diff --git "a/checkpoint-5840/trainer_state.json" "b/checkpoint-5840/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-5840/trainer_state.json"
@@ -0,0 +1,4834 @@
+{
+  "best_metric": 0.6604002714157104,
+  "best_model_checkpoint": "./vit-base-beans/checkpoint-1160",
+  "epoch": 14.974358974358974,
+  "global_step": 5840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019965811965811967,
+      "loss": 1.7468,
+      "step": 10
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019931623931623932,
+      "loss": 1.653,
+      "step": 20
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019897435897435898,
+      "loss": 1.5026,
+      "step": 30
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019863247863247864,
+      "loss": 1.4296,
+      "step": 40
+    },
+    {
+      "epoch": 0.1,
+      "eval_accuracy": 0.41975308641975306,
+      "eval_loss": 1.4691457748413086,
+      "eval_runtime": 61.6215,
+      "eval_samples_per_second": 57.902,
+      "eval_steps_per_second": 7.238,
+      "step": 40
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001982905982905983,
+      "loss": 1.369,
+      "step": 50
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019794871794871796,
+      "loss": 1.2948,
+      "step": 60
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001976068376068376,
+      "loss": 1.3074,
+      "step": 70
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019726495726495727,
+      "loss": 1.3074,
+      "step": 80
+    },
+    {
+      "epoch": 0.21,
+      "eval_accuracy": 0.5549943883277216,
+      "eval_loss": 1.1999692916870117,
+      "eval_runtime": 33.0603,
+      "eval_samples_per_second": 107.924,
+      "eval_steps_per_second": 13.491,
+      "step": 80
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019692307692307696,
+      "loss": 1.1725,
+      "step": 90
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019658119658119659,
+      "loss": 1.1944,
+      "step": 100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019623931623931624,
+      "loss": 1.18,
+      "step": 110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001958974358974359,
+      "loss": 1.1731,
+      "step": 120
+    },
+    {
+      "epoch": 0.31,
+      "eval_accuracy": 0.6689113355780022,
+      "eval_loss": 0.978410005569458,
+      "eval_runtime": 37.0531,
+      "eval_samples_per_second": 96.294,
+      "eval_steps_per_second": 12.037,
+      "step": 120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019555555555555556,
+      "loss": 0.9441,
+      "step": 130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019521367521367522,
+      "loss": 1.0459,
+      "step": 140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019487179487179487,
+      "loss": 1.1109,
+      "step": 150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019452991452991453,
+      "loss": 1.0094,
+      "step": 160
+    },
+    {
+      "epoch": 0.41,
+      "eval_accuracy": 0.6616161616161617,
+      "eval_loss": 0.9483916759490967,
+      "eval_runtime": 32.6783,
+      "eval_samples_per_second": 109.185,
+      "eval_steps_per_second": 13.648,
+      "step": 160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941880341880342,
+      "loss": 0.9165,
+      "step": 170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019384615384615385,
+      "loss": 1.0904,
+      "step": 180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001935042735042735,
+      "loss": 0.9846,
+      "step": 190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019316239316239316,
+      "loss": 1.0908,
+      "step": 200
+    },
+    {
+      "epoch": 0.51,
+      "eval_accuracy": 0.5454545454545454,
+      "eval_loss": 1.2091434001922607,
+      "eval_runtime": 32.1465,
+      "eval_samples_per_second": 110.992,
+      "eval_steps_per_second": 13.874,
+      "step": 200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019282051282051282,
+      "loss": 1.0193,
+      "step": 210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001924786324786325,
+      "loss": 0.9809,
+      "step": 220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00019213675213675216,
+      "loss": 0.9673,
+      "step": 230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00019179487179487182,
+      "loss": 0.9402,
+      "step": 240
+    },
+    {
+      "epoch": 0.62,
+      "eval_accuracy": 0.6481481481481481,
+      "eval_loss": 0.9721583724021912,
+      "eval_runtime": 33.9462,
+      "eval_samples_per_second": 105.107,
+      "eval_steps_per_second": 13.138,
+      "step": 240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00019145299145299148,
+      "loss": 0.8103,
+      "step": 250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019111111111111114,
+      "loss": 0.8564,
+      "step": 260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001907692307692308,
+      "loss": 1.1074,
+      "step": 270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019042735042735042,
+      "loss": 1.044,
+      "step": 280
+    },
+    {
+      "epoch": 0.72,
+      "eval_accuracy": 0.5897867564534232,
+      "eval_loss": 1.1147791147232056,
+      "eval_runtime": 32.6846,
+      "eval_samples_per_second": 109.164,
+      "eval_steps_per_second": 13.646,
+      "step": 280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019008547008547008,
+      "loss": 1.016,
+      "step": 290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018974358974358974,
+      "loss": 0.8815,
+      "step": 300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001894017094017094,
+      "loss": 0.8718,
+      "step": 310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018905982905982906,
+      "loss": 0.6886,
+      "step": 320
+    },
+    {
+      "epoch": 0.82,
+      "eval_accuracy": 0.7272727272727273,
+      "eval_loss": 0.7878678441047668,
+      "eval_runtime": 31.1497,
+      "eval_samples_per_second": 114.544,
+      "eval_steps_per_second": 14.318,
+      "step": 320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001887179487179487,
+      "loss": 0.8122,
+      "step": 330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00018837606837606837,
+      "loss": 0.6107,
+      "step": 340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00018803418803418803,
+      "loss": 0.7783,
+      "step": 350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001876923076923077,
+      "loss": 0.8432,
+      "step": 360
+    },
+    {
+      "epoch": 0.92,
+      "eval_accuracy": 0.7480359147025814,
+      "eval_loss": 0.7056980729103088,
+      "eval_runtime": 32.2533,
+      "eval_samples_per_second": 110.624,
+      "eval_steps_per_second": 13.828,
+      "step": 360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00018735042735042737,
+      "loss": 0.8314,
+      "step": 370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00018700854700854703,
+      "loss": 0.8488,
+      "step": 380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8547,
+      "step": 390
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00018632478632478634,
+      "loss": 0.7394,
+      "step": 400
+    },
+    {
+      "epoch": 1.03,
+      "eval_accuracy": 0.6835016835016835,
+      "eval_loss": 0.9139666557312012,
+      "eval_runtime": 42.9913,
+      "eval_samples_per_second": 82.994,
+      "eval_steps_per_second": 10.374,
+      "step": 400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.000185982905982906,
+      "loss": 0.5767,
+      "step": 410
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00018564102564102566,
+      "loss": 0.5304,
+      "step": 420
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00018529914529914532,
+      "loss": 0.6296,
+      "step": 430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00018495726495726497,
+      "loss": 0.6004,
+      "step": 440
+    },
+    {
+      "epoch": 1.13,
+      "eval_accuracy": 0.7255892255892256,
+      "eval_loss": 0.7451665997505188,
+      "eval_runtime": 31.8253,
+      "eval_samples_per_second": 112.112,
+      "eval_steps_per_second": 14.014,
+      "step": 440
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00018461538461538463,
+      "loss": 0.518,
+      "step": 450
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00018427350427350426,
+      "loss": 0.5825,
+      "step": 460
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00018393162393162392,
+      "loss": 0.5465,
+      "step": 470
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00018358974358974358,
+      "loss": 0.5201,
+      "step": 480
+    },
+    {
+      "epoch": 1.23,
+      "eval_accuracy": 0.6879910213243546,
+      "eval_loss": 0.8496339917182922,
+      "eval_runtime": 33.617,
+      "eval_samples_per_second": 106.137,
+      "eval_steps_per_second": 13.267,
+      "step": 480
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00018324786324786324,
+      "loss": 0.5016,
+      "step": 490
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00018290598290598292,
+      "loss": 0.6101,
+      "step": 500
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00018256410256410258,
+      "loss": 0.5572,
+      "step": 510
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00018222222222222224,
+      "loss": 0.4039,
+      "step": 520
+    },
+    {
+      "epoch": 1.33,
+      "eval_accuracy": 0.7312008978675645,
+      "eval_loss": 0.7843908071517944,
+      "eval_runtime": 31.7575,
+      "eval_samples_per_second": 112.351,
+      "eval_steps_per_second": 14.044,
+      "step": 520
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001818803418803419,
+      "loss": 0.4831,
+      "step": 530
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00018153846153846155,
+      "loss": 0.5493,
+      "step": 540
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001811965811965812,
+      "loss": 0.4942,
+      "step": 550
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00018085470085470087,
+      "loss": 0.4475,
+      "step": 560
+    },
+    {
+      "epoch": 1.44,
+      "eval_accuracy": 0.6593714927048261,
+      "eval_loss": 0.9572438597679138,
+      "eval_runtime": 31.2916,
+      "eval_samples_per_second": 114.024,
+      "eval_steps_per_second": 14.253,
+      "step": 560
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00018051282051282052,
+      "loss": 0.4667,
+      "step": 570
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00018017094017094018,
+      "loss": 0.4685,
+      "step": 580
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00017982905982905984,
+      "loss": 0.5028,
+      "step": 590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001794871794871795,
+      "loss": 0.6071,
+      "step": 600
+    },
+    {
+      "epoch": 1.54,
+      "eval_accuracy": 0.7676767676767676,
+      "eval_loss": 0.6666165590286255,
+      "eval_runtime": 32.2041,
+      "eval_samples_per_second": 110.793,
+      "eval_steps_per_second": 13.849,
+      "step": 600
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00017914529914529916,
+      "loss": 0.5134,
+      "step": 610
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001788034188034188,
+      "loss": 0.658,
+      "step": 620
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00017846153846153847,
+      "loss": 0.606,
+      "step": 630
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00017811965811965813,
+      "loss": 0.5004,
+      "step": 640
+    },
+    {
+      "epoch": 1.64,
+      "eval_accuracy": 0.7373737373737373,
+      "eval_loss": 0.7380097508430481,
+      "eval_runtime": 32.5063,
+      "eval_samples_per_second": 109.763,
+      "eval_steps_per_second": 13.72,
+      "step": 640
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.5404,
+      "step": 650
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00017743589743589744,
+      "loss": 0.5775,
+      "step": 660
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001770940170940171,
+      "loss": 0.5013,
+      "step": 670
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00017675213675213676,
+      "loss": 0.6024,
+      "step": 680
+    },
+    {
+      "epoch": 1.74,
+      "eval_accuracy": 0.7407407407407407,
+      "eval_loss": 0.7545998692512512,
+      "eval_runtime": 36.6802,
+      "eval_samples_per_second": 97.273,
+      "eval_steps_per_second": 12.159,
+      "step": 680
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00017641025641025642,
+      "loss": 0.4815,
+      "step": 690
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00017606837606837607,
+      "loss": 0.4727,
+      "step": 700
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00017572649572649573,
+      "loss": 0.4698,
+      "step": 710
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001753846153846154,
+      "loss": 0.4813,
+      "step": 720
+    },
+    {
+      "epoch": 1.85,
+      "eval_accuracy": 0.7609427609427609,
+      "eval_loss": 0.7190226912498474,
+      "eval_runtime": 32.2305,
+      "eval_samples_per_second": 110.702,
+      "eval_steps_per_second": 13.838,
+      "step": 720
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00017504273504273505,
+      "loss": 0.5407,
