      "loss": 0.2319,
+      "step": 142000
+    },
+    {
+      "epoch": 916.19,
+      "learning_rate": 4.655913978494624e-06,
+      "loss": 0.1795,
+      "step": 142010
+    },
+    {
+      "epoch": 916.26,
+      "learning_rate": 4.652329749103943e-06,
+      "loss": 0.2165,
+      "step": 142020
+    },
+    {
+      "epoch": 916.32,
+      "learning_rate": 4.648745519713262e-06,
+      "loss": 0.2255,
+      "step": 142030
+    },
+    {
+      "epoch": 916.39,
+      "learning_rate": 4.6451612903225815e-06,
+      "loss": 0.2394,
+      "step": 142040
+    },
+    {
+      "epoch": 916.45,
+      "learning_rate": 4.6415770609319e-06,
+      "loss": 0.1564,
+      "step": 142050
+    },
+    {
+      "epoch": 916.51,
+      "learning_rate": 4.637992831541219e-06,
+      "loss": 0.1419,
+      "step": 142060
+    },
+    {
+      "epoch": 916.58,
+      "learning_rate": 4.634408602150538e-06,
+      "loss": 0.2388,
+      "step": 142070
+    },
+    {
+      "epoch": 916.64,
+      "learning_rate": 4.630824372759857e-06,
+      "loss": 0.2012,
+      "step": 142080
+    },
+    {
+      "epoch": 916.71,
+      "learning_rate": 4.627240143369176e-06,
+      "loss": 0.2058,
+      "step": 142090
+    },
+    {
+      "epoch": 916.77,
+      "learning_rate": 4.623655913978495e-06,
+      "loss": 0.1662,
+      "step": 142100
+    },
+    {
+      "epoch": 916.84,
+      "learning_rate": 4.620071684587814e-06,
+      "loss": 0.184,
+      "step": 142110
+    },
+    {
+      "epoch": 916.9,
+      "learning_rate": 4.616487455197133e-06,
+      "loss": 0.2175,
+      "step": 142120
+    },
+    {
+      "epoch": 916.96,
+      "learning_rate": 4.612903225806452e-06,
+      "loss": 0.1876,
+      "step": 142130
+    },
+    {
+      "epoch": 917.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.8939963579177856,
+      "eval_runtime": 0.8819,
+      "eval_samples_per_second": 227.926,
+      "eval_steps_per_second": 7.938,
+      "step": 142135
+    },
+    {
+      "epoch": 917.03,
+      "learning_rate": 4.60931899641577e-06,
+      "loss": 0.1988,
+      "step": 142140
+    },
+    {
+      "epoch": 917.1,
+      "learning_rate": 4.605734767025089e-06,
+      "loss": 0.1579,
+      "step": 142150
+    },
+    {
+      "epoch": 917.16,
+      "learning_rate": 4.602150537634408e-06,
+      "loss": 0.1942,
+      "step": 142160
+    },
+    {
+      "epoch": 917.23,
+      "learning_rate": 4.5985663082437275e-06,
+      "loss": 0.2081,
+      "step": 142170
+    },
+    {
+      "epoch": 917.29,
+      "learning_rate": 4.5949820788530465e-06,
+      "loss": 0.1748,
+      "step": 142180
+    },
+    {
+      "epoch": 917.35,
+      "learning_rate": 4.5913978494623656e-06,
+      "loss": 0.1858,
+      "step": 142190
+    },
+    {
+      "epoch": 917.42,
+      "learning_rate": 4.587813620071685e-06,
+      "loss": 0.2092,
+      "step": 142200
+    },
+    {
+      "epoch": 917.48,
+      "learning_rate": 4.584229390681004e-06,
+      "loss": 0.1806,
+      "step": 142210
+    },
+    {
+      "epoch": 917.55,
+      "learning_rate": 4.580645161290323e-06,
+      "loss": 0.2214,
+      "step": 142220
+    },
+    {
+      "epoch": 917.61,
+      "learning_rate": 4.577060931899642e-06,
+      "loss": 0.2091,
+      "step": 142230
+    },
+    {
+      "epoch": 917.68,
+      "learning_rate": 4.573476702508961e-06,
+      "loss": 0.2075,
+      "step": 142240
+    },
+    {
+      "epoch": 917.74,
+      "learning_rate": 4.56989247311828e-06,
+      "loss": 0.2304,
+      "step": 142250
+    },
+    {
+      "epoch": 917.8,
+      "learning_rate": 4.566308243727599e-06,
+      "loss": 0.2183,
+      "step": 142260
+    },
+    {
+      "epoch": 917.87,
+      "learning_rate": 4.562724014336918e-06,
+      "loss": 0.2473,
+      "step": 142270
+    },
+    {
+      "epoch": 917.93,
+      "learning_rate": 4.559139784946237e-06,
+      "loss": 0.2038,
+      "step": 142280
+    },
+    {
+      "epoch": 918.0,
+      "learning_rate": 4.555555555555556e-06,
+      "loss": 0.2047,
+      "step": 142290
+    },
+    {
+      "epoch": 918.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9192139506340027,
+      "eval_runtime": 0.8683,
+      "eval_samples_per_second": 231.474,
+      "eval_steps_per_second": 8.061,
+      "step": 142290
+    },
+    {
+      "epoch": 918.06,
+      "learning_rate": 4.551971326164875e-06,
+      "loss": 0.1901,
+      "step": 142300
+    },
+    {
+      "epoch": 918.13,
+      "learning_rate": 4.548387096774194e-06,
+      "loss": 0.1665,
+      "step": 142310
+    },
+    {
+      "epoch": 918.19,
+      "learning_rate": 4.544802867383513e-06,
+      "loss": 0.2094,
+      "step": 142320
+    },
+    {
+      "epoch": 918.26,
+      "learning_rate": 4.541218637992832e-06,
+      "loss": 0.1697,
+      "step": 142330
+    },
+    {
+      "epoch": 918.32,
+      "learning_rate": 4.537634408602151e-06,
+      "loss": 0.2009,
+      "step": 142340
+    },
+    {
+      "epoch": 918.39,
+      "learning_rate": 4.53405017921147e-06,
+      "loss": 0.1898,
+      "step": 142350
+    },
+    {
+      "epoch": 918.45,
+      "learning_rate": 4.5304659498207894e-06,
+      "loss": 0.1795,
+      "step": 142360
+    },
+    {
+      "epoch": 918.51,
+      "learning_rate": 4.526881720430108e-06,
+      "loss": 0.2009,
+      "step": 142370
+    },
+    {
+      "epoch": 918.58,
+      "learning_rate": 4.523297491039427e-06,
+      "loss": 0.1977,
+      "step": 142380
+    },
+    {
+      "epoch": 918.64,
+      "learning_rate": 4.519713261648746e-06,
+      "loss": 0.1986,
+      "step": 142390
+    },
+    {
+      "epoch": 918.71,
+      "learning_rate": 4.516129032258065e-06,
+      "loss": 0.2142,
+      "step": 142400
+    },
+    {
+      "epoch": 918.77,
+      "learning_rate": 4.512544802867384e-06,
+      "loss": 0.1934,
+      "step": 142410
+    },
+    {
+      "epoch": 918.84,
+      "learning_rate": 4.508960573476703e-06,
+      "loss": 0.1923,
+      "step": 142420
+    },
+    {
+      "epoch": 918.9,
+      "learning_rate": 4.505376344086021e-06,
+      "loss": 0.1747,
+      "step": 142430
+    },
+    {
+      "epoch": 918.96,
+      "learning_rate": 4.50179211469534e-06,
+      "loss": 0.1744,
+      "step": 142440
+    },
+    {
+      "epoch": 919.0,
+      "eval_accuracy": 0.7910447761194029,
+      "eval_loss": 0.90620356798172,
+      "eval_runtime": 0.9005,
+      "eval_samples_per_second": 223.22,
+      "eval_steps_per_second": 7.774,
+      "step": 142445
+    },
+    {
+      "epoch": 919.03,
+      "learning_rate": 4.498207885304659e-06,
+      "loss": 0.2416,
+      "step": 142450
+    },
+    {
+      "epoch": 919.1,
+      "learning_rate": 4.494623655913978e-06,
+      "loss": 0.2456,
+      "step": 142460
+    },
+    {
+      "epoch": 919.16,
+      "learning_rate": 4.491039426523297e-06,
+      "loss": 0.1592,
+      "step": 142470
+    },
+    {
+      "epoch": 919.23,
+      "learning_rate": 4.487455197132616e-06,
+      "loss": 0.1453,
+      "step": 142480
+    },
+    {
+      "epoch": 919.29,
+      "learning_rate": 4.4838709677419354e-06,
+      "loss": 0.2026,
+      "step": 142490
+    },
+    {
+      "epoch": 919.35,
+      "learning_rate": 4.4802867383512545e-06,
+      "loss": 0.2489,
+      "step": 142500
+    },
+    {
+      "epoch": 919.42,
+      "learning_rate": 4.4767025089605735e-06,
+      "loss": 0.1712,
+      "step": 142510
+    },
+    {
+      "epoch": 919.48,
+      "learning_rate": 4.473118279569893e-06,
+      "loss": 0.1791,
+      "step": 142520
+    },
+    {
+      "epoch": 919.55,
+      "learning_rate": 4.469534050179212e-06,
+      "loss": 0.1652,
+      "step": 142530
+    },
+    {
+      "epoch": 919.61,
+      "learning_rate": 4.465949820788531e-06,
+      "loss": 0.1863,
+      "step": 142540
+    },
+    {
+      "epoch": 919.68,
+      "learning_rate": 4.46236559139785e-06,
+      "loss": 0.2284,
+      "step": 142550
+    },
+    {
+      "epoch": 919.74,
+      "learning_rate": 4.458781362007169e-06,
+      "loss": 0.2317,
+      "step": 142560
+    },
+    {
+      "epoch": 919.8,
+      "learning_rate": 4.455197132616488e-06,
+      "loss": 0.1946,
+      "step": 142570
+    },
+    {
+      "epoch": 919.87,
+      "learning_rate": 4.451612903225807e-06,
+      "loss": 0.1743,
+      "step": 142580
+    },
+    {
+      "epoch": 919.93,
+      "learning_rate": 4.448028673835126e-06,
+      "loss": 0.2161,
+      "step": 142590
+    },
+    {
+      "epoch": 920.0,
+      "learning_rate": 4.444444444444445e-06,
+      "loss": 0.2091,
+      "step": 142600
+    },
+    {
+      "epoch": 920.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9265903234481812,
+      "eval_runtime": 0.8898,
+      "eval_samples_per_second": 225.902,
+      "eval_steps_per_second": 7.867,
+      "step": 142600
+    },
+    {
+      "epoch": 920.06,
+      "learning_rate": 4.440860215053764e-06,
+      "loss": 0.2367,
+      "step": 142610
+    },
+    {
+      "epoch": 920.13,
+      "learning_rate": 4.437275985663083e-06,
+      "loss": 0.1804,
+      "step": 142620
+    },
+    {
+      "epoch": 920.19,
+      "learning_rate": 4.433691756272402e-06,
+      "loss": 0.2111,
+      "step": 142630
+    },
+    {
+      "epoch": 920.26,
+      "learning_rate": 4.430107526881721e-06,
+      "loss": 0.2235,
+      "step": 142640
+    },
+    {
+      "epoch": 920.32,
+      "learning_rate": 4.42652329749104e-06,
+      "loss": 0.2035,
+      "step": 142650
+    },
+    {
+      "epoch": 920.39,
+      "learning_rate": 4.422939068100359e-06,
+      "loss": 0.2158,
+      "step": 142660
+    },
+    {
+      "epoch": 920.45,
+      "learning_rate": 4.419354838709678e-06,
+      "loss": 0.1631,
+      "step": 142670
+    },
+    {
+      "epoch": 920.51,
+      "learning_rate": 4.4157706093189966e-06,
+      "loss": 0.2057,
+      "step": 142680
+    },
+    {
+      "epoch": 920.58,
+      "learning_rate": 4.412186379928316e-06,
+      "loss": 0.2,
+      "step": 142690
+    },
+    {
+      "epoch": 920.64,
+      "learning_rate": 4.408602150537635e-06,
+      "loss": 0.2091,
+      "step": 142700
+    },
+    {
+      "epoch": 920.71,
+      "learning_rate": 4.405017921146954e-06,
+      "loss": 0.1879,
+      "step": 142710
+    },
+    {
+      "epoch": 920.77,
+      "learning_rate": 4.401433691756272e-06,
+      "loss": 0.1773,
+      "step": 142720
+    },
+    {
+      "epoch": 920.84,
+      "learning_rate": 4.397849462365591e-06,
+      "loss": 0.1646,
+      "step": 142730
+    },
+    {
+      "epoch": 920.9,
+      "learning_rate": 4.39426523297491e-06,
+      "loss": 0.1508,
+      "step": 142740
+    },
+    {
+      "epoch": 920.96,
+      "learning_rate": 4.390681003584229e-06,
+      "loss": 0.2237,
+      "step": 142750
+    },
+    {
+      "epoch": 921.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9291101694107056,
+      "eval_runtime": 0.8646,
+      "eval_samples_per_second": 232.473,
+      "eval_steps_per_second": 8.096,
+      "step": 142755
+    },
+    {
+      "epoch": 921.03,
+      "learning_rate": 4.387096774193548e-06,
+      "loss": 0.2353,
+      "step": 142760
+    },
+    {
+      "epoch": 921.1,
+      "learning_rate": 4.383512544802867e-06,
+      "loss": 0.174,
+      "step": 142770
+    },
+    {
+      "epoch": 921.16,
+      "learning_rate": 4.379928315412186e-06,
+      "loss": 0.1656,
+      "step": 142780
+    },
+    {
+      "epoch": 921.23,
+      "learning_rate": 4.376344086021505e-06,
+      "loss": 0.2179,
+      "step": 142790
+    },
+    {
+      "epoch": 921.29,
+      "learning_rate": 4.372759856630824e-06,
+      "loss": 0.2421,
+      "step": 142800
+    },
+    {
+      "epoch": 921.35,
+      "learning_rate": 4.369175627240143e-06,
+      "loss": 0.23,
+      "step": 142810
+    },
+    {
+      "epoch": 921.42,
+      "learning_rate": 4.3655913978494624e-06,
+      "loss": 0.1829,
+      "step": 142820
+    },
+    {
+      "epoch": 921.48,
+      "learning_rate": 4.3620071684587815e-06,
+      "loss": 0.1895,
+      "step": 142830
+    },
+    {
+      "epoch": 921.55,
+      "learning_rate": 4.3584229390681005e-06,
+      "loss": 0.2387,
+      "step": 142840
+    },
+    {
+      "epoch": 921.61,
+      "learning_rate": 4.35483870967742e-06,
+      "loss": 0.2261,
+      "step": 142850
+    },
+    {
+      "epoch": 921.68,
+      "learning_rate": 4.351254480286739e-06,
+      "loss": 0.1582,
+      "step": 142860
+    },
+    {
+      "epoch": 921.74,
+      "learning_rate": 4.347670250896058e-06,
+      "loss": 0.1892,
+      "step": 142870
+    },
+    {
+      "epoch": 921.8,
+      "learning_rate": 4.344086021505377e-06,
+      "loss": 0.1854,
+      "step": 142880
+    },
+    {
+      "epoch": 921.87,
+      "learning_rate": 4.340501792114696e-06,
+      "loss": 0.203,
+      "step": 142890
+    },
+    {
+      "epoch": 921.93,
+      "learning_rate": 4.336917562724015e-06,
+      "loss": 0.2011,
+      "step": 142900
+    },
+    {
+      "epoch": 922.0,
+      "learning_rate": 4.333333333333334e-06,
+      "loss": 0.1442,
+      "step": 142910
+    },
+    {
+      "epoch": 922.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9226329326629639,
+      "eval_runtime": 0.913,
+      "eval_samples_per_second": 220.161,
+      "eval_steps_per_second": 7.667,
+      "step": 142910
+    },
+    {
+      "epoch": 922.06,
+      "learning_rate": 4.329749103942653e-06,
+      "loss": 0.1892,
+      "step": 142920
+    },
+    {
+      "epoch": 922.13,
+      "learning_rate": 4.326164874551972e-06,
+      "loss": 0.1999,
+      "step": 142930
+    },
+    {
+      "epoch": 922.19,
+      "learning_rate": 4.322580645161291e-06,
+      "loss": 0.1968,
+      "step": 142940
+    },
+    {
+      "epoch": 922.26,
+      "learning_rate": 4.31899641577061e-06,
+      "loss": 0.152,
+      "step": 142950
+    },
+    {
+      "epoch": 922.32,
+      "learning_rate": 4.315412186379929e-06,
+      "loss": 0.1871,
+      "step": 142960
+    },
+    {
+      "epoch": 922.39,
+      "learning_rate": 4.311827956989247e-06,
+      "loss": 0.2144,
+      "step": 142970
+    },
+    {
+      "epoch": 922.45,
+      "learning_rate": 4.3082437275985664e-06,
+      "loss": 0.2187,
+      "step": 142980
+    },
+    {
+      "epoch": 922.51,
+      "learning_rate": 4.3046594982078855e-06,
+      "loss": 0.1985,
+      "step": 142990
+    },
+    {
+      "epoch": 922.58,
+      "learning_rate": 4.3010752688172045e-06,
+      "loss": 0.2069,
+      "step": 143000
+    },
+    {
+      "epoch": 922.64,
+      "learning_rate": 4.297491039426524e-06,
+      "loss": 0.2077,
+      "step": 143010
+    },
+    {
+      "epoch": 922.71,
+      "learning_rate": 4.293906810035842e-06,
+      "loss": 0.2102,
+      "step": 143020
+    },
+    {
+      "epoch": 922.77,
+      "learning_rate": 4.290322580645161e-06,
+      "loss": 0.2217,
+      "step": 143030
+    },
+    {
+      "epoch": 922.84,
+      "learning_rate": 4.28673835125448e-06,
+      "loss": 0.1894,
+      "step": 143040
+    },
+    {
+      "epoch": 922.9,
+      "learning_rate": 4.283154121863799e-06,
+      "loss": 0.2387,
+      "step": 143050
+    },
+    {
+      "epoch": 922.96,
+      "learning_rate": 4.279569892473118e-06,
+      "loss": 0.1334,
+      "step": 143060
+    },
+    {
+      "epoch": 923.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9204243421554565,
+      "eval_runtime": 0.8983,
+      "eval_samples_per_second": 223.747,
+      "eval_steps_per_second": 7.792,
+      "step": 143065
+    },
+    {
+      "epoch": 923.03,
+      "learning_rate": 4.275985663082437e-06,
+      "loss": 0.2249,
+      "step": 143070
+    },
+    {
+      "epoch": 923.1,
+      "learning_rate": 4.272401433691756e-06,
+      "loss": 0.1918,
+      "step": 143080
+    },
+    {
+      "epoch": 923.16,
+      "learning_rate": 4.268817204301075e-06,
+      "loss": 0.2002,
+      "step": 143090
+    },
+    {
+      "epoch": 923.23,
+      "learning_rate": 4.265232974910394e-06,
+      "loss": 0.1985,
+      "step": 143100
+    },
+    {
+      "epoch": 923.29,
+      "learning_rate": 4.261648745519713e-06,
+      "loss": 0.2198,
+      "step": 143110
+    },
+    {
+      "epoch": 923.35,
+      "learning_rate": 4.258064516129032e-06,
+      "loss": 0.2396,
+      "step": 143120
+    },
+    {
+      "epoch": 923.42,
+      "learning_rate": 4.254480286738351e-06,
+      "loss": 0.1705,
+      "step": 143130
+    },
+    {
+      "epoch": 923.48,
+      "learning_rate": 4.25089605734767e-06,
+      "loss": 0.2364,
+      "step": 143140
+    },
+    {
+      "epoch": 923.55,
+      "learning_rate": 4.2473118279569895e-06,
+      "loss": 0.1733,
+      "step": 143150
+    },
+    {
+      "epoch": 923.61,
+      "learning_rate": 4.2437275985663085e-06,
+      "loss": 0.1887,
+      "step": 143160
+    },
+    {
+      "epoch": 923.68,
+      "learning_rate": 4.2401433691756276e-06,
+      "loss": 0.1798,
+      "step": 143170
+    },
+    {
+      "epoch": 923.74,
+      "learning_rate": 4.236559139784947e-06,
+      "loss": 0.1846,
+      "step": 143180
+    },
+    {
+      "epoch": 923.8,
+      "learning_rate": 4.232974910394266e-06,
+      "loss": 0.1795,
+      "step": 143190
+    },
+    {
+      "epoch": 923.87,
+      "learning_rate": 4.229390681003585e-06,
+      "loss": 0.1957,
+      "step": 143200
+    },
+    {
+      "epoch": 923.93,
+      "learning_rate": 4.225806451612904e-06,
+      "loss": 0.1859,
+      "step": 143210
+    },
+    {
+      "epoch": 924.0,
+      "learning_rate": 4.222222222222223e-06,
+      "loss": 0.1525,
+      "step": 143220
+    },
+    {
+      "epoch": 924.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9542303085327148,
+      "eval_runtime": 0.8867,
+      "eval_samples_per_second": 226.681,
+      "eval_steps_per_second": 7.894,
+      "step": 143220
+    },
+    {
+      "epoch": 924.06,
+      "learning_rate": 4.218637992831542e-06,
+      "loss": 0.203,
+      "step": 143230
+    },
+    {
+      "epoch": 924.13,
+      "learning_rate": 4.215053763440861e-06,
+      "loss": 0.2067,
+      "step": 143240
+    },
+    {
+      "epoch": 924.19,
+      "learning_rate": 4.21146953405018e-06,
+      "loss": 0.1992,
+      "step": 143250
+    },
+    {
+      "epoch": 924.26,
+      "learning_rate": 4.207885304659499e-06,
+      "loss": 0.18,
+      "step": 143260
+    },
+    {
+      "epoch": 924.32,
+      "learning_rate": 4.204301075268817e-06,
+      "loss": 0.1961,
+      "step": 143270
+    },
+    {
+      "epoch": 924.39,
+      "learning_rate": 4.200716845878136e-06,
+      "loss": 0.242,
+      "step": 143280
+    },
+    {
+      "epoch": 924.45,
+      "learning_rate": 4.197132616487455e-06,
+      "loss": 0.1931,
+      "step": 143290
+    },
+    {
+      "epoch": 924.51,
+      "learning_rate": 4.193548387096774e-06,
+      "loss": 0.198,
+      "step": 143300
+    },
+    {
+      "epoch": 924.58,
+      "learning_rate": 4.1899641577060935e-06,
+      "loss": 0.2104,
+      "step": 143310
+    },
+    {
+      "epoch": 924.64,
+      "learning_rate": 4.1863799283154125e-06,
+      "loss": 0.1695,
+      "step": 143320
+    },
+    {
+      "epoch": 924.71,
+      "learning_rate": 4.182795698924731e-06,
+      "loss": 0.2134,
+      "step": 143330
+    },
+    {
+      "epoch": 924.77,
+      "learning_rate": 4.17921146953405e-06,
+      "loss": 0.2403,
+      "step": 143340
+    },
+    {
+      "epoch": 924.84,
+      "learning_rate": 4.175627240143369e-06,
+      "loss": 0.1958,
+      "step": 143350
+    },
+    {
+      "epoch": 924.9,
+      "learning_rate": 4.172043010752688e-06,
+      "loss": 0.2102,
+      "step": 143360
+    },
+    {
+      "epoch": 924.96,
+      "learning_rate": 4.168458781362007e-06,
+      "loss": 0.2168,
+      "step": 143370
+    },
+    {
+      "epoch": 925.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9341832399368286,
+      "eval_runtime": 0.876,
+      "eval_samples_per_second": 229.451,
+      "eval_steps_per_second": 7.991,
+      "step": 143375
+    },
+    {
+      "epoch": 925.03,
+      "learning_rate": 4.164874551971326e-06,
+      "loss": 0.1906,
+      "step": 143380
+    },
+    {
+      "epoch": 925.1,
+      "learning_rate": 4.161290322580645e-06,
+      "loss": 0.1983,
+      "step": 143390
+    },
+    {
+      "epoch": 925.16,
+      "learning_rate": 4.157706093189964e-06,
+      "loss": 0.2283,
+      "step": 143400
+    },
+    {
+      "epoch": 925.23,
+      "learning_rate": 4.154121863799283e-06,
+      "loss": 0.1938,
+      "step": 143410
+    },
+    {
+      "epoch": 925.29,
+      "learning_rate": 4.150537634408602e-06,
+      "loss": 0.1803,
+      "step": 143420
+    },
+    {
+      "epoch": 925.35,
+      "learning_rate": 4.146953405017921e-06,
+      "loss": 0.2015,
+      "step": 143430
+    },
+    {
+      "epoch": 925.42,
+      "learning_rate": 4.14336917562724e-06,
+      "loss": 0.1988,
+      "step": 143440
+    },
+    {
+      "epoch": 925.48,
+      "learning_rate": 4.139784946236559e-06,
+      "loss": 0.1906,
+      "step": 143450
+    },
+    {
+      "epoch": 925.55,
+      "learning_rate": 4.136200716845878e-06,
+      "loss": 0.2495,
+      "step": 143460
+    },
+    {
+      "epoch": 925.61,
+      "learning_rate": 4.1326164874551974e-06,
+      "loss": 0.1922,
+      "step": 143470
+    },
+    {
+      "epoch": 925.68,
+      "learning_rate": 4.1290322580645165e-06,
+      "loss": 0.231,
+      "step": 143480
+    },
+    {
+      "epoch": 925.74,
+      "learning_rate": 4.1254480286738355e-06,
+      "loss": 0.1928,
+      "step": 143490
+    },
+    {
+      "epoch": 925.8,
+      "learning_rate": 4.121863799283155e-06,
+      "loss": 0.1731,
+      "step": 143500
+    },
+    {
+      "epoch": 925.87,
+      "learning_rate": 4.118279569892474e-06,
+      "loss": 0.1794,
+      "step": 143510
+    },
+    {
+      "epoch": 925.93,
+      "learning_rate": 4.114695340501793e-06,
+      "loss": 0.2241,
+      "step": 143520
+    },
+    {
+      "epoch": 926.0,
+      "learning_rate": 4.111111111111112e-06,
+      "loss": 0.21,
+      "step": 143530
+    },
+    {
+      "epoch": 926.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9533934593200684,
+      "eval_runtime": 0.9022,
+      "eval_samples_per_second": 222.781,
+      "eval_steps_per_second": 7.759,
+      "step": 143530
+    },
+    {
+      "epoch": 926.06,
+      "learning_rate": 4.107526881720431e-06,
+      "loss": 0.1825,
+      "step": 143540
+    },
+    {
+      "epoch": 926.13,
+      "learning_rate": 4.10394265232975e-06,
+      "loss": 0.2006,
+      "step": 143550
+    },
+    {
+      "epoch": 926.19,
+      "learning_rate": 4.100358422939068e-06,
+      "loss": 0.1655,
+      "step": 143560
+    },
+    {
+      "epoch": 926.26,
+      "learning_rate": 4.096774193548387e-06,
+      "loss": 0.1964,
+      "step": 143570
+    },
+    {
+      "epoch": 926.32,
+      "learning_rate": 4.093189964157706e-06,
+      "loss": 0.2142,
+      "step": 143580
+    },
+    {
+      "epoch": 926.39,
+      "learning_rate": 4.089605734767025e-06,
+      "loss": 0.2272,
+      "step": 143590
+    },
+    {
+      "epoch": 926.45,
+      "learning_rate": 4.086021505376344e-06,
+      "loss": 0.2251,
+      "step": 143600
+    },
+    {
+      "epoch": 926.51,
+      "learning_rate": 4.082437275985663e-06,
+      "loss": 0.2478,
+      "step": 143610
+    },
+    {
+      "epoch": 926.58,
+      "learning_rate": 4.078853046594982e-06,
+      "loss": 0.186,
+      "step": 143620
+    },
+    {
+      "epoch": 926.64,
+      "learning_rate": 4.075268817204301e-06,
+      "loss": 0.1981,
+      "step": 143630
+    },
+    {
+      "epoch": 926.71,
+      "learning_rate": 4.0716845878136205e-06,
+      "loss": 0.1934,