+      "step": 730
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001747008547008547,
+      "loss": 0.4137,
+      "step": 740
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00017435897435897436,
+      "loss": 0.5252,
+      "step": 750
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00017401709401709402,
+      "loss": 0.5608,
+      "step": 760
+    },
+    {
+      "epoch": 1.95,
+      "eval_accuracy": 0.7283950617283951,
+      "eval_loss": 0.7672268748283386,
+      "eval_runtime": 34.3059,
+      "eval_samples_per_second": 104.006,
+      "eval_steps_per_second": 13.001,
+      "step": 760
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00017367521367521368,
+      "loss": 0.4692,
+      "step": 770
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.488,
+      "step": 780
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.000172991452991453,
+      "loss": 0.3798,
+      "step": 790
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00017264957264957268,
+      "loss": 0.3272,
+      "step": 800
+    },
+    {
+      "epoch": 2.05,
+      "eval_accuracy": 0.7497194163860831,
+      "eval_loss": 0.7055391073226929,
+      "eval_runtime": 34.4041,
+      "eval_samples_per_second": 103.709,
+      "eval_steps_per_second": 12.964,
+      "step": 800
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00017230769230769234,
+      "loss": 0.2257,
+      "step": 810
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.000171965811965812,
+      "loss": 0.238,
+      "step": 820
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00017162393162393162,
+      "loss": 0.3225,
+      "step": 830
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00017128205128205128,
+      "loss": 0.2803,
+      "step": 840
+    },
+    {
+      "epoch": 2.15,
+      "eval_accuracy": 0.7514029180695847,
+      "eval_loss": 0.7669840455055237,
+      "eval_runtime": 32.953,
+      "eval_samples_per_second": 108.275,
+      "eval_steps_per_second": 13.534,
+      "step": 840
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00017094017094017094,
+      "loss": 0.234,
+      "step": 850
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0001705982905982906,
+      "loss": 0.2201,
+      "step": 860
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00017025641025641026,
+      "loss": 0.4651,
+      "step": 870
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0001699145299145299,
+      "loss": 0.4051,
+      "step": 880
+    },
+    {
+      "epoch": 2.26,
+      "eval_accuracy": 0.7620650953984287,
+      "eval_loss": 0.691470205783844,
+      "eval_runtime": 37.2444,
+      "eval_samples_per_second": 95.8,
+      "eval_steps_per_second": 11.975,
+      "step": 880
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00016957264957264957,
+      "loss": 0.3069,
+      "step": 890
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00016923076923076923,
+      "loss": 0.2787,
+      "step": 900
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00016888888888888889,
+      "loss": 0.2794,
+      "step": 910
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00016854700854700854,
+      "loss": 0.3436,
+      "step": 920
+    },
+    {
+      "epoch": 2.36,
+      "eval_accuracy": 0.7631874298540965,
+      "eval_loss": 0.6948888301849365,
+      "eval_runtime": 32.1442,
+      "eval_samples_per_second": 111.0,
+      "eval_steps_per_second": 13.875,
+      "step": 920
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00016820512820512823,
+      "loss": 0.2964,
+      "step": 930
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00016786324786324789,
+      "loss": 0.1973,
+      "step": 940
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00016752136752136754,
+      "loss": 0.2257,
+      "step": 950
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0001671794871794872,
+      "loss": 0.2759,
+      "step": 960
+    },
+    {
+      "epoch": 2.46,
+      "eval_accuracy": 0.7356902356902357,
+      "eval_loss": 0.8428576588630676,
+      "eval_runtime": 32.8186,
+      "eval_samples_per_second": 108.719,
+      "eval_steps_per_second": 13.59,
+      "step": 960
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00016683760683760686,
+      "loss": 0.272,
+      "step": 970
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00016649572649572652,
+      "loss": 0.2721,
+      "step": 980
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00016615384615384617,
+      "loss": 0.1414,
+      "step": 990
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00016581196581196583,
+      "loss": 0.5024,
+      "step": 1000
+    },
+    {
+      "epoch": 2.56,
+      "eval_accuracy": 0.7502805836139169,
+      "eval_loss": 0.8222711086273193,
+      "eval_runtime": 32.7463,
+      "eval_samples_per_second": 108.959,
+      "eval_steps_per_second": 13.62,
+      "step": 1000
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00016547008547008546,
+      "loss": 0.3348,
+      "step": 1010
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00016512820512820512,
+      "loss": 0.3692,
+      "step": 1020
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00016478632478632478,
+      "loss": 0.3981,
+      "step": 1030
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00016444444444444444,
+      "loss": 0.3571,
+      "step": 1040
+    },
+    {
+      "epoch": 2.67,
+      "eval_accuracy": 0.7289562289562289,
+      "eval_loss": 0.8546615839004517,
+      "eval_runtime": 33.6376,
+      "eval_samples_per_second": 106.072,
+      "eval_steps_per_second": 13.259,
+      "step": 1040
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0001641025641025641,
+      "loss": 0.3647,
+      "step": 1050
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00016376068376068375,
+      "loss": 0.2547,
+      "step": 1060
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00016341880341880344,
+      "loss": 0.3708,
+      "step": 1070
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0001630769230769231,
+      "loss": 0.3197,
+      "step": 1080
+    },
+    {
+      "epoch": 2.77,
+      "eval_accuracy": 0.754769921436588,
+      "eval_loss": 0.7111819982528687,
+      "eval_runtime": 33.1915,
+      "eval_samples_per_second": 107.497,
+      "eval_steps_per_second": 13.437,
+      "step": 1080
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00016273504273504275,
+      "loss": 0.2805,
+      "step": 1090
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0001623931623931624,
+      "loss": 0.2792,
+      "step": 1100
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00016205128205128207,
+      "loss": 0.3781,
+      "step": 1110
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00016170940170940172,
+      "loss": 0.3872,
+      "step": 1120
+    },
+    {
+      "epoch": 2.87,
+      "eval_accuracy": 0.7485970819304153,
+      "eval_loss": 0.7666726112365723,
+      "eval_runtime": 33.3793,
+      "eval_samples_per_second": 106.892,
+      "eval_steps_per_second": 13.362,
+      "step": 1120
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00016136752136752138,
+      "loss": 0.3489,
+      "step": 1130
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00016102564102564104,
+      "loss": 0.3436,
+      "step": 1140
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0001606837606837607,
+      "loss": 0.2608,
+      "step": 1150
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00016034188034188036,
+      "loss": 0.2699,
+      "step": 1160
+    },
+    {
+      "epoch": 2.97,
+      "eval_accuracy": 0.7867564534231201,
+      "eval_loss": 0.6604002714157104,
+      "eval_runtime": 42.2756,
+      "eval_samples_per_second": 84.398,
+      "eval_steps_per_second": 10.55,
+      "step": 1160
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00016,
+      "loss": 0.1998,
+      "step": 1170
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00015965811965811967,
+      "loss": 0.1375,
+      "step": 1180
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0001593162393162393,
+      "loss": 0.1511,
+      "step": 1190
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00015897435897435896,
+      "loss": 0.237,
+      "step": 1200
+    },
+    {
+      "epoch": 3.08,
+      "eval_accuracy": 0.7570145903479237,
+      "eval_loss": 0.8170732855796814,
+      "eval_runtime": 34.8004,
+      "eval_samples_per_second": 102.528,
+      "eval_steps_per_second": 12.816,
+      "step": 1200
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00015863247863247864,
+      "loss": 0.1256,
+      "step": 1210
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0001582905982905983,
+      "loss": 0.1109,
+      "step": 1220
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00015794871794871796,
+      "loss": 0.1218,
+      "step": 1230
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00015760683760683762,
+      "loss": 0.2952,
+      "step": 1240
+    },
+    {
+      "epoch": 3.18,
+      "eval_accuracy": 0.7598204264870931,
+      "eval_loss": 0.8052437901496887,
+      "eval_runtime": 33.2814,
+      "eval_samples_per_second": 107.207,
+      "eval_steps_per_second": 13.401,
+      "step": 1240
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00015726495726495727,
+      "loss": 0.1483,
+      "step": 1250
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00015692307692307693,
+      "loss": 0.1639,
+      "step": 1260
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0001565811965811966,
+      "loss": 0.1832,
+      "step": 1270
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00015623931623931625,
+      "loss": 0.217,
+      "step": 1280
+    },
+    {
+      "epoch": 3.28,
+      "eval_accuracy": 0.7822671156004489,
+      "eval_loss": 0.7558466196060181,
+      "eval_runtime": 33.6999,
+      "eval_samples_per_second": 105.876,
+      "eval_steps_per_second": 13.234,
+      "step": 1280
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0001558974358974359,
+      "loss": 0.1555,
+      "step": 1290
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.1288,
+      "step": 1300
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00015521367521367522,
+      "loss": 0.1276,
+      "step": 1310
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00015487179487179488,
+      "loss": 0.1804,
+      "step": 1320
+    },
+    {
+      "epoch": 3.38,
+      "eval_accuracy": 0.7631874298540965,
+      "eval_loss": 0.8506298065185547,
+      "eval_runtime": 35.5517,
+      "eval_samples_per_second": 100.361,
+      "eval_steps_per_second": 12.545,
+      "step": 1320
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00015452991452991454,
+      "loss": 0.1482,
+      "step": 1330
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0001541880341880342,
+      "loss": 0.1177,
+      "step": 1340
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00015384615384615385,
+      "loss": 0.137,
+      "step": 1350
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0001535042735042735,
+      "loss": 0.234,
+      "step": 1360
+    },
+    {
+      "epoch": 3.49,
+      "eval_accuracy": 0.7974186307519641,
+      "eval_loss": 0.7378016114234924,
+      "eval_runtime": 33.7299,
+      "eval_samples_per_second": 105.782,
+      "eval_steps_per_second": 13.223,
+      "step": 1360
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0001531623931623932,
+      "loss": 0.2669,
+      "step": 1370
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00015282051282051282,
+      "loss": 0.1098,
+      "step": 1380
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00015247863247863248,
+      "loss": 0.1492,
+      "step": 1390
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00015213675213675214,
+      "loss": 0.1799,
+      "step": 1400
+    },
+    {
+      "epoch": 3.59,
+      "eval_accuracy": 0.7749719416386083,
+      "eval_loss": 0.7870209813117981,
+      "eval_runtime": 33.1,
+      "eval_samples_per_second": 107.795,
+      "eval_steps_per_second": 13.474,
+      "step": 1400
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0001517948717948718,
+      "loss": 0.2032,
+      "step": 1410
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00015145299145299146,
+      "loss": 0.3042,
+      "step": 1420
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0001511111111111111,
+      "loss": 0.2242,
+      "step": 1430
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00015076923076923077,
+      "loss": 0.2508,
+      "step": 1440
+    },
+    {
+      "epoch": 3.69,
+      "eval_accuracy": 0.7283950617283951,
+      "eval_loss": 1.016114354133606,
+      "eval_runtime": 33.0909,
+      "eval_samples_per_second": 107.824,
+      "eval_steps_per_second": 13.478,
+      "step": 1440
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00015042735042735043,
+      "loss": 0.3502,
+      "step": 1450
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00015008547008547009,
+      "loss": 0.1808,
+      "step": 1460
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00014974358974358974,
+      "loss": 0.1187,
+      "step": 1470
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0001494017094017094,
+      "loss": 0.2714,
+      "step": 1480
+    },
+    {
+      "epoch": 3.79,
+      "eval_accuracy": 0.7721661054994389,
+      "eval_loss": 0.8105702996253967,
+      "eval_runtime": 33.8257,
+      "eval_samples_per_second": 105.482,
+      "eval_steps_per_second": 13.185,
+      "step": 1480
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00014905982905982906,
+      "loss": 0.1947,
+      "step": 1490
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00014871794871794872,
+      "loss": 0.2269,
+      "step": 1500
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0001483760683760684,
+      "loss": 0.2676,
+      "step": 1510
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00014803418803418806,
+      "loss": 0.1342,
+      "step": 1520
+    },
+    {
+      "epoch": 3.9,
+      "eval_accuracy": 0.7665544332210998,
+      "eval_loss": 0.8097100853919983,
+      "eval_runtime": 32.9851,
+      "eval_samples_per_second": 108.17,
+      "eval_steps_per_second": 13.521,
+      "step": 1520
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00014769230769230772,
+      "loss": 0.1616,
+      "step": 1530
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00014735042735042737,
+      "loss": 0.1682,
+      "step": 1540
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00014700854700854703,
+      "loss": 0.1335,
+      "step": 1550
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.2337,
+      "step": 1560
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.7805836139169473,
+      "eval_loss": 0.7859733700752258,
+      "eval_runtime": 32.3155,
+      "eval_samples_per_second": 110.411,
+      "eval_steps_per_second": 13.801,
+      "step": 1560
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00014632478632478632,
+      "loss": 0.0954,
+      "step": 1570
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00014598290598290598,
+      "loss": 0.1129,
+      "step": 1580
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00014564102564102564,
+      "loss": 0.085,
+      "step": 1590
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.0001452991452991453,
+      "loss": 0.052,
+      "step": 1600
+    },
+    {
+      "epoch": 4.1,
+      "eval_accuracy": 0.7654320987654321,
+      "eval_loss": 0.8651421666145325,
+      "eval_runtime": 37.0818,
+      "eval_samples_per_second": 96.22,
+      "eval_steps_per_second": 12.027,
+      "step": 1600
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00014495726495726495,
+      "loss": 0.0705,
+      "step": 1610
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0001446153846153846,
+      "loss": 0.0467,
+      "step": 1620
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00014427350427350427,
+      "loss": 0.0585,
+      "step": 1630
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00014393162393162392,
+      "loss": 0.2216,
+      "step": 1640
+    },
+    {
+      "epoch": 4.21,
+      "eval_accuracy": 0.7800224466891134,
+      "eval_loss": 0.8762761950492859,
+      "eval_runtime": 32.976,
+      "eval_samples_per_second": 108.2,
+      "eval_steps_per_second": 13.525,
+      "step": 1640
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0001435897435897436,
+      "loss": 0.0832,
+      "step": 1650
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00014324786324786327,
+      "loss": 0.0812,
+      "step": 1660
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00014290598290598292,
+      "loss": 0.1584,
+      "step": 1670
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00014256410256410258,
+      "loss": 0.1194,
+      "step": 1680
+    },
+    {
+      "epoch": 4.31,
+      "eval_accuracy": 0.7822671156004489,
+      "eval_loss": 0.848014771938324,
+      "eval_runtime": 33.7757,
+      "eval_samples_per_second": 105.638,
+      "eval_steps_per_second": 13.205,
+      "step": 1680
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00014222222222222224,
+      "loss": 0.1615,
+      "step": 1690
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0001418803418803419,
+      "loss": 0.1423,
+      "step": 1700
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00014153846153846156,
+      "loss": 0.0727,
+      "step": 1710
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0001411965811965812,
+      "loss": 0.1959,
+      "step": 1720
+    },
+    {
+      "epoch": 4.41,
+      "eval_accuracy": 0.7581369248035915,
+      "eval_loss": 0.9282656908035278,
+      "eval_runtime": 33.2614,
+      "eval_samples_per_second": 107.272,
+      "eval_steps_per_second": 13.409,
+      "step": 1720
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00014085470085470087,
+      "loss": 0.1193,
+      "step": 1730
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0001405128205128205,
+      "loss": 0.1106,
+      "step": 1740
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00014017094017094016,
+      "loss": 0.0624,
+      "step": 1750
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.00013982905982905982,
+      "loss": 0.095,
+      "step": 1760
+    },
+    {
+      "epoch": 4.51,
+      "eval_accuracy": 0.7648709315375982,
+      "eval_loss": 0.9624485969543457,
+      "eval_runtime": 32.5961,
+      "eval_samples_per_second": 109.461,
+      "eval_steps_per_second": 13.683,
+      "step": 1760
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00013948717948717947,
+      "loss": 0.0659,
+      "step": 1770
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.00013914529914529916,
+      "loss": 0.1178,
+      "step": 1780
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00013880341880341882,
+      "loss": 0.0864,
+      "step": 1790
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00013846153846153847,
+      "loss": 0.0727,
+      "step": 1800
+    },
+    {
+      "epoch": 4.62,
+      "eval_accuracy": 0.7615039281705949,
+      "eval_loss": 0.9454267621040344,
+      "eval_runtime": 33.1697,
+      "eval_samples_per_second": 107.568,
+      "eval_steps_per_second": 13.446,
+      "step": 1800
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00013811965811965813,
+      "loss": 0.0658,
+      "step": 1810
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0001377777777777778,
+      "loss": 0.1133,
+      "step": 1820
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00013743589743589745,
+      "loss": 0.0978,
+      "step": 1830
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0001370940170940171,
+      "loss": 0.0989,
+      "step": 1840
+    },
+    {
+      "epoch": 4.72,
+      "eval_accuracy": 0.7878787878787878,
+      "eval_loss": 0.8532549142837524,
+      "eval_runtime": 33.6969,
+      "eval_samples_per_second": 105.885,
+      "eval_steps_per_second": 13.236,
+      "step": 1840
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00013675213675213676,
+      "loss": 0.1147,
+      "step": 1850
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00013641025641025642,
+      "loss": 0.187,
+      "step": 1860
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00013606837606837608,
+      "loss": 0.1036,
+      "step": 1870
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00013572649572649574,
+      "loss": 0.0785,
+      "step": 1880
+    },
+    {
+      "epoch": 4.82,
+      "eval_accuracy": 0.7890011223344556,
+      "eval_loss": 0.8057104349136353,
+      "eval_runtime": 31.6741,
+      "eval_samples_per_second": 112.647,
+      "eval_steps_per_second": 14.081,
+      "step": 1880
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0001353846153846154,
+      "loss": 0.0838,
+      "step": 1890
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00013504273504273505,
+      "loss": 0.0703,
+      "step": 1900
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0001347008547008547,
+      "loss": 0.0543,
+      "step": 1910
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00013435897435897437,
+      "loss": 0.1162,
+      "step": 1920
+    },
+    {
+      "epoch": 4.92,
+      "eval_accuracy": 0.7631874298540965,
+      "eval_loss": 0.9254993200302124,
+      "eval_runtime": 32.3255,
+      "eval_samples_per_second": 110.377,
+      "eval_steps_per_second": 13.797,
+      "step": 1920
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00013401709401709402,
+      "loss": 0.2486,
+      "step": 1930
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.00013367521367521368,
+      "loss": 0.1092,
+      "step": 1940
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.1394,
+      "step": 1950
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.000132991452991453,
+      "loss": 0.0824,
+      "step": 1960
+    },
+    {
+      "epoch": 5.03,
+      "eval_accuracy": 0.7665544332210998,
+      "eval_loss": 0.9655769467353821,
+      "eval_runtime": 33.4626,
+      "eval_samples_per_second": 106.627,
+      "eval_steps_per_second": 13.328,
+      "step": 1960
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00013264957264957266,
+      "loss": 0.0459,
+      "step": 1970
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0001323076923076923,
+      "loss": 0.0497,
+      "step": 1980
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00013196581196581197,
+      "loss": 0.1127,
+      "step": 1990
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00013162393162393163,
+      "loss": 0.117,
+      "step": 2000
+    },
+    {
+      "epoch": 5.13,
+      "eval_accuracy": 0.7850729517396184,
+      "eval_loss": 0.9316745400428772,
+      "eval_runtime": 33.9606,
+      "eval_samples_per_second": 105.063,
+      "eval_steps_per_second": 13.133,
+      "step": 2000
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00013128205128205129,
+      "loss": 0.0828,
+      "step": 2010
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00013094017094017094,
+      "loss": 0.1093,
+      "step": 2020
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.0001305982905982906,
+      "loss": 0.0873,
+      "step": 2030
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00013025641025641026,
+      "loss": 0.0714,
+      "step": 2040
+    },
+    {
+      "epoch": 5.23,
+      "eval_accuracy": 0.7811447811447811,
+      "eval_loss": 0.881272554397583,
+      "eval_runtime": 32.119,
+      "eval_samples_per_second": 111.087,
+      "eval_steps_per_second": 13.886,
+      "step": 2040
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00012991452991452992,
+      "loss": 0.0285,
+      "step": 2050
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00012957264957264957,
+      "loss": 0.0122,
+      "step": 2060
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00012923076923076923,
+      "loss": 0.0407,
+      "step": 2070
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00012888888888888892,
+      "loss": 0.086,
+      "step": 2080
+    },
+    {
+      "epoch": 5.33,
+      "eval_accuracy": 0.7901234567901234,
+      "eval_loss": 0.9098324775695801,
+      "eval_runtime": 31.6914,
+      "eval_samples_per_second": 112.586,
+      "eval_steps_per_second": 14.073,
+      "step": 2080
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00012854700854700857,
+      "loss": 0.0616,
+      "step": 2090
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00012820512820512823,
+      "loss": 0.1591,
+      "step": 2100
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00012786324786324786,
+      "loss": 0.1219,
+      "step": 2110
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00012752136752136752,
+      "loss": 0.1538,
+      "step": 2120
+    },
+    {
+      "epoch": 5.44,
+      "eval_accuracy": 0.7934904601571269,
+      "eval_loss": 0.8712666630744934,
+      "eval_runtime": 31.735,
+      "eval_samples_per_second": 112.431,
+      "eval_steps_per_second": 14.054,
+      "step": 2120
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00012717948717948718,
+      "loss": 0.0529,
+      "step": 2130
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00012683760683760684,
+      "loss": 0.1755,
+      "step": 2140
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.0001264957264957265,
+      "loss": 0.0202,
+      "step": 2150
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00012615384615384615,
+      "loss": 0.0522,
+      "step": 2160
+    },
+    {
+      "epoch": 5.54,
+      "eval_accuracy": 0.7744107744107744,
+      "eval_loss": 0.9605388641357422,
+      "eval_runtime": 31.6059,
+      "eval_samples_per_second": 112.89,
+      "eval_steps_per_second": 14.111,
+      "step": 2160
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.0001258119658119658,
+      "loss": 0.0517,
+      "step": 2170
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00012547008547008547,
+      "loss": 0.0873,
+      "step": 2180
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00012512820512820512,
+      "loss": 0.1013,
+      "step": 2190
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00012478632478632478,
+      "loss": 0.0485,
+      "step": 2200
+    },
+    {
+      "epoch": 5.64,
+      "eval_accuracy": 0.7856341189674523,
+      "eval_loss": 0.881409227848053,
+      "eval_runtime": 36.6436,