+      "step": 143640
+    },
+    {
+      "epoch": 926.77,
+      "learning_rate": 4.068100358422939e-06,
+      "loss": 0.2421,
+      "step": 143650
+    },
+    {
+      "epoch": 926.84,
+      "learning_rate": 4.064516129032258e-06,
+      "loss": 0.2011,
+      "step": 143660
+    },
+    {
+      "epoch": 926.9,
+      "learning_rate": 4.060931899641577e-06,
+      "loss": 0.1799,
+      "step": 143670
+    },
+    {
+      "epoch": 926.96,
+      "learning_rate": 4.057347670250896e-06,
+      "loss": 0.1929,
+      "step": 143680
+    },
+    {
+      "epoch": 927.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9137667417526245,
+      "eval_runtime": 0.8781,
+      "eval_samples_per_second": 228.897,
+      "eval_steps_per_second": 7.972,
+      "step": 143685
+    },
+    {
+      "epoch": 927.03,
+      "learning_rate": 4.053763440860215e-06,
+      "loss": 0.1959,
+      "step": 143690
+    },
+    {
+      "epoch": 927.1,
+      "learning_rate": 4.050179211469534e-06,
+      "loss": 0.1902,
+      "step": 143700
+    },
+    {
+      "epoch": 927.16,
+      "learning_rate": 4.046594982078853e-06,
+      "loss": 0.1798,
+      "step": 143710
+    },
+    {
+      "epoch": 927.23,
+      "learning_rate": 4.043010752688172e-06,
+      "loss": 0.1778,
+      "step": 143720
+    },
+    {
+      "epoch": 927.29,
+      "learning_rate": 4.039426523297491e-06,
+      "loss": 0.2192,
+      "step": 143730
+    },
+    {
+      "epoch": 927.35,
+      "learning_rate": 4.03584229390681e-06,
+      "loss": 0.1683,
+      "step": 143740
+    },
+    {
+      "epoch": 927.42,
+      "learning_rate": 4.032258064516129e-06,
+      "loss": 0.2052,
+      "step": 143750
+    },
+    {
+      "epoch": 927.48,
+      "learning_rate": 4.028673835125448e-06,
+      "loss": 0.1424,
+      "step": 143760
+    },
+    {
+      "epoch": 927.55,
+      "learning_rate": 4.025089605734767e-06,
+      "loss": 0.1951,
+      "step": 143770
+    },
+    {
+      "epoch": 927.61,
+      "learning_rate": 4.021505376344086e-06,
+      "loss": 0.1536,
+      "step": 143780
+    },
+    {
+      "epoch": 927.68,
+      "learning_rate": 4.017921146953405e-06,
+      "loss": 0.1624,
+      "step": 143790
+    },
+    {
+      "epoch": 927.74,
+      "learning_rate": 4.0143369175627245e-06,
+      "loss": 0.1735,
+      "step": 143800
+    },
+    {
+      "epoch": 927.8,
+      "learning_rate": 4.0107526881720435e-06,
+      "loss": 0.1901,
+      "step": 143810
+    },
+    {
+      "epoch": 927.87,
+      "learning_rate": 4.0071684587813626e-06,
+      "loss": 0.2021,
+      "step": 143820
+    },
+    {
+      "epoch": 927.93,
+      "learning_rate": 4.003584229390682e-06,
+      "loss": 0.1717,
+      "step": 143830
+    },
+    {
+      "epoch": 928.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1875,
+      "step": 143840
+    },
+    {
+      "epoch": 928.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9482174515724182,
+      "eval_runtime": 0.9177,
+      "eval_samples_per_second": 219.022,
+      "eval_steps_per_second": 7.628,
+      "step": 143840
+    },
+    {
+      "epoch": 928.06,
+      "learning_rate": 3.996415770609319e-06,
+      "loss": 0.2301,
+      "step": 143850
+    },
+    {
+      "epoch": 928.13,
+      "learning_rate": 3.992831541218638e-06,
+      "loss": 0.2044,
+      "step": 143860
+    },
+    {
+      "epoch": 928.19,
+      "learning_rate": 3.989247311827957e-06,
+      "loss": 0.2301,
+      "step": 143870
+    },
+    {
+      "epoch": 928.26,
+      "learning_rate": 3.985663082437276e-06,
+      "loss": 0.2283,
+      "step": 143880
+    },
+    {
+      "epoch": 928.32,
+      "learning_rate": 3.982078853046595e-06,
+      "loss": 0.206,
+      "step": 143890
+    },
+    {
+      "epoch": 928.39,
+      "learning_rate": 3.978494623655914e-06,
+      "loss": 0.1689,
+      "step": 143900
+    },
+    {
+      "epoch": 928.45,
+      "learning_rate": 3.974910394265233e-06,
+      "loss": 0.1636,
+      "step": 143910
+    },
+    {
+      "epoch": 928.51,
+      "learning_rate": 3.971326164874552e-06,
+      "loss": 0.2243,
+      "step": 143920
+    },
+    {
+      "epoch": 928.58,
+      "learning_rate": 3.967741935483871e-06,
+      "loss": 0.193,
+      "step": 143930
+    },
+    {
+      "epoch": 928.64,
+      "learning_rate": 3.96415770609319e-06,
+      "loss": 0.2391,
+      "step": 143940
+    },
+    {
+      "epoch": 928.71,
+      "learning_rate": 3.960573476702509e-06,
+      "loss": 0.1726,
+      "step": 143950
+    },
+    {
+      "epoch": 928.77,
+      "learning_rate": 3.9569892473118284e-06,
+      "loss": 0.2178,
+      "step": 143960
+    },
+    {
+      "epoch": 928.84,
+      "learning_rate": 3.953405017921147e-06,
+      "loss": 0.2162,
+      "step": 143970
+    },
+    {
+      "epoch": 928.9,
+      "learning_rate": 3.949820788530466e-06,
+      "loss": 0.1749,
+      "step": 143980
+    },
+    {
+      "epoch": 928.96,
+      "learning_rate": 3.946236559139785e-06,
+      "loss": 0.1401,
+      "step": 143990
+    },
+    {
+      "epoch": 929.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9125292301177979,
+      "eval_runtime": 0.8625,
+      "eval_samples_per_second": 233.035,
+      "eval_steps_per_second": 8.116,
+      "step": 143995
+    },
+    {
+      "epoch": 929.03,
+      "learning_rate": 3.942652329749104e-06,
+      "loss": 0.2348,
+      "step": 144000
+    },
+    {
+      "epoch": 929.1,
+      "learning_rate": 3.939068100358423e-06,
+      "loss": 0.2073,
+      "step": 144010
+    },
+    {
+      "epoch": 929.16,
+      "learning_rate": 3.935483870967742e-06,
+      "loss": 0.1958,
+      "step": 144020
+    },
+    {
+      "epoch": 929.23,
+      "learning_rate": 3.931899641577061e-06,
+      "loss": 0.2206,
+      "step": 144030
+    },
+    {
+      "epoch": 929.29,
+      "learning_rate": 3.92831541218638e-06,
+      "loss": 0.1946,
+      "step": 144040
+    },
+    {
+      "epoch": 929.35,
+      "learning_rate": 3.924731182795699e-06,
+      "loss": 0.1969,
+      "step": 144050
+    },
+    {
+      "epoch": 929.42,
+      "learning_rate": 3.921146953405018e-06,
+      "loss": 0.1733,
+      "step": 144060
+    },
+    {
+      "epoch": 929.48,
+      "learning_rate": 3.917562724014337e-06,
+      "loss": 0.2458,
+      "step": 144070
+    },
+    {
+      "epoch": 929.55,
+      "learning_rate": 3.913978494623656e-06,
+      "loss": 0.251,
+      "step": 144080
+    },
+    {
+      "epoch": 929.61,
+      "learning_rate": 3.910394265232975e-06,
+      "loss": 0.152,
+      "step": 144090
+    },
+    {
+      "epoch": 929.68,
+      "learning_rate": 3.906810035842294e-06,
+      "loss": 0.206,
+      "step": 144100
+    },
+    {
+      "epoch": 929.74,
+      "learning_rate": 3.903225806451613e-06,
+      "loss": 0.1757,
+      "step": 144110
+    },
+    {
+      "epoch": 929.8,
+      "learning_rate": 3.8996415770609324e-06,
+      "loss": 0.1847,
+      "step": 144120
+    },
+    {
+      "epoch": 929.87,
+      "learning_rate": 3.8960573476702515e-06,
+      "loss": 0.1966,
+      "step": 144130
+    },
+    {
+      "epoch": 929.93,
+      "learning_rate": 3.8924731182795705e-06,
+      "loss": 0.2076,
+      "step": 144140
+    },
+    {
+      "epoch": 930.0,
+      "learning_rate": 3.888888888888889e-06,
+      "loss": 0.1832,
+      "step": 144150
+    },
+    {
+      "epoch": 930.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9022831916809082,
+      "eval_runtime": 0.9267,
+      "eval_samples_per_second": 216.897,
+      "eval_steps_per_second": 7.554,
+      "step": 144150
+    },
+    {
+      "epoch": 930.06,
+      "learning_rate": 3.885304659498208e-06,
+      "loss": 0.1904,
+      "step": 144160
+    },
+    {
+      "epoch": 930.13,
+      "learning_rate": 3.881720430107527e-06,
+      "loss": 0.2046,
+      "step": 144170
+    },
+    {
+      "epoch": 930.19,
+      "learning_rate": 3.878136200716846e-06,
+      "loss": 0.2033,
+      "step": 144180
+    },
+    {
+      "epoch": 930.26,
+      "learning_rate": 3.874551971326165e-06,
+      "loss": 0.2047,
+      "step": 144190
+    },
+    {
+      "epoch": 930.32,
+      "learning_rate": 3.870967741935484e-06,
+      "loss": 0.1809,
+      "step": 144200
+    },
+    {
+      "epoch": 930.39,
+      "learning_rate": 3.867383512544803e-06,
+      "loss": 0.2238,
+      "step": 144210
+    },
+    {
+      "epoch": 930.45,
+      "learning_rate": 3.863799283154122e-06,
+      "loss": 0.1509,
+      "step": 144220
+    },
+    {
+      "epoch": 930.51,
+      "learning_rate": 3.860215053763441e-06,
+      "loss": 0.2236,
+      "step": 144230
+    },
+    {
+      "epoch": 930.58,
+      "learning_rate": 3.85663082437276e-06,
+      "loss": 0.2245,
+      "step": 144240
+    },
+    {
+      "epoch": 930.64,
+      "learning_rate": 3.853046594982079e-06,
+      "loss": 0.2327,
+      "step": 144250
+    },
+    {
+      "epoch": 930.71,
+      "learning_rate": 3.849462365591398e-06,
+      "loss": 0.2174,
+      "step": 144260
+    },
+    {
+      "epoch": 930.77,
+      "learning_rate": 3.845878136200717e-06,
+      "loss": 0.2183,
+      "step": 144270
+    },
+    {
+      "epoch": 930.84,
+      "learning_rate": 3.842293906810036e-06,
+      "loss": 0.1699,
+      "step": 144280
+    },
+    {
+      "epoch": 930.9,
+      "learning_rate": 3.838709677419355e-06,
+      "loss": 0.1876,
+      "step": 144290
+    },
+    {
+      "epoch": 930.96,
+      "learning_rate": 3.835125448028674e-06,
+      "loss": 0.1962,
+      "step": 144300
+    },
+    {
+      "epoch": 931.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9277967214584351,
+      "eval_runtime": 0.8833,
+      "eval_samples_per_second": 227.553,
+      "eval_steps_per_second": 7.925,
+      "step": 144305
+    },
+    {
+      "epoch": 931.03,
+      "learning_rate": 3.831541218637993e-06,
+      "loss": 0.1904,
+      "step": 144310
+    },
+    {
+      "epoch": 931.1,
+      "learning_rate": 3.827956989247312e-06,
+      "loss": 0.1764,
+      "step": 144320
+    },
+    {
+      "epoch": 931.16,
+      "learning_rate": 3.824372759856631e-06,
+      "loss": 0.1849,
+      "step": 144330
+    },
+    {
+      "epoch": 931.23,
+      "learning_rate": 3.82078853046595e-06,
+      "loss": 0.219,
+      "step": 144340
+    },
+    {
+      "epoch": 931.29,
+      "learning_rate": 3.817204301075269e-06,
+      "loss": 0.197,
+      "step": 144350
+    },
+    {
+      "epoch": 931.35,
+      "learning_rate": 3.813620071684588e-06,
+      "loss": 0.1699,
+      "step": 144360
+    },
+    {
+      "epoch": 931.42,
+      "learning_rate": 3.810035842293907e-06,
+      "loss": 0.2083,
+      "step": 144370
+    },
+    {
+      "epoch": 931.48,
+      "learning_rate": 3.8064516129032257e-06,
+      "loss": 0.2057,
+      "step": 144380
+    },
+    {
+      "epoch": 931.55,
+      "learning_rate": 3.8028673835125447e-06,
+      "loss": 0.1884,
+      "step": 144390
+    },
+    {
+      "epoch": 931.61,
+      "learning_rate": 3.7992831541218638e-06,
+      "loss": 0.1889,
+      "step": 144400
+    },
+    {
+      "epoch": 931.68,
+      "learning_rate": 3.795698924731183e-06,
+      "loss": 0.199,
+      "step": 144410
+    },
+    {
+      "epoch": 931.74,
+      "learning_rate": 3.792114695340502e-06,
+      "loss": 0.1686,
+      "step": 144420
+    },
+    {
+      "epoch": 931.8,
+      "learning_rate": 3.788530465949821e-06,
+      "loss": 0.1883,
+      "step": 144430
+    },
+    {
+      "epoch": 931.87,
+      "learning_rate": 3.78494623655914e-06,
+      "loss": 0.2024,
+      "step": 144440
+    },
+    {
+      "epoch": 931.93,
+      "learning_rate": 3.781362007168459e-06,
+      "loss": 0.2356,
+      "step": 144450
+    },
+    {
+      "epoch": 932.0,
+      "learning_rate": 3.777777777777778e-06,
+      "loss": 0.2099,
+      "step": 144460
+    },
+    {
+      "epoch": 932.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9378855228424072,
+      "eval_runtime": 0.9912,
+      "eval_samples_per_second": 202.775,
+      "eval_steps_per_second": 7.062,
+      "step": 144460
+    },
+    {
+      "epoch": 932.06,
+      "learning_rate": 3.774193548387097e-06,
+      "loss": 0.1722,
+      "step": 144470
+    },
+    {
+      "epoch": 932.13,
+      "learning_rate": 3.770609318996416e-06,
+      "loss": 0.2026,
+      "step": 144480
+    },
+    {
+      "epoch": 932.19,
+      "learning_rate": 3.7670250896057352e-06,
+      "loss": 0.1557,
+      "step": 144490
+    },
+    {
+      "epoch": 932.26,
+      "learning_rate": 3.763440860215054e-06,
+      "loss": 0.1882,
+      "step": 144500
+    },
+    {
+      "epoch": 932.32,
+      "learning_rate": 3.759856630824373e-06,
+      "loss": 0.1795,
+      "step": 144510
+    },
+    {
+      "epoch": 932.39,
+      "learning_rate": 3.756272401433692e-06,
+      "loss": 0.1572,
+      "step": 144520
+    },
+    {
+      "epoch": 932.45,
+      "learning_rate": 3.752688172043011e-06,
+      "loss": 0.1713,
+      "step": 144530
+    },
+    {
+      "epoch": 932.51,
+      "learning_rate": 3.74910394265233e-06,
+      "loss": 0.2058,
+      "step": 144540
+    },
+    {
+      "epoch": 932.58,
+      "learning_rate": 3.745519713261649e-06,
+      "loss": 0.2321,
+      "step": 144550
+    },
+    {
+      "epoch": 932.64,
+      "learning_rate": 3.741935483870968e-06,
+      "loss": 0.1827,
+      "step": 144560
+    },
+    {
+      "epoch": 932.71,
+      "learning_rate": 3.7383512544802872e-06,
+      "loss": 0.2115,
+      "step": 144570
+    },
+    {
+      "epoch": 932.77,
+      "learning_rate": 3.7347670250896063e-06,
+      "loss": 0.1822,
+      "step": 144580
+    },
+    {
+      "epoch": 932.84,
+      "learning_rate": 3.7311827956989253e-06,
+      "loss": 0.1579,
+      "step": 144590
+    },
+    {
+      "epoch": 932.9,
+      "learning_rate": 3.7275985663082444e-06,
+      "loss": 0.2104,
+      "step": 144600
+    },
+    {
+      "epoch": 932.96,
+      "learning_rate": 3.7240143369175626e-06,
+      "loss": 0.1885,
+      "step": 144610
+    },
+    {
+      "epoch": 933.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9122596979141235,
+      "eval_runtime": 0.8876,
+      "eval_samples_per_second": 226.465,
+      "eval_steps_per_second": 7.887,
+      "step": 144615
+    },
+    {
+      "epoch": 933.03,
+      "learning_rate": 3.7204301075268816e-06,
+      "loss": 0.23,
+      "step": 144620
+    },
+    {
+      "epoch": 933.1,
+      "learning_rate": 3.7168458781362007e-06,
+      "loss": 0.1837,
+      "step": 144630
+    },
+    {
+      "epoch": 933.16,
+      "learning_rate": 3.7132616487455197e-06,
+      "loss": 0.2025,
+      "step": 144640
+    },
+    {
+      "epoch": 933.23,
+      "learning_rate": 3.709677419354839e-06,
+      "loss": 0.1606,
+      "step": 144650
+    },
+    {
+      "epoch": 933.29,
+      "learning_rate": 3.706093189964158e-06,
+      "loss": 0.2062,
+      "step": 144660
+    },
+    {
+      "epoch": 933.35,
+      "learning_rate": 3.7025089605734765e-06,
+      "loss": 0.2096,
+      "step": 144670
+    },
+    {
+      "epoch": 933.42,
+      "learning_rate": 3.6989247311827955e-06,
+      "loss": 0.1834,
+      "step": 144680
+    },
+    {
+      "epoch": 933.48,
+      "learning_rate": 3.6953405017921146e-06,
+      "loss": 0.1523,
+      "step": 144690
+    },
+    {
+      "epoch": 933.55,
+      "learning_rate": 3.6917562724014336e-06,
+      "loss": 0.1533,
+      "step": 144700
+    },
+    {
+      "epoch": 933.61,
+      "learning_rate": 3.6881720430107527e-06,
+      "loss": 0.1902,
+      "step": 144710
+    },
+    {
+      "epoch": 933.68,
+      "learning_rate": 3.6845878136200717e-06,
+      "loss": 0.213,
+      "step": 144720
+    },
+    {
+      "epoch": 933.74,
+      "learning_rate": 3.6810035842293908e-06,
+      "loss": 0.1933,
+      "step": 144730
+    },
+    {
+      "epoch": 933.8,
+      "learning_rate": 3.67741935483871e-06,
+      "loss": 0.234,
+      "step": 144740
+    },
+    {
+      "epoch": 933.87,
+      "learning_rate": 3.673835125448029e-06,
+      "loss": 0.1995,
+      "step": 144750
+    },
+    {
+      "epoch": 933.93,
+      "learning_rate": 3.670250896057348e-06,
+      "loss": 0.1315,
+      "step": 144760
+    },
+    {
+      "epoch": 934.0,
+      "learning_rate": 3.666666666666667e-06,
+      "loss": 0.2045,
+      "step": 144770
+    },
+    {
+      "epoch": 934.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9185261726379395,
+      "eval_runtime": 0.8919,
+      "eval_samples_per_second": 225.369,
+      "eval_steps_per_second": 7.849,
+      "step": 144770
+    },
+    {
+      "epoch": 934.06,
+      "learning_rate": 3.663082437275986e-06,
+      "loss": 0.2212,
+      "step": 144780
+    },
+    {
+      "epoch": 934.13,
+      "learning_rate": 3.659498207885305e-06,
+      "loss": 0.2037,
+      "step": 144790
+    },
+    {
+      "epoch": 934.19,
+      "learning_rate": 3.6559139784946237e-06,
+      "loss": 0.206,
+      "step": 144800
+    },
+    {
+      "epoch": 934.26,
+      "learning_rate": 3.6523297491039428e-06,
+      "loss": 0.1724,
+      "step": 144810
+    },
+    {
+      "epoch": 934.32,
+      "learning_rate": 3.648745519713262e-06,
+      "loss": 0.1843,
+      "step": 144820
+    },
+    {
+      "epoch": 934.39,
+      "learning_rate": 3.645161290322581e-06,
+      "loss": 0.1751,
+      "step": 144830
+    },
+    {
+      "epoch": 934.45,
+      "learning_rate": 3.6415770609319e-06,
+      "loss": 0.1733,
+      "step": 144840
+    },
+    {
+      "epoch": 934.51,
+      "learning_rate": 3.637992831541219e-06,
+      "loss": 0.2237,
+      "step": 144850
+    },
+    {
+      "epoch": 934.58,
+      "learning_rate": 3.634408602150538e-06,
+      "loss": 0.1884,
+      "step": 144860
+    },
+    {
+      "epoch": 934.64,
+      "learning_rate": 3.630824372759857e-06,
+      "loss": 0.1985,
+      "step": 144870
+    },
+    {
+      "epoch": 934.71,
+      "learning_rate": 3.627240143369176e-06,
+      "loss": 0.1914,
+      "step": 144880
+    },
+    {
+      "epoch": 934.77,
+      "learning_rate": 3.623655913978495e-06,
+      "loss": 0.1785,
+      "step": 144890
+    },
+    {
+      "epoch": 934.84,
+      "learning_rate": 3.6200716845878142e-06,
+      "loss": 0.1859,
+      "step": 144900
+    },
+    {
+      "epoch": 934.9,
+      "learning_rate": 3.6164874551971333e-06,
+      "loss": 0.1674,
+      "step": 144910
+    },
+    {
+      "epoch": 934.96,
+      "learning_rate": 3.612903225806452e-06,
+      "loss": 0.1961,
+      "step": 144920
+    },
+    {
+      "epoch": 935.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9308218359947205,
+      "eval_runtime": 0.8746,
+      "eval_samples_per_second": 229.824,
+      "eval_steps_per_second": 8.004,
+      "step": 144925
+    },
+    {
+      "epoch": 935.03,
+      "learning_rate": 3.6093189964157706e-06,
+      "loss": 0.1947,
+      "step": 144930
+    },
+    {
+      "epoch": 935.1,
+      "learning_rate": 3.6057347670250896e-06,
+      "loss": 0.1624,
+      "step": 144940
+    },
+    {
+      "epoch": 935.16,
+      "learning_rate": 3.6021505376344087e-06,
+      "loss": 0.2041,
+      "step": 144950
+    },
+    {
+      "epoch": 935.23,
+      "learning_rate": 3.5985663082437277e-06,
+      "loss": 0.1971,
+      "step": 144960
+    },
+    {
+      "epoch": 935.29,
+      "learning_rate": 3.5949820788530463e-06,
+      "loss": 0.195,
+      "step": 144970
+    },
+    {
+      "epoch": 935.35,
+      "learning_rate": 3.5913978494623654e-06,
+      "loss": 0.1882,
+      "step": 144980
+    },
+    {
+      "epoch": 935.42,
+      "learning_rate": 3.5878136200716844e-06,
+      "loss": 0.1674,
+      "step": 144990
+    },
+    {
+      "epoch": 935.48,
+      "learning_rate": 3.5842293906810035e-06,
+      "loss": 0.2547,
+      "step": 145000
+    },
+    {
+      "epoch": 935.55,
+      "learning_rate": 3.5806451612903225e-06,
+      "loss": 0.2247,
+      "step": 145010
+    },
+    {
+      "epoch": 935.61,
+      "learning_rate": 3.5770609318996416e-06,
+      "loss": 0.1825,
+      "step": 145020
+    },
+    {
+      "epoch": 935.68,
+      "learning_rate": 3.5734767025089607e-06,
+      "loss": 0.2152,
+      "step": 145030
+    },
+    {
+      "epoch": 935.74,
+      "learning_rate": 3.5698924731182797e-06,
+      "loss": 0.2235,
+      "step": 145040
+    },
+    {
+      "epoch": 935.8,
+      "learning_rate": 3.5663082437275988e-06,
+      "loss": 0.178,
+      "step": 145050
+    },
+    {
+      "epoch": 935.87,
+      "learning_rate": 3.562724014336918e-06,
+      "loss": 0.16,
+      "step": 145060
+    },
+    {
+      "epoch": 935.93,
+      "learning_rate": 3.559139784946237e-06,
+      "loss": 0.2058,
+      "step": 145070
+    },
+    {
+      "epoch": 936.0,
+      "learning_rate": 3.555555555555556e-06,
+      "loss": 0.2282,
+      "step": 145080
+    },
+    {
+      "epoch": 936.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9122019410133362,
+      "eval_runtime": 0.8969,
+      "eval_samples_per_second": 224.096,
+      "eval_steps_per_second": 7.804,
+      "step": 145080
+    },
+    {
+      "epoch": 936.06,
+      "learning_rate": 3.5519713261648745e-06,
+      "loss": 0.2262,
+      "step": 145090
+    },
+    {
+      "epoch": 936.13,
+      "learning_rate": 3.5483870967741936e-06,
+      "loss": 0.2147,
+      "step": 145100
+    },
+    {
+      "epoch": 936.19,
+      "learning_rate": 3.5448028673835126e-06,
+      "loss": 0.1856,
+      "step": 145110
+    },
+    {
+      "epoch": 936.26,
+      "learning_rate": 3.5412186379928317e-06,
+      "loss": 0.1911,
+      "step": 145120
+    },
+    {
+      "epoch": 936.32,
+      "learning_rate": 3.5376344086021507e-06,
+      "loss": 0.205,
+      "step": 145130
+    },
+    {
+      "epoch": 936.39,
+      "learning_rate": 3.53405017921147e-06,
+      "loss": 0.2126,
+      "step": 145140
+    },
+    {
+      "epoch": 936.45,
+      "learning_rate": 3.530465949820789e-06,
+      "loss": 0.1764,
+      "step": 145150
+    },
+    {
+      "epoch": 936.51,
+      "learning_rate": 3.526881720430108e-06,
+      "loss": 0.2009,
+      "step": 145160
+    },
+    {
+      "epoch": 936.58,
+      "learning_rate": 3.523297491039427e-06,
+      "loss": 0.219,
+      "step": 145170
+    },
+    {
+      "epoch": 936.64,
+      "learning_rate": 3.519713261648746e-06,
+      "loss": 0.2081,
+      "step": 145180
+    },
+    {
+      "epoch": 936.71,
+      "learning_rate": 3.516129032258065e-06,
+      "loss": 0.1955,
+      "step": 145190
+    },
+    {
+      "epoch": 936.77,
+      "learning_rate": 3.512544802867384e-06,
+      "loss": 0.2324,
+      "step": 145200
+    },
+    {
+      "epoch": 936.84,
+      "learning_rate": 3.508960573476703e-06,
+      "loss": 0.2238,
+      "step": 145210
+    },
+    {
+      "epoch": 936.9,
+      "learning_rate": 3.505376344086022e-06,
+      "loss": 0.171,
+      "step": 145220
+    },
+    {
+      "epoch": 936.96,
+      "learning_rate": 3.501792114695341e-06,
+      "loss": 0.2124,
+      "step": 145230
+    },
+    {
+      "epoch": 937.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9183009266853333,
+      "eval_runtime": 0.8701,
+      "eval_samples_per_second": 231.011,
+      "eval_steps_per_second": 8.045,
+      "step": 145235
+    },
+    {
+      "epoch": 937.03,
+      "learning_rate": 3.49820788530466e-06,
+      "loss": 0.2465,
+      "step": 145240
+    },
+    {
+      "epoch": 937.1,
+      "learning_rate": 3.4946236559139785e-06,
+      "loss": 0.1801,
+      "step": 145250
+    },
+    {
+      "epoch": 937.16,
+      "learning_rate": 3.491039426523297e-06,
+      "loss": 0.1728,
+      "step": 145260
+    },
+    {
+      "epoch": 937.23,