+      "eval_samples_per_second": 97.37,
+      "eval_steps_per_second": 12.171,
+      "step": 2200
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00012444444444444444,
+      "loss": 0.0432,
+      "step": 2210
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00012410256410256412,
+      "loss": 0.0601,
+      "step": 2220
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00012376068376068378,
+      "loss": 0.0593,
+      "step": 2230
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00012341880341880344,
+      "loss": 0.0374,
+      "step": 2240
+    },
+    {
+      "epoch": 5.74,
+      "eval_accuracy": 0.7946127946127947,
+      "eval_loss": 0.8851107358932495,
+      "eval_runtime": 31.8935,
+      "eval_samples_per_second": 111.872,
+      "eval_steps_per_second": 13.984,
+      "step": 2240
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0001230769230769231,
+      "loss": 0.106,
+      "step": 2250
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00012273504273504276,
+      "loss": 0.1051,
+      "step": 2260
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0001223931623931624,
+      "loss": 0.0107,
+      "step": 2270
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00012205128205128207,
+      "loss": 0.019,
+      "step": 2280
+    },
+    {
+      "epoch": 5.85,
+      "eval_accuracy": 0.7850729517396184,
+      "eval_loss": 0.9623528122901917,
+      "eval_runtime": 32.6648,
+      "eval_samples_per_second": 109.231,
+      "eval_steps_per_second": 13.654,
+      "step": 2280
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0001217094017094017,
+      "loss": 0.0304,
+      "step": 2290
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00012136752136752136,
+      "loss": 0.0269,
+      "step": 2300
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00012102564102564103,
+      "loss": 0.1185,
+      "step": 2310
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00012068376068376069,
+      "loss": 0.1115,
+      "step": 2320
+    },
+    {
+      "epoch": 5.95,
+      "eval_accuracy": 0.7811447811447811,
+      "eval_loss": 0.9703367948532104,
+      "eval_runtime": 40.8069,
+      "eval_samples_per_second": 87.436,
+      "eval_steps_per_second": 10.93,
+      "step": 2320
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00012034188034188035,
+      "loss": 0.0199,
+      "step": 2330
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00012,
+      "loss": 0.0796,
+      "step": 2340
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00011965811965811966,
+      "loss": 0.0881,
+      "step": 2350
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00011931623931623932,
+      "loss": 0.0579,
+      "step": 2360
+    },
+    {
+      "epoch": 6.05,
+      "eval_accuracy": 0.7901234567901234,
+      "eval_loss": 0.9741098880767822,
+      "eval_runtime": 34.3499,
+      "eval_samples_per_second": 103.872,
+      "eval_steps_per_second": 12.984,
+      "step": 2360
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00011897435897435898,
+      "loss": 0.0237,
+      "step": 2370
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00011863247863247863,
+      "loss": 0.0536,
+      "step": 2380
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00011829059829059829,
+      "loss": 0.0947,
+      "step": 2390
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00011794871794871796,
+      "loss": 0.027,
+      "step": 2400
+    },
+    {
+      "epoch": 6.15,
+      "eval_accuracy": 0.77665544332211,
+      "eval_loss": 1.0114481449127197,
+      "eval_runtime": 32.0411,
+      "eval_samples_per_second": 111.357,
+      "eval_steps_per_second": 13.92,
+      "step": 2400
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00011760683760683762,
+      "loss": 0.0638,
+      "step": 2410
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00011726495726495728,
+      "loss": 0.015,
+      "step": 2420
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00011692307692307694,
+      "loss": 0.0663,
+      "step": 2430
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.0001165811965811966,
+      "loss": 0.0202,
+      "step": 2440
+    },
+    {
+      "epoch": 6.26,
+      "eval_accuracy": 0.7929292929292929,
+      "eval_loss": 0.9184174537658691,
+      "eval_runtime": 32.1376,
+      "eval_samples_per_second": 111.022,
+      "eval_steps_per_second": 13.878,
+      "step": 2440
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00011623931623931625,
+      "loss": 0.0905,
+      "step": 2450
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00011589743589743591,
+      "loss": 0.0114,
+      "step": 2460
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00011555555555555555,
+      "loss": 0.0427,
+      "step": 2470
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00011521367521367521,
+      "loss": 0.0023,
+      "step": 2480
+    },
+    {
+      "epoch": 6.36,
+      "eval_accuracy": 0.7755331088664422,
+      "eval_loss": 1.037878155708313,
+      "eval_runtime": 32.6535,
+      "eval_samples_per_second": 109.269,
+      "eval_steps_per_second": 13.659,
+      "step": 2480
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00011487179487179487,
+      "loss": 0.0371,
+      "step": 2490
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00011452991452991453,
+      "loss": 0.0473,
+      "step": 2500
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00011418803418803418,
+      "loss": 0.0203,
+      "step": 2510
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00011384615384615384,
+      "loss": 0.1106,
+      "step": 2520
+    },
+    {
+      "epoch": 6.46,
+      "eval_accuracy": 0.7839506172839507,
+      "eval_loss": 0.9894450902938843,
+      "eval_runtime": 34.0314,
+      "eval_samples_per_second": 104.844,
+      "eval_steps_per_second": 13.106,
+      "step": 2520
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00011350427350427351,
+      "loss": 0.048,
+      "step": 2530
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00011316239316239317,
+      "loss": 0.0615,
+      "step": 2540
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00011282051282051283,
+      "loss": 0.0127,
+      "step": 2550
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00011247863247863249,
+      "loss": 0.0032,
+      "step": 2560
+    },
+    {
+      "epoch": 6.56,
+      "eval_accuracy": 0.8063973063973064,
+      "eval_loss": 0.912738025188446,
+      "eval_runtime": 36.8685,
+      "eval_samples_per_second": 96.776,
+      "eval_steps_per_second": 12.097,
+      "step": 2560
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00011213675213675214,
+      "loss": 0.0161,
+      "step": 2570
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0001117948717948718,
+      "loss": 0.0356,
+      "step": 2580
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00011145299145299146,
+      "loss": 0.0041,
+      "step": 2590
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.0305,
+      "step": 2600
+    },
+    {
+      "epoch": 6.67,
+      "eval_accuracy": 0.803030303030303,
+      "eval_loss": 0.920154333114624,
+      "eval_runtime": 34.2244,
+      "eval_samples_per_second": 104.253,
+      "eval_steps_per_second": 13.032,
+      "step": 2600
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00011076923076923077,
+      "loss": 0.0022,
+      "step": 2610
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00011042735042735045,
+      "loss": 0.0211,
+      "step": 2620
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0001100854700854701,
+      "loss": 0.0031,
+      "step": 2630
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00010974358974358976,
+      "loss": 0.021,
+      "step": 2640
+    },
+    {
+      "epoch": 6.77,
+      "eval_accuracy": 0.7985409652076318,
+      "eval_loss": 0.9700907468795776,
+      "eval_runtime": 34.5087,
+      "eval_samples_per_second": 103.394,
+      "eval_steps_per_second": 12.924,
+      "step": 2640
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00010940170940170942,
+      "loss": 0.0434,
+      "step": 2650
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00010905982905982905,
+      "loss": 0.0163,
+      "step": 2660
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00010871794871794872,
+      "loss": 0.0366,
+      "step": 2670
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00010837606837606838,
+      "loss": 0.044,
+      "step": 2680
+    },
+    {
+      "epoch": 6.87,
+      "eval_accuracy": 0.792368125701459,
+      "eval_loss": 1.1099859476089478,
+      "eval_runtime": 34.3976,
+      "eval_samples_per_second": 103.728,
+      "eval_steps_per_second": 12.966,
+      "step": 2680
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00010803418803418804,
+      "loss": 0.0224,
+      "step": 2690
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0001076923076923077,
+      "loss": 0.0068,
+      "step": 2700
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00010735042735042735,
+      "loss": 0.0419,
+      "step": 2710
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00010700854700854701,
+      "loss": 0.0097,
+      "step": 2720
+    },
+    {
+      "epoch": 6.97,
+      "eval_accuracy": 0.7631874298540965,
+      "eval_loss": 1.2310634851455688,
+      "eval_runtime": 37.3556,
+      "eval_samples_per_second": 95.514,
+      "eval_steps_per_second": 11.939,
+      "step": 2720
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.0381,
+      "step": 2730
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00010632478632478632,
+      "loss": 0.0017,
+      "step": 2740
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.000105982905982906,
+      "loss": 0.0077,
+      "step": 2750
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00010564102564102565,
+      "loss": 0.0015,
+      "step": 2760
+    },
+    {
+      "epoch": 7.08,
+      "eval_accuracy": 0.8052749719416387,
+      "eval_loss": 0.9980599880218506,
+      "eval_runtime": 38.1071,
+      "eval_samples_per_second": 93.631,
+      "eval_steps_per_second": 11.704,
+      "step": 2760
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00010529914529914531,
+      "loss": 0.0017,
+      "step": 2770
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00010495726495726497,
+      "loss": 0.0041,
+      "step": 2780
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00010461538461538463,
+      "loss": 0.0348,
+      "step": 2790
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00010427350427350428,
+      "loss": 0.0034,
+      "step": 2800
+    },
+    {
+      "epoch": 7.18,
+      "eval_accuracy": 0.7929292929292929,
+      "eval_loss": 1.0722278356552124,
+      "eval_runtime": 34.0426,
+      "eval_samples_per_second": 104.81,
+      "eval_steps_per_second": 13.101,
+      "step": 2800
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00010393162393162394,
+      "loss": 0.0296,
+      "step": 2810
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0001035897435897436,
+      "loss": 0.0614,
+      "step": 2820
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00010324786324786326,
+      "loss": 0.0014,
+      "step": 2830
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0001029059829059829,
+      "loss": 0.0317,
+      "step": 2840
+    },
+    {
+      "epoch": 7.28,
+      "eval_accuracy": 0.7968574635241302,
+      "eval_loss": 1.049239158630371,
+      "eval_runtime": 34.4577,
+      "eval_samples_per_second": 103.547,
+      "eval_steps_per_second": 12.943,
+      "step": 2840
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00010256410256410256,
+      "loss": 0.0122,
+      "step": 2850
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00010222222222222222,
+      "loss": 0.0019,
+      "step": 2860
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00010188034188034187,
+      "loss": 0.0055,
+      "step": 2870
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00010153846153846153,
+      "loss": 0.0013,
+      "step": 2880
+    },
+    {
+      "epoch": 7.38,
+      "eval_accuracy": 0.792368125701459,
+      "eval_loss": 1.101838231086731,
+      "eval_runtime": 38.1047,
+      "eval_samples_per_second": 93.637,
+      "eval_steps_per_second": 11.705,
+      "step": 2880
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001011965811965812,
+      "loss": 0.0027,
+      "step": 2890
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00010085470085470086,
+      "loss": 0.1045,
+      "step": 2900
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00010051282051282052,
+      "loss": 0.003,
+      "step": 2910
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00010017094017094018,
+      "loss": 0.0145,
+      "step": 2920
+    },
+    {
+      "epoch": 7.49,
+      "eval_accuracy": 0.7890011223344556,
+      "eval_loss": 1.1084059476852417,
+      "eval_runtime": 32.7356,
+      "eval_samples_per_second": 108.995,
+      "eval_steps_per_second": 13.624,
+      "step": 2920
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 9.982905982905983e-05,
+      "loss": 0.0023,
+      "step": 2930
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 9.948717948717949e-05,
+      "loss": 0.003,
+      "step": 2940
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 9.914529914529915e-05,
+      "loss": 0.0292,
+      "step": 2950
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 9.88034188034188e-05,
+      "loss": 0.0171,
+      "step": 2960
+    },
+    {
+      "epoch": 7.59,
+      "eval_accuracy": 0.8069584736251403,
+      "eval_loss": 1.0281527042388916,
+      "eval_runtime": 32.3307,
+      "eval_samples_per_second": 110.359,
+      "eval_steps_per_second": 13.795,
+      "step": 2960
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 9.846153846153848e-05,
+      "loss": 0.0433,
+      "step": 2970
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 9.811965811965812e-05,
+      "loss": 0.0178,
+      "step": 2980
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 9.777777777777778e-05,
+      "loss": 0.0218,
+      "step": 2990
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 9.743589743589744e-05,
+      "loss": 0.0429,
+      "step": 3000
+    },
+    {
+      "epoch": 7.69,
+      "eval_accuracy": 0.7901234567901234,
+      "eval_loss": 1.107517123222351,
+      "eval_runtime": 32.4213,
+      "eval_samples_per_second": 110.051,
+      "eval_steps_per_second": 13.756,
+      "step": 3000
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 9.70940170940171e-05,
+      "loss": 0.0068,
+      "step": 3010
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 9.675213675213675e-05,
+      "loss": 0.0084,
+      "step": 3020
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 9.641025641025641e-05,
+      "loss": 0.0019,
+      "step": 3030
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 9.606837606837608e-05,
+      "loss": 0.0012,
+      "step": 3040
+    },
+    {
+      "epoch": 7.79,
+      "eval_accuracy": 0.7951739618406285,
+      "eval_loss": 1.1173745393753052,
+      "eval_runtime": 33.3697,
+      "eval_samples_per_second": 106.923,
+      "eval_steps_per_second": 13.365,
+      "step": 3040
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 9.572649572649574e-05,
+      "loss": 0.0012,
+      "step": 3050
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 9.53846153846154e-05,
+      "loss": 0.0148,
+      "step": 3060
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 9.504273504273504e-05,
+      "loss": 0.0396,
+      "step": 3070
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 9.47008547008547e-05,
+      "loss": 0.0267,
+      "step": 3080
+    },
+    {
+      "epoch": 7.9,
+      "eval_accuracy": 0.7895622895622896,
+      "eval_loss": 1.1118639707565308,
+      "eval_runtime": 39.7858,
+      "eval_samples_per_second": 89.68,
+      "eval_steps_per_second": 11.21,
+      "step": 3080
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 9.435897435897436e-05,
+      "loss": 0.0531,
+      "step": 3090
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 9.401709401709401e-05,
+      "loss": 0.0025,
+      "step": 3100
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.367521367521369e-05,
+      "loss": 0.0388,
+      "step": 3110
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.0313,
+      "step": 3120
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7951739618406285,
+      "eval_loss": 1.1246873140335083,
+      "eval_runtime": 33.2457,
+      "eval_samples_per_second": 107.322,
+      "eval_steps_per_second": 13.415,
+      "step": 3120
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 9.2991452991453e-05,
+      "loss": 0.0015,
+      "step": 3130
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 9.264957264957266e-05,
+      "loss": 0.001,
+      "step": 3140
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 9.230769230769232e-05,
+      "loss": 0.0048,
+      "step": 3150
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 9.196581196581196e-05,
+      "loss": 0.001,
+      "step": 3160
+    },
+    {
+      "epoch": 8.1,
+      "eval_accuracy": 0.7861952861952862,
+      "eval_loss": 1.1706517934799194,
+      "eval_runtime": 34.9683,
+      "eval_samples_per_second": 102.035,
+      "eval_steps_per_second": 12.754,
+      "step": 3160
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 9.162393162393162e-05,
+      "loss": 0.0203,
+      "step": 3170
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 9.128205128205129e-05,
+      "loss": 0.001,
+      "step": 3180
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 9.094017094017095e-05,
+      "loss": 0.0019,
+      "step": 3190
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 9.05982905982906e-05,
+      "loss": 0.0372,
+      "step": 3200
+    },
+    {
+      "epoch": 8.21,
+      "eval_accuracy": 0.8024691358024691,
+      "eval_loss": 1.1111818552017212,
+      "eval_runtime": 33.7848,
+      "eval_samples_per_second": 105.61,
+      "eval_steps_per_second": 13.201,
+      "step": 3200
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 9.025641025641026e-05,
+      "loss": 0.0031,
+      "step": 3210
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 8.991452991452992e-05,
+      "loss": 0.002,
+      "step": 3220
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 8.957264957264958e-05,
+      "loss": 0.0009,
+      "step": 3230
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 8.923076923076924e-05,
+      "loss": 0.0088,
+      "step": 3240
+    },
+    {
+      "epoch": 8.31,
+      "eval_accuracy": 0.8092031425364759,
+      "eval_loss": 1.0615540742874146,
+      "eval_runtime": 32.6537,
+      "eval_samples_per_second": 109.268,
+      "eval_steps_per_second": 13.658,
+      "step": 3240
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.0009,
+      "step": 3250
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 8.854700854700855e-05,
+      "loss": 0.0389,
+      "step": 3260
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 8.820512820512821e-05,
+      "loss": 0.0024,
+      "step": 3270
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 8.786324786324787e-05,
+      "loss": 0.0155,
+      "step": 3280
+    },
+    {
+      "epoch": 8.41,
+      "eval_accuracy": 0.7968574635241302,
+      "eval_loss": 1.1376060247421265,
+      "eval_runtime": 34.6264,
+      "eval_samples_per_second": 103.043,
+      "eval_steps_per_second": 12.88,
+      "step": 3280
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 8.752136752136752e-05,
+      "loss": 0.0011,
+      "step": 3290
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 8.717948717948718e-05,
+      "loss": 0.0013,
+      "step": 3300
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 8.683760683760684e-05,
+      "loss": 0.0017,
+      "step": 3310
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 8.64957264957265e-05,
+      "loss": 0.0095,
+      "step": 3320
+    },
+    {
+      "epoch": 8.51,
+      "eval_accuracy": 0.7794612794612794,
+      "eval_loss": 1.210555911064148,
+      "eval_runtime": 33.508,
+      "eval_samples_per_second": 106.482,
+      "eval_steps_per_second": 13.31,
+      "step": 3320
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 8.615384615384617e-05,
+      "loss": 0.0133,
+      "step": 3330
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 8.581196581196581e-05,
+      "loss": 0.0007,
+      "step": 3340
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 8.547008547008547e-05,
+      "loss": 0.0008,
+      "step": 3350
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 8.512820512820513e-05,
+      "loss": 0.0008,
+      "step": 3360
+    },
+    {
+      "epoch": 8.62,
+      "eval_accuracy": 0.797979797979798,
+      "eval_loss": 1.1278263330459595,
+      "eval_runtime": 33.3484,
+      "eval_samples_per_second": 106.992,
+      "eval_steps_per_second": 13.374,
+      "step": 3360
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 8.478632478632479e-05,
+      "loss": 0.0062,
+      "step": 3370
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 8.444444444444444e-05,
+      "loss": 0.0464,
+      "step": 3380
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 8.410256410256411e-05,
+      "loss": 0.0314,
+      "step": 3390
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 8.376068376068377e-05,
+      "loss": 0.0068,
+      "step": 3400
+    },
+    {
+      "epoch": 8.72,
+      "eval_accuracy": 0.8108866442199776,
+      "eval_loss": 1.0497138500213623,
+      "eval_runtime": 34.5617,
+      "eval_samples_per_second": 103.236,
+      "eval_steps_per_second": 12.904,
+      "step": 3400
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 8.341880341880343e-05,
+      "loss": 0.0007,
+      "step": 3410
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 8.307692307692309e-05,
+      "loss": 0.0009,
+      "step": 3420
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 8.273504273504273e-05,
+      "loss": 0.0235,
+      "step": 3430
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 8.239316239316239e-05,
+      "loss": 0.0446,
+      "step": 3440
+    },
+    {
+      "epoch": 8.82,
+      "eval_accuracy": 0.8024691358024691,
+      "eval_loss": 1.1239783763885498,
+      "eval_runtime": 35.6118,
+      "eval_samples_per_second": 100.192,
+      "eval_steps_per_second": 12.524,
+      "step": 3440
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 8.205128205128205e-05,
+      "loss": 0.0454,
+      "step": 3450
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 8.170940170940172e-05,
+      "loss": 0.0012,
+      "step": 3460
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 8.136752136752138e-05,
+      "loss": 0.0222,
+      "step": 3470
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 8.102564102564103e-05,
+      "loss": 0.0026,
+      "step": 3480
+    },
+    {
+      "epoch": 8.92,
+      "eval_accuracy": 0.8120089786756454,
+      "eval_loss": 1.0241341590881348,
+      "eval_runtime": 44.9153,
+      "eval_samples_per_second": 79.438,
+      "eval_steps_per_second": 9.93,
+      "step": 3480
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 8.068376068376069e-05,
+      "loss": 0.0112,
+      "step": 3490
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 8.034188034188035e-05,
+      "loss": 0.0044,
+      "step": 3500
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 8e-05,
+      "loss": 0.0125,
+      "step": 3510
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 7.965811965811965e-05,
+      "loss": 0.0315,
+      "step": 3520
+    },
+    {
+      "epoch": 9.03,
+      "eval_accuracy": 0.8024691358024691,
+      "eval_loss": 1.0902154445648193,
+      "eval_runtime": 33.6468,
+      "eval_samples_per_second": 106.043,
+      "eval_steps_per_second": 13.255,
+      "step": 3520
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 7.931623931623932e-05,
+      "loss": 0.0312,
+      "step": 3530
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 7.897435897435898e-05,
+      "loss": 0.0011,
+      "step": 3540
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 7.863247863247864e-05,
+      "loss": 0.0101,
+      "step": 3550
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 7.82905982905983e-05,
+      "loss": 0.0262,
+      "step": 3560
+    },
+    {
+      "epoch": 9.13,
+      "eval_accuracy": 0.8035914702581369,
+      "eval_loss": 1.0600807666778564,
+      "eval_runtime": 34.7033,
+      "eval_samples_per_second": 102.814,
+      "eval_steps_per_second": 12.852,
+      "step": 3560
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 7.794871794871795e-05,
+      "loss": 0.0007,
+      "step": 3570
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 7.760683760683761e-05,
+      "loss": 0.0008,
+      "step": 3580
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.726495726495727e-05,
+      "loss": 0.0008,
+      "step": 3590
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.692307692307693e-05,
+      "loss": 0.0014,
+      "step": 3600
+    },
+    {
+      "epoch": 9.23,
+      "eval_accuracy": 0.7946127946127947,
+      "eval_loss": 1.1410149335861206,
+      "eval_runtime": 36.6565,
+      "eval_samples_per_second": 97.336,
+      "eval_steps_per_second": 12.167,
+      "step": 3600
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.65811965811966e-05,
+      "loss": 0.021,
+      "step": 3610
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.623931623931624e-05,
+      "loss": 0.0006,
+      "step": 3620
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 7.58974358974359e-05,
+      "loss": 0.0007,
+      "step": 3630
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 7.555555555555556e-05,
+      "loss": 0.0006,
+      "step": 3640
+    },
+    {
+      "epoch": 9.33,
+      "eval_accuracy": 0.8103254769921436,
+      "eval_loss": 1.0399866104125977,
+      "eval_runtime": 33.4528,
+      "eval_samples_per_second": 106.658,
+      "eval_steps_per_second": 13.332,
+      "step": 3640
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 7.521367521367521e-05,
+      "loss": 0.0006,
+      "step": 3650
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 7.487179487179487e-05,
+      "loss": 0.0009,
+      "step": 3660
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 7.452991452991453e-05,