+      "learning_rate": 3.487455197132616e-06,
+      "loss": 0.1756,
+      "step": 145270
+    },
+    {
+      "epoch": 937.29,
+      "learning_rate": 3.4838709677419353e-06,
+      "loss": 0.1891,
+      "step": 145280
+    },
+    {
+      "epoch": 937.35,
+      "learning_rate": 3.4802867383512543e-06,
+      "loss": 0.2054,
+      "step": 145290
+    },
+    {
+      "epoch": 937.42,
+      "learning_rate": 3.4767025089605734e-06,
+      "loss": 0.1991,
+      "step": 145300
+    },
+    {
+      "epoch": 937.48,
+      "learning_rate": 3.4731182795698924e-06,
+      "loss": 0.2089,
+      "step": 145310
+    },
+    {
+      "epoch": 937.55,
+      "learning_rate": 3.4695340501792115e-06,
+      "loss": 0.1691,
+      "step": 145320
+    },
+    {
+      "epoch": 937.61,
+      "learning_rate": 3.4659498207885305e-06,
+      "loss": 0.2211,
+      "step": 145330
+    },
+    {
+      "epoch": 937.68,
+      "learning_rate": 3.4623655913978496e-06,
+      "loss": 0.1886,
+      "step": 145340
+    },
+    {
+      "epoch": 937.74,
+      "learning_rate": 3.4587813620071686e-06,
+      "loss": 0.2106,
+      "step": 145350
+    },
+    {
+      "epoch": 937.8,
+      "learning_rate": 3.4551971326164877e-06,
+      "loss": 0.154,
+      "step": 145360
+    },
+    {
+      "epoch": 937.87,
+      "learning_rate": 3.4516129032258067e-06,
+      "loss": 0.1959,
+      "step": 145370
+    },
+    {
+      "epoch": 937.93,
+      "learning_rate": 3.4480286738351258e-06,
+      "loss": 0.141,
+      "step": 145380
+    },
+    {
+      "epoch": 938.0,
+      "learning_rate": 3.4444444444444444e-06,
+      "loss": 0.1811,
+      "step": 145390
+    },
+    {
+      "epoch": 938.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9246042966842651,
+      "eval_runtime": 0.911,
+      "eval_samples_per_second": 220.645,
+      "eval_steps_per_second": 7.684,
+      "step": 145390
+    },
+    {
+      "epoch": 938.06,
+      "learning_rate": 3.4408602150537635e-06,
+      "loss": 0.1754,
+      "step": 145400
+    },
+    {
+      "epoch": 938.13,
+      "learning_rate": 3.4372759856630825e-06,
+      "loss": 0.1695,
+      "step": 145410
+    },
+    {
+      "epoch": 938.19,
+      "learning_rate": 3.4336917562724016e-06,
+      "loss": 0.2125,
+      "step": 145420
+    },
+    {
+      "epoch": 938.26,
+      "learning_rate": 3.4301075268817206e-06,
+      "loss": 0.1956,
+      "step": 145430
+    },
+    {
+      "epoch": 938.32,
+      "learning_rate": 3.4265232974910397e-06,
+      "loss": 0.1983,
+      "step": 145440
+    },
+    {
+      "epoch": 938.39,
+      "learning_rate": 3.4229390681003587e-06,
+      "loss": 0.1847,
+      "step": 145450
+    },
+    {
+      "epoch": 938.45,
+      "learning_rate": 3.4193548387096778e-06,
+      "loss": 0.2279,
+      "step": 145460
+    },
+    {
+      "epoch": 938.51,
+      "learning_rate": 3.415770609318997e-06,
+      "loss": 0.2294,
+      "step": 145470
+    },
+    {
+      "epoch": 938.58,
+      "learning_rate": 3.412186379928316e-06,
+      "loss": 0.215,
+      "step": 145480
+    },
+    {
+      "epoch": 938.64,
+      "learning_rate": 3.408602150537635e-06,
+      "loss": 0.2325,
+      "step": 145490
+    },
+    {
+      "epoch": 938.71,
+      "learning_rate": 3.405017921146954e-06,
+      "loss": 0.199,
+      "step": 145500
+    },
+    {
+      "epoch": 938.77,
+      "learning_rate": 3.4014336917562726e-06,
+      "loss": 0.1888,
+      "step": 145510
+    },
+    {
+      "epoch": 938.84,
+      "learning_rate": 3.3978494623655917e-06,
+      "loss": 0.1823,
+      "step": 145520
+    },
+    {
+      "epoch": 938.9,
+      "learning_rate": 3.3942652329749107e-06,
+      "loss": 0.1777,
+      "step": 145530
+    },
+    {
+      "epoch": 938.96,
+      "learning_rate": 3.3906810035842298e-06,
+      "loss": 0.1959,
+      "step": 145540
+    },
+    {
+      "epoch": 939.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9318833947181702,
+      "eval_runtime": 0.926,
+      "eval_samples_per_second": 217.074,
+      "eval_steps_per_second": 7.56,
+      "step": 145545
+    },
+    {
+      "epoch": 939.03,
+      "learning_rate": 3.387096774193549e-06,
+      "loss": 0.1534,
+      "step": 145550
+    },
+    {
+      "epoch": 939.1,
+      "learning_rate": 3.383512544802868e-06,
+      "loss": 0.1868,
+      "step": 145560
+    },
+    {
+      "epoch": 939.16,
+      "learning_rate": 3.379928315412186e-06,
+      "loss": 0.2114,
+      "step": 145570
+    },
+    {
+      "epoch": 939.23,
+      "learning_rate": 3.376344086021505e-06,
+      "loss": 0.199,
+      "step": 145580
+    },
+    {
+      "epoch": 939.29,
+      "learning_rate": 3.372759856630824e-06,
+      "loss": 0.2609,
+      "step": 145590
+    },
+    {
+      "epoch": 939.35,
+      "learning_rate": 3.3691756272401432e-06,
+      "loss": 0.2511,
+      "step": 145600
+    },
+    {
+      "epoch": 939.42,
+      "learning_rate": 3.3655913978494623e-06,
+      "loss": 0.2152,
+      "step": 145610
+    },
+    {
+      "epoch": 939.48,
+      "learning_rate": 3.3620071684587813e-06,
+      "loss": 0.1519,
+      "step": 145620
+    },
+    {
+      "epoch": 939.55,
+      "learning_rate": 3.3584229390681004e-06,
+      "loss": 0.1802,
+      "step": 145630
+    },
+    {
+      "epoch": 939.61,
+      "learning_rate": 3.3548387096774194e-06,
+      "loss": 0.2258,
+      "step": 145640
+    },
+    {
+      "epoch": 939.68,
+      "learning_rate": 3.3512544802867385e-06,
+      "loss": 0.1927,
+      "step": 145650
+    },
+    {
+      "epoch": 939.74,
+      "learning_rate": 3.3476702508960575e-06,
+      "loss": 0.2023,
+      "step": 145660
+    },
+    {
+      "epoch": 939.8,
+      "learning_rate": 3.3440860215053766e-06,
+      "loss": 0.1883,
+      "step": 145670
+    },
+    {
+      "epoch": 939.87,
+      "learning_rate": 3.3405017921146952e-06,
+      "loss": 0.2406,
+      "step": 145680
+    },
+    {
+      "epoch": 939.93,
+      "learning_rate": 3.3369175627240143e-06,
+      "loss": 0.2077,
+      "step": 145690
+    },
+    {
+      "epoch": 940.0,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.216,
+      "step": 145700
+    },
+    {
+      "epoch": 940.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9471909403800964,
+      "eval_runtime": 0.8824,
+      "eval_samples_per_second": 227.777,
+      "eval_steps_per_second": 7.933,
+      "step": 145700
+    },
+    {
+      "epoch": 940.06,
+      "learning_rate": 3.3297491039426524e-06,
+      "loss": 0.2235,
+      "step": 145710
+    },
+    {
+      "epoch": 940.13,
+      "learning_rate": 3.3261648745519714e-06,
+      "loss": 0.1795,
+      "step": 145720
+    },
+    {
+      "epoch": 940.19,
+      "learning_rate": 3.3225806451612905e-06,
+      "loss": 0.1778,
+      "step": 145730
+    },
+    {
+      "epoch": 940.26,
+      "learning_rate": 3.3189964157706095e-06,
+      "loss": 0.1597,
+      "step": 145740
+    },
+    {
+      "epoch": 940.32,
+      "learning_rate": 3.3154121863799286e-06,
+      "loss": 0.2125,
+      "step": 145750
+    },
+    {
+      "epoch": 940.39,
+      "learning_rate": 3.3118279569892476e-06,
+      "loss": 0.2051,
+      "step": 145760
+    },
+    {
+      "epoch": 940.45,
+      "learning_rate": 3.3082437275985667e-06,
+      "loss": 0.2148,
+      "step": 145770
+    },
+    {
+      "epoch": 940.51,
+      "learning_rate": 3.3046594982078857e-06,
+      "loss": 0.1904,
+      "step": 145780
+    },
+    {
+      "epoch": 940.58,
+      "learning_rate": 3.301075268817205e-06,
+      "loss": 0.1801,
+      "step": 145790
+    },
+    {
+      "epoch": 940.64,
+      "learning_rate": 3.2974910394265234e-06,
+      "loss": 0.1832,
+      "step": 145800
+    },
+    {
+      "epoch": 940.71,
+      "learning_rate": 3.2939068100358425e-06,
+      "loss": 0.219,
+      "step": 145810
+    },
+    {
+      "epoch": 940.77,
+      "learning_rate": 3.2903225806451615e-06,
+      "loss": 0.2081,
+      "step": 145820
+    },
+    {
+      "epoch": 940.84,
+      "learning_rate": 3.2867383512544806e-06,
+      "loss": 0.1989,
+      "step": 145830
+    },
+    {
+      "epoch": 940.9,
+      "learning_rate": 3.2831541218637996e-06,
+      "loss": 0.2031,
+      "step": 145840
+    },
+    {
+      "epoch": 940.96,
+      "learning_rate": 3.2795698924731187e-06,
+      "loss": 0.1888,
+      "step": 145850
+    },
+    {
+      "epoch": 941.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9192982912063599,
+      "eval_runtime": 0.8835,
+      "eval_samples_per_second": 227.5,
+      "eval_steps_per_second": 7.923,
+      "step": 145855
+    },
+    {
+      "epoch": 941.03,
+      "learning_rate": 3.2759856630824377e-06,
+      "loss": 0.2418,
+      "step": 145860
+    },
+    {
+      "epoch": 941.1,
+      "learning_rate": 3.2724014336917568e-06,
+      "loss": 0.1919,
+      "step": 145870
+    },
+    {
+      "epoch": 941.16,
+      "learning_rate": 3.268817204301076e-06,
+      "loss": 0.1773,
+      "step": 145880
+    },
+    {
+      "epoch": 941.23,
+      "learning_rate": 3.265232974910394e-06,
+      "loss": 0.2176,
+      "step": 145890
+    },
+    {
+      "epoch": 941.29,
+      "learning_rate": 3.261648745519713e-06,
+      "loss": 0.2021,
+      "step": 145900
+    },
+    {
+      "epoch": 941.35,
+      "learning_rate": 3.258064516129032e-06,
+      "loss": 0.186,
+      "step": 145910
+    },
+    {
+      "epoch": 941.42,
+      "learning_rate": 3.254480286738351e-06,
+      "loss": 0.207,
+      "step": 145920
+    },
+    {
+      "epoch": 941.48,
+      "learning_rate": 3.2508960573476702e-06,
+      "loss": 0.2142,
+      "step": 145930
+    },
+    {
+      "epoch": 941.55,
+      "learning_rate": 3.2473118279569893e-06,
+      "loss": 0.1937,
+      "step": 145940
+    },
+    {
+      "epoch": 941.61,
+      "learning_rate": 3.2437275985663084e-06,
+      "loss": 0.1957,
+      "step": 145950
+    },
+    {
+      "epoch": 941.68,
+      "learning_rate": 3.2401433691756274e-06,
+      "loss": 0.2104,
+      "step": 145960
+    },
+    {
+      "epoch": 941.74,
+      "learning_rate": 3.236559139784946e-06,
+      "loss": 0.1979,
+      "step": 145970
+    },
+    {
+      "epoch": 941.8,
+      "learning_rate": 3.232974910394265e-06,
+      "loss": 0.2185,
+      "step": 145980
+    },
+    {
+      "epoch": 941.87,
+      "learning_rate": 3.229390681003584e-06,
+      "loss": 0.1771,
+      "step": 145990
+    },
+    {
+      "epoch": 941.93,
+      "learning_rate": 3.225806451612903e-06,
+      "loss": 0.2202,
+      "step": 146000
+    },
+    {
+      "epoch": 942.0,
+      "learning_rate": 3.2222222222222222e-06,
+      "loss": 0.1784,
+      "step": 146010
+    },
+    {
+      "epoch": 942.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9371645450592041,
+      "eval_runtime": 0.9136,
+      "eval_samples_per_second": 220.017,
+      "eval_steps_per_second": 7.662,
+      "step": 146010
+    },
+    {
+      "epoch": 942.06,
+      "learning_rate": 3.2186379928315413e-06,
+      "loss": 0.1882,
+      "step": 146020
+    },
+    {
+      "epoch": 942.13,
+      "learning_rate": 3.2150537634408603e-06,
+      "loss": 0.146,
+      "step": 146030
+    },
+    {
+      "epoch": 942.19,
+      "learning_rate": 3.2114695340501794e-06,
+      "loss": 0.2067,
+      "step": 146040
+    },
+    {
+      "epoch": 942.26,
+      "learning_rate": 3.2078853046594984e-06,
+      "loss": 0.2379,
+      "step": 146050
+    },
+    {
+      "epoch": 942.32,
+      "learning_rate": 3.2043010752688175e-06,
+      "loss": 0.2186,
+      "step": 146060
+    },
+    {
+      "epoch": 942.39,
+      "learning_rate": 3.2007168458781366e-06,
+      "loss": 0.2231,
+      "step": 146070
+    },
+    {
+      "epoch": 942.45,
+      "learning_rate": 3.1971326164874556e-06,
+      "loss": 0.1809,
+      "step": 146080
+    },
+    {
+      "epoch": 942.51,
+      "learning_rate": 3.1935483870967747e-06,
+      "loss": 0.2054,
+      "step": 146090
+    },
+    {
+      "epoch": 942.58,
+      "learning_rate": 3.1899641577060933e-06,
+      "loss": 0.1994,
+      "step": 146100
+    },
+    {
+      "epoch": 942.64,
+      "learning_rate": 3.1863799283154123e-06,
+      "loss": 0.2094,
+      "step": 146110
+    },
+    {
+      "epoch": 942.71,
+      "learning_rate": 3.1827956989247314e-06,
+      "loss": 0.2297,
+      "step": 146120
+    },
+    {
+      "epoch": 942.77,
+      "learning_rate": 3.1792114695340504e-06,
+      "loss": 0.1958,
+      "step": 146130
+    },
+    {
+      "epoch": 942.84,
+      "learning_rate": 3.1756272401433695e-06,
+      "loss": 0.2241,
+      "step": 146140
+    },
+    {
+      "epoch": 942.9,
+      "learning_rate": 3.1720430107526885e-06,
+      "loss": 0.2132,
+      "step": 146150
+    },
+    {
+      "epoch": 942.96,
+      "learning_rate": 3.1684587813620076e-06,
+      "loss": 0.2027,
+      "step": 146160
+    },
+    {
+      "epoch": 943.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.966864287853241,
+      "eval_runtime": 0.8725,
+      "eval_samples_per_second": 230.381,
+      "eval_steps_per_second": 8.023,
+      "step": 146165
+    },
+    {
+      "epoch": 943.03,
+      "learning_rate": 3.1648745519713266e-06,
+      "loss": 0.181,
+      "step": 146170
+    },
+    {
+      "epoch": 943.1,
+      "learning_rate": 3.1612903225806457e-06,
+      "loss": 0.1837,
+      "step": 146180
+    },
+    {
+      "epoch": 943.16,
+      "learning_rate": 3.1577060931899648e-06,
+      "loss": 0.1891,
+      "step": 146190
+    },
+    {
+      "epoch": 943.23,
+      "learning_rate": 3.154121863799284e-06,
+      "loss": 0.1914,
+      "step": 146200
+    },
+    {
+      "epoch": 943.29,
+      "learning_rate": 3.150537634408602e-06,
+      "loss": 0.1679,
+      "step": 146210
+    },
+    {
+      "epoch": 943.35,
+      "learning_rate": 3.146953405017921e-06,
+      "loss": 0.2089,
+      "step": 146220
+    },
+    {
+      "epoch": 943.42,
+      "learning_rate": 3.14336917562724e-06,
+      "loss": 0.1835,
+      "step": 146230
+    },
+    {
+      "epoch": 943.48,
+      "learning_rate": 3.139784946236559e-06,
+      "loss": 0.1608,
+      "step": 146240
+    },
+    {
+      "epoch": 943.55,
+      "learning_rate": 3.1362007168458782e-06,
+      "loss": 0.2394,
+      "step": 146250
+    },
+    {
+      "epoch": 943.61,
+      "learning_rate": 3.1326164874551973e-06,
+      "loss": 0.2537,
+      "step": 146260
+    },
+    {
+      "epoch": 943.68,
+      "learning_rate": 3.129032258064516e-06,
+      "loss": 0.2007,
+      "step": 146270
+    },
+    {
+      "epoch": 943.74,
+      "learning_rate": 3.125448028673835e-06,
+      "loss": 0.2047,
+      "step": 146280
+    },
+    {
+      "epoch": 943.8,
+      "learning_rate": 3.121863799283154e-06,
+      "loss": 0.1931,
+      "step": 146290
+    },
+    {
+      "epoch": 943.87,
+      "learning_rate": 3.118279569892473e-06,
+      "loss": 0.1574,
+      "step": 146300
+    },
+    {
+      "epoch": 943.93,
+      "learning_rate": 3.114695340501792e-06,
+      "loss": 0.209,
+      "step": 146310
+    },
+    {
+      "epoch": 944.0,
+      "learning_rate": 3.111111111111111e-06,
+      "loss": 0.2435,
+      "step": 146320
+    },
+    {
+      "epoch": 944.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9413461089134216,
+      "eval_runtime": 0.8814,
+      "eval_samples_per_second": 228.047,
+      "eval_steps_per_second": 7.942,
+      "step": 146320
+    },
+    {
+      "epoch": 944.06,
+      "learning_rate": 3.10752688172043e-06,
+      "loss": 0.2066,
+      "step": 146330
+    },
+    {
+      "epoch": 944.13,
+      "learning_rate": 3.1039426523297493e-06,
+      "loss": 0.1773,
+      "step": 146340
+    },
+    {
+      "epoch": 944.19,
+      "learning_rate": 3.1003584229390683e-06,
+      "loss": 0.1849,
+      "step": 146350
+    },
+    {
+      "epoch": 944.26,
+      "learning_rate": 3.0967741935483874e-06,
+      "loss": 0.1808,
+      "step": 146360
+    },
+    {
+      "epoch": 944.32,
+      "learning_rate": 3.0931899641577064e-06,
+      "loss": 0.1686,
+      "step": 146370
+    },
+    {
+      "epoch": 944.39,
+      "learning_rate": 3.0896057347670255e-06,
+      "loss": 0.1963,
+      "step": 146380
+    },
+    {
+      "epoch": 944.45,
+      "learning_rate": 3.086021505376344e-06,
+      "loss": 0.1922,
+      "step": 146390
+    },
+    {
+      "epoch": 944.51,
+      "learning_rate": 3.082437275985663e-06,
+      "loss": 0.2311,
+      "step": 146400
+    },
+    {
+      "epoch": 944.58,
+      "learning_rate": 3.078853046594982e-06,
+      "loss": 0.2398,
+      "step": 146410
+    },
+    {
+      "epoch": 944.64,
+      "learning_rate": 3.0752688172043013e-06,
+      "loss": 0.2284,
+      "step": 146420
+    },
+    {
+      "epoch": 944.71,
+      "learning_rate": 3.0716845878136203e-06,
+      "loss": 0.2103,
+      "step": 146430
+    },
+    {
+      "epoch": 944.77,
+      "learning_rate": 3.0681003584229394e-06,
+      "loss": 0.1838,
+      "step": 146440
+    },
+    {
+      "epoch": 944.84,
+      "learning_rate": 3.064516129032258e-06,
+      "loss": 0.2133,
+      "step": 146450
+    },
+    {
+      "epoch": 944.9,
+      "learning_rate": 3.060931899641577e-06,
+      "loss": 0.1594,
+      "step": 146460
+    },
+    {
+      "epoch": 944.96,
+      "learning_rate": 3.057347670250896e-06,
+      "loss": 0.1945,
+      "step": 146470
+    },
+    {
+      "epoch": 945.0,
+      "eval_accuracy": 0.7562189054726368,
+      "eval_loss": 0.978330671787262,
+      "eval_runtime": 0.8996,
+      "eval_samples_per_second": 223.432,
+      "eval_steps_per_second": 7.781,
+      "step": 146475
+    },
+    {
+      "epoch": 945.03,
+      "learning_rate": 3.053763440860215e-06,
+      "loss": 0.2036,
+      "step": 146480
+    },
+    {
+      "epoch": 945.1,
+      "learning_rate": 3.050179211469534e-06,
+      "loss": 0.2205,
+      "step": 146490
+    },
+    {
+      "epoch": 945.16,
+      "learning_rate": 3.0465949820788532e-06,
+      "loss": 0.1843,
+      "step": 146500
+    },
+    {
+      "epoch": 945.23,
+      "learning_rate": 3.0430107526881723e-06,
+      "loss": 0.1745,
+      "step": 146510
+    },
+    {
+      "epoch": 945.29,
+      "learning_rate": 3.0394265232974914e-06,
+      "loss": 0.2122,
+      "step": 146520
+    },
+    {
+      "epoch": 945.35,
+      "learning_rate": 3.0358422939068104e-06,
+      "loss": 0.157,
+      "step": 146530
+    },
+    {
+      "epoch": 945.42,
+      "learning_rate": 3.032258064516129e-06,
+      "loss": 0.1694,
+      "step": 146540
+    },
+    {
+      "epoch": 945.48,
+      "learning_rate": 3.028673835125448e-06,
+      "loss": 0.1974,
+      "step": 146550
+    },
+    {
+      "epoch": 945.55,
+      "learning_rate": 3.025089605734767e-06,
+      "loss": 0.1684,
+      "step": 146560
+    },
+    {
+      "epoch": 945.61,
+      "learning_rate": 3.021505376344086e-06,
+      "loss": 0.2125,
+      "step": 146570
+    },
+    {
+      "epoch": 945.68,
+      "learning_rate": 3.0179211469534052e-06,
+      "loss": 0.1473,
+      "step": 146580
+    },
+    {
+      "epoch": 945.74,
+      "learning_rate": 3.0143369175627243e-06,
+      "loss": 0.2129,
+      "step": 146590
+    },
+    {
+      "epoch": 945.8,
+      "learning_rate": 3.0107526881720433e-06,
+      "loss": 0.203,
+      "step": 146600
+    },
+    {
+      "epoch": 945.87,
+      "learning_rate": 3.007168458781362e-06,
+      "loss": 0.2102,
+      "step": 146610
+    },
+    {
+      "epoch": 945.93,
+      "learning_rate": 3.003584229390681e-06,
+      "loss": 0.2314,
+      "step": 146620
+    },
+    {
+      "epoch": 946.0,
+      "learning_rate": 3e-06,
+      "loss": 0.1869,
+      "step": 146630
+    },
+    {
+      "epoch": 946.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.940397322177887,
+      "eval_runtime": 0.8761,
+      "eval_samples_per_second": 229.437,
+      "eval_steps_per_second": 7.99,
+      "step": 146630
+    },
+    {
+      "epoch": 946.06,
+      "learning_rate": 2.996415770609319e-06,
+      "loss": 0.2319,
+      "step": 146640
+    },
+    {
+      "epoch": 946.13,
+      "learning_rate": 2.992831541218638e-06,
+      "loss": 0.1869,
+      "step": 146650
+    },
+    {
+      "epoch": 946.19,
+      "learning_rate": 2.9892473118279572e-06,
+      "loss": 0.2237,
+      "step": 146660
+    },
+    {
+      "epoch": 946.26,
+      "learning_rate": 2.9856630824372763e-06,
+      "loss": 0.2184,
+      "step": 146670
+    },
+    {
+      "epoch": 946.32,
+      "learning_rate": 2.982078853046595e-06,
+      "loss": 0.1599,
+      "step": 146680
+    },
+    {
+      "epoch": 946.39,
+      "learning_rate": 2.978494623655914e-06,
+      "loss": 0.1791,
+      "step": 146690
+    },
+    {
+      "epoch": 946.45,
+      "learning_rate": 2.974910394265233e-06,
+      "loss": 0.2184,
+      "step": 146700
+    },
+    {
+      "epoch": 946.51,
+      "learning_rate": 2.971326164874552e-06,
+      "loss": 0.2547,
+      "step": 146710
+    },
+    {
+      "epoch": 946.58,
+      "learning_rate": 2.967741935483871e-06,
+      "loss": 0.2023,
+      "step": 146720
+    },
+    {
+      "epoch": 946.64,
+      "learning_rate": 2.96415770609319e-06,
+      "loss": 0.2061,
+      "step": 146730
+    },
+    {
+      "epoch": 946.71,
+      "learning_rate": 2.9605734767025092e-06,
+      "loss": 0.193,
+      "step": 146740
+    },
+    {
+      "epoch": 946.77,
+      "learning_rate": 2.9569892473118283e-06,
+      "loss": 0.1957,
+      "step": 146750
+    },
+    {
+      "epoch": 946.84,
+      "learning_rate": 2.9534050179211473e-06,
+      "loss": 0.1834,
+      "step": 146760
+    },
+    {
+      "epoch": 946.9,
+      "learning_rate": 2.949820788530466e-06,
+      "loss": 0.1868,
+      "step": 146770
+    },
+    {
+      "epoch": 946.96,
+      "learning_rate": 2.946236559139785e-06,
+      "loss": 0.1597,
+      "step": 146780
+    },
+    {
+      "epoch": 947.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9727910757064819,
+      "eval_runtime": 0.866,
+      "eval_samples_per_second": 232.108,
+      "eval_steps_per_second": 8.083,
+      "step": 146785
+    },
+    {
+      "epoch": 947.03,
+      "learning_rate": 2.942652329749104e-06,
+      "loss": 0.2111,
+      "step": 146790
+    },
+    {
+      "epoch": 947.1,
+      "learning_rate": 2.939068100358423e-06,
+      "loss": 0.2149,
+      "step": 146800
+    },
+    {
+      "epoch": 947.16,
+      "learning_rate": 2.935483870967742e-06,
+      "loss": 0.1987,
+      "step": 146810
+    },
+    {
+      "epoch": 947.23,
+      "learning_rate": 2.9318996415770612e-06,
+      "loss": 0.1599,
+      "step": 146820
+    },
+    {
+      "epoch": 947.29,
+      "learning_rate": 2.92831541218638e-06,
+      "loss": 0.1791,
+      "step": 146830
+    },
+    {
+      "epoch": 947.35,
+      "learning_rate": 2.924731182795699e-06,
+      "loss": 0.2599,
+      "step": 146840
+    },
+    {
+      "epoch": 947.42,
+      "learning_rate": 2.921146953405018e-06,
+      "loss": 0.2196,
+      "step": 146850
+    },
+    {
+      "epoch": 947.48,
+      "learning_rate": 2.917562724014337e-06,
+      "loss": 0.1669,
+      "step": 146860
+    },
+    {
+      "epoch": 947.55,
+      "learning_rate": 2.913978494623656e-06,
+      "loss": 0.1912,
+      "step": 146870
+    },
+    {
+      "epoch": 947.61,
+      "learning_rate": 2.910394265232975e-06,
+      "loss": 0.2075,
+      "step": 146880
+    },
+    {
+      "epoch": 947.68,
+      "learning_rate": 2.906810035842294e-06,
+      "loss": 0.1939,
+      "step": 146890
+    },
+    {
+      "epoch": 947.74,
+      "learning_rate": 2.903225806451613e-06,
+      "loss": 0.1558,
+      "step": 146900
+    },
+    {
+      "epoch": 947.8,
+      "learning_rate": 2.8996415770609323e-06,
+      "loss": 0.2044,
+      "step": 146910
+    },
+    {
+      "epoch": 947.87,
+      "learning_rate": 2.8960573476702513e-06,
+      "loss": 0.2043,
+      "step": 146920
+    },
+    {
+      "epoch": 947.93,
+      "learning_rate": 2.89247311827957e-06,
+      "loss": 0.206,
+      "step": 146930
+    },
+    {
+      "epoch": 948.0,
+      "learning_rate": 2.888888888888889e-06,
+      "loss": 0.1763,
+      "step": 146940
+    },
+    {
+      "epoch": 948.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9586777687072754,
+      "eval_runtime": 0.8617,
+      "eval_samples_per_second": 233.252,
+      "eval_steps_per_second": 8.123,
+      "step": 146940
+    },
+    {
+      "epoch": 948.06,
+      "learning_rate": 2.885304659498208e-06,
+      "loss": 0.1986,
+      "step": 146950
+    },
+    {
+      "epoch": 948.13,
+      "learning_rate": 2.881720430107527e-06,
+      "loss": 0.2204,
+      "step": 146960
+    },
+    {
+      "epoch": 948.19,
+      "learning_rate": 2.878136200716846e-06,
+      "loss": 0.2216,
+      "step": 146970
+    },
+    {
+      "epoch": 948.26,
+      "learning_rate": 2.8745519713261648e-06,
+      "loss": 0.1728,
+      "step": 146980
+    },
+    {
+      "epoch": 948.32,
+      "learning_rate": 2.870967741935484e-06,
+      "loss": 0.2256,
+      "step": 146990
+    },
+    {
+      "epoch": 948.39,
+      "learning_rate": 2.867383512544803e-06,
+      "loss": 0.2293,
+      "step": 147000
+    },
+    {
+      "epoch": 948.45,
+      "learning_rate": 2.863799283154122e-06,
+      "loss": 0.1959,
+      "step": 147010
+    },
+    {
+      "epoch": 948.51,
+      "learning_rate": 2.860215053763441e-06,
+      "loss": 0.211,
+      "step": 147020
+    },
+    {
+      "epoch": 948.58,
+      "learning_rate": 2.85663082437276e-06,
+      "loss": 0.204,
+      "step": 147030
+    },
+    {
+      "epoch": 948.64,
+      "learning_rate": 2.853046594982079e-06,
+      "loss": 0.1695,
+      "step": 147040
+    },
+    {
+      "epoch": 948.71,
+      "learning_rate": 2.849462365591398e-06,
+      "loss": 0.1694,
+      "step": 147050
+    },
+    {
+      "epoch": 948.77,
+      "learning_rate": 2.845878136200717e-06,
+      "loss": 0.1873,
+      "step": 147060
+    },
+    {
+      "epoch": 948.84,
+      "learning_rate": 2.8422939068100362e-06,
+      "loss": 0.1991,
+      "step": 147070
+    },
+    {
+      "epoch": 948.9,
+      "learning_rate": 2.8387096774193553e-06,
+      "loss": 0.185,
+      "step": 147080
+    },
+    {
+      "epoch": 948.96,
+      "learning_rate": 2.835125448028674e-06,
+      "loss": 0.1611,
+      "step": 147090
+    },
+    {
+      "epoch": 949.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9417126178741455,
+      "eval_runtime": 0.8863,
+      "eval_samples_per_second": 226.793,
+      "eval_steps_per_second": 7.898,
+      "step": 147095
+    },
+    {
+      "epoch": 949.03,
+      "learning_rate": 2.831541218637993e-06,
+      "loss": 0.2284,
+      "step": 147100
+    },
+    {
+      "epoch": 949.1,
+      "learning_rate": 2.827956989247312e-06,
+      "loss": 0.2114,
+      "step": 147110
+    },
+    {
+      "epoch": 949.16,
+      "learning_rate": 2.824372759856631e-06,
+      "loss": 0.216,
+      "step": 147120
+    },
+    {
+      "epoch": 949.23,
+      "learning_rate": 2.8207885304659497e-06,
+      "loss": 0.2295,
+      "step": 147130
+    },
+    {
+      "epoch": 949.29,
+      "learning_rate": 2.8172043010752688e-06,
+      "loss": 0.1587,
+      "step": 147140
+    },
+    {
+      "epoch": 949.35,
+      "learning_rate": 2.813620071684588e-06,
+      "loss": 0.21,
+      "step": 147150
+    },
+    {
+      "epoch": 949.42,
+      "learning_rate": 2.810035842293907e-06,
+      "loss": 0.1963,
+      "step": 147160
+    },
+    {
+      "epoch": 949.48,
+      "learning_rate": 2.806451612903226e-06,
+      "loss": 0.1396,
+      "step": 147170
+    },
+    {
+      "epoch": 949.55,
+      "learning_rate": 2.802867383512545e-06,
+      "loss": 0.188,
+      "step": 147180
+    },
+    {
+      "epoch": 949.61,
+      "learning_rate": 2.799283154121864e-06,
+      "loss": 0.2051,
+      "step": 147190
+    },
+    {
+      "epoch": 949.68,
+      "learning_rate": 2.795698924731183e-06,
+      "loss": 0.2196,
+      "step": 147200
+    },
+    {
+      "epoch": 949.74,
+      "learning_rate": 2.792114695340502e-06,
+      "loss": 0.1945,
+      "step": 147210
+    },
+    {
+      "epoch": 949.8,
+      "learning_rate": 2.788530465949821e-06,
+      "loss": 0.1779,
+      "step": 147220
+    },
+    {
+      "epoch": 949.87,
+      "learning_rate": 2.7849462365591402e-06,
+      "loss": 0.2019,
+      "step": 147230
+    },
+    {
+      "epoch": 949.93,
+      "learning_rate": 2.7813620071684593e-06,
+      "loss": 0.2086,
+      "step": 147240
+    },
+    {
+      "epoch": 950.0,
+      "learning_rate": 2.777777777777778e-06,
+      "loss": 0.2021,
+      "step": 147250
+    },
+    {
+      "epoch": 950.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9322615265846252,
+      "eval_runtime": 0.8759,
+      "eval_samples_per_second": 229.477,
+      "eval_steps_per_second": 7.992,
+      "step": 147250
+    },
+    {
+      "epoch": 950.06,
+      "learning_rate": 2.774193548387097e-06,
+      "loss": 0.2063,
+      "step": 147260
+    },
+    {
+      "epoch": 950.13,
+      "learning_rate": 2.7706093189964156e-06,
+      "loss": 0.2134,
+      "step": 147270
+    },
+    {
+      "epoch": 950.19,
+      "learning_rate": 2.7670250896057346e-06,
+      "loss": 0.1937,
+      "step": 147280
+    },
+    {
+      "epoch": 950.26,
+      "learning_rate": 2.7634408602150537e-06,
+      "loss": 0.1737,
+      "step": 147290
+    },
+    {
+      "epoch": 950.32,
+      "learning_rate": 2.7598566308243727e-06,
+      "loss": 0.1909,
+      "step": 147300
+    },
+    {
+      "epoch": 950.39,
+      "learning_rate": 2.756272401433692e-06,
+      "loss": 0.2246,
+      "step": 147310
+    },
+    {
+      "epoch": 950.45,
+      "learning_rate": 2.752688172043011e-06,
+      "loss": 0.1878,
+      "step": 147320
+    },
+    {
+      "epoch": 950.51,
+      "learning_rate": 2.74910394265233e-06,
+      "loss": 0.2093,
+      "step": 147330
+    },
+    {
+      "epoch": 950.58,
+      "learning_rate": 2.745519713261649e-06,
+      "loss": 0.2179,
+      "step": 147340
+    },
+    {
+      "epoch": 950.64,
+      "learning_rate": 2.741935483870968e-06,
+      "loss": 0.1873,
+      "step": 147350
+    },
+    {
+      "epoch": 950.71,
+      "learning_rate": 2.738351254480287e-06,
+      "loss": 0.1874,
+      "step": 147360
+    },
+    {
+      "epoch": 950.77,
+      "learning_rate": 2.734767025089606e-06,
+      "loss": 0.1597,
+      "step": 147370
+    },
+    {
+      "epoch": 950.84,
+      "learning_rate": 2.731182795698925e-06,
+      "loss": 0.198,
+      "step": 147380
+    },
+    {
+      "epoch": 950.9,
+      "learning_rate": 2.7275985663082442e-06,
+      "loss": 0.1813,
+      "step": 147390
+    },
+    {
+      "epoch": 950.96,
+      "learning_rate": 2.724014336917563e-06,
+      "loss": 0.1678,
+      "step": 147400
+    },
+    {
+      "epoch": 951.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.925751805305481,
+      "eval_runtime": 0.8766,
+      "eval_samples_per_second": 229.29,
+      "eval_steps_per_second": 7.985,
+      "step": 147405
+    },
+    {
+      "epoch": 951.03,
+      "learning_rate": 2.720430107526882e-06,
+      "loss": 0.2064,
+      "step": 147410
+    },
+    {
+      "epoch": 951.1,
+      "learning_rate": 2.7168458781362005e-06,
+      "loss": 0.1535,
+      "step": 147420
+    },
+    {
+      "epoch": 951.16,
+      "learning_rate": 2.7132616487455196e-06,
+      "loss": 0.1755,
+      "step": 147430
+    },
+    {
+      "epoch": 951.23,
+      "learning_rate": 2.7096774193548386e-06,
+      "loss": 0.1988,
+      "step": 147440
+    },
+    {
+      "epoch": 951.29,
+      "learning_rate": 2.7060931899641577e-06,
+      "loss": 0.236,
+      "step": 147450
+    },
+    {
+      "epoch": 951.35,
+      "learning_rate": 2.7025089605734767e-06,
+      "loss": 0.2167,
+      "step": 147460
+    },
+    {
+      "epoch": 951.42,
+      "learning_rate": 2.6989247311827958e-06,
+      "loss": 0.2004,
+      "step": 147470
+    },
+    {
+      "epoch": 951.48,
+      "learning_rate": 2.695340501792115e-06,
+      "loss": 0.2285,
+      "step": 147480
+    },
+    {
+      "epoch": 951.55,
+      "learning_rate": 2.691756272401434e-06,
+      "loss": 0.1845,
+      "step": 147490
+    },
+    {
+      "epoch": 951.61,
+      "learning_rate": 2.688172043010753e-06,
+      "loss": 0.1746,
+      "step": 147500
+    },
+    {
+      "epoch": 951.68,
+      "learning_rate": 2.684587813620072e-06,
+      "loss": 0.1894,
+      "step": 147510
+    },
+    {
+      "epoch": 951.74,
+      "learning_rate": 2.681003584229391e-06,
+      "loss": 0.2389,
+      "step": 147520
+    },
+    {
+      "epoch": 951.8,
+      "learning_rate": 2.67741935483871e-06,
+      "loss": 0.1995,
+      "step": 147530
+    },
+    {
+      "epoch": 951.87,
+      "learning_rate": 2.6738351254480287e-06,
+      "loss": 0.1777,
+      "step": 147540
+    },
+    {
+      "epoch": 951.93,
+      "learning_rate": 2.6702508960573478e-06,
+      "loss": 0.2262,
+      "step": 147550
+    },
+    {
+      "epoch": 952.0,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 0.2221,
+      "step": 147560
+    },
+    {
+      "epoch": 952.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9536035060882568,
+      "eval_runtime": 0.8904,
+      "eval_samples_per_second": 225.736,
+      "eval_steps_per_second": 7.861,
+      "step": 147560
+    },
+    {
+      "epoch": 952.06,
+      "learning_rate": 2.6630824372759855e-06,
+      "loss": 0.1916,
+      "step": 147570
+    },
+    {
+      "epoch": 952.13,
+      "learning_rate": 2.6594982078853045e-06,
+      "loss": 0.1954,
+      "step": 147580
+    },
+    {
+      "epoch": 952.19,
+      "learning_rate": 2.6559139784946236e-06,
+      "loss": 0.148,
+      "step": 147590
+    },
+    {
+      "epoch": 952.26,
+      "learning_rate": 2.6523297491039426e-06,
+      "loss": 0.1756,
+      "step": 147600
+    },
+    {
+      "epoch": 952.32,
+      "learning_rate": 2.6487455197132617e-06,
+      "loss": 0.2486,
+      "step": 147610
+    },
+    {
+      "epoch": 952.39,
+      "learning_rate": 2.6451612903225807e-06,
+      "loss": 0.1787,
+      "step": 147620
+    },
+    {
+      "epoch": 952.45,
+      "learning_rate": 2.6415770609318998e-06,
+      "loss": 0.1709,
+      "step": 147630
+    },
+    {
+      "epoch": 952.51,
+      "learning_rate": 2.637992831541219e-06,
+      "loss": 0.2193,
+      "step": 147640
+    },
+    {
+      "epoch": 952.58,
+      "learning_rate": 2.634408602150538e-06,
+      "loss": 0.2493,
+      "step": 147650
+    },
+    {
+      "epoch": 952.64,
+      "learning_rate": 2.630824372759857e-06,
+      "loss": 0.1854,
+      "step": 147660
+    },
+    {
+      "epoch": 952.71,
+      "learning_rate": 2.627240143369176e-06,
+      "loss": 0.176,
+      "step": 147670
+    },
+    {
+      "epoch": 952.77,
+      "learning_rate": 2.623655913978495e-06,
+      "loss": 0.2044,
+      "step": 147680
+    },
+    {
+      "epoch": 952.84,
+      "learning_rate": 2.6200716845878137e-06,
+      "loss": 0.157,
+      "step": 147690
+    },
+    {
+      "epoch": 952.9,
+      "learning_rate": 2.6164874551971327e-06,
+      "loss": 0.1996,
+      "step": 147700
+    },
+    {
+      "epoch": 952.96,
+      "learning_rate": 2.6129032258064518e-06,
+      "loss": 0.152,
+      "step": 147710
+    },
+    {
+      "epoch": 953.0,
+      "eval_accuracy": 0.7860696517412935,
+      "eval_loss": 0.9189280271530151,
+      "eval_runtime": 0.876,
+      "eval_samples_per_second": 229.461,
+      "eval_steps_per_second": 7.991,
+      "step": 147715
+    },
+    {
+      "epoch": 953.03,
+      "learning_rate": 2.609318996415771e-06,
+      "loss": 0.2735,
+      "step": 147720
+    },
+    {
+      "epoch": 953.1,
+      "learning_rate": 2.6057347670250894e-06,
+      "loss": 0.2124,
+      "step": 147730
+    },
+    {
+      "epoch": 953.16,
+      "learning_rate": 2.6021505376344085e-06,
+      "loss": 0.1887,
+      "step": 147740
+    },
+    {
+      "epoch": 953.23,
+      "learning_rate": 2.5985663082437275e-06,
+      "loss": 0.1548,
+      "step": 147750
+    },
+    {
+      "epoch": 953.29,
+      "learning_rate": 2.5949820788530466e-06,
+      "loss": 0.2076,
+      "step": 147760
+    },
+    {
+      "epoch": 953.35,
+      "learning_rate": 2.5913978494623656e-06,
+      "loss": 0.1867,
+      "step": 147770
+    },
+    {
+      "epoch": 953.42,
+      "learning_rate": 2.5878136200716847e-06,
+      "loss": 0.1862,
+      "step": 147780
+    },
+    {
+      "epoch": 953.48,
+      "learning_rate": 2.5842293906810038e-06,
+      "loss": 0.1734,
+      "step": 147790
+    },
+    {
+      "epoch": 953.55,
+      "learning_rate": 2.580645161290323e-06,
+      "loss": 0.17,
+      "step": 147800
+    },
+    {
+      "epoch": 953.61,
+      "learning_rate": 2.577060931899642e-06,
+      "loss": 0.2207,
+      "step": 147810
+    },
+    {
+      "epoch": 953.68,
+      "learning_rate": 2.573476702508961e-06,
+      "loss": 0.2085,
+      "step": 147820
+    },
+    {
+      "epoch": 953.74,
+      "learning_rate": 2.56989247311828e-06,
+      "loss": 0.1764,
+      "step": 147830
+    },
+    {
+      "epoch": 953.8,
+      "learning_rate": 2.5663082437275986e-06,
+      "loss": 0.2572,
+      "step": 147840
+    },
+    {
+      "epoch": 953.87,
+      "learning_rate": 2.5627240143369176e-06,
+      "loss": 0.1729,
+      "step": 147850
+    },
+    {
+      "epoch": 953.93,
+      "learning_rate": 2.5591397849462367e-06,
+      "loss": 0.2166,
+      "step": 147860
+    },
+    {
+      "epoch": 954.0,
+      "learning_rate": 2.5555555555555557e-06,
+      "loss": 0.202,
+      "step": 147870
+    },
+    {
+      "epoch": 954.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9137061834335327,
+      "eval_runtime": 0.9527,
+      "eval_samples_per_second": 210.97,
+      "eval_steps_per_second": 7.347,
+      "step": 147870
+    },
+    {
+      "epoch": 954.06,
+      "learning_rate": 2.551971326164875e-06,
+      "loss": 0.1938,
+      "step": 147880
+    },
+    {
+      "epoch": 954.13,
+      "learning_rate": 2.5483870967741934e-06,
+      "loss": 0.1717,
+      "step": 147890
+    },
+    {
+      "epoch": 954.19,
+      "learning_rate": 2.5448028673835125e-06,
+      "loss": 0.2235,
+      "step": 147900
+    },
+    {
+      "epoch": 954.26,
+      "learning_rate": 2.5412186379928315e-06,
+      "loss": 0.1959,
+      "step": 147910
+    },
+    {
+      "epoch": 954.32,
+      "learning_rate": 2.5376344086021506e-06,
+      "loss": 0.2102,
+      "step": 147920
+    },
+    {
+      "epoch": 954.39,
+      "learning_rate": 2.5340501792114696e-06,
+      "loss": 0.1949,
+      "step": 147930
+    },
+    {
+      "epoch": 954.45,
+      "learning_rate": 2.5304659498207887e-06,
+      "loss": 0.2251,
+      "step": 147940
+    },
+    {
+      "epoch": 954.51,
+      "learning_rate": 2.5268817204301077e-06,
+      "loss": 0.2271,
+      "step": 147950
+    },
+    {
+      "epoch": 954.58,
+      "learning_rate": 2.523297491039427e-06,
+      "loss": 0.2057,
+      "step": 147960
+    },
+    {
+      "epoch": 954.64,
+      "learning_rate": 2.519713261648746e-06,
+      "loss": 0.2126,
+      "step": 147970
+    },
+    {
+      "epoch": 954.71,
+      "learning_rate": 2.5161290322580645e-06,
+      "loss": 0.2275,
+      "step": 147980
+    },
+    {
+      "epoch": 954.77,
+      "learning_rate": 2.5125448028673835e-06,
+      "loss": 0.1752,
+      "step": 147990
+    },
+    {
+      "epoch": 954.84,
+      "learning_rate": 2.5089605734767026e-06,
+      "loss": 0.1959,
+      "step": 148000
+    },
+    {
+      "epoch": 954.9,
+      "learning_rate": 2.5053763440860216e-06,
+      "loss": 0.1902,
+      "step": 148010
+    },
+    {
+      "epoch": 954.96,
+      "learning_rate": 2.5017921146953407e-06,
+      "loss": 0.2276,
+      "step": 148020
+    },
+    {
+      "epoch": 955.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9099946022033691,
+      "eval_runtime": 0.9128,
+      "eval_samples_per_second": 220.21,
+      "eval_steps_per_second": 7.669,
+      "step": 148025
+    },
+    {
+      "epoch": 955.03,
+      "learning_rate": 2.4982078853046597e-06,
+      "loss": 0.2357,
+      "step": 148030
+    },
+    {
+      "epoch": 955.1,
+      "learning_rate": 2.4946236559139788e-06,
+      "loss": 0.2128,
+      "step": 148040
+    },
+    {
+      "epoch": 955.16,
+      "learning_rate": 2.4910394265232974e-06,
+      "loss": 0.2335,
+      "step": 148050
+    },
+    {
+      "epoch": 955.23,
+      "learning_rate": 2.4874551971326165e-06,
+      "loss": 0.2042,
+      "step": 148060
+    },
+    {
+      "epoch": 955.29,
+      "learning_rate": 2.4838709677419355e-06,
+      "loss": 0.2233,
+      "step": 148070
+    },
+    {
+      "epoch": 955.35,
+      "learning_rate": 2.4802867383512546e-06,
+      "loss": 0.1881,
+      "step": 148080
+    },
+    {
+      "epoch": 955.42,
+      "learning_rate": 2.4767025089605736e-06,
+      "loss": 0.1711,
+      "step": 148090
+    },
+    {
+      "epoch": 955.48,
+      "learning_rate": 2.4731182795698927e-06,
+      "loss": 0.178,
+      "step": 148100
+    },
+    {
+      "epoch": 955.55,
+      "learning_rate": 2.4695340501792117e-06,
+      "loss": 0.1813,
+      "step": 148110
+    },
+    {
+      "epoch": 955.61,
+      "learning_rate": 2.4659498207885308e-06,
+      "loss": 0.2022,
+      "step": 148120
+    },
+    {
+      "epoch": 955.68,
+      "learning_rate": 2.4623655913978494e-06,
+      "loss": 0.1929,
+      "step": 148130
+    },
+    {
+      "epoch": 955.74,
+      "learning_rate": 2.4587813620071685e-06,
+      "loss": 0.2257,
+      "step": 148140
+    },
+    {
+      "epoch": 955.8,
+      "learning_rate": 2.4551971326164875e-06,
+      "loss": 0.1849,
+      "step": 148150
+    },
+    {
+      "epoch": 955.87,
+      "learning_rate": 2.4516129032258066e-06,
+      "loss": 0.1818,
+      "step": 148160
+    },
+    {
+      "epoch": 955.93,
+      "learning_rate": 2.4480286738351256e-06,
+      "loss": 0.2334,
+      "step": 148170
+    },
+    {
+      "epoch": 956.0,
+      "learning_rate": 2.4444444444444447e-06,
+      "loss": 0.172,
+      "step": 148180
+    },
+    {
+      "epoch": 956.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9008236527442932,
+      "eval_runtime": 0.8741,
+      "eval_samples_per_second": 229.939,
+      "eval_steps_per_second": 8.008,
+      "step": 148180
+    },
+    {
+      "epoch": 956.06,
+      "learning_rate": 2.4408602150537637e-06,
+      "loss": 0.2316,
+      "step": 148190
+    },
+    {
+      "epoch": 956.13,
+      "learning_rate": 2.4372759856630828e-06,
+      "loss": 0.2099,
+      "step": 148200
+    },
+    {
+      "epoch": 956.19,
+      "learning_rate": 2.4336917562724014e-06,
+      "loss": 0.159,
+      "step": 148210
+    },
+    {
+      "epoch": 956.26,
+      "learning_rate": 2.4301075268817204e-06,
+      "loss": 0.1597,
+      "step": 148220
+    },
+    {
+      "epoch": 956.32,
+      "learning_rate": 2.4265232974910395e-06,
+      "loss": 0.1734,
+      "step": 148230
+    },
+    {
+      "epoch": 956.39,
+      "learning_rate": 2.4229390681003586e-06,
+      "loss": 0.2209,
+      "step": 148240
+    },
+    {
+      "epoch": 956.45,
+      "learning_rate": 2.4193548387096776e-06,
+      "loss": 0.2079,
+      "step": 148250
+    },
+    {
+      "epoch": 956.51,
+      "learning_rate": 2.4157706093189967e-06,
+      "loss": 0.2305,
+      "step": 148260
+    },
+    {
+      "epoch": 956.58,
+      "learning_rate": 2.4121863799283157e-06,
+      "loss": 0.2214,
+      "step": 148270
+    },
+    {
+      "epoch": 956.64,
+      "learning_rate": 2.4086021505376343e-06,
+      "loss": 0.1368,
+      "step": 148280
+    },
+    {
+      "epoch": 956.71,
+      "learning_rate": 2.4050179211469534e-06,
+      "loss": 0.217,
+      "step": 148290
+    },
+    {
+      "epoch": 956.77,
+      "learning_rate": 2.4014336917562724e-06,
+      "loss": 0.1818,
+      "step": 148300
+    },
+    {
+      "epoch": 956.84,
+      "learning_rate": 2.3978494623655915e-06,
+      "loss": 0.1266,
+      "step": 148310
+    },
+    {
+      "epoch": 956.9,
+      "learning_rate": 2.3942652329749105e-06,
+      "loss": 0.2219,
+      "step": 148320
+    },
+    {
+      "epoch": 956.96,
+      "learning_rate": 2.3906810035842296e-06,
+      "loss": 0.2113,
+      "step": 148330
+    },
+    {
+      "epoch": 957.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9492504596710205,
+      "eval_runtime": 0.8889,
+      "eval_samples_per_second": 226.117,
+      "eval_steps_per_second": 7.875,
+      "step": 148335
+    },
+    {
+      "epoch": 957.03,
+      "learning_rate": 2.3870967741935486e-06,
+      "loss": 0.1994,
+      "step": 148340
+    },
+    {
+      "epoch": 957.1,
+      "learning_rate": 2.3835125448028677e-06,
+      "loss": 0.1655,
+      "step": 148350
+    },
+    {
+      "epoch": 957.16,
+      "learning_rate": 2.3799283154121868e-06,
+      "loss": 0.1947,
+      "step": 148360
+    },
+    {
+      "epoch": 957.23,
+      "learning_rate": 2.3763440860215054e-06,
+      "loss": 0.1875,
+      "step": 148370
+    },
+    {
+      "epoch": 957.29,
+      "learning_rate": 2.3727598566308244e-06,
+      "loss": 0.1705,
+      "step": 148380
+    },
+    {
+      "epoch": 957.35,
+      "learning_rate": 2.3691756272401435e-06,
+      "loss": 0.197,
+      "step": 148390
+    },
+    {
+      "epoch": 957.42,
+      "learning_rate": 2.3655913978494625e-06,
+      "loss": 0.1907,
+      "step": 148400
+    },
+    {
+      "epoch": 957.48,
+      "learning_rate": 2.3620071684587816e-06,
+      "loss": 0.1825,
+      "step": 148410
+    },
+    {
+      "epoch": 957.55,
+      "learning_rate": 2.3584229390681002e-06,
+      "loss": 0.1578,
+      "step": 148420
+    },
+    {
+      "epoch": 957.61,
+      "learning_rate": 2.3548387096774193e-06,
+      "loss": 0.17,
+      "step": 148430
+    },
+    {
+      "epoch": 957.68,
+      "learning_rate": 2.3512544802867383e-06,
+      "loss": 0.1655,
+      "step": 148440
+    },
+    {
+      "epoch": 957.74,
+      "learning_rate": 2.3476702508960574e-06,
+      "loss": 0.1961,
+      "step": 148450
+    },
+    {
+      "epoch": 957.8,
+      "learning_rate": 2.3440860215053764e-06,
+      "loss": 0.1995,
+      "step": 148460
+    },
+    {
+      "epoch": 957.87,
+      "learning_rate": 2.3405017921146955e-06,
+      "loss": 0.2249,
+      "step": 148470
+    },
+    {
+      "epoch": 957.93,
+      "learning_rate": 2.3369175627240145e-06,
+      "loss": 0.22,
+      "step": 148480
+    },
+    {
+      "epoch": 958.0,
+      "learning_rate": 2.3333333333333336e-06,
+      "loss": 0.1913,
+      "step": 148490
+    },
+    {
+      "epoch": 958.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9278009533882141,
+      "eval_runtime": 0.8863,
+      "eval_samples_per_second": 226.782,
+      "eval_steps_per_second": 7.898,
+      "step": 148490
+    },
+    {
+      "epoch": 958.06,
+      "learning_rate": 2.3297491039426526e-06,
+      "loss": 0.2289,
+      "step": 148500
+    },
+    {
+      "epoch": 958.13,
+      "learning_rate": 2.3261648745519717e-06,
+      "loss": 0.2064,
+      "step": 148510
+    },
+    {
+      "epoch": 958.19,
+      "learning_rate": 2.3225806451612907e-06,
+      "loss": 0.2133,