+      "loss": 0.0368,
+      "step": 3670
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 7.41880341880342e-05,
+      "loss": 0.0063,
+      "step": 3680
+    },
+    {
+      "epoch": 9.44,
+      "eval_accuracy": 0.8058361391694725,
+      "eval_loss": 1.107193112373352,
+      "eval_runtime": 41.6725,
+      "eval_samples_per_second": 85.62,
+      "eval_steps_per_second": 10.702,
+      "step": 3680
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 7.384615384615386e-05,
+      "loss": 0.0207,
+      "step": 3690
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 7.350427350427352e-05,
+      "loss": 0.0006,
+      "step": 3700
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 7.316239316239316e-05,
+      "loss": 0.0025,
+      "step": 3710
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 7.282051282051282e-05,
+      "loss": 0.0006,
+      "step": 3720
+    },
+    {
+      "epoch": 9.54,
+      "eval_accuracy": 0.8013468013468014,
+      "eval_loss": 1.1391396522521973,
+      "eval_runtime": 32.0452,
+      "eval_samples_per_second": 111.343,
+      "eval_steps_per_second": 13.918,
+      "step": 3720
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 7.247863247863248e-05,
+      "loss": 0.0031,
+      "step": 3730
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 7.213675213675213e-05,
+      "loss": 0.001,
+      "step": 3740
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 7.17948717948718e-05,
+      "loss": 0.0297,
+      "step": 3750
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 7.145299145299146e-05,
+      "loss": 0.0007,
+      "step": 3760
+    },
+    {
+      "epoch": 9.64,
+      "eval_accuracy": 0.8013468013468014,
+      "eval_loss": 1.124064326286316,
+      "eval_runtime": 32.9836,
+      "eval_samples_per_second": 108.175,
+      "eval_steps_per_second": 13.522,
+      "step": 3760
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 7.111111111111112e-05,
+      "loss": 0.0006,
+      "step": 3770
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 7.076923076923078e-05,
+      "loss": 0.0006,
+      "step": 3780
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 7.042735042735044e-05,
+      "loss": 0.0216,
+      "step": 3790
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.008547008547008e-05,
+      "loss": 0.0006,
+      "step": 3800
+    },
+    {
+      "epoch": 9.74,
+      "eval_accuracy": 0.7789001122334456,
+      "eval_loss": 1.2622841596603394,
+      "eval_runtime": 34.3515,
+      "eval_samples_per_second": 103.867,
+      "eval_steps_per_second": 12.983,
+      "step": 3800
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.974358974358974e-05,
+      "loss": 0.0007,
+      "step": 3810
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.940170940170941e-05,
+      "loss": 0.0015,
+      "step": 3820
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 6.905982905982907e-05,
+      "loss": 0.0009,
+      "step": 3830
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 6.871794871794872e-05,
+      "loss": 0.0006,
+      "step": 3840
+    },
+    {
+      "epoch": 9.85,
+      "eval_accuracy": 0.7991021324354658,
+      "eval_loss": 1.1737140417099,
+      "eval_runtime": 43.7608,
+      "eval_samples_per_second": 81.534,
+      "eval_steps_per_second": 10.192,
+      "step": 3840
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 6.837606837606838e-05,
+      "loss": 0.0005,
+      "step": 3850
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 6.803418803418804e-05,
+      "loss": 0.0014,
+      "step": 3860
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 6.76923076923077e-05,
+      "loss": 0.0006,
+      "step": 3870
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 6.735042735042735e-05,
+      "loss": 0.0011,
+      "step": 3880
+    },
+    {
+      "epoch": 9.95,
+      "eval_accuracy": 0.7884399551066218,
+      "eval_loss": 1.2243778705596924,
+      "eval_runtime": 36.7265,
+      "eval_samples_per_second": 97.15,
+      "eval_steps_per_second": 12.144,
+      "step": 3880
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 6.700854700854701e-05,
+      "loss": 0.0015,
+      "step": 3890
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.0117,
+      "step": 3900
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 6.632478632478633e-05,
+      "loss": 0.0007,
+      "step": 3910
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 6.598290598290599e-05,
+      "loss": 0.0047,
+      "step": 3920
+    },
+    {
+      "epoch": 10.05,
+      "eval_accuracy": 0.8002244668911336,
+      "eval_loss": 1.185262680053711,
+      "eval_runtime": 32.7462,
+      "eval_samples_per_second": 108.959,
+      "eval_steps_per_second": 13.62,
+      "step": 3920
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 6.564102564102564e-05,
+      "loss": 0.0006,
+      "step": 3930
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 6.52991452991453e-05,
+      "loss": 0.0288,
+      "step": 3940
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 6.495726495726496e-05,
+      "loss": 0.0005,
+      "step": 3950
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 6.461538461538462e-05,
+      "loss": 0.0018,
+      "step": 3960
+    },
+    {
+      "epoch": 10.15,
+      "eval_accuracy": 0.8092031425364759,
+      "eval_loss": 1.1372754573822021,
+      "eval_runtime": 34.7934,
+      "eval_samples_per_second": 102.548,
+      "eval_steps_per_second": 12.819,
+      "step": 3960
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 6.427350427350429e-05,
+      "loss": 0.0005,
+      "step": 3970
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 6.393162393162393e-05,
+      "loss": 0.0005,
+      "step": 3980
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 6.358974358974359e-05,
+      "loss": 0.0005,
+      "step": 3990
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 6.324786324786325e-05,
+      "loss": 0.0006,
+      "step": 4000
+    },
+    {
+      "epoch": 10.26,
+      "eval_accuracy": 0.8069584736251403,
+      "eval_loss": 1.1416363716125488,
+      "eval_runtime": 33.7666,
+      "eval_samples_per_second": 105.667,
+      "eval_steps_per_second": 13.208,
+      "step": 4000
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 6.29059829059829e-05,
+      "loss": 0.0005,
+      "step": 4010
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 6.256410256410256e-05,
+      "loss": 0.012,
+      "step": 4020
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 6.222222222222222e-05,
+      "loss": 0.0013,
+      "step": 4030
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 6.188034188034189e-05,
+      "loss": 0.007,
+      "step": 4040
+    },
+    {
+      "epoch": 10.36,
+      "eval_accuracy": 0.8047138047138047,
+      "eval_loss": 1.205945372581482,
+      "eval_runtime": 32.5065,
+      "eval_samples_per_second": 109.763,
+      "eval_steps_per_second": 13.72,
+      "step": 4040
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 6.153846153846155e-05,
+      "loss": 0.0005,
+      "step": 4050
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 6.11965811965812e-05,
+      "loss": 0.0096,
+      "step": 4060
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 6.085470085470085e-05,
+      "loss": 0.0008,
+      "step": 4070
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 6.0512820512820515e-05,
+      "loss": 0.0005,
+      "step": 4080
+    },
+    {
+      "epoch": 10.46,
+      "eval_accuracy": 0.8170594837261503,
+      "eval_loss": 1.131420612335205,
+      "eval_runtime": 36.6164,
+      "eval_samples_per_second": 97.443,
+      "eval_steps_per_second": 12.18,
+      "step": 4080
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 6.017094017094017e-05,
+      "loss": 0.0005,
+      "step": 4090
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 5.982905982905983e-05,
+      "loss": 0.0006,
+      "step": 4100
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 5.948717948717949e-05,
+      "loss": 0.0005,
+      "step": 4110
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 5.9145299145299146e-05,
+      "loss": 0.0005,
+      "step": 4120
+    },
+    {
+      "epoch": 10.56,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.1375317573547363,
+      "eval_runtime": 32.4301,
+      "eval_samples_per_second": 110.021,
+      "eval_steps_per_second": 13.753,
+      "step": 4120
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 5.880341880341881e-05,
+      "loss": 0.0004,
+      "step": 4130
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 5.846153846153847e-05,
+      "loss": 0.0005,
+      "step": 4140
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 5.8119658119658126e-05,
+      "loss": 0.0005,
+      "step": 4150
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 5.7777777777777776e-05,
+      "loss": 0.0004,
+      "step": 4160
+    },
+    {
+      "epoch": 10.67,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.136337161064148,
+      "eval_runtime": 31.869,
+      "eval_samples_per_second": 111.959,
+      "eval_steps_per_second": 13.995,
+      "step": 4160
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 5.7435897435897434e-05,
+      "loss": 0.0004,
+      "step": 4170
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 5.709401709401709e-05,
+      "loss": 0.0004,
+      "step": 4180
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 5.6752136752136756e-05,
+      "loss": 0.0005,
+      "step": 4190
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 5.6410256410256414e-05,
+      "loss": 0.0004,
+      "step": 4200
+    },
+    {
+      "epoch": 10.77,
+      "eval_accuracy": 0.819304152637486,
+      "eval_loss": 1.1374398469924927,
+      "eval_runtime": 34.7543,
+      "eval_samples_per_second": 102.664,
+      "eval_steps_per_second": 12.833,
+      "step": 4200
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 5.606837606837607e-05,
+      "loss": 0.0007,
+      "step": 4210
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 5.572649572649573e-05,
+      "loss": 0.0004,
+      "step": 4220
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 5.538461538461539e-05,
+      "loss": 0.0005,
+      "step": 4230
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 5.504273504273505e-05,
+      "loss": 0.0009,
+      "step": 4240
+    },
+    {
+      "epoch": 10.87,
+      "eval_accuracy": 0.8069584736251403,
+      "eval_loss": 1.1739455461502075,
+      "eval_runtime": 33.5639,
+      "eval_samples_per_second": 106.305,
+      "eval_steps_per_second": 13.288,
+      "step": 4240
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 5.470085470085471e-05,
+      "loss": 0.0005,
+      "step": 4250
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 5.435897435897436e-05,
+      "loss": 0.0005,
+      "step": 4260
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 5.401709401709402e-05,
+      "loss": 0.0004,
+      "step": 4270
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 5.3675213675213675e-05,
+      "loss": 0.0004,
+      "step": 4280
+    },
+    {
+      "epoch": 10.97,
+      "eval_accuracy": 0.8131313131313131,
+      "eval_loss": 1.145166277885437,
+      "eval_runtime": 32.5186,
+      "eval_samples_per_second": 109.722,
+      "eval_steps_per_second": 13.715,
+      "step": 4280
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 0.0004,
+      "step": 4290
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 5.2991452991453e-05,
+      "loss": 0.0004,
+      "step": 4300
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 5.2649572649572655e-05,
+      "loss": 0.0011,
+      "step": 4310
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 5.230769230769231e-05,
+      "loss": 0.0004,
+      "step": 4320
+    },
+    {
+      "epoch": 11.08,
+      "eval_accuracy": 0.797979797979798,
+      "eval_loss": 1.2409276962280273,
+      "eval_runtime": 32.9205,
+      "eval_samples_per_second": 108.382,
+      "eval_steps_per_second": 13.548,
+      "step": 4320
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 5.196581196581197e-05,
+      "loss": 0.0005,
+      "step": 4330
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 5.162393162393163e-05,
+      "loss": 0.0004,
+      "step": 4340
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 5.128205128205128e-05,
+      "loss": 0.0074,
+      "step": 4350
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 5.094017094017094e-05,
+      "loss": 0.0004,
+      "step": 4360
+    },
+    {
+      "epoch": 11.18,
+      "eval_accuracy": 0.8063973063973064,
+      "eval_loss": 1.1721081733703613,
+      "eval_runtime": 33.0524,
+      "eval_samples_per_second": 107.95,
+      "eval_steps_per_second": 13.494,
+      "step": 4360
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 5.05982905982906e-05,
+      "loss": 0.0004,
+      "step": 4370
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 5.025641025641026e-05,
+      "loss": 0.0004,
+      "step": 4380
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 4.991452991452992e-05,
+      "loss": 0.0021,
+      "step": 4390