+      "step": 148520
+    },
+    {
+      "epoch": 958.26,
+      "learning_rate": 2.3189964157706094e-06,
+      "loss": 0.2082,
+      "step": 148530
+    },
+    {
+      "epoch": 958.32,
+      "learning_rate": 2.3154121863799284e-06,
+      "loss": 0.2366,
+      "step": 148540
+    },
+    {
+      "epoch": 958.39,
+      "learning_rate": 2.3118279569892475e-06,
+      "loss": 0.1756,
+      "step": 148550
+    },
+    {
+      "epoch": 958.45,
+      "learning_rate": 2.3082437275985665e-06,
+      "loss": 0.2313,
+      "step": 148560
+    },
+    {
+      "epoch": 958.51,
+      "learning_rate": 2.304659498207885e-06,
+      "loss": 0.1635,
+      "step": 148570
+    },
+    {
+      "epoch": 958.58,
+      "learning_rate": 2.301075268817204e-06,
+      "loss": 0.2019,
+      "step": 148580
+    },
+    {
+      "epoch": 958.64,
+      "learning_rate": 2.2974910394265233e-06,
+      "loss": 0.173,
+      "step": 148590
+    },
+    {
+      "epoch": 958.71,
+      "learning_rate": 2.2939068100358423e-06,
+      "loss": 0.2027,
+      "step": 148600
+    },
+    {
+      "epoch": 958.77,
+      "learning_rate": 2.2903225806451614e-06,
+      "loss": 0.1854,
+      "step": 148610
+    },
+    {
+      "epoch": 958.84,
+      "learning_rate": 2.2867383512544804e-06,
+      "loss": 0.2012,
+      "step": 148620
+    },
+    {
+      "epoch": 958.9,
+      "learning_rate": 2.2831541218637995e-06,
+      "loss": 0.2206,
+      "step": 148630
+    },
+    {
+      "epoch": 958.96,
+      "learning_rate": 2.2795698924731185e-06,
+      "loss": 0.2195,
+      "step": 148640
+    },
+    {
+      "epoch": 959.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9002814888954163,
+      "eval_runtime": 0.8795,
+      "eval_samples_per_second": 228.543,
+      "eval_steps_per_second": 7.959,
+      "step": 148645
+    },
+    {
+      "epoch": 959.03,
+      "learning_rate": 2.2759856630824376e-06,
+      "loss": 0.2368,
+      "step": 148650
+    },
+    {
+      "epoch": 959.1,
+      "learning_rate": 2.2724014336917566e-06,
+      "loss": 0.1683,
+      "step": 148660
+    },
+    {
+      "epoch": 959.16,
+      "learning_rate": 2.2688172043010757e-06,
+      "loss": 0.2205,
+      "step": 148670
+    },
+    {
+      "epoch": 959.23,
+      "learning_rate": 2.2652329749103947e-06,
+      "loss": 0.1708,
+      "step": 148680
+    },
+    {
+      "epoch": 959.29,
+      "learning_rate": 2.2616487455197133e-06,
+      "loss": 0.2082,
+      "step": 148690
+    },
+    {
+      "epoch": 959.35,
+      "learning_rate": 2.2580645161290324e-06,
+      "loss": 0.2339,
+      "step": 148700
+    },
+    {
+      "epoch": 959.42,
+      "learning_rate": 2.2544802867383515e-06,
+      "loss": 0.1537,
+      "step": 148710
+    },
+    {
+      "epoch": 959.48,
+      "learning_rate": 2.25089605734767e-06,
+      "loss": 0.1762,
+      "step": 148720
+    },
+    {
+      "epoch": 959.55,
+      "learning_rate": 2.247311827956989e-06,
+      "loss": 0.1859,
+      "step": 148730
+    },
+    {
+      "epoch": 959.61,
+      "learning_rate": 2.243727598566308e-06,
+      "loss": 0.1798,
+      "step": 148740
+    },
+    {
+      "epoch": 959.68,
+      "learning_rate": 2.2401433691756272e-06,
+      "loss": 0.1862,
+      "step": 148750
+    },
+    {
+      "epoch": 959.74,
+      "learning_rate": 2.2365591397849463e-06,
+      "loss": 0.1663,
+      "step": 148760
+    },
+    {
+      "epoch": 959.8,
+      "learning_rate": 2.2329749103942653e-06,
+      "loss": 0.2068,
+      "step": 148770
+    },
+    {
+      "epoch": 959.87,
+      "learning_rate": 2.2293906810035844e-06,
+      "loss": 0.2094,
+      "step": 148780
+    },
+    {
+      "epoch": 959.93,
+      "learning_rate": 2.2258064516129034e-06,
+      "loss": 0.1617,
+      "step": 148790
+    },
+    {
+      "epoch": 960.0,
+      "learning_rate": 2.2222222222222225e-06,
+      "loss": 0.1542,
+      "step": 148800
+    },
+    {
+      "epoch": 960.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9458550810813904,
+      "eval_runtime": 0.8632,
+      "eval_samples_per_second": 232.851,
+      "eval_steps_per_second": 8.109,
+      "step": 148800
+    },
+    {
+      "epoch": 960.06,
+      "learning_rate": 2.2186379928315416e-06,
+      "loss": 0.1678,
+      "step": 148810
+    },
+    {
+      "epoch": 960.13,
+      "learning_rate": 2.2150537634408606e-06,
+      "loss": 0.2083,
+      "step": 148820
+    },
+    {
+      "epoch": 960.19,
+      "learning_rate": 2.2114695340501797e-06,
+      "loss": 0.1818,
+      "step": 148830
+    },
+    {
+      "epoch": 960.26,
+      "learning_rate": 2.2078853046594983e-06,
+      "loss": 0.2144,
+      "step": 148840
+    },
+    {
+      "epoch": 960.32,
+      "learning_rate": 2.2043010752688173e-06,
+      "loss": 0.1795,
+      "step": 148850
+    },
+    {
+      "epoch": 960.39,
+      "learning_rate": 2.200716845878136e-06,
+      "loss": 0.1716,
+      "step": 148860
+    },
+    {
+      "epoch": 960.45,
+      "learning_rate": 2.197132616487455e-06,
+      "loss": 0.196,
+      "step": 148870
+    },
+    {
+      "epoch": 960.51,
+      "learning_rate": 2.193548387096774e-06,
+      "loss": 0.1528,
+      "step": 148880
+    },
+    {
+      "epoch": 960.58,
+      "learning_rate": 2.189964157706093e-06,
+      "loss": 0.1773,
+      "step": 148890
+    },
+    {
+      "epoch": 960.64,
+      "learning_rate": 2.186379928315412e-06,
+      "loss": 0.2204,
+      "step": 148900
+    },
+    {
+      "epoch": 960.71,
+      "learning_rate": 2.1827956989247312e-06,
+      "loss": 0.1793,
+      "step": 148910
+    },
+    {
+      "epoch": 960.77,
+      "learning_rate": 2.1792114695340503e-06,
+      "loss": 0.1971,
+      "step": 148920
+    },
+    {
+      "epoch": 960.84,
+      "learning_rate": 2.1756272401433693e-06,
+      "loss": 0.2292,
+      "step": 148930
+    },
+    {
+      "epoch": 960.9,
+      "learning_rate": 2.1720430107526884e-06,
+      "loss": 0.1757,
+      "step": 148940
+    },
+    {
+      "epoch": 960.96,
+      "learning_rate": 2.1684587813620074e-06,
+      "loss": 0.2059,
+      "step": 148950
+    },
+    {
+      "epoch": 961.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9220042824745178,
+      "eval_runtime": 0.8827,
+      "eval_samples_per_second": 227.711,
+      "eval_steps_per_second": 7.93,
+      "step": 148955
+    },
+    {
+      "epoch": 961.03,
+      "learning_rate": 2.1648745519713265e-06,
+      "loss": 0.1846,
+      "step": 148960
+    },
+    {
+      "epoch": 961.1,
+      "learning_rate": 2.1612903225806455e-06,
+      "loss": 0.2167,
+      "step": 148970
+    },
+    {
+      "epoch": 961.16,
+      "learning_rate": 2.1577060931899646e-06,
+      "loss": 0.1622,
+      "step": 148980
+    },
+    {
+      "epoch": 961.23,
+      "learning_rate": 2.1541218637992832e-06,
+      "loss": 0.1771,
+      "step": 148990
+    },
+    {
+      "epoch": 961.29,
+      "learning_rate": 2.1505376344086023e-06,
+      "loss": 0.223,
+      "step": 149000
+    },
+    {
+      "epoch": 961.35,
+      "learning_rate": 2.146953405017921e-06,
+      "loss": 0.188,
+      "step": 149010
+    },
+    {
+      "epoch": 961.42,
+      "learning_rate": 2.14336917562724e-06,
+      "loss": 0.2096,
+      "step": 149020
+    },
+    {
+      "epoch": 961.48,
+      "learning_rate": 2.139784946236559e-06,
+      "loss": 0.1827,
+      "step": 149030
+    },
+    {
+      "epoch": 961.55,
+      "learning_rate": 2.136200716845878e-06,
+      "loss": 0.1856,
+      "step": 149040
+    },
+    {
+      "epoch": 961.61,
+      "learning_rate": 2.132616487455197e-06,
+      "loss": 0.1829,
+      "step": 149050
+    },
+    {
+      "epoch": 961.68,
+      "learning_rate": 2.129032258064516e-06,
+      "loss": 0.1939,
+      "step": 149060
+    },
+    {
+      "epoch": 961.74,
+      "learning_rate": 2.125448028673835e-06,
+      "loss": 0.1848,
+      "step": 149070
+    },
+    {
+      "epoch": 961.8,
+      "learning_rate": 2.1218637992831543e-06,
+      "loss": 0.2179,
+      "step": 149080
+    },
+    {
+      "epoch": 961.87,
+      "learning_rate": 2.1182795698924733e-06,
+      "loss": 0.1549,
+      "step": 149090
+    },
+    {
+      "epoch": 961.93,
+      "learning_rate": 2.1146953405017924e-06,
+      "loss": 0.195,
+      "step": 149100
+    },
+    {
+      "epoch": 962.0,
+      "learning_rate": 2.1111111111111114e-06,
+      "loss": 0.1995,
+      "step": 149110
+    },
+    {
+      "epoch": 962.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9517995119094849,
+      "eval_runtime": 0.8604,
+      "eval_samples_per_second": 233.612,
+      "eval_steps_per_second": 8.136,
+      "step": 149110
+    },
+    {
+      "epoch": 962.06,
+      "learning_rate": 2.1075268817204305e-06,
+      "loss": 0.227,
+      "step": 149120
+    },
+    {
+      "epoch": 962.13,
+      "learning_rate": 2.1039426523297495e-06,
+      "loss": 0.1984,
+      "step": 149130
+    },
+    {
+      "epoch": 962.19,
+      "learning_rate": 2.100358422939068e-06,
+      "loss": 0.2271,
+      "step": 149140
+    },
+    {
+      "epoch": 962.26,
+      "learning_rate": 2.096774193548387e-06,
+      "loss": 0.1744,
+      "step": 149150
+    },
+    {
+      "epoch": 962.32,
+      "learning_rate": 2.0931899641577063e-06,
+      "loss": 0.1784,
+      "step": 149160
+    },
+    {
+      "epoch": 962.39,
+      "learning_rate": 2.089605734767025e-06,
+      "loss": 0.2481,
+      "step": 149170
+    },
+    {
+      "epoch": 962.45,
+      "learning_rate": 2.086021505376344e-06,
+      "loss": 0.1683,
+      "step": 149180
+    },
+    {
+      "epoch": 962.51,
+      "learning_rate": 2.082437275985663e-06,
+      "loss": 0.1992,
+      "step": 149190
+    },
+    {
+      "epoch": 962.58,
+      "learning_rate": 2.078853046594982e-06,
+      "loss": 0.167,
+      "step": 149200
+    },
+    {
+      "epoch": 962.64,
+      "learning_rate": 2.075268817204301e-06,
+      "loss": 0.1891,
+      "step": 149210
+    },
+    {
+      "epoch": 962.71,
+      "learning_rate": 2.07168458781362e-06,
+      "loss": 0.1933,
+      "step": 149220
+    },
+    {
+      "epoch": 962.77,
+      "learning_rate": 2.068100358422939e-06,
+      "loss": 0.2183,
+      "step": 149230
+    },
+    {
+      "epoch": 962.84,
+      "learning_rate": 2.0645161290322582e-06,
+      "loss": 0.1568,
+      "step": 149240
+    },
+    {
+      "epoch": 962.9,
+      "learning_rate": 2.0609318996415773e-06,
+      "loss": 0.2074,
+      "step": 149250
+    },
+    {
+      "epoch": 962.96,
+      "learning_rate": 2.0573476702508963e-06,
+      "loss": 0.1785,
+      "step": 149260
+    },
+    {
+      "epoch": 963.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9563015699386597,
+      "eval_runtime": 0.9375,
+      "eval_samples_per_second": 214.406,
+      "eval_steps_per_second": 7.467,
+      "step": 149265
+    },
+    {
+      "epoch": 963.03,
+      "learning_rate": 2.0537634408602154e-06,
+      "loss": 0.2412,
+      "step": 149270
+    },
+    {
+      "epoch": 963.1,
+      "learning_rate": 2.050179211469534e-06,
+      "loss": 0.1723,
+      "step": 149280
+    },
+    {
+      "epoch": 963.16,
+      "learning_rate": 2.046594982078853e-06,
+      "loss": 0.1827,
+      "step": 149290
+    },
+    {
+      "epoch": 963.23,
+      "learning_rate": 2.043010752688172e-06,
+      "loss": 0.1988,
+      "step": 149300
+    },
+    {
+      "epoch": 963.29,
+      "learning_rate": 2.039426523297491e-06,
+      "loss": 0.2128,
+      "step": 149310
+    },
+    {
+      "epoch": 963.35,
+      "learning_rate": 2.0358422939068102e-06,
+      "loss": 0.2359,
+      "step": 149320
+    },
+    {
+      "epoch": 963.42,
+      "learning_rate": 2.032258064516129e-06,
+      "loss": 0.2614,
+      "step": 149330
+    },
+    {
+      "epoch": 963.48,
+      "learning_rate": 2.028673835125448e-06,
+      "loss": 0.24,
+      "step": 149340
+    },
+    {
+      "epoch": 963.55,
+      "learning_rate": 2.025089605734767e-06,
+      "loss": 0.2003,
+      "step": 149350
+    },
+    {
+      "epoch": 963.61,
+      "learning_rate": 2.021505376344086e-06,
+      "loss": 0.1975,
+      "step": 149360
+    },
+    {
+      "epoch": 963.68,
+      "learning_rate": 2.017921146953405e-06,
+      "loss": 0.1811,
+      "step": 149370
+    },
+    {
+      "epoch": 963.74,
+      "learning_rate": 2.014336917562724e-06,
+      "loss": 0.1908,
+      "step": 149380
+    },
+    {
+      "epoch": 963.8,
+      "learning_rate": 2.010752688172043e-06,
+      "loss": 0.1857,
+      "step": 149390
+    },
+    {
+      "epoch": 963.87,
+      "learning_rate": 2.0071684587813622e-06,
+      "loss": 0.2152,
+      "step": 149400
+    },
+    {
+      "epoch": 963.93,
+      "learning_rate": 2.0035842293906813e-06,
+      "loss": 0.1652,
+      "step": 149410
+    },
+    {
+      "epoch": 964.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.1913,
+      "step": 149420
+    },
+    {
+      "epoch": 964.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9031722545623779,
+      "eval_runtime": 0.8743,
+      "eval_samples_per_second": 229.909,
+      "eval_steps_per_second": 8.007,
+      "step": 149420
+    },
+    {
+      "epoch": 964.06,
+      "learning_rate": 1.996415770609319e-06,
+      "loss": 0.2575,
+      "step": 149430
+    },
+    {
+      "epoch": 964.13,
+      "learning_rate": 1.992831541218638e-06,
+      "loss": 0.171,
+      "step": 149440
+    },
+    {
+      "epoch": 964.19,
+      "learning_rate": 1.989247311827957e-06,
+      "loss": 0.2573,
+      "step": 149450
+    },
+    {
+      "epoch": 964.26,
+      "learning_rate": 1.985663082437276e-06,
+      "loss": 0.1674,
+      "step": 149460
+    },
+    {
+      "epoch": 964.32,
+      "learning_rate": 1.982078853046595e-06,
+      "loss": 0.1762,
+      "step": 149470
+    },
+    {
+      "epoch": 964.39,
+      "learning_rate": 1.9784946236559142e-06,
+      "loss": 0.2315,
+      "step": 149480
+    },
+    {
+      "epoch": 964.45,
+      "learning_rate": 1.974910394265233e-06,
+      "loss": 0.2096,
+      "step": 149490
+    },
+    {
+      "epoch": 964.51,
+      "learning_rate": 1.971326164874552e-06,
+      "loss": 0.1861,
+      "step": 149500
+    },
+    {
+      "epoch": 964.58,
+      "learning_rate": 1.967741935483871e-06,
+      "loss": 0.175,
+      "step": 149510
+    },
+    {
+      "epoch": 964.64,
+      "learning_rate": 1.96415770609319e-06,
+      "loss": 0.1824,
+      "step": 149520
+    },
+    {
+      "epoch": 964.71,
+      "learning_rate": 1.960573476702509e-06,
+      "loss": 0.2175,
+      "step": 149530
+    },
+    {
+      "epoch": 964.77,
+      "learning_rate": 1.956989247311828e-06,
+      "loss": 0.2014,
+      "step": 149540
+    },
+    {
+      "epoch": 964.84,
+      "learning_rate": 1.953405017921147e-06,
+      "loss": 0.1957,
+      "step": 149550
+    },
+    {
+      "epoch": 964.9,
+      "learning_rate": 1.9498207885304662e-06,
+      "loss": 0.1909,
+      "step": 149560
+    },
+    {
+      "epoch": 964.96,
+      "learning_rate": 1.9462365591397853e-06,
+      "loss": 0.184,
+      "step": 149570
+    },
+    {
+      "epoch": 965.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9493474960327148,
+      "eval_runtime": 0.8805,
+      "eval_samples_per_second": 228.268,
+      "eval_steps_per_second": 7.95,
+      "step": 149575
+    },
+    {
+      "epoch": 965.03,
+      "learning_rate": 1.942652329749104e-06,
+      "loss": 0.2077,
+      "step": 149580
+    },
+    {
+      "epoch": 965.1,
+      "learning_rate": 1.939068100358423e-06,
+      "loss": 0.1789,
+      "step": 149590
+    },
+    {
+      "epoch": 965.16,
+      "learning_rate": 1.935483870967742e-06,
+      "loss": 0.1982,
+      "step": 149600
+    },
+    {
+      "epoch": 965.23,
+      "learning_rate": 1.931899641577061e-06,
+      "loss": 0.1858,
+      "step": 149610
+    },
+    {
+      "epoch": 965.29,
+      "learning_rate": 1.92831541218638e-06,
+      "loss": 0.1918,
+      "step": 149620
+    },
+    {
+      "epoch": 965.35,
+      "learning_rate": 1.924731182795699e-06,
+      "loss": 0.1801,
+      "step": 149630
+    },
+    {
+      "epoch": 965.42,
+      "learning_rate": 1.921146953405018e-06,
+      "loss": 0.1607,
+      "step": 149640
+    },
+    {
+      "epoch": 965.48,
+      "learning_rate": 1.917562724014337e-06,
+      "loss": 0.2084,
+      "step": 149650
+    },
+    {
+      "epoch": 965.55,
+      "learning_rate": 1.913978494623656e-06,
+      "loss": 0.2192,
+      "step": 149660
+    },
+    {
+      "epoch": 965.61,
+      "learning_rate": 1.910394265232975e-06,
+      "loss": 0.1847,
+      "step": 149670
+    },
+    {
+      "epoch": 965.68,
+      "learning_rate": 1.906810035842294e-06,
+      "loss": 0.2229,
+      "step": 149680
+    },
+    {
+      "epoch": 965.74,
+      "learning_rate": 1.9032258064516128e-06,
+      "loss": 0.2519,
+      "step": 149690
+    },
+    {
+      "epoch": 965.8,
+      "learning_rate": 1.8996415770609319e-06,
+      "loss": 0.1914,
+      "step": 149700
+    },
+    {
+      "epoch": 965.87,
+      "learning_rate": 1.896057347670251e-06,
+      "loss": 0.2032,
+      "step": 149710
+    },
+    {
+      "epoch": 965.93,
+      "learning_rate": 1.89247311827957e-06,
+      "loss": 0.1884,
+      "step": 149720
+    },
+    {
+      "epoch": 966.0,
+      "learning_rate": 1.888888888888889e-06,
+      "loss": 0.2222,
+      "step": 149730
+    },
+    {
+      "epoch": 966.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9463191032409668,
+      "eval_runtime": 0.8813,
+      "eval_samples_per_second": 228.08,
+      "eval_steps_per_second": 7.943,
+      "step": 149730
+    },
+    {
+      "epoch": 966.06,
+      "learning_rate": 1.885304659498208e-06,
+      "loss": 0.2194,
+      "step": 149740
+    },
+    {
+      "epoch": 966.13,
+      "learning_rate": 1.881720430107527e-06,
+      "loss": 0.1999,
+      "step": 149750
+    },
+    {
+      "epoch": 966.19,
+      "learning_rate": 1.878136200716846e-06,
+      "loss": 0.1993,
+      "step": 149760
+    },
+    {
+      "epoch": 966.26,
+      "learning_rate": 1.874551971326165e-06,
+      "loss": 0.2054,
+      "step": 149770
+    },
+    {
+      "epoch": 966.32,
+      "learning_rate": 1.870967741935484e-06,
+      "loss": 0.2087,
+      "step": 149780
+    },
+    {
+      "epoch": 966.39,
+      "learning_rate": 1.8673835125448031e-06,
+      "loss": 0.2109,
+      "step": 149790
+    },
+    {
+      "epoch": 966.45,
+      "learning_rate": 1.8637992831541222e-06,
+      "loss": 0.2046,
+      "step": 149800
+    },
+    {
+      "epoch": 966.51,
+      "learning_rate": 1.8602150537634408e-06,
+      "loss": 0.1438,
+      "step": 149810
+    },
+    {
+      "epoch": 966.58,
+      "learning_rate": 1.8566308243727599e-06,
+      "loss": 0.2284,
+      "step": 149820
+    },
+    {
+      "epoch": 966.64,
+      "learning_rate": 1.853046594982079e-06,
+      "loss": 0.2194,
+      "step": 149830
+    },
+    {
+      "epoch": 966.71,
+      "learning_rate": 1.8494623655913978e-06,
+      "loss": 0.1983,
+      "step": 149840
+    },
+    {
+      "epoch": 966.77,
+      "learning_rate": 1.8458781362007168e-06,
+      "loss": 0.1828,
+      "step": 149850
+    },
+    {
+      "epoch": 966.84,
+      "learning_rate": 1.8422939068100359e-06,
+      "loss": 0.1928,
+      "step": 149860
+    },
+    {
+      "epoch": 966.9,
+      "learning_rate": 1.838709677419355e-06,
+      "loss": 0.2533,
+      "step": 149870
+    },
+    {
+      "epoch": 966.96,
+      "learning_rate": 1.835125448028674e-06,
+      "loss": 0.1983,
+      "step": 149880
+    },
+    {
+      "epoch": 967.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9336521029472351,
+      "eval_runtime": 0.874,
+      "eval_samples_per_second": 229.968,
+      "eval_steps_per_second": 8.009,
+      "step": 149885
+    },
+    {
+      "epoch": 967.03,
+      "learning_rate": 1.831541218637993e-06,
+      "loss": 0.1905,
+      "step": 149890
+    },
+    {
+      "epoch": 967.1,
+      "learning_rate": 1.8279569892473119e-06,
+      "loss": 0.2511,
+      "step": 149900
+    },
+    {
+      "epoch": 967.16,
+      "learning_rate": 1.824372759856631e-06,
+      "loss": 0.195,
+      "step": 149910
+    },
+    {
+      "epoch": 967.23,
+      "learning_rate": 1.82078853046595e-06,
+      "loss": 0.1933,
+      "step": 149920
+    },
+    {
+      "epoch": 967.29,
+      "learning_rate": 1.817204301075269e-06,
+      "loss": 0.2261,
+      "step": 149930
+    },
+    {
+      "epoch": 967.35,
+      "learning_rate": 1.813620071684588e-06,
+      "loss": 0.2301,
+      "step": 149940
+    },
+    {
+      "epoch": 967.42,
+      "learning_rate": 1.8100358422939071e-06,
+      "loss": 0.1988,
+      "step": 149950
+    },
+    {
+      "epoch": 967.48,
+      "learning_rate": 1.806451612903226e-06,
+      "loss": 0.195,
+      "step": 149960
+    },
+    {
+      "epoch": 967.55,
+      "learning_rate": 1.8028673835125448e-06,
+      "loss": 0.1893,
+      "step": 149970
+    },
+    {
+      "epoch": 967.61,
+      "learning_rate": 1.7992831541218639e-06,
+      "loss": 0.2076,
+      "step": 149980
+    },
+    {
+      "epoch": 967.68,
+      "learning_rate": 1.7956989247311827e-06,
+      "loss": 0.164,
+      "step": 149990
+    },
+    {
+      "epoch": 967.74,
+      "learning_rate": 1.7921146953405017e-06,
+      "loss": 0.1871,
+      "step": 150000
+    },
+    {
+      "epoch": 967.8,
+      "learning_rate": 1.7885304659498208e-06,
+      "loss": 0.2057,
+      "step": 150010
+    },
+    {
+      "epoch": 967.87,
+      "learning_rate": 1.7849462365591399e-06,
+      "loss": 0.1773,
+      "step": 150020
+    },
+    {
+      "epoch": 967.93,
+      "learning_rate": 1.781362007168459e-06,
+      "loss": 0.1827,
+      "step": 150030
+    },
+    {
+      "epoch": 968.0,
+      "learning_rate": 1.777777777777778e-06,
+      "loss": 0.2338,
+      "step": 150040
+    },
+    {
+      "epoch": 968.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.938439667224884,
+      "eval_runtime": 0.8738,
+      "eval_samples_per_second": 230.035,
+      "eval_steps_per_second": 8.011,
+      "step": 150040
+    },
+    {
+      "epoch": 968.06,
+      "learning_rate": 1.7741935483870968e-06,
+      "loss": 0.2303,
+      "step": 150050
+    },
+    {
+      "epoch": 968.13,
+      "learning_rate": 1.7706093189964158e-06,
+      "loss": 0.2238,
+      "step": 150060
+    },
+    {
+      "epoch": 968.19,
+      "learning_rate": 1.767025089605735e-06,
+      "loss": 0.2196,
+      "step": 150070
+    },
+    {
+      "epoch": 968.26,
+      "learning_rate": 1.763440860215054e-06,
+      "loss": 0.1911,
+      "step": 150080
+    },
+    {
+      "epoch": 968.32,
+      "learning_rate": 1.759856630824373e-06,
+      "loss": 0.1696,
+      "step": 150090
+    },
+    {
+      "epoch": 968.39,
+      "learning_rate": 1.756272401433692e-06,
+      "loss": 0.2029,
+      "step": 150100
+    },
+    {
+      "epoch": 968.45,
+      "learning_rate": 1.752688172043011e-06,
+      "loss": 0.2202,
+      "step": 150110
+    },
+    {
+      "epoch": 968.51,
+      "learning_rate": 1.74910394265233e-06,
+      "loss": 0.2031,
+      "step": 150120
+    },
+    {
+      "epoch": 968.58,
+      "learning_rate": 1.7455197132616486e-06,
+      "loss": 0.1993,
+      "step": 150130
+    },
+    {
+      "epoch": 968.64,
+      "learning_rate": 1.7419354838709676e-06,
+      "loss": 0.2277,
+      "step": 150140
+    },
+    {
+      "epoch": 968.71,
+      "learning_rate": 1.7383512544802867e-06,
+      "loss": 0.2198,
+      "step": 150150
+    },
+    {
+      "epoch": 968.77,
+      "learning_rate": 1.7347670250896057e-06,
+      "loss": 0.1739,
+      "step": 150160
+    },
+    {
+      "epoch": 968.84,
+      "learning_rate": 1.7311827956989248e-06,
+      "loss": 0.2293,
+      "step": 150170
+    },
+    {
+      "epoch": 968.9,
+      "learning_rate": 1.7275985663082438e-06,
+      "loss": 0.2165,
+      "step": 150180
+    },
+    {
+      "epoch": 968.96,
+      "learning_rate": 1.7240143369175629e-06,
+      "loss": 0.2123,
+      "step": 150190
+    },
+    {
+      "epoch": 969.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9630251526832581,
+      "eval_runtime": 0.9155,
+      "eval_samples_per_second": 219.548,
+      "eval_steps_per_second": 7.646,
+      "step": 150195
+    },
+    {
+      "epoch": 969.03,
+      "learning_rate": 1.7204301075268817e-06,
+      "loss": 0.2063,
+      "step": 150200
+    },
+    {
+      "epoch": 969.1,
+      "learning_rate": 1.7168458781362008e-06,
+      "loss": 0.1597,
+      "step": 150210
+    },
+    {
+      "epoch": 969.16,
+      "learning_rate": 1.7132616487455198e-06,
+      "loss": 0.1931,
+      "step": 150220
+    },
+    {
+      "epoch": 969.23,
+      "learning_rate": 1.7096774193548389e-06,
+      "loss": 0.1942,
+      "step": 150230
+    },
+    {
+      "epoch": 969.29,
+      "learning_rate": 1.706093189964158e-06,
+      "loss": 0.1998,
+      "step": 150240
+    },
+    {
+      "epoch": 969.35,
+      "learning_rate": 1.702508960573477e-06,
+      "loss": 0.1957,
+      "step": 150250
+    },
+    {
+      "epoch": 969.42,
+      "learning_rate": 1.6989247311827958e-06,
+      "loss": 0.1795,
+      "step": 150260
+    },
+    {
+      "epoch": 969.48,
+      "learning_rate": 1.6953405017921149e-06,
+      "loss": 0.2144,
+      "step": 150270
+    },
+    {
+      "epoch": 969.55,
+      "learning_rate": 1.691756272401434e-06,
+      "loss": 0.24,
+      "step": 150280
+    },
+    {
+      "epoch": 969.61,
+      "learning_rate": 1.6881720430107526e-06,
+      "loss": 0.1793,
+      "step": 150290
+    },
+    {
+      "epoch": 969.68,
+      "learning_rate": 1.6845878136200716e-06,
+      "loss": 0.1886,
+      "step": 150300
+    },
+    {
+      "epoch": 969.74,
+      "learning_rate": 1.6810035842293907e-06,
+      "loss": 0.195,
+      "step": 150310
+    },
+    {
+      "epoch": 969.8,
+      "learning_rate": 1.6774193548387097e-06,
+      "loss": 0.2146,
+      "step": 150320
+    },
+    {
+      "epoch": 969.87,
+      "learning_rate": 1.6738351254480288e-06,
+      "loss": 0.1636,
+      "step": 150330
+    },
+    {
+      "epoch": 969.93,
+      "learning_rate": 1.6702508960573476e-06,
+      "loss": 0.188,
+      "step": 150340
+    },
+    {
+      "epoch": 970.0,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.2056,
+      "step": 150350
+    },
+    {
+      "epoch": 970.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9294943809509277,
+      "eval_runtime": 0.8941,
+      "eval_samples_per_second": 224.796,
+      "eval_steps_per_second": 7.829,
+      "step": 150350
+    },
+    {
+      "epoch": 970.06,
+      "learning_rate": 1.6630824372759857e-06,
+      "loss": 0.2349,
+      "step": 150360
+    },
+    {
+      "epoch": 970.13,
+      "learning_rate": 1.6594982078853048e-06,
+      "loss": 0.186,
+      "step": 150370
+    },
+    {
+      "epoch": 970.19,
+      "learning_rate": 1.6559139784946238e-06,
+      "loss": 0.2063,
+      "step": 150380
+    },
+    {
+      "epoch": 970.26,
+      "learning_rate": 1.6523297491039429e-06,
+      "loss": 0.224,
+      "step": 150390
+    },
+    {
+      "epoch": 970.32,
+      "learning_rate": 1.6487455197132617e-06,
+      "loss": 0.1947,
+      "step": 150400
+    },
+    {
+      "epoch": 970.39,
+      "learning_rate": 1.6451612903225808e-06,
+      "loss": 0.1956,
+      "step": 150410
+    },
+    {
+      "epoch": 970.45,
+      "learning_rate": 1.6415770609318998e-06,
+      "loss": 0.2242,
+      "step": 150420
+    },
+    {
+      "epoch": 970.51,
+      "learning_rate": 1.6379928315412189e-06,
+      "loss": 0.1746,
+      "step": 150430
+    },
+    {
+      "epoch": 970.58,
+      "learning_rate": 1.634408602150538e-06,
+      "loss": 0.1826,
+      "step": 150440
+    },
+    {
+      "epoch": 970.64,
+      "learning_rate": 1.6308243727598565e-06,
+      "loss": 0.1932,
+      "step": 150450
+    },
+    {
+      "epoch": 970.71,
+      "learning_rate": 1.6272401433691756e-06,
+      "loss": 0.1896,
+      "step": 150460
+    },
+    {
+      "epoch": 970.77,
+      "learning_rate": 1.6236559139784947e-06,
+      "loss": 0.2236,
+      "step": 150470
+    },
+    {
+      "epoch": 970.84,
+      "learning_rate": 1.6200716845878137e-06,
+      "loss": 0.1831,
+      "step": 150480
+    },
+    {
+      "epoch": 970.9,
+      "learning_rate": 1.6164874551971325e-06,
+      "loss": 0.1828,
+      "step": 150490
+    },
+    {
+      "epoch": 970.96,
+      "learning_rate": 1.6129032258064516e-06,
+      "loss": 0.2148,
+      "step": 150500
+    },
+    {
+      "epoch": 971.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9239917993545532,
+      "eval_runtime": 0.894,
+      "eval_samples_per_second": 224.827,
+      "eval_steps_per_second": 7.83,
+      "step": 150505
+    },
+    {
+      "epoch": 971.03,
+      "learning_rate": 1.6093189964157706e-06,
+      "loss": 0.2562,
+      "step": 150510
+    },
+    {
+      "epoch": 971.1,
+      "learning_rate": 1.6057347670250897e-06,
+      "loss": 0.2121,
+      "step": 150520
+    },
+    {
+      "epoch": 971.16,
+      "learning_rate": 1.6021505376344088e-06,
+      "loss": 0.2298,
+      "step": 150530
+    },
+    {
+      "epoch": 971.23,
+      "learning_rate": 1.5985663082437278e-06,
+      "loss": 0.1839,
+      "step": 150540
+    },
+    {
+      "epoch": 971.29,
+      "learning_rate": 1.5949820788530466e-06,
+      "loss": 0.1703,
+      "step": 150550
+    },
+    {
+      "epoch": 971.35,
+      "learning_rate": 1.5913978494623657e-06,
+      "loss": 0.1724,
+      "step": 150560
+    },
+    {
+      "epoch": 971.42,
+      "learning_rate": 1.5878136200716847e-06,
+      "loss": 0.2021,
+      "step": 150570
+    },
+    {
+      "epoch": 971.48,
+      "learning_rate": 1.5842293906810038e-06,
+      "loss": 0.1903,
+      "step": 150580
+    },
+    {
+      "epoch": 971.55,
+      "learning_rate": 1.5806451612903229e-06,
+      "loss": 0.1659,
+      "step": 150590
+    },
+    {
+      "epoch": 971.61,
+      "learning_rate": 1.577060931899642e-06,
+      "loss": 0.1879,
+      "step": 150600
+    },
+    {
+      "epoch": 971.68,
+      "learning_rate": 1.5734767025089605e-06,
+      "loss": 0.1908,
+      "step": 150610
+    },
+    {
+      "epoch": 971.74,
+      "learning_rate": 1.5698924731182796e-06,
+      "loss": 0.1474,
+      "step": 150620
+    },
+    {
+      "epoch": 971.8,
+      "learning_rate": 1.5663082437275986e-06,
+      "loss": 0.1851,
+      "step": 150630
+    },
+    {
+      "epoch": 971.87,
+      "learning_rate": 1.5627240143369175e-06,
+      "loss": 0.1787,
+      "step": 150640
+    },
+    {
+      "epoch": 971.93,
+      "learning_rate": 1.5591397849462365e-06,
+      "loss": 0.2086,
+      "step": 150650
+    },
+    {
+      "epoch": 972.0,
+      "learning_rate": 1.5555555555555556e-06,
+      "loss": 0.1916,
+      "step": 150660
+    },
+    {
+      "epoch": 972.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9597251415252686,
+      "eval_runtime": 0.8695,
+      "eval_samples_per_second": 231.156,
+      "eval_steps_per_second": 8.05,
+      "step": 150660
+    },
+    {
+      "epoch": 972.06,
+      "learning_rate": 1.5519713261648746e-06,
+      "loss": 0.1952,
+      "step": 150670
+    },
+    {
+      "epoch": 972.13,
+      "learning_rate": 1.5483870967741937e-06,
+      "loss": 0.1398,
+      "step": 150680
+    },
+    {
+      "epoch": 972.19,
+      "learning_rate": 1.5448028673835127e-06,
+      "loss": 0.2122,
+      "step": 150690
+    },
+    {
+      "epoch": 972.26,
+      "learning_rate": 1.5412186379928316e-06,
+      "loss": 0.1403,
+      "step": 150700
+    },
+    {
+      "epoch": 972.32,
+      "learning_rate": 1.5376344086021506e-06,
+      "loss": 0.207,
+      "step": 150710
+    },
+    {
+      "epoch": 972.39,
+      "learning_rate": 1.5340501792114697e-06,
+      "loss": 0.1872,
+      "step": 150720
+    },
+    {
+      "epoch": 972.45,
+      "learning_rate": 1.5304659498207885e-06,
+      "loss": 0.1978,
+      "step": 150730
+    },
+    {
+      "epoch": 972.51,
+      "learning_rate": 1.5268817204301076e-06,
+      "loss": 0.212,
+      "step": 150740
+    },
+    {
+      "epoch": 972.58,
+      "learning_rate": 1.5232974910394266e-06,
+      "loss": 0.2003,
+      "step": 150750
+    },
+    {
+      "epoch": 972.64,
+      "learning_rate": 1.5197132616487457e-06,
+      "loss": 0.2509,
+      "step": 150760
+    },
+    {
+      "epoch": 972.71,
+      "learning_rate": 1.5161290322580645e-06,
+      "loss": 0.208,
+      "step": 150770
+    },
+    {
+      "epoch": 972.77,
+      "learning_rate": 1.5125448028673836e-06,
+      "loss": 0.1943,
+      "step": 150780
+    },
+    {
+      "epoch": 972.84,
+      "learning_rate": 1.5089605734767026e-06,
+      "loss": 0.1618,
+      "step": 150790
+    },
+    {
+      "epoch": 972.9,
+      "learning_rate": 1.5053763440860217e-06,
+      "loss": 0.1896,
+      "step": 150800
+    },
+    {
+      "epoch": 972.96,
+      "learning_rate": 1.5017921146953405e-06,
+      "loss": 0.1717,
+      "step": 150810
+    },
+    {
+      "epoch": 973.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9347867965698242,
+      "eval_runtime": 0.8847,
+      "eval_samples_per_second": 227.207,
+      "eval_steps_per_second": 7.913,
+      "step": 150815
+    },
+    {
+      "epoch": 973.03,
+      "learning_rate": 1.4982078853046596e-06,
+      "loss": 0.1807,
+      "step": 150820
+    },
+    {
+      "epoch": 973.1,
+      "learning_rate": 1.4946236559139786e-06,
+      "loss": 0.2126,
+      "step": 150830
+    },
+    {
+      "epoch": 973.16,
+      "learning_rate": 1.4910394265232975e-06,
+      "loss": 0.2072,
+      "step": 150840
+    },
+    {
+      "epoch": 973.23,
+      "learning_rate": 1.4874551971326165e-06,
+      "loss": 0.1747,
+      "step": 150850
+    },
+    {
+      "epoch": 973.29,
+      "learning_rate": 1.4838709677419356e-06,
+      "loss": 0.1755,
+      "step": 150860
+    },
+    {
+      "epoch": 973.35,
+      "learning_rate": 1.4802867383512546e-06,
+      "loss": 0.2087,
+      "step": 150870
+    },
+    {
+      "epoch": 973.42,
+      "learning_rate": 1.4767025089605737e-06,
+      "loss": 0.1863,
+      "step": 150880
+    },
+    {
+      "epoch": 973.48,
+      "learning_rate": 1.4731182795698925e-06,
+      "loss": 0.1983,
+      "step": 150890
+    },
+    {
+      "epoch": 973.55,
+      "learning_rate": 1.4695340501792116e-06,
+      "loss": 0.2113,
+      "step": 150900
+    },
+    {
+      "epoch": 973.61,
+      "learning_rate": 1.4659498207885306e-06,
+      "loss": 0.1871,
+      "step": 150910
+    },
+    {
+      "epoch": 973.68,
+      "learning_rate": 1.4623655913978494e-06,
+      "loss": 0.2064,
+      "step": 150920
+    },
+    {
+      "epoch": 973.74,
+      "learning_rate": 1.4587813620071685e-06,
+      "loss": 0.2077,
+      "step": 150930
+    },
+    {
+      "epoch": 973.8,
+      "learning_rate": 1.4551971326164876e-06,
+      "loss": 0.1884,
+      "step": 150940
+    },
+    {
+      "epoch": 973.87,
+      "learning_rate": 1.4516129032258066e-06,
+      "loss": 0.2315,
+      "step": 150950
+    },
+    {
+      "epoch": 973.93,
+      "learning_rate": 1.4480286738351257e-06,
+      "loss": 0.2377,
+      "step": 150960
+    },
+    {
+      "epoch": 974.0,
+      "learning_rate": 1.4444444444444445e-06,
+      "loss": 0.1886,
+      "step": 150970
+    },
+    {
+      "epoch": 974.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9386924505233765,
+      "eval_runtime": 0.8655,
+      "eval_samples_per_second": 232.235,
+      "eval_steps_per_second": 8.088,
+      "step": 150970
+    },
+    {
+      "epoch": 974.06,
+      "learning_rate": 1.4408602150537635e-06,
+      "loss": 0.1802,
+      "step": 150980
+    },
+    {
+      "epoch": 974.13,
+      "learning_rate": 1.4372759856630824e-06,
+      "loss": 0.1912,
+      "step": 150990
+    },
+    {
+      "epoch": 974.19,
+      "learning_rate": 1.4336917562724014e-06,
+      "loss": 0.2216,
+      "step": 151000
+    },
+    {
+      "epoch": 974.26,
+      "learning_rate": 1.4301075268817205e-06,
+      "loss": 0.1754,
+      "step": 151010
+    },
+    {
+      "epoch": 974.32,
+      "learning_rate": 1.4265232974910395e-06,
+      "loss": 0.192,
+      "step": 151020
+    },
+    {
+      "epoch": 974.39,
+      "learning_rate": 1.4229390681003586e-06,
+      "loss": 0.2186,
+      "step": 151030
+    },
+    {
+      "epoch": 974.45,
+      "learning_rate": 1.4193548387096776e-06,
+      "loss": 0.1784,
+      "step": 151040
+    },
+    {
+      "epoch": 974.51,
+      "learning_rate": 1.4157706093189965e-06,
+      "loss": 0.1979,
+      "step": 151050
+    },
+    {
+      "epoch": 974.58,
+      "learning_rate": 1.4121863799283155e-06,
+      "loss": 0.1867,
+      "step": 151060
+    },
+    {
+      "epoch": 974.64,
+      "learning_rate": 1.4086021505376344e-06,
+      "loss": 0.2433,
+      "step": 151070
+    },
+    {
+      "epoch": 974.71,
+      "learning_rate": 1.4050179211469534e-06,
+      "loss": 0.1811,
+      "step": 151080
+    },
+    {
+      "epoch": 974.77,
+      "learning_rate": 1.4014336917562725e-06,
+      "loss": 0.1961,
+      "step": 151090
+    },
+    {
+      "epoch": 974.84,
+      "learning_rate": 1.3978494623655915e-06,
+      "loss": 0.1946,
+      "step": 151100
+    },
+    {
+      "epoch": 974.9,
+      "learning_rate": 1.3942652329749106e-06,
+      "loss": 0.2143,
+      "step": 151110
+    },
+    {
+      "epoch": 974.96,
+      "learning_rate": 1.3906810035842296e-06,
+      "loss": 0.2068,
+      "step": 151120
+    },
+    {
+      "epoch": 975.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.946816086769104,
+      "eval_runtime": 0.8562,
+      "eval_samples_per_second": 234.747,
+      "eval_steps_per_second": 8.175,
+      "step": 151125
+    },
+    {
+      "epoch": 975.03,
+      "learning_rate": 1.3870967741935485e-06,
+      "loss": 0.2163,
+      "step": 151130
+    },
+    {
+      "epoch": 975.1,
+      "learning_rate": 1.3835125448028673e-06,
+      "loss": 0.182,
+      "step": 151140
+    },
+    {
+      "epoch": 975.16,
+      "learning_rate": 1.3799283154121864e-06,
+      "loss": 0.2184,
+      "step": 151150
+    },
+    {
+      "epoch": 975.23,
+      "learning_rate": 1.3763440860215054e-06,
+      "loss": 0.2295,
+      "step": 151160
+    },
+    {
+      "epoch": 975.29,
+      "learning_rate": 1.3727598566308245e-06,
+      "loss": 0.1737,
+      "step": 151170
+    },
+    {
+      "epoch": 975.35,
+      "learning_rate": 1.3691756272401435e-06,
+      "loss": 0.1893,
+      "step": 151180
+    },
+    {
+      "epoch": 975.42,
+      "learning_rate": 1.3655913978494626e-06,
+      "loss": 0.1847,
+      "step": 151190
+    },
+    {
+      "epoch": 975.48,
+      "learning_rate": 1.3620071684587814e-06,
+      "loss": 0.1867,
+      "step": 151200
+    },
+    {
+      "epoch": 975.55,
+      "learning_rate": 1.3584229390681003e-06,
+      "loss": 0.1655,
+      "step": 151210
+    },
+    {
+      "epoch": 975.61,
+      "learning_rate": 1.3548387096774193e-06,
+      "loss": 0.1758,
+      "step": 151220
+    },
+    {
+      "epoch": 975.68,
+      "learning_rate": 1.3512544802867384e-06,
+      "loss": 0.2065,
+      "step": 151230
+    },
+    {
+      "epoch": 975.74,
+      "learning_rate": 1.3476702508960574e-06,
+      "loss": 0.2155,
+      "step": 151240
+    },
+    {
+      "epoch": 975.8,
+      "learning_rate": 1.3440860215053765e-06,
+      "loss": 0.1855,
+      "step": 151250
+    },
+    {
+      "epoch": 975.87,
+      "learning_rate": 1.3405017921146955e-06,
+      "loss": 0.1879,
+      "step": 151260
+    },
+    {
+      "epoch": 975.93,
+      "learning_rate": 1.3369175627240144e-06,
+      "loss": 0.2525,
+      "step": 151270
+    },
+    {
+      "epoch": 976.0,
+      "learning_rate": 1.3333333333333334e-06,
+      "loss": 0.1812,
+      "step": 151280
+    },
+    {
+      "epoch": 976.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.941020131111145,
+      "eval_runtime": 0.8805,
+      "eval_samples_per_second": 228.273,
+      "eval_steps_per_second": 7.95,
+      "step": 151280
+    },
+    {
+      "epoch": 976.06,
+      "learning_rate": 1.3297491039426523e-06,
+      "loss": 0.2173,
+      "step": 151290
+    },
+    {
+      "epoch": 976.13,
+      "learning_rate": 1.3261648745519713e-06,
+      "loss": 0.1932,
+      "step": 151300
+    },
+    {
+      "epoch": 976.19,
+      "learning_rate": 1.3225806451612904e-06,
+      "loss": 0.2151,
+      "step": 151310
+    },
+    {
+      "epoch": 976.26,
+      "learning_rate": 1.3189964157706094e-06,
+      "loss": 0.1876,
+      "step": 151320
+    },
+    {
+      "epoch": 976.32,
+      "learning_rate": 1.3154121863799285e-06,
+      "loss": 0.1629,
+      "step": 151330
+    },
+    {
+      "epoch": 976.39,
+      "learning_rate": 1.3118279569892475e-06,
+      "loss": 0.1695,
+      "step": 151340
+    },
+    {
+      "epoch": 976.45,
+      "learning_rate": 1.3082437275985664e-06,
+      "loss": 0.1624,
+      "step": 151350
+    },
+    {
+      "epoch": 976.51,
+      "learning_rate": 1.3046594982078854e-06,
+      "loss": 0.1731,
+      "step": 151360
+    },
+    {
+      "epoch": 976.58,
+      "learning_rate": 1.3010752688172042e-06,
+      "loss": 0.1695,
+      "step": 151370
+    },
+    {
+      "epoch": 976.64,
+      "learning_rate": 1.2974910394265233e-06,
+      "loss": 0.1737,
+      "step": 151380
+    },
+    {
+      "epoch": 976.71,
+      "learning_rate": 1.2939068100358424e-06,
+      "loss": 0.1811,
+      "step": 151390
+    },
+    {
+      "epoch": 976.77,
+      "learning_rate": 1.2903225806451614e-06,
+      "loss": 0.1879,
+      "step": 151400
+    },
+    {
+      "epoch": 976.84,
+      "learning_rate": 1.2867383512544805e-06,
+      "loss": 0.218,
+      "step": 151410
+    },
+    {
+      "epoch": 976.9,
+      "learning_rate": 1.2831541218637993e-06,
+      "loss": 0.201,
+      "step": 151420
+    },
+    {
+      "epoch": 976.96,
+      "learning_rate": 1.2795698924731183e-06,
+      "loss": 0.2144,
+      "step": 151430
+    },
+    {
+      "epoch": 977.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9194276928901672,
+      "eval_runtime": 0.8821,
+      "eval_samples_per_second": 227.876,
+      "eval_steps_per_second": 7.936,
+      "step": 151435
+    },
+    {
+      "epoch": 977.03,
+      "learning_rate": 1.2759856630824374e-06,
+      "loss": 0.2096,
+      "step": 151440
+    },
+    {
+      "epoch": 977.1,
+      "learning_rate": 1.2724014336917562e-06,
+      "loss": 0.1705,
+      "step": 151450
+    },
+    {
+      "epoch": 977.16,
+      "learning_rate": 1.2688172043010753e-06,
+      "loss": 0.1937,
+      "step": 151460
+    },
+    {
+      "epoch": 977.23,
+      "learning_rate": 1.2652329749103943e-06,
+      "loss": 0.1896,
+      "step": 151470
+    },
+    {
+      "epoch": 977.29,
+      "learning_rate": 1.2616487455197134e-06,
+      "loss": 0.1855,
+      "step": 151480
+    },
+    {
+      "epoch": 977.35,
+      "learning_rate": 1.2580645161290322e-06,
+      "loss": 0.1988,
+      "step": 151490
+    },
+    {
+      "epoch": 977.42,
+      "learning_rate": 1.2544802867383513e-06,
+      "loss": 0.2064,
+      "step": 151500
+    },
+    {
+      "epoch": 977.48,
+      "learning_rate": 1.2508960573476703e-06,
+      "loss": 0.1538,
+      "step": 151510
+    },
+    {
+      "epoch": 977.55,
+      "learning_rate": 1.2473118279569894e-06,
+      "loss": 0.1962,
+      "step": 151520
+    },
+    {
+      "epoch": 977.61,
+      "learning_rate": 1.2437275985663082e-06,
+      "loss": 0.1777,
+      "step": 151530
+    },
+    {
+      "epoch": 977.68,
+      "learning_rate": 1.2401433691756273e-06,
+      "loss": 0.217,
+      "step": 151540
+    },
+    {
+      "epoch": 977.74,
+      "learning_rate": 1.2365591397849463e-06,
+      "loss": 0.2117,
+      "step": 151550
+    },
+    {
+      "epoch": 977.8,
+      "learning_rate": 1.2329749103942654e-06,
+      "loss": 0.1941,
+      "step": 151560
+    },
+    {
+      "epoch": 977.87,
+      "learning_rate": 1.2293906810035842e-06,
+      "loss": 0.1816,
+      "step": 151570
+    },
+    {
+      "epoch": 977.93,
+      "learning_rate": 1.2258064516129033e-06,
+      "loss": 0.2015,
+      "step": 151580
+    },
+    {
+      "epoch": 978.0,
+      "learning_rate": 1.2222222222222223e-06,
+      "loss": 0.1813,
+      "step": 151590
+    },
+    {
+      "epoch": 978.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9239837527275085,
+      "eval_runtime": 0.8812,
+      "eval_samples_per_second": 228.105,
+      "eval_steps_per_second": 7.944,
+      "step": 151590
+    },
+    {
+      "epoch": 978.06,
+      "learning_rate": 1.2186379928315414e-06,
+      "loss": 0.251,
+      "step": 151600
+    },
+    {
+      "epoch": 978.13,
+      "learning_rate": 1.2150537634408602e-06,
+      "loss": 0.1953,
+      "step": 151610
+    },
+    {
+      "epoch": 978.19,
+      "learning_rate": 1.2114695340501793e-06,
+      "loss": 0.2271,
+      "step": 151620
+    },
+    {
+      "epoch": 978.26,
+      "learning_rate": 1.2078853046594983e-06,
+      "loss": 0.2133,
+      "step": 151630
+    },
+    {
+      "epoch": 978.32,
+      "learning_rate": 1.2043010752688172e-06,
+      "loss": 0.1765,
+      "step": 151640
+    },
+    {
+      "epoch": 978.39,
+      "learning_rate": 1.2007168458781362e-06,
+      "loss": 0.1913,
+      "step": 151650
+    },
+    {
+      "epoch": 978.45,
+      "learning_rate": 1.1971326164874553e-06,
+      "loss": 0.1849,
+      "step": 151660
+    },
+    {
+      "epoch": 978.51,
+      "learning_rate": 1.1935483870967743e-06,
+      "loss": 0.186,
+      "step": 151670
+    },
+    {
+      "epoch": 978.58,
+      "learning_rate": 1.1899641577060934e-06,
+      "loss": 0.179,
+      "step": 151680
+    },
+    {
+      "epoch": 978.64,
+      "learning_rate": 1.1863799283154122e-06,
+      "loss": 0.2239,
+      "step": 151690
+    },
+    {
+      "epoch": 978.71,
+      "learning_rate": 1.1827956989247313e-06,
+      "loss": 0.1846,
+      "step": 151700
+    },
+    {
+      "epoch": 978.77,
+      "learning_rate": 1.1792114695340501e-06,
+      "loss": 0.2147,
+      "step": 151710
+    },
+    {
+      "epoch": 978.84,
+      "learning_rate": 1.1756272401433692e-06,
+      "loss": 0.147,
+      "step": 151720
+    },
+    {
+      "epoch": 978.9,
+      "learning_rate": 1.1720430107526882e-06,
+      "loss": 0.1925,
+      "step": 151730
+    },
+    {
+      "epoch": 978.96,
+      "learning_rate": 1.1684587813620073e-06,
+      "loss": 0.1759,
+      "step": 151740
+    },
+    {
+      "epoch": 979.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9289430975914001,
+      "eval_runtime": 0.8784,
+      "eval_samples_per_second": 228.819,
+      "eval_steps_per_second": 7.969,
+      "step": 151745
+    },
+    {
+      "epoch": 979.03,
+      "learning_rate": 1.1648745519713263e-06,
+      "loss": 0.2462,
+      "step": 151750
+    },
+    {
+      "epoch": 979.1,
+      "learning_rate": 1.1612903225806454e-06,
+      "loss": 0.172,
+      "step": 151760
+    },
+    {
+      "epoch": 979.16,
+      "learning_rate": 1.1577060931899642e-06,
+      "loss": 0.191,
+      "step": 151770
+    },
+    {