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 4.9572649572649575e-05,
+      "loss": 0.0004,
+      "step": 4400
+    },
+    {
+      "epoch": 11.28,
+      "eval_accuracy": 0.8041526374859708,
+      "eval_loss": 1.195190191268921,
+      "eval_runtime": 34.4526,
+      "eval_samples_per_second": 103.563,
+      "eval_steps_per_second": 12.945,
+      "step": 4400
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 4.923076923076924e-05,
+      "loss": 0.0004,
+      "step": 4410
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 0.0004,
+      "step": 4420
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 4.854700854700855e-05,
+      "loss": 0.0004,
+      "step": 4430
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 4.8205128205128205e-05,
+      "loss": 0.0004,
+      "step": 4440
+    },
+    {
+      "epoch": 11.38,
+      "eval_accuracy": 0.8024691358024691,
+      "eval_loss": 1.2087780237197876,
+      "eval_runtime": 37.4802,
+      "eval_samples_per_second": 95.197,
+      "eval_steps_per_second": 11.9,
+      "step": 4440
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 4.786324786324787e-05,
+      "loss": 0.0004,
+      "step": 4450
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 4.752136752136752e-05,
+      "loss": 0.0023,
+      "step": 4460
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 4.717948717948718e-05,
+      "loss": 0.0004,
+      "step": 4470
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 4.683760683760684e-05,
+      "loss": 0.0004,
+      "step": 4480
+    },
+    {
+      "epoch": 11.49,
+      "eval_accuracy": 0.8170594837261503,
+      "eval_loss": 1.1391326189041138,
+      "eval_runtime": 33.734,
+      "eval_samples_per_second": 105.769,
+      "eval_steps_per_second": 13.221,
+      "step": 4480
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 4.64957264957265e-05,
+      "loss": 0.0004,
+      "step": 4490
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 4.615384615384616e-05,
+      "loss": 0.0004,
+      "step": 4500
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 4.581196581196581e-05,
+      "loss": 0.0007,
+      "step": 4510
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 4.5470085470085474e-05,
+      "loss": 0.0004,
+      "step": 4520
+    },
+    {
+      "epoch": 11.59,
+      "eval_accuracy": 0.8148148148148148,
+      "eval_loss": 1.1593190431594849,
+      "eval_runtime": 32.1112,
+      "eval_samples_per_second": 111.114,
+      "eval_steps_per_second": 13.889,
+      "step": 4520
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 4.512820512820513e-05,
+      "loss": 0.0004,
+      "step": 4530
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 4.478632478632479e-05,
+      "loss": 0.0004,
+      "step": 4540
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.0003,
+      "step": 4550
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 4.4102564102564104e-05,
+      "loss": 0.0004,
+      "step": 4560
+    },
+    {
+      "epoch": 11.69,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1428256034851074,
+      "eval_runtime": 34.6547,
+      "eval_samples_per_second": 102.959,
+      "eval_steps_per_second": 12.87,
+      "step": 4560
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 4.376068376068376e-05,
+      "loss": 0.0004,
+      "step": 4570
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 4.341880341880342e-05,
+      "loss": 0.0003,
+      "step": 4580
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 4.3076923076923084e-05,
+      "loss": 0.0003,
+      "step": 4590
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 4.2735042735042735e-05,
+      "loss": 0.0003,
+      "step": 4600
+    },
+    {
+      "epoch": 11.79,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1417738199234009,
+      "eval_runtime": 33.7453,
+      "eval_samples_per_second": 105.733,
+      "eval_steps_per_second": 13.217,
+      "step": 4600
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 4.239316239316239e-05,
+      "loss": 0.0003,
+      "step": 4610
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 4.205128205128206e-05,
+      "loss": 0.0003,
+      "step": 4620
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 4.1709401709401715e-05,
+      "loss": 0.0003,
+      "step": 4630
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 4.1367521367521366e-05,
+      "loss": 0.0003,
+      "step": 4640
+    },
+    {
+      "epoch": 11.9,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1414417028427124,
+      "eval_runtime": 31.736,
+      "eval_samples_per_second": 112.427,
+      "eval_steps_per_second": 14.053,
+      "step": 4640
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 4.1025641025641023e-05,
+      "loss": 0.0003,
+      "step": 4650
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 4.068376068376069e-05,
+      "loss": 0.0003,
+      "step": 4660
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 4.0341880341880346e-05,
+      "loss": 0.0005,
+      "step": 4670
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 4e-05,
+      "loss": 0.0006,
+      "step": 4680
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.8215488215488216,
+      "eval_loss": 1.160527229309082,
+      "eval_runtime": 32.9127,
+      "eval_samples_per_second": 108.408,
+      "eval_steps_per_second": 13.551,
+      "step": 4680
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 3.965811965811966e-05,
+      "loss": 0.0003,
+      "step": 4690
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 3.931623931623932e-05,
+      "loss": 0.0004,
+      "step": 4700
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 3.8974358974358976e-05,
+      "loss": 0.0003,
+      "step": 4710
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 3.8632478632478634e-05,
+      "loss": 0.0003,
+      "step": 4720
+    },
+    {
+      "epoch": 12.1,
+      "eval_accuracy": 0.8080808080808081,
+      "eval_loss": 1.209052562713623,
+      "eval_runtime": 32.6946,
+      "eval_samples_per_second": 109.131,
+      "eval_steps_per_second": 13.641,
+      "step": 4720
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 3.82905982905983e-05,
+      "loss": 0.0004,
+      "step": 4730
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 3.794871794871795e-05,
+      "loss": 0.0003,
+      "step": 4740
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 3.760683760683761e-05,
+      "loss": 0.0004,
+      "step": 4750
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 3.7264957264957265e-05,
+      "loss": 0.0005,
+      "step": 4760
+    },
+    {
+      "epoch": 12.21,
+      "eval_accuracy": 0.8232323232323232,
+      "eval_loss": 1.1692516803741455,
+      "eval_runtime": 33.4085,
+      "eval_samples_per_second": 106.799,
+      "eval_steps_per_second": 13.35,
+      "step": 4760
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 3.692307692307693e-05,
+      "loss": 0.0003,
+      "step": 4770
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 3.658119658119658e-05,
+      "loss": 0.0003,
+      "step": 4780
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 3.623931623931624e-05,
+      "loss": 0.0004,
+      "step": 4790
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 3.58974358974359e-05,
+      "loss": 0.0003,
+      "step": 4800
+    },
+    {
+      "epoch": 12.31,
+      "eval_accuracy": 0.8215488215488216,
+      "eval_loss": 1.152122974395752,
+      "eval_runtime": 34.5292,
+      "eval_samples_per_second": 103.333,
+      "eval_steps_per_second": 12.917,
+      "step": 4800
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 3.555555555555556e-05,
+      "loss": 0.0005,
+      "step": 4810
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 3.521367521367522e-05,
+      "loss": 0.0003,
+      "step": 4820
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 3.487179487179487e-05,
+      "loss": 0.0003,
+      "step": 4830
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 3.452991452991453e-05,
+      "loss": 0.0003,
+      "step": 4840
+    },
+    {
+      "epoch": 12.41,
+      "eval_accuracy": 0.8125701459034792,
+      "eval_loss": 1.1672531366348267,
+      "eval_runtime": 32.9525,
+      "eval_samples_per_second": 108.277,
+      "eval_steps_per_second": 13.535,
+      "step": 4840
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 3.418803418803419e-05,
+      "loss": 0.0003,
+      "step": 4850
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 3.384615384615385e-05,
+      "loss": 0.0003,
+      "step": 4860
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 3.3504273504273506e-05,
+      "loss": 0.0003,
+      "step": 4870
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 3.3162393162393164e-05,
+      "loss": 0.0003,
+      "step": 4880
+    },
+    {
+      "epoch": 12.51,
+      "eval_accuracy": 0.8125701459034792,
+      "eval_loss": 1.1674937009811401,
+      "eval_runtime": 34.1043,
+      "eval_samples_per_second": 104.62,
+      "eval_steps_per_second": 13.078,
+      "step": 4880
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 3.282051282051282e-05,
+      "loss": 0.0003,
+      "step": 4890
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 3.247863247863248e-05,
+      "loss": 0.0003,
+      "step": 4900
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 3.2136752136752144e-05,
+      "loss": 0.0003,
+      "step": 4910
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 3.1794871794871795e-05,
+      "loss": 0.0003,
+      "step": 4920
+    },
+    {
+      "epoch": 12.62,
+      "eval_accuracy": 0.8148148148148148,
+      "eval_loss": 1.1658987998962402,
+      "eval_runtime": 34.5805,
+      "eval_samples_per_second": 103.179,
+      "eval_steps_per_second": 12.897,
+      "step": 4920
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 3.145299145299145e-05,
+      "loss": 0.0003,
+      "step": 4930
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 0.0003,
+      "step": 4940
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 3.0769230769230774e-05,
+      "loss": 0.0003,
+      "step": 4950
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 3.0427350427350425e-05,
+      "loss": 0.0003,
+      "step": 4960
+    },
+    {
+      "epoch": 12.72,
+      "eval_accuracy": 0.8142536475869809,
+      "eval_loss": 1.1660560369491577,
+      "eval_runtime": 33.7463,
+      "eval_samples_per_second": 105.73,
+      "eval_steps_per_second": 13.216,
+      "step": 4960
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 3.0085470085470086e-05,
+      "loss": 0.0003,
+      "step": 4970
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.9743589743589744e-05,
+      "loss": 0.0003,
+      "step": 4980
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.9401709401709405e-05,
+      "loss": 0.0003,
+      "step": 4990
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.9059829059829063e-05,
+      "loss": 0.0003,
+      "step": 5000
+    },
+    {
+      "epoch": 12.82,
+      "eval_accuracy": 0.8159371492704826,
+      "eval_loss": 1.1644172668457031,
+      "eval_runtime": 32.5332,
+      "eval_samples_per_second": 109.672,
+      "eval_steps_per_second": 13.709,
+      "step": 5000
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.8717948717948717e-05,
+      "loss": 0.0003,
+      "step": 5010
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.8376068376068378e-05,
+      "loss": 0.0003,
+      "step": 5020
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.8034188034188036e-05,
+      "loss": 0.0003,
+      "step": 5030
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.7692307692307694e-05,
+      "loss": 0.0003,
+      "step": 5040
+    },
+    {
+      "epoch": 12.92,
+      "eval_accuracy": 0.8148148148148148,
+      "eval_loss": 1.1629841327667236,
+      "eval_runtime": 34.2224,
+      "eval_samples_per_second": 104.259,
+      "eval_steps_per_second": 13.032,
+      "step": 5040
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.7350427350427355e-05,
+      "loss": 0.0003,
+      "step": 5050
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.700854700854701e-05,
+      "loss": 0.0003,
+      "step": 5060
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 0.0003,
+      "step": 5070
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.6324786324786328e-05,
+      "loss": 0.0003,
+      "step": 5080
+    },
+    {
+      "epoch": 13.03,
+      "eval_accuracy": 0.8159371492704826,
+      "eval_loss": 1.1632208824157715,
+      "eval_runtime": 41.2446,
+      "eval_samples_per_second": 86.508,
+      "eval_steps_per_second": 10.814,
+      "step": 5080
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.5982905982905985e-05,
+      "loss": 0.0003,
+      "step": 5090
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.564102564102564e-05,
+      "loss": 0.0003,
+      "step": 5100
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.52991452991453e-05,
+      "loss": 0.0003,
+      "step": 5110
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.495726495726496e-05,
+      "loss": 0.0003,
+      "step": 5120
+    },
+    {
+      "epoch": 13.13,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.1635642051696777,