+      "epoch": 979.23,
+      "learning_rate": 1.1541218637992833e-06,
+      "loss": 0.1901,
+      "step": 151780
+    },
+    {
+      "epoch": 979.29,
+      "learning_rate": 1.150537634408602e-06,
+      "loss": 0.168,
+      "step": 151790
+    },
+    {
+      "epoch": 979.35,
+      "learning_rate": 1.1469534050179212e-06,
+      "loss": 0.1619,
+      "step": 151800
+    },
+    {
+      "epoch": 979.42,
+      "learning_rate": 1.1433691756272402e-06,
+      "loss": 0.2087,
+      "step": 151810
+    },
+    {
+      "epoch": 979.48,
+      "learning_rate": 1.1397849462365593e-06,
+      "loss": 0.1727,
+      "step": 151820
+    },
+    {
+      "epoch": 979.55,
+      "learning_rate": 1.1362007168458783e-06,
+      "loss": 0.1817,
+      "step": 151830
+    },
+    {
+      "epoch": 979.61,
+      "learning_rate": 1.1326164874551974e-06,
+      "loss": 0.2073,
+      "step": 151840
+    },
+    {
+      "epoch": 979.68,
+      "learning_rate": 1.1290322580645162e-06,
+      "loss": 0.2438,
+      "step": 151850
+    },
+    {
+      "epoch": 979.74,
+      "learning_rate": 1.125448028673835e-06,
+      "loss": 0.1862,
+      "step": 151860
+    },
+    {
+      "epoch": 979.8,
+      "learning_rate": 1.121863799283154e-06,
+      "loss": 0.183,
+      "step": 151870
+    },
+    {
+      "epoch": 979.87,
+      "learning_rate": 1.1182795698924731e-06,
+      "loss": 0.1861,
+      "step": 151880
+    },
+    {
+      "epoch": 979.93,
+      "learning_rate": 1.1146953405017922e-06,
+      "loss": 0.197,
+      "step": 151890
+    },
+    {
+      "epoch": 980.0,
+      "learning_rate": 1.1111111111111112e-06,
+      "loss": 0.2373,
+      "step": 151900
+    },
+    {
+      "epoch": 980.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.921204686164856,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.378,
+      "eval_steps_per_second": 8.058,
+      "step": 151900
+    },
+    {
+      "epoch": 980.06,
+      "learning_rate": 1.1075268817204303e-06,
+      "loss": 0.1639,
+      "step": 151910
+    },
+    {
+      "epoch": 980.13,
+      "learning_rate": 1.1039426523297491e-06,
+      "loss": 0.1945,
+      "step": 151920
+    },
+    {
+      "epoch": 980.19,
+      "learning_rate": 1.100358422939068e-06,
+      "loss": 0.1469,
+      "step": 151930
+    },
+    {
+      "epoch": 980.26,
+      "learning_rate": 1.096774193548387e-06,
+      "loss": 0.2337,
+      "step": 151940
+    },
+    {
+      "epoch": 980.32,
+      "learning_rate": 1.093189964157706e-06,
+      "loss": 0.2062,
+      "step": 151950
+    },
+    {
+      "epoch": 980.39,
+      "learning_rate": 1.0896057347670251e-06,
+      "loss": 0.2039,
+      "step": 151960
+    },
+    {
+      "epoch": 980.45,
+      "learning_rate": 1.0860215053763442e-06,
+      "loss": 0.2089,
+      "step": 151970
+    },
+    {
+      "epoch": 980.51,
+      "learning_rate": 1.0824372759856632e-06,
+      "loss": 0.201,
+      "step": 151980
+    },
+    {
+      "epoch": 980.58,
+      "learning_rate": 1.0788530465949823e-06,
+      "loss": 0.1878,
+      "step": 151990
+    },
+    {
+      "epoch": 980.64,
+      "learning_rate": 1.0752688172043011e-06,
+      "loss": 0.1978,
+      "step": 152000
+    },
+    {
+      "epoch": 980.71,
+      "learning_rate": 1.07168458781362e-06,
+      "loss": 0.2093,
+      "step": 152010
+    },
+    {
+      "epoch": 980.77,
+      "learning_rate": 1.068100358422939e-06,
+      "loss": 0.1706,
+      "step": 152020
+    },
+    {
+      "epoch": 980.84,
+      "learning_rate": 1.064516129032258e-06,
+      "loss": 0.1988,
+      "step": 152030
+    },
+    {
+      "epoch": 980.9,
+      "learning_rate": 1.0609318996415771e-06,
+      "loss": 0.2062,
+      "step": 152040
+    },
+    {
+      "epoch": 980.96,
+      "learning_rate": 1.0573476702508962e-06,
+      "loss": 0.2157,
+      "step": 152050
+    },
+    {
+      "epoch": 981.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9387524127960205,
+      "eval_runtime": 0.9024,
+      "eval_samples_per_second": 222.745,
+      "eval_steps_per_second": 7.757,
+      "step": 152055
+    },
+    {
+      "epoch": 981.03,
+      "learning_rate": 1.0537634408602152e-06,
+      "loss": 0.2021,
+      "step": 152060
+    },
+    {
+      "epoch": 981.1,
+      "learning_rate": 1.050179211469534e-06,
+      "loss": 0.225,
+      "step": 152070
+    },
+    {
+      "epoch": 981.16,
+      "learning_rate": 1.0465949820788531e-06,
+      "loss": 0.1995,
+      "step": 152080
+    },
+    {
+      "epoch": 981.23,
+      "learning_rate": 1.043010752688172e-06,
+      "loss": 0.1636,
+      "step": 152090
+    },
+    {
+      "epoch": 981.29,
+      "learning_rate": 1.039426523297491e-06,
+      "loss": 0.1984,
+      "step": 152100
+    },
+    {
+      "epoch": 981.35,
+      "learning_rate": 1.03584229390681e-06,
+      "loss": 0.2232,
+      "step": 152110
+    },
+    {
+      "epoch": 981.42,
+      "learning_rate": 1.0322580645161291e-06,
+      "loss": 0.1211,
+      "step": 152120
+    },
+    {
+      "epoch": 981.48,
+      "learning_rate": 1.0286738351254482e-06,
+      "loss": 0.1979,
+      "step": 152130
+    },
+    {
+      "epoch": 981.55,
+      "learning_rate": 1.025089605734767e-06,
+      "loss": 0.2034,
+      "step": 152140
+    },
+    {
+      "epoch": 981.61,
+      "learning_rate": 1.021505376344086e-06,
+      "loss": 0.21,
+      "step": 152150
+    },
+    {
+      "epoch": 981.68,
+      "learning_rate": 1.0179211469534051e-06,
+      "loss": 0.1648,
+      "step": 152160
+    },
+    {
+      "epoch": 981.74,
+      "learning_rate": 1.014336917562724e-06,
+      "loss": 0.154,
+      "step": 152170
+    },
+    {
+      "epoch": 981.8,
+      "learning_rate": 1.010752688172043e-06,
+      "loss": 0.2158,
+      "step": 152180
+    },
+    {
+      "epoch": 981.87,
+      "learning_rate": 1.007168458781362e-06,
+      "loss": 0.2092,
+      "step": 152190
+    },
+    {
+      "epoch": 981.93,
+      "learning_rate": 1.0035842293906811e-06,
+      "loss": 0.2149,
+      "step": 152200
+    },
+    {
+      "epoch": 982.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.2019,
+      "step": 152210
+    },
+    {
+      "epoch": 982.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9329764246940613,
+      "eval_runtime": 0.8784,
+      "eval_samples_per_second": 228.823,
+      "eval_steps_per_second": 7.969,
+      "step": 152210
+    },
+    {
+      "epoch": 982.06,
+      "learning_rate": 9.96415770609319e-07,
+      "loss": 0.2491,
+      "step": 152220
+    },
+    {
+      "epoch": 982.13,
+      "learning_rate": 9.92831541218638e-07,
+      "loss": 0.1959,
+      "step": 152230
+    },
+    {
+      "epoch": 982.19,
+      "learning_rate": 9.892473118279571e-07,
+      "loss": 0.174,
+      "step": 152240
+    },
+    {
+      "epoch": 982.26,
+      "learning_rate": 9.85663082437276e-07,
+      "loss": 0.1831,
+      "step": 152250
+    },
+    {
+      "epoch": 982.32,
+      "learning_rate": 9.82078853046595e-07,
+      "loss": 0.207,
+      "step": 152260
+    },
+    {
+      "epoch": 982.39,
+      "learning_rate": 9.78494623655914e-07,
+      "loss": 0.16,
+      "step": 152270
+    },
+    {
+      "epoch": 982.45,
+      "learning_rate": 9.749103942652331e-07,
+      "loss": 0.1913,
+      "step": 152280
+    },
+    {
+      "epoch": 982.51,
+      "learning_rate": 9.71326164874552e-07,
+      "loss": 0.1605,
+      "step": 152290
+    },
+    {
+      "epoch": 982.58,
+      "learning_rate": 9.67741935483871e-07,
+      "loss": 0.1704,
+      "step": 152300
+    },
+    {
+      "epoch": 982.64,
+      "learning_rate": 9.6415770609319e-07,
+      "loss": 0.1819,
+      "step": 152310
+    },
+    {
+      "epoch": 982.71,
+      "learning_rate": 9.60573476702509e-07,
+      "loss": 0.1902,
+      "step": 152320
+    },
+    {
+      "epoch": 982.77,
+      "learning_rate": 9.56989247311828e-07,
+      "loss": 0.1657,
+      "step": 152330
+    },
+    {
+      "epoch": 982.84,
+      "learning_rate": 9.53405017921147e-07,
+      "loss": 0.1986,
+      "step": 152340
+    },
+    {
+      "epoch": 982.9,
+      "learning_rate": 9.498207885304659e-07,
+      "loss": 0.2091,
+      "step": 152350
+    },
+    {
+      "epoch": 982.96,
+      "learning_rate": 9.46236559139785e-07,
+      "loss": 0.1988,
+      "step": 152360
+    },
+    {
+      "epoch": 983.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9611635804176331,
+      "eval_runtime": 0.8849,
+      "eval_samples_per_second": 227.14,
+      "eval_steps_per_second": 7.91,
+      "step": 152365
+    },
+    {
+      "epoch": 983.03,
+      "learning_rate": 9.42652329749104e-07,
+      "loss": 0.2051,
+      "step": 152370
+    },
+    {
+      "epoch": 983.1,
+      "learning_rate": 9.39068100358423e-07,
+      "loss": 0.1923,
+      "step": 152380
+    },
+    {
+      "epoch": 983.16,
+      "learning_rate": 9.35483870967742e-07,
+      "loss": 0.2005,
+      "step": 152390
+    },
+    {
+      "epoch": 983.23,
+      "learning_rate": 9.318996415770611e-07,
+      "loss": 0.1531,
+      "step": 152400
+    },
+    {
+      "epoch": 983.29,
+      "learning_rate": 9.283154121863799e-07,
+      "loss": 0.1648,
+      "step": 152410
+    },
+    {
+      "epoch": 983.35,
+      "learning_rate": 9.247311827956989e-07,
+      "loss": 0.1987,
+      "step": 152420
+    },
+    {
+      "epoch": 983.42,
+      "learning_rate": 9.211469534050179e-07,
+      "loss": 0.2063,
+      "step": 152430
+    },
+    {
+      "epoch": 983.48,
+      "learning_rate": 9.17562724014337e-07,
+      "loss": 0.1696,
+      "step": 152440
+    },
+    {
+      "epoch": 983.55,
+      "learning_rate": 9.139784946236559e-07,
+      "loss": 0.192,
+      "step": 152450
+    },
+    {
+      "epoch": 983.61,
+      "learning_rate": 9.10394265232975e-07,
+      "loss": 0.1915,
+      "step": 152460
+    },
+    {
+      "epoch": 983.68,
+      "learning_rate": 9.06810035842294e-07,
+      "loss": 0.22,
+      "step": 152470
+    },
+    {
+      "epoch": 983.74,
+      "learning_rate": 9.03225806451613e-07,
+      "loss": 0.2137,
+      "step": 152480
+    },
+    {
+      "epoch": 983.8,
+      "learning_rate": 8.996415770609319e-07,
+      "loss": 0.2049,
+      "step": 152490
+    },
+    {
+      "epoch": 983.87,
+      "learning_rate": 8.960573476702509e-07,
+      "loss": 0.1682,
+      "step": 152500
+    },
+    {
+      "epoch": 983.93,
+      "learning_rate": 8.924731182795699e-07,
+      "loss": 0.2108,
+      "step": 152510
+    },
+    {
+      "epoch": 984.0,
+      "learning_rate": 8.88888888888889e-07,
+      "loss": 0.1909,
+      "step": 152520
+    },
+    {
+      "epoch": 984.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9438244104385376,
+      "eval_runtime": 0.8637,
+      "eval_samples_per_second": 232.729,
+      "eval_steps_per_second": 8.105,
+      "step": 152520
+    },
+    {
+      "epoch": 984.06,
+      "learning_rate": 8.853046594982079e-07,
+      "loss": 0.2182,
+      "step": 152530
+    },
+    {
+      "epoch": 984.13,
+      "learning_rate": 8.81720430107527e-07,
+      "loss": 0.2796,
+      "step": 152540
+    },
+    {
+      "epoch": 984.19,
+      "learning_rate": 8.78136200716846e-07,
+      "loss": 0.1993,
+      "step": 152550
+    },
+    {
+      "epoch": 984.26,
+      "learning_rate": 8.74551971326165e-07,
+      "loss": 0.2426,
+      "step": 152560
+    },
+    {
+      "epoch": 984.32,
+      "learning_rate": 8.709677419354838e-07,
+      "loss": 0.1688,
+      "step": 152570
+    },
+    {
+      "epoch": 984.39,
+      "learning_rate": 8.673835125448029e-07,
+      "loss": 0.1722,
+      "step": 152580
+    },
+    {
+      "epoch": 984.45,
+      "learning_rate": 8.637992831541219e-07,
+      "loss": 0.2287,
+      "step": 152590
+    },
+    {
+      "epoch": 984.51,
+      "learning_rate": 8.602150537634409e-07,
+      "loss": 0.1968,
+      "step": 152600
+    },
+    {
+      "epoch": 984.58,
+      "learning_rate": 8.566308243727599e-07,
+      "loss": 0.1978,
+      "step": 152610
+    },
+    {
+      "epoch": 984.64,
+      "learning_rate": 8.53046594982079e-07,
+      "loss": 0.1861,
+      "step": 152620
+    },
+    {
+      "epoch": 984.71,
+      "learning_rate": 8.494623655913979e-07,
+      "loss": 0.2179,
+      "step": 152630
+    },
+    {
+      "epoch": 984.77,
+      "learning_rate": 8.45878136200717e-07,
+      "loss": 0.1696,
+      "step": 152640
+    },
+    {
+      "epoch": 984.84,
+      "learning_rate": 8.422939068100358e-07,
+      "loss": 0.1991,
+      "step": 152650
+    },
+    {
+      "epoch": 984.9,
+      "learning_rate": 8.387096774193549e-07,
+      "loss": 0.2183,
+      "step": 152660
+    },
+    {
+      "epoch": 984.96,
+      "learning_rate": 8.351254480286738e-07,
+      "loss": 0.2089,
+      "step": 152670
+    },
+    {
+      "epoch": 985.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9126525521278381,
+      "eval_runtime": 0.9092,
+      "eval_samples_per_second": 221.07,
+      "eval_steps_per_second": 7.699,
+      "step": 152675
+    },
+    {
+      "epoch": 985.03,
+      "learning_rate": 8.315412186379929e-07,
+      "loss": 0.2107,
+      "step": 152680
+    },
+    {
+      "epoch": 985.1,
+      "learning_rate": 8.279569892473119e-07,
+      "loss": 0.2075,
+      "step": 152690
+    },
+    {
+      "epoch": 985.16,
+      "learning_rate": 8.243727598566309e-07,
+      "loss": 0.1852,
+      "step": 152700
+    },
+    {
+      "epoch": 985.23,
+      "learning_rate": 8.207885304659499e-07,
+      "loss": 0.2188,
+      "step": 152710
+    },
+    {
+      "epoch": 985.29,
+      "learning_rate": 8.17204301075269e-07,
+      "loss": 0.1856,
+      "step": 152720
+    },
+    {
+      "epoch": 985.35,
+      "learning_rate": 8.136200716845878e-07,
+      "loss": 0.1811,
+      "step": 152730
+    },
+    {
+      "epoch": 985.42,
+      "learning_rate": 8.100358422939069e-07,
+      "loss": 0.2065,
+      "step": 152740
+    },
+    {
+      "epoch": 985.48,
+      "learning_rate": 8.064516129032258e-07,
+      "loss": 0.228,
+      "step": 152750
+    },
+    {
+      "epoch": 985.55,
+      "learning_rate": 8.028673835125448e-07,
+      "loss": 0.2457,
+      "step": 152760
+    },
+    {
+      "epoch": 985.61,
+      "learning_rate": 7.992831541218639e-07,
+      "loss": 0.2543,
+      "step": 152770
+    },
+    {
+      "epoch": 985.68,
+      "learning_rate": 7.956989247311828e-07,
+      "loss": 0.1796,
+      "step": 152780
+    },
+    {
+      "epoch": 985.74,
+      "learning_rate": 7.921146953405019e-07,
+      "loss": 0.2381,
+      "step": 152790
+    },
+    {
+      "epoch": 985.8,
+      "learning_rate": 7.88530465949821e-07,
+      "loss": 0.2302,
+      "step": 152800
+    },
+    {
+      "epoch": 985.87,
+      "learning_rate": 7.849462365591398e-07,
+      "loss": 0.1609,
+      "step": 152810
+    },
+    {
+      "epoch": 985.93,
+      "learning_rate": 7.813620071684587e-07,
+      "loss": 0.2352,
+      "step": 152820
+    },
+    {
+      "epoch": 986.0,
+      "learning_rate": 7.777777777777778e-07,
+      "loss": 0.1903,
+      "step": 152830
+    },
+    {
+      "epoch": 986.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.919438898563385,
+      "eval_runtime": 0.8851,
+      "eval_samples_per_second": 227.086,
+      "eval_steps_per_second": 7.908,
+      "step": 152830
+    },
+    {
+      "epoch": 986.06,
+      "learning_rate": 7.741935483870968e-07,
+      "loss": 0.2372,
+      "step": 152840
+    },
+    {
+      "epoch": 986.13,
+      "learning_rate": 7.706093189964158e-07,
+      "loss": 0.1842,
+      "step": 152850
+    },
+    {
+      "epoch": 986.19,
+      "learning_rate": 7.670250896057348e-07,
+      "loss": 0.179,
+      "step": 152860
+    },
+    {
+      "epoch": 986.26,
+      "learning_rate": 7.634408602150538e-07,
+      "loss": 0.1802,
+      "step": 152870
+    },
+    {
+      "epoch": 986.32,
+      "learning_rate": 7.598566308243728e-07,
+      "loss": 0.1649,
+      "step": 152880
+    },
+    {
+      "epoch": 986.39,
+      "learning_rate": 7.562724014336918e-07,
+      "loss": 0.2398,
+      "step": 152890
+    },
+    {
+      "epoch": 986.45,
+      "learning_rate": 7.526881720430108e-07,
+      "loss": 0.2158,
+      "step": 152900
+    },
+    {
+      "epoch": 986.51,
+      "learning_rate": 7.491039426523298e-07,
+      "loss": 0.2019,
+      "step": 152910
+    },
+    {
+      "epoch": 986.58,
+      "learning_rate": 7.455197132616487e-07,
+      "loss": 0.177,
+      "step": 152920
+    },
+    {
+      "epoch": 986.64,
+      "learning_rate": 7.419354838709678e-07,
+      "loss": 0.2284,
+      "step": 152930
+    },
+    {
+      "epoch": 986.71,
+      "learning_rate": 7.383512544802868e-07,
+      "loss": 0.1687,
+      "step": 152940
+    },
+    {
+      "epoch": 986.77,
+      "learning_rate": 7.347670250896058e-07,
+      "loss": 0.1837,
+      "step": 152950
+    },
+    {
+      "epoch": 986.84,
+      "learning_rate": 7.311827956989247e-07,
+      "loss": 0.1686,
+      "step": 152960
+    },
+    {
+      "epoch": 986.9,
+      "learning_rate": 7.275985663082438e-07,
+      "loss": 0.1938,
+      "step": 152970
+    },
+    {
+      "epoch": 986.96,
+      "learning_rate": 7.240143369175628e-07,
+      "loss": 0.2146,
+      "step": 152980
+    },
+    {
+      "epoch": 987.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9454371333122253,
+      "eval_runtime": 0.8851,
+      "eval_samples_per_second": 227.092,
+      "eval_steps_per_second": 7.909,
+      "step": 152985
+    },
+    {
+      "epoch": 987.03,
+      "learning_rate": 7.204301075268818e-07,
+      "loss": 0.2093,
+      "step": 152990
+    },
+    {
+      "epoch": 987.1,
+      "learning_rate": 7.168458781362007e-07,
+      "loss": 0.1888,
+      "step": 153000
+    },
+    {
+      "epoch": 987.16,
+      "learning_rate": 7.132616487455198e-07,
+      "loss": 0.1923,
+      "step": 153010
+    },
+    {
+      "epoch": 987.23,
+      "learning_rate": 7.096774193548388e-07,
+      "loss": 0.2077,
+      "step": 153020
+    },
+    {
+      "epoch": 987.29,
+      "learning_rate": 7.060931899641578e-07,
+      "loss": 0.19,
+      "step": 153030
+    },
+    {
+      "epoch": 987.35,
+      "learning_rate": 7.025089605734767e-07,
+      "loss": 0.1797,
+      "step": 153040
+    },
+    {
+      "epoch": 987.42,
+      "learning_rate": 6.989247311827958e-07,
+      "loss": 0.2004,
+      "step": 153050
+    },
+    {
+      "epoch": 987.48,
+      "learning_rate": 6.953405017921148e-07,
+      "loss": 0.1933,
+      "step": 153060
+    },
+    {
+      "epoch": 987.55,
+      "learning_rate": 6.917562724014337e-07,
+      "loss": 0.1699,
+      "step": 153070
+    },
+    {
+      "epoch": 987.61,
+      "learning_rate": 6.881720430107527e-07,
+      "loss": 0.1541,
+      "step": 153080
+    },
+    {
+      "epoch": 987.68,
+      "learning_rate": 6.845878136200718e-07,
+      "loss": 0.1959,
+      "step": 153090
+    },
+    {
+      "epoch": 987.74,
+      "learning_rate": 6.810035842293907e-07,
+      "loss": 0.1315,
+      "step": 153100
+    },
+    {
+      "epoch": 987.8,
+      "learning_rate": 6.774193548387097e-07,
+      "loss": 0.1574,
+      "step": 153110
+    },
+    {
+      "epoch": 987.87,
+      "learning_rate": 6.738351254480287e-07,
+      "loss": 0.1885,
+      "step": 153120
+    },
+    {
+      "epoch": 987.93,
+      "learning_rate": 6.702508960573478e-07,
+      "loss": 0.1791,
+      "step": 153130
+    },
+    {
+      "epoch": 988.0,
+      "learning_rate": 6.666666666666667e-07,
+      "loss": 0.1716,
+      "step": 153140
+    },
+    {
+      "epoch": 988.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9619142413139343,
+      "eval_runtime": 0.8687,
+      "eval_samples_per_second": 231.387,
+      "eval_steps_per_second": 8.058,
+      "step": 153140
+    },
+    {
+      "epoch": 988.06,
+      "learning_rate": 6.630824372759857e-07,
+      "loss": 0.2036,
+      "step": 153150
+    },
+    {
+      "epoch": 988.13,
+      "learning_rate": 6.594982078853047e-07,
+      "loss": 0.1715,
+      "step": 153160
+    },
+    {
+      "epoch": 988.19,
+      "learning_rate": 6.559139784946238e-07,
+      "loss": 0.1588,
+      "step": 153170
+    },
+    {
+      "epoch": 988.26,
+      "learning_rate": 6.523297491039427e-07,
+      "loss": 0.1788,
+      "step": 153180
+    },
+    {
+      "epoch": 988.32,
+      "learning_rate": 6.487455197132616e-07,
+      "loss": 0.2115,
+      "step": 153190
+    },
+    {
+      "epoch": 988.39,
+      "learning_rate": 6.451612903225807e-07,
+      "loss": 0.1813,
+      "step": 153200
+    },
+    {
+      "epoch": 988.45,
+      "learning_rate": 6.415770609318996e-07,
+      "loss": 0.1602,
+      "step": 153210
+    },
+    {
+      "epoch": 988.51,
+      "learning_rate": 6.379928315412187e-07,
+      "loss": 0.1606,
+      "step": 153220
+    },
+    {
+      "epoch": 988.58,
+      "learning_rate": 6.344086021505376e-07,
+      "loss": 0.1801,
+      "step": 153230
+    },
+    {
+      "epoch": 988.64,
+      "learning_rate": 6.308243727598567e-07,
+      "loss": 0.1759,
+      "step": 153240
+    },
+    {
+      "epoch": 988.71,
+      "learning_rate": 6.272401433691756e-07,
+      "loss": 0.2332,
+      "step": 153250
+    },
+    {
+      "epoch": 988.77,
+      "learning_rate": 6.236559139784947e-07,
+      "loss": 0.173,
+      "step": 153260
+    },
+    {
+      "epoch": 988.84,
+      "learning_rate": 6.200716845878136e-07,
+      "loss": 0.1651,
+      "step": 153270
+    },
+    {
+      "epoch": 988.9,
+      "learning_rate": 6.164874551971327e-07,
+      "loss": 0.1887,
+      "step": 153280
+    },
+    {
+      "epoch": 988.96,
+      "learning_rate": 6.129032258064516e-07,
+      "loss": 0.1798,
+      "step": 153290
+    },
+    {
+      "epoch": 989.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9590880274772644,
+      "eval_runtime": 0.9342,
+      "eval_samples_per_second": 215.167,
+      "eval_steps_per_second": 7.493,
+      "step": 153295
+    },
+    {
+      "epoch": 989.03,
+      "learning_rate": 6.093189964157707e-07,
+      "loss": 0.2148,
+      "step": 153300
+    },
+    {
+      "epoch": 989.1,
+      "learning_rate": 6.057347670250896e-07,
+      "loss": 0.1849,
+      "step": 153310
+    },
+    {
+      "epoch": 989.16,
+      "learning_rate": 6.021505376344086e-07,
+      "loss": 0.1686,
+      "step": 153320
+    },
+    {
+      "epoch": 989.23,
+      "learning_rate": 5.985663082437276e-07,
+      "loss": 0.1478,
+      "step": 153330
+    },
+    {
+      "epoch": 989.29,
+      "learning_rate": 5.949820788530467e-07,
+      "loss": 0.1702,
+      "step": 153340
+    },
+    {
+      "epoch": 989.35,
+      "learning_rate": 5.913978494623656e-07,
+      "loss": 0.1884,
+      "step": 153350
+    },
+    {
+      "epoch": 989.42,
+      "learning_rate": 5.878136200716846e-07,
+      "loss": 0.1723,
+      "step": 153360
+    },
+    {
+      "epoch": 989.48,
+      "learning_rate": 5.842293906810036e-07,
+      "loss": 0.2001,
+      "step": 153370
+    },
+    {
+      "epoch": 989.55,
+      "learning_rate": 5.806451612903227e-07,
+      "loss": 0.1431,
+      "step": 153380
+    },
+    {
+      "epoch": 989.61,
+      "learning_rate": 5.770609318996416e-07,
+      "loss": 0.1908,
+      "step": 153390
+    },
+    {
+      "epoch": 989.68,
+      "learning_rate": 5.734767025089606e-07,
+      "loss": 0.1561,
+      "step": 153400
+    },
+    {
+      "epoch": 989.74,
+      "learning_rate": 5.698924731182796e-07,
+      "loss": 0.2116,
+      "step": 153410
+    },
+    {
+      "epoch": 989.8,
+      "learning_rate": 5.663082437275987e-07,
+      "loss": 0.2103,
+      "step": 153420
+    },
+    {
+      "epoch": 989.87,
+      "learning_rate": 5.627240143369175e-07,
+      "loss": 0.2066,
+      "step": 153430
+    },
+    {
+      "epoch": 989.93,
+      "learning_rate": 5.591397849462366e-07,
+      "loss": 0.212,
+      "step": 153440
+    },
+    {
+      "epoch": 990.0,
+      "learning_rate": 5.555555555555556e-07,
+      "loss": 0.1998,
+      "step": 153450
+    },
+    {
+      "epoch": 990.0,
+      "eval_accuracy": 0.7810945273631841,
+      "eval_loss": 0.9308421611785889,
+      "eval_runtime": 0.8673,
+      "eval_samples_per_second": 231.741,
+      "eval_steps_per_second": 8.071,
+      "step": 153450
+    },
+    {
+      "epoch": 990.06,
+      "learning_rate": 5.519713261648746e-07,
+      "loss": 0.2205,
+      "step": 153460
+    },
+    {
+      "epoch": 990.13,
+      "learning_rate": 5.483870967741935e-07,
+      "loss": 0.2052,
+      "step": 153470
+    },
+    {
+      "epoch": 990.19,
+      "learning_rate": 5.448028673835126e-07,
+      "loss": 0.1972,
+      "step": 153480
+    },
+    {
+      "epoch": 990.26,
+      "learning_rate": 5.412186379928316e-07,
+      "loss": 0.2134,
+      "step": 153490
+    },
+    {
+      "epoch": 990.32,
+      "learning_rate": 5.376344086021506e-07,
+      "loss": 0.172,
+      "step": 153500
+    },
+    {
+      "epoch": 990.39,
+      "learning_rate": 5.340501792114695e-07,
+      "loss": 0.1803,
+      "step": 153510
+    },
+    {
+      "epoch": 990.45,
+      "learning_rate": 5.304659498207886e-07,
+      "loss": 0.2623,
+      "step": 153520
+    },
+    {
+      "epoch": 990.51,
+      "learning_rate": 5.268817204301076e-07,
+      "loss": 0.2215,
+      "step": 153530
+    },
+    {
+      "epoch": 990.58,
+      "learning_rate": 5.232974910394266e-07,
+      "loss": 0.1661,
+      "step": 153540
+    },
+    {
+      "epoch": 990.64,
+      "learning_rate": 5.197132616487455e-07,
+      "loss": 0.2057,
+      "step": 153550
+    },
+    {
+      "epoch": 990.71,
+      "learning_rate": 5.161290322580646e-07,
+      "loss": 0.184,
+      "step": 153560
+    },
+    {
+      "epoch": 990.77,
+      "learning_rate": 5.125448028673835e-07,
+      "loss": 0.189,
+      "step": 153570
+    },
+    {
+      "epoch": 990.84,
+      "learning_rate": 5.089605734767026e-07,
+      "loss": 0.1732,
+      "step": 153580
+    },
+    {
+      "epoch": 990.9,
+      "learning_rate": 5.053763440860215e-07,
+      "loss": 0.2024,
+      "step": 153590
+    },
+    {
+      "epoch": 990.96,
+      "learning_rate": 5.017921146953406e-07,
+      "loss": 0.1873,
+      "step": 153600
+    },
+    {
+      "epoch": 991.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.914587676525116,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 231.244,
+      "eval_steps_per_second": 8.053,
+      "step": 153605
+    },
+    {
+      "epoch": 991.03,
+      "learning_rate": 4.982078853046595e-07,
+      "loss": 0.2218,
+      "step": 153610
+    },
+    {
+      "epoch": 991.1,
+      "learning_rate": 4.946236559139786e-07,
+      "loss": 0.1845,
+      "step": 153620
+    },
+    {
+      "epoch": 991.16,
+      "learning_rate": 4.910394265232975e-07,
+      "loss": 0.1666,
+      "step": 153630
+    },
+    {
+      "epoch": 991.23,
+      "learning_rate": 4.874551971326166e-07,
+      "loss": 0.1833,
+      "step": 153640
+    },
+    {
+      "epoch": 991.29,
+      "learning_rate": 4.838709677419355e-07,
+      "loss": 0.2077,
+      "step": 153650
+    },
+    {
+      "epoch": 991.35,
+      "learning_rate": 4.802867383512546e-07,
+      "loss": 0.1647,
+      "step": 153660
+    },
+    {
+      "epoch": 991.42,
+      "learning_rate": 4.767025089605735e-07,
+      "loss": 0.232,
+      "step": 153670
+    },
+    {
+      "epoch": 991.48,
+      "learning_rate": 4.731182795698925e-07,
+      "loss": 0.2014,
+      "step": 153680
+    },
+    {
+      "epoch": 991.55,
+      "learning_rate": 4.695340501792115e-07,
+      "loss": 0.1795,
+      "step": 153690
+    },
+    {
+      "epoch": 991.61,
+      "learning_rate": 4.6594982078853055e-07,
+      "loss": 0.2068,
+      "step": 153700
+    },
+    {
+      "epoch": 991.68,
+      "learning_rate": 4.6236559139784944e-07,
+      "loss": 0.2329,
+      "step": 153710
+    },
+    {
+      "epoch": 991.74,
+      "learning_rate": 4.587813620071685e-07,
+      "loss": 0.1858,
+      "step": 153720
+    },
+    {
+      "epoch": 991.8,
+      "learning_rate": 4.551971326164875e-07,
+      "loss": 0.2084,
+      "step": 153730
+    },
+    {
+      "epoch": 991.87,
+      "learning_rate": 4.516129032258065e-07,
+      "loss": 0.2136,
+      "step": 153740
+    },
+    {
+      "epoch": 991.93,
+      "learning_rate": 4.4802867383512544e-07,
+      "loss": 0.2114,
+      "step": 153750
+    },
+    {
+      "epoch": 992.0,
+      "learning_rate": 4.444444444444445e-07,
+      "loss": 0.1557,
+      "step": 153760
+    },
+    {
+      "epoch": 992.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9523319005966187,
+      "eval_runtime": 0.8679,
+      "eval_samples_per_second": 231.592,
+      "eval_steps_per_second": 8.065,
+      "step": 153760
+    },
+    {
+      "epoch": 992.06,
+      "learning_rate": 4.408602150537635e-07,
+      "loss": 0.1914,
+      "step": 153770
+    },
+    {
+      "epoch": 992.13,
+      "learning_rate": 4.372759856630825e-07,
+      "loss": 0.1714,
+      "step": 153780
+    },
+    {
+      "epoch": 992.19,
+      "learning_rate": 4.3369175627240143e-07,
+      "loss": 0.1603,
+      "step": 153790
+    },
+    {
+      "epoch": 992.26,
+      "learning_rate": 4.3010752688172043e-07,
+      "loss": 0.2097,
+      "step": 153800
+    },
+    {
+      "epoch": 992.32,
+      "learning_rate": 4.265232974910395e-07,
+      "loss": 0.2116,
+      "step": 153810
+    },
+    {
+      "epoch": 992.39,
+      "learning_rate": 4.229390681003585e-07,
+      "loss": 0.2042,
+      "step": 153820
+    },
+    {
+      "epoch": 992.45,
+      "learning_rate": 4.1935483870967743e-07,
+      "loss": 0.2198,
+      "step": 153830
+    },
+    {
+      "epoch": 992.51,
+      "learning_rate": 4.1577060931899643e-07,
+      "loss": 0.191,
+      "step": 153840
+    },
+    {
+      "epoch": 992.58,
+      "learning_rate": 4.1218637992831543e-07,
+      "loss": 0.226,
+      "step": 153850
+    },
+    {
+      "epoch": 992.64,
+      "learning_rate": 4.086021505376345e-07,
+      "loss": 0.2019,
+      "step": 153860
+    },
+    {
+      "epoch": 992.71,
+      "learning_rate": 4.050179211469534e-07,
+      "loss": 0.2024,
+      "step": 153870
+    },
+    {
+      "epoch": 992.77,
+      "learning_rate": 4.014336917562724e-07,
+      "loss": 0.1711,
+      "step": 153880
+    },
+    {
+      "epoch": 992.84,
+      "learning_rate": 3.978494623655914e-07,
+      "loss": 0.2001,
+      "step": 153890
+    },
+    {
+      "epoch": 992.9,
+      "learning_rate": 3.942652329749105e-07,
+      "loss": 0.1873,
+      "step": 153900
+    },
+    {
+      "epoch": 992.96,
+      "learning_rate": 3.9068100358422937e-07,
+      "loss": 0.1635,
+      "step": 153910
+    },
+    {
+      "epoch": 993.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9340052008628845,
+      "eval_runtime": 0.8988,
+      "eval_samples_per_second": 223.633,
+      "eval_steps_per_second": 7.788,
+      "step": 153915
+    },
+    {
+      "epoch": 993.03,
+      "learning_rate": 3.870967741935484e-07,
+      "loss": 0.1986,
+      "step": 153920
+    },
+    {
+      "epoch": 993.1,
+      "learning_rate": 3.835125448028674e-07,
+      "loss": 0.1695,
+      "step": 153930
+    },
+    {
+      "epoch": 993.16,
+      "learning_rate": 3.799283154121864e-07,
+      "loss": 0.1626,
+      "step": 153940
+    },
+    {
+      "epoch": 993.23,
+      "learning_rate": 3.763440860215054e-07,
+      "loss": 0.2231,
+      "step": 153950
+    },
+    {
+      "epoch": 993.29,
+      "learning_rate": 3.7275985663082436e-07,
+      "loss": 0.2072,
+      "step": 153960
+    },
+    {
+      "epoch": 993.35,
+      "learning_rate": 3.691756272401434e-07,
+      "loss": 0.2023,
+      "step": 153970
+    },
+    {
+      "epoch": 993.42,
+      "learning_rate": 3.6559139784946236e-07,
+      "loss": 0.2445,
+      "step": 153980
+    },
+    {
+      "epoch": 993.48,
+      "learning_rate": 3.620071684587814e-07,
+      "loss": 0.205,
+      "step": 153990
+    },
+    {
+      "epoch": 993.55,
+      "learning_rate": 3.5842293906810036e-07,
+      "loss": 0.2046,
+      "step": 154000
+    },
+    {
+      "epoch": 993.61,
+      "learning_rate": 3.548387096774194e-07,
+      "loss": 0.1743,
+      "step": 154010
+    },
+    {
+      "epoch": 993.68,
+      "learning_rate": 3.5125448028673836e-07,
+      "loss": 0.1723,
+      "step": 154020
+    },
+    {
+      "epoch": 993.74,
+      "learning_rate": 3.476702508960574e-07,
+      "loss": 0.2189,
+      "step": 154030
+    },
+    {
+      "epoch": 993.8,
+      "learning_rate": 3.4408602150537636e-07,
+      "loss": 0.1769,
+      "step": 154040
+    },
+    {
+      "epoch": 993.87,
+      "learning_rate": 3.4050179211469536e-07,
+      "loss": 0.1782,
+      "step": 154050
+    },
+    {
+      "epoch": 993.93,
+      "learning_rate": 3.3691756272401435e-07,
+      "loss": 0.1887,
+      "step": 154060
+    },
+    {
+      "epoch": 994.0,
+      "learning_rate": 3.3333333333333335e-07,
+      "loss": 0.1817,
+      "step": 154070
+    },
+    {
+      "epoch": 994.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9365586638450623,
+      "eval_runtime": 0.901,
+      "eval_samples_per_second": 223.077,
+      "eval_steps_per_second": 7.769,
+      "step": 154070
+    },
+    {
+      "epoch": 994.06,
+      "learning_rate": 3.2974910394265235e-07,
+      "loss": 0.2558,
+      "step": 154080
+    },
+    {
+      "epoch": 994.13,
+      "learning_rate": 3.2616487455197135e-07,
+      "loss": 0.149,
+      "step": 154090
+    },
+    {
+      "epoch": 994.19,
+      "learning_rate": 3.2258064516129035e-07,
+      "loss": 0.2281,
+      "step": 154100
+    },
+    {
+      "epoch": 994.26,
+      "learning_rate": 3.1899641577060935e-07,
+      "loss": 0.1331,
+      "step": 154110
+    },
+    {
+      "epoch": 994.32,
+      "learning_rate": 3.1541218637992835e-07,
+      "loss": 0.2259,
+      "step": 154120
+    },
+    {
+      "epoch": 994.39,
+      "learning_rate": 3.1182795698924735e-07,
+      "loss": 0.1865,
+      "step": 154130
+    },
+    {
+      "epoch": 994.45,
+      "learning_rate": 3.0824372759856635e-07,
+      "loss": 0.1947,
+      "step": 154140
+    },
+    {
+      "epoch": 994.51,
+      "learning_rate": 3.0465949820788535e-07,
+      "loss": 0.1912,
+      "step": 154150
+    },
+    {
+      "epoch": 994.58,
+      "learning_rate": 3.010752688172043e-07,
+      "loss": 0.2343,
+      "step": 154160
+    },
+    {
+      "epoch": 994.64,
+      "learning_rate": 2.9749103942652334e-07,
+      "loss": 0.1921,
+      "step": 154170
+    },
+    {
+      "epoch": 994.71,
+      "learning_rate": 2.939068100358423e-07,
+      "loss": 0.1827,
+      "step": 154180
+    },
+    {
+      "epoch": 994.77,
+      "learning_rate": 2.9032258064516134e-07,
+      "loss": 0.2021,
+      "step": 154190
+    },
+    {
+      "epoch": 994.84,
+      "learning_rate": 2.867383512544803e-07,
+      "loss": 0.1676,
+      "step": 154200
+    },
+    {
+      "epoch": 994.9,
+      "learning_rate": 2.8315412186379934e-07,
+      "loss": 0.1817,
+      "step": 154210
+    },
+    {
+      "epoch": 994.96,
+      "learning_rate": 2.795698924731183e-07,
+      "loss": 0.1747,
+      "step": 154220
+    },
+    {
+      "epoch": 995.0,
+      "eval_accuracy": 0.7611940298507462,
+      "eval_loss": 0.9430158734321594,
+      "eval_runtime": 0.8889,
+      "eval_samples_per_second": 226.114,
+      "eval_steps_per_second": 7.875,
+      "step": 154225
+    },
+    {
+      "epoch": 995.03,
+      "learning_rate": 2.759856630824373e-07,
+      "loss": 0.1804,
+      "step": 154230
+    },
+    {
+      "epoch": 995.1,
+      "learning_rate": 2.724014336917563e-07,
+      "loss": 0.185,
+      "step": 154240
+    },
+    {
+      "epoch": 995.16,
+      "learning_rate": 2.688172043010753e-07,
+      "loss": 0.2049,
+      "step": 154250
+    },
+    {
+      "epoch": 995.23,
+      "learning_rate": 2.652329749103943e-07,
+      "loss": 0.1933,
+      "step": 154260
+    },
+    {
+      "epoch": 995.29,
+      "learning_rate": 2.616487455197133e-07,
+      "loss": 0.1946,
+      "step": 154270
+    },
+    {
+      "epoch": 995.35,
+      "learning_rate": 2.580645161290323e-07,
+      "loss": 0.1973,
+      "step": 154280
+    },
+    {
+      "epoch": 995.42,
+      "learning_rate": 2.544802867383513e-07,
+      "loss": 0.2302,
+      "step": 154290
+    },
+    {
+      "epoch": 995.48,
+      "learning_rate": 2.508960573476703e-07,
+      "loss": 0.1624,
+      "step": 154300
+    },
+    {
+      "epoch": 995.55,
+      "learning_rate": 2.473118279569893e-07,
+      "loss": 0.2013,
+      "step": 154310
+    },
+    {
+      "epoch": 995.61,
+      "learning_rate": 2.437275985663083e-07,
+      "loss": 0.2087,
+      "step": 154320
+    },
+    {
+      "epoch": 995.68,
+      "learning_rate": 2.401433691756273e-07,
+      "loss": 0.2225,
+      "step": 154330
+    },
+    {
+      "epoch": 995.74,
+      "learning_rate": 2.3655913978494625e-07,
+      "loss": 0.1864,
+      "step": 154340
+    },
+    {
+      "epoch": 995.8,
+      "learning_rate": 2.3297491039426527e-07,
+      "loss": 0.2121,
+      "step": 154350
+    },
+    {
+      "epoch": 995.87,
+      "learning_rate": 2.2939068100358425e-07,
+      "loss": 0.2109,
+      "step": 154360
+    },
+    {
+      "epoch": 995.93,
+      "learning_rate": 2.2580645161290325e-07,
+      "loss": 0.2029,
+      "step": 154370
+    },
+    {
+      "epoch": 996.0,
+      "learning_rate": 2.2222222222222224e-07,
+      "loss": 0.1956,
+      "step": 154380
+    },
+    {
+      "epoch": 996.0,
+      "eval_accuracy": 0.7661691542288557,
+      "eval_loss": 0.9223435521125793,
+      "eval_runtime": 0.9479,
+      "eval_samples_per_second": 212.046,
+      "eval_steps_per_second": 7.385,
+      "step": 154380
+    },
+    {
+      "epoch": 996.06,
+      "learning_rate": 2.1863799283154124e-07,
+      "loss": 0.2023,
+      "step": 154390
+    },
+    {
+      "epoch": 996.13,
+      "learning_rate": 2.1505376344086022e-07,
+      "loss": 0.1778,
+      "step": 154400
+    },
+    {
+      "epoch": 996.19,
+      "learning_rate": 2.1146953405017924e-07,
+      "loss": 0.2002,
+      "step": 154410
+    },
+    {
+      "epoch": 996.26,
+      "learning_rate": 2.0788530465949821e-07,
+      "loss": 0.1511,
+      "step": 154420
+    },
+    {
+      "epoch": 996.32,
+      "learning_rate": 2.0430107526881724e-07,
+      "loss": 0.1844,
+      "step": 154430
+    },
+    {
+      "epoch": 996.39,
+      "learning_rate": 2.007168458781362e-07,
+      "loss": 0.1482,
+      "step": 154440
+    },
+    {
+      "epoch": 996.45,
+      "learning_rate": 1.9713261648745524e-07,
+      "loss": 0.2039,
+      "step": 154450
+    },
+    {
+      "epoch": 996.51,
+      "learning_rate": 1.935483870967742e-07,
+      "loss": 0.212,
+      "step": 154460
+    },
+    {
+      "epoch": 996.58,
+      "learning_rate": 1.899641577060932e-07,
+      "loss": 0.1782,
+      "step": 154470
+    },
+    {
+      "epoch": 996.64,
+      "learning_rate": 1.8637992831541218e-07,
+      "loss": 0.2069,
+      "step": 154480
+    },
+    {
+      "epoch": 996.71,
+      "learning_rate": 1.8279569892473118e-07,
+      "loss": 0.2713,
+      "step": 154490
+    },
+    {
+      "epoch": 996.77,
+      "learning_rate": 1.7921146953405018e-07,
+      "loss": 0.2269,
+      "step": 154500
+    },
+    {
+      "epoch": 996.84,
+      "learning_rate": 1.7562724014336918e-07,
+      "loss": 0.205,
+      "step": 154510
+    },
+    {
+      "epoch": 996.9,
+      "learning_rate": 1.7204301075268818e-07,
+      "loss": 0.2193,
+      "step": 154520
+    },
+    {
+      "epoch": 996.96,
+      "learning_rate": 1.6845878136200718e-07,
+      "loss": 0.2481,
+      "step": 154530
+    },
+    {
+      "epoch": 997.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9252193570137024,
+      "eval_runtime": 0.8671,
+      "eval_samples_per_second": 231.819,
+      "eval_steps_per_second": 8.073,
+      "step": 154535
+    },
+    {
+      "epoch": 997.03,
+      "learning_rate": 1.6487455197132618e-07,
+      "loss": 0.2173,
+      "step": 154540
+    },
+    {
+      "epoch": 997.1,
+      "learning_rate": 1.6129032258064518e-07,
+      "loss": 0.2005,
+      "step": 154550
+    },
+    {
+      "epoch": 997.16,
+      "learning_rate": 1.5770609318996417e-07,
+      "loss": 0.2362,
+      "step": 154560
+    },
+    {
+      "epoch": 997.23,
+      "learning_rate": 1.5412186379928317e-07,
+      "loss": 0.1712,
+      "step": 154570
+    },
+    {
+      "epoch": 997.29,
+      "learning_rate": 1.5053763440860215e-07,
+      "loss": 0.1791,
+      "step": 154580
+    },
+    {
+      "epoch": 997.35,
+      "learning_rate": 1.4695340501792115e-07,
+      "loss": 0.2222,
+      "step": 154590
+    },
+    {
+      "epoch": 997.42,
+      "learning_rate": 1.4336917562724014e-07,
+      "loss": 0.2059,
+      "step": 154600
+    },
+    {
+      "epoch": 997.48,
+      "learning_rate": 1.3978494623655914e-07,
+      "loss": 0.2226,
+      "step": 154610
+    },
+    {
+      "epoch": 997.55,
+      "learning_rate": 1.3620071684587814e-07,
+      "loss": 0.1757,
+      "step": 154620
+    },
+    {
+      "epoch": 997.61,
+      "learning_rate": 1.3261648745519714e-07,
+      "loss": 0.1671,
+      "step": 154630
+    },
+    {
+      "epoch": 997.68,
+      "learning_rate": 1.2903225806451614e-07,
+      "loss": 0.2029,
+      "step": 154640
+    },
+    {
+      "epoch": 997.74,
+      "learning_rate": 1.2544802867383514e-07,
+      "loss": 0.1822,
+      "step": 154650
+    },
+    {
+      "epoch": 997.8,
+      "learning_rate": 1.2186379928315414e-07,
+      "loss": 0.1843,
+      "step": 154660
+    },
+    {
+      "epoch": 997.87,
+      "learning_rate": 1.1827956989247312e-07,
+      "loss": 0.217,
+      "step": 154670
+    },
+    {
+      "epoch": 997.93,
+      "learning_rate": 1.1469534050179212e-07,
+      "loss": 0.2203,
+      "step": 154680
+    },
+    {
+      "epoch": 998.0,
+      "learning_rate": 1.1111111111111112e-07,
+      "loss": 0.178,
+      "step": 154690
+    },
+    {
+      "epoch": 998.0,
+      "eval_accuracy": 0.7761194029850746,
+      "eval_loss": 0.9372466206550598,
+      "eval_runtime": 0.8545,
+      "eval_samples_per_second": 235.232,
+      "eval_steps_per_second": 8.192,
+      "step": 154690
+    },
+    {
+      "epoch": 998.06,
+      "learning_rate": 1.0752688172043011e-07,
+      "loss": 0.2365,
+      "step": 154700
+    },
+    {
+      "epoch": 998.13,
+      "learning_rate": 1.0394265232974911e-07,
+      "loss": 0.2561,
+      "step": 154710
+    },
+    {
+      "epoch": 998.19,
+      "learning_rate": 1.003584229390681e-07,
+      "loss": 0.1499,
+      "step": 154720
+    },
+    {
+      "epoch": 998.26,
+      "learning_rate": 9.67741935483871e-08,
+      "loss": 0.2194,
+      "step": 154730
+    },
+    {
+      "epoch": 998.32,
+      "learning_rate": 9.318996415770609e-08,
+      "loss": 0.2395,
+      "step": 154740
+    },
+    {
+      "epoch": 998.39,
+      "learning_rate": 8.960573476702509e-08,
+      "loss": 0.2081,
+      "step": 154750
+    },
+    {
+      "epoch": 998.45,
+      "learning_rate": 8.602150537634409e-08,
+      "loss": 0.2059,
+      "step": 154760
+    },
+    {
+      "epoch": 998.51,
+      "learning_rate": 8.243727598566309e-08,
+      "loss": 0.1717,
+      "step": 154770
+    },
+    {
+      "epoch": 998.58,
+      "learning_rate": 7.885304659498209e-08,
+      "loss": 0.1999,
+      "step": 154780
+    },
+    {
+      "epoch": 998.64,
+      "learning_rate": 7.526881720430107e-08,
+      "loss": 0.1533,
+      "step": 154790
+    },
+    {
+      "epoch": 998.71,
+      "learning_rate": 7.168458781362007e-08,
+      "loss": 0.2231,
+      "step": 154800
+    },
+    {
+      "epoch": 998.77,
+      "learning_rate": 6.810035842293907e-08,
+      "loss": 0.2162,
+      "step": 154810
+    },
+    {
+      "epoch": 998.84,
+      "learning_rate": 6.451612903225807e-08,
+      "loss": 0.1585,
+      "step": 154820
+    },
+    {
+      "epoch": 998.9,
+      "learning_rate": 6.093189964157707e-08,
+      "loss": 0.1921,
+      "step": 154830
+    },
+    {
+      "epoch": 998.96,
+      "learning_rate": 5.734767025089606e-08,
+      "loss": 0.1913,
+      "step": 154840
+    },
+    {
+      "epoch": 999.0,
+      "eval_accuracy": 0.7512437810945274,
+      "eval_loss": 0.9621124267578125,
+      "eval_runtime": 0.8791,
+      "eval_samples_per_second": 228.646,
+      "eval_steps_per_second": 7.963,
+      "step": 154845
+    },
+    {
+      "epoch": 999.03,
+      "learning_rate": 5.3763440860215054e-08,
+      "loss": 0.2102,
+      "step": 154850
+    },
+    {
+      "epoch": 999.1,
+      "learning_rate": 5.017921146953405e-08,
+      "loss": 0.1814,
+      "step": 154860
+    },
+    {
+      "epoch": 999.16,
+      "learning_rate": 4.6594982078853046e-08,
+      "loss": 0.2166,
+      "step": 154870
+    },
+    {
+      "epoch": 999.23,
+      "learning_rate": 4.3010752688172045e-08,
+      "loss": 0.1674,
+      "step": 154880
+    },
+    {
+      "epoch": 999.29,
+      "learning_rate": 3.9426523297491044e-08,
+      "loss": 0.2365,
+      "step": 154890
+    },
+    {
+      "epoch": 999.35,
+      "learning_rate": 3.5842293906810036e-08,
+      "loss": 0.2005,
+      "step": 154900
+    },
+    {
+      "epoch": 999.42,
+      "learning_rate": 3.2258064516129035e-08,
+      "loss": 0.1999,
+      "step": 154910
+    },
+    {
+      "epoch": 999.48,
+      "learning_rate": 2.867383512544803e-08,
+      "loss": 0.1933,
+      "step": 154920
+    },
+    {
+      "epoch": 999.55,
+      "learning_rate": 2.5089605734767027e-08,
+      "loss": 0.2152,
+      "step": 154930
+    },
+    {
+      "epoch": 999.61,
+      "learning_rate": 2.1505376344086022e-08,
+      "loss": 0.2056,
+      "step": 154940
+    },
+    {
+      "epoch": 999.68,
+      "learning_rate": 1.7921146953405018e-08,
+      "loss": 0.1654,
+      "step": 154950
+    },
+    {
+      "epoch": 999.74,
+      "learning_rate": 1.4336917562724015e-08,
+      "loss": 0.1897,
+      "step": 154960
+    },
+    {
+      "epoch": 999.8,
+      "learning_rate": 1.0752688172043011e-08,
+      "loss": 0.2238,
+      "step": 154970
+    },
+    {
+      "epoch": 999.87,
+      "learning_rate": 7.168458781362008e-09,
+      "loss": 0.1719,
+      "step": 154980
+    },
+    {
+      "epoch": 999.93,
+      "learning_rate": 3.584229390681004e-09,
+      "loss": 0.24,
+      "step": 154990
+    },
+    {
+      "epoch": 1000.0,
+      "learning_rate": 0.0,
+      "loss": 0.2168,
+      "step": 155000
+    },
+    {
+      "epoch": 1000.0,
+      "eval_accuracy": 0.7711442786069652,
+      "eval_loss": 0.9338966608047485,
+      "eval_runtime": 0.8982,
+      "eval_samples_per_second": 223.784,
+      "eval_steps_per_second": 7.793,
+      "step": 155000
+    },
+    {
+      "epoch": 1000.0,
+      "step": 155000,
+      "total_flos": 3.844484891203777e+20,
+      "train_loss": 0.26615277438779034,
+      "train_runtime": 35181.852,
+      "train_samples_per_second": 565.036,
+      "train_steps_per_second": 4.406
+    }
+  ],
+  "max_steps": 155000,
+  "num_train_epochs": 1000,
+  "total_flos": 3.844484891203777e+20,
+  "trial_name": null,
+  "trial_params": null
+}