+      "eval_runtime": 33.779,
+      "eval_samples_per_second": 105.628,
+      "eval_steps_per_second": 13.203,
+      "step": 5120
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.461538461538462e-05,
+      "loss": 0.0003,
+      "step": 5130
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.4273504273504274e-05,
+      "loss": 0.0003,
+      "step": 5140
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.3931623931623935e-05,
+      "loss": 0.0003,
+      "step": 5150
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.358974358974359e-05,
+      "loss": 0.0003,
+      "step": 5160
+    },
+    {
+      "epoch": 13.23,
+      "eval_accuracy": 0.8170594837261503,
+      "eval_loss": 1.1645357608795166,
+      "eval_runtime": 34.4551,
+      "eval_samples_per_second": 103.555,
+      "eval_steps_per_second": 12.944,
+      "step": 5160
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.324786324786325e-05,
+      "loss": 0.0003,
+      "step": 5170
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2905982905982905e-05,
+      "loss": 0.0003,
+      "step": 5180
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2564102564102566e-05,
+      "loss": 0.0003,
+      "step": 5190
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.0003,
+      "step": 5200
+    },
+    {
+      "epoch": 13.33,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.1655912399291992,
+      "eval_runtime": 33.1731,
+      "eval_samples_per_second": 107.557,
+      "eval_steps_per_second": 13.445,
+      "step": 5200
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.188034188034188e-05,
+      "loss": 0.0003,
+      "step": 5210
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.1538461538461542e-05,
+      "loss": 0.0003,
+      "step": 5220
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.1196581196581196e-05,
+      "loss": 0.0003,
+      "step": 5230
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.0854700854700857e-05,
+      "loss": 0.0003,
+      "step": 5240
+    },
+    {
+      "epoch": 13.44,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.1652870178222656,
+      "eval_runtime": 32.1186,
+      "eval_samples_per_second": 111.088,
+      "eval_steps_per_second": 13.886,
+      "step": 5240
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.0512820512820512e-05,
+      "loss": 0.0003,
+      "step": 5250
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.0170940170940173e-05,
+      "loss": 0.0003,
+      "step": 5260
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.982905982905983e-05,
+      "loss": 0.0003,
+      "step": 5270
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.9487179487179488e-05,
+      "loss": 0.0003,
+      "step": 5280
+    },
+    {
+      "epoch": 13.54,
+      "eval_accuracy": 0.8164983164983165,
+      "eval_loss": 1.1658546924591064,
+      "eval_runtime": 35.35,
+      "eval_samples_per_second": 100.933,
+      "eval_steps_per_second": 12.617,
+      "step": 5280
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.914529914529915e-05,
+      "loss": 0.0003,
+      "step": 5290
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.8803418803418804e-05,
+      "loss": 0.0003,
+      "step": 5300
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.8461538461538465e-05,
+      "loss": 0.0003,
+      "step": 5310
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.811965811965812e-05,
+      "loss": 0.0003,
+      "step": 5320
+    },
+    {
+      "epoch": 13.64,
+      "eval_accuracy": 0.8170594837261503,
+      "eval_loss": 1.166174054145813,
+      "eval_runtime": 33.5318,
+      "eval_samples_per_second": 106.406,
+      "eval_steps_per_second": 13.301,
+      "step": 5320
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.777777777777778e-05,
+      "loss": 0.0003,
+      "step": 5330
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.7435897435897434e-05,
+      "loss": 0.0003,
+      "step": 5340
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.7094017094017095e-05,
+      "loss": 0.0003,
+      "step": 5350
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.6752136752136753e-05,
+      "loss": 0.0003,
+      "step": 5360
+    },
+    {
+      "epoch": 13.74,
+      "eval_accuracy": 0.8176206509539843,
+      "eval_loss": 1.1661781072616577,
+      "eval_runtime": 32.3616,
+      "eval_samples_per_second": 110.254,
+      "eval_steps_per_second": 13.782,
+      "step": 5360
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.641025641025641e-05,
+      "loss": 0.0003,
+      "step": 5370
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.6068376068376072e-05,
+      "loss": 0.0003,
+      "step": 5380
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5726495726495726e-05,
+      "loss": 0.0003,
+      "step": 5390
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5384615384615387e-05,
+      "loss": 0.0003,
+      "step": 5400
+    },
+    {
+      "epoch": 13.85,
+      "eval_accuracy": 0.8176206509539843,
+      "eval_loss": 1.1663531064987183,
+      "eval_runtime": 35.1534,
+      "eval_samples_per_second": 101.498,
+      "eval_steps_per_second": 12.687,
+      "step": 5400
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5042735042735043e-05,
+      "loss": 0.0003,
+      "step": 5410
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.4700854700854703e-05,
+      "loss": 0.0003,
+      "step": 5420
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.4358974358974359e-05,
+      "loss": 0.0003,
+      "step": 5430
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.4017094017094018e-05,
+      "loss": 0.0002,
+      "step": 5440
+    },
+    {
+      "epoch": 13.95,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.1670769453048706,
+      "eval_runtime": 33.5689,
+      "eval_samples_per_second": 106.289,
+      "eval_steps_per_second": 13.286,
+      "step": 5440
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.3675213675213677e-05,
+      "loss": 0.0003,
+      "step": 5450
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.0003,
+      "step": 5460
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.2991452991452993e-05,
+      "loss": 0.0003,
+      "step": 5470
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.264957264957265e-05,
+      "loss": 0.0003,
+      "step": 5480
+    },
+    {
+      "epoch": 14.05,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.1676580905914307,
+      "eval_runtime": 42.4697,
+      "eval_samples_per_second": 84.013,
+      "eval_steps_per_second": 10.502,
+      "step": 5480
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.230769230769231e-05,
+      "loss": 0.0003,
+      "step": 5490
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.1965811965811967e-05,
+      "loss": 0.0003,
+      "step": 5500
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.1623931623931625e-05,
+      "loss": 0.0003,
+      "step": 5510
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.1282051282051283e-05,
+      "loss": 0.0003,
+      "step": 5520
+    },
+    {
+      "epoch": 14.15,
+      "eval_accuracy": 0.8181818181818182,
+      "eval_loss": 1.1681442260742188,
+      "eval_runtime": 33.5678,
+      "eval_samples_per_second": 106.292,
+      "eval_steps_per_second": 13.287,
+      "step": 5520
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.094017094017094e-05,
+      "loss": 0.0003,
+      "step": 5530
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.0598290598290598e-05,
+      "loss": 0.0003,
+      "step": 5540
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.0256410256410256e-05,
+      "loss": 0.0003,
+      "step": 5550
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.914529914529915e-06,
+      "loss": 0.0003,
+      "step": 5560
+    },
+    {
+      "epoch": 14.26,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1684342622756958,
+      "eval_runtime": 32.4759,
+      "eval_samples_per_second": 109.866,
+      "eval_steps_per_second": 13.733,
+      "step": 5560
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.572649572649575e-06,
+      "loss": 0.0003,
+      "step": 5570
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 0.0003,
+      "step": 5580
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 8.88888888888889e-06,
+      "loss": 0.0003,
+      "step": 5590
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.547008547008548e-06,
+      "loss": 0.0003,
+      "step": 5600
+    },
+    {
+      "epoch": 14.36,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1688289642333984,
+      "eval_runtime": 32.3951,
+      "eval_samples_per_second": 110.14,
+      "eval_steps_per_second": 13.768,
+      "step": 5600
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 8.205128205128205e-06,
+      "loss": 0.0002,
+      "step": 5610
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 7.863247863247863e-06,
+      "loss": 0.0002,
+      "step": 5620
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 7.521367521367522e-06,
+      "loss": 0.0003,
+      "step": 5630
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 7.179487179487179e-06,
+      "loss": 0.0003,
+      "step": 5640
+    },
+    {
+      "epoch": 14.46,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.168882131576538,
+      "eval_runtime": 31.9999,
+      "eval_samples_per_second": 111.5,
+      "eval_steps_per_second": 13.938,
+      "step": 5640
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 6.837606837606839e-06,
+      "loss": 0.0003,
+      "step": 5650
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 6.495726495726496e-06,
+      "loss": 0.0003,
+      "step": 5660
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 6.153846153846155e-06,
+      "loss": 0.0003,
+      "step": 5670
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 5.8119658119658126e-06,
+      "loss": 0.0002,
+      "step": 5680
+    },
+    {
+      "epoch": 14.56,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.1692728996276855,
+      "eval_runtime": 33.4951,
+      "eval_samples_per_second": 106.523,
+      "eval_steps_per_second": 13.315,
+      "step": 5680
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 5.47008547008547e-06,
+      "loss": 0.0002,
+      "step": 5690
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 5.128205128205128e-06,
+      "loss": 0.0003,
+      "step": 5700
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 4.786324786324787e-06,
+      "loss": 0.0002,
+      "step": 5710
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 4.444444444444445e-06,
+      "loss": 0.0002,
+      "step": 5720
+    },
+    {
+      "epoch": 14.67,
+      "eval_accuracy": 0.8187429854096521,
+      "eval_loss": 1.169299602508545,
+      "eval_runtime": 32.5001,
+      "eval_samples_per_second": 109.784,
+      "eval_steps_per_second": 13.723,
+      "step": 5720
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.102564102564103e-06,
+      "loss": 0.0003,
+      "step": 5730
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 3.760683760683761e-06,
+      "loss": 0.0003,
+      "step": 5740
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 3.4188034188034193e-06,
+      "loss": 0.0003,
+      "step": 5750
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 3.0769230769230774e-06,
+      "loss": 0.0002,
+      "step": 5760
+    },
+    {
+      "epoch": 14.77,
+      "eval_accuracy": 0.819304152637486,
+      "eval_loss": 1.1695406436920166,
+      "eval_runtime": 33.5775,
+      "eval_samples_per_second": 106.262,
+      "eval_steps_per_second": 13.283,
+      "step": 5760
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.735042735042735e-06,
+      "loss": 0.0003,
+      "step": 5770
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.3931623931623937e-06,
+      "loss": 0.0002,
+      "step": 5780
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.0512820512820513e-06,
+      "loss": 0.0003,
+      "step": 5790
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.7094017094017097e-06,
+      "loss": 0.0003,
+      "step": 5800
+    },
+    {
+      "epoch": 14.87,
+      "eval_accuracy": 0.819304152637486,
+      "eval_loss": 1.1696120500564575,
+      "eval_runtime": 34.2314,
+      "eval_samples_per_second": 104.232,
+      "eval_steps_per_second": 13.029,
+      "step": 5800
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.3675213675213676e-06,
+      "loss": 0.0003,
+      "step": 5810
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.0256410256410257e-06,
+      "loss": 0.0003,
+      "step": 5820
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 6.837606837606838e-07,
+      "loss": 0.0002,
+      "step": 5830
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 3.418803418803419e-07,
+      "loss": 0.0002,
+      "step": 5840
+    },
+    {
+      "epoch": 14.97,
+      "eval_accuracy": 0.819304152637486,
+      "eval_loss": 1.1696137189865112,
+      "eval_runtime": 43.2941,
+      "eval_samples_per_second": 82.413,
+      "eval_steps_per_second": 10.302,
+      "step": 5840
+    }
+  ],
+  "max_steps": 5850,
+  "num_train_epochs": 15,
+  "total_flos": 7.231346758696919e+18,
+  "trial_name": null,
+  "trial_params": null
+}