diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,9957 @@
+{
+  "best_metric": 0.833830714225769,
+  "best_model_checkpoint": "/home/nlplab12/Desktop/roberta-pretrain/ckpt/roberta/pretrain/medium/512/checkpoint-12420",
+  "epoch": 10.559424425487993,
+  "eval_steps": 90,
+  "global_step": 12420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.008500045535958229,
+      "grad_norm": 514775.65625,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 7.2556,
+      "step": 10
+    },
+    {
+      "epoch": 0.017000091071916457,
+      "grad_norm": 399649.5,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 6.7984,
+      "step": 20
+    },
+    {
+      "epoch": 0.025500136607874684,
+      "grad_norm": 137177.0,
+      "learning_rate": 1.25e-05,
+      "loss": 6.2165,
+      "step": 30
+    },
+    {
+      "epoch": 0.034000182143832915,
+      "grad_norm": 151481.765625,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 5.9275,
+      "step": 40
+    },
+    {
+      "epoch": 0.04250022767979114,
+      "grad_norm": 92555.8359375,
+      "learning_rate": 2.0833333333333333e-05,
+      "loss": 5.7498,
+      "step": 50
+    },
+    {
+      "epoch": 0.05100027321574937,
+      "grad_norm": 86679.46875,
+      "learning_rate": 2.5e-05,
+      "loss": 5.6096,
+      "step": 60
+    },
+    {
+      "epoch": 0.059500318751707595,
+      "grad_norm": 66003.234375,
+      "learning_rate": 2.9166666666666666e-05,
+      "loss": 5.4786,
+      "step": 70
+    },
+    {
+      "epoch": 0.06800036428766583,
+      "grad_norm": 57082.796875,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 5.3565,
+      "step": 80
+    },
+    {
+      "epoch": 0.07650040982362405,
+      "grad_norm": 46905.7734375,
+      "learning_rate": 3.75e-05,
+      "loss": 5.2382,
+      "step": 90
+    },
+    {
+      "epoch": 0.07650040982362405,
+      "eval_accuracy": 0.2240983525108359,
+      "eval_loss": 5.1665496826171875,
+      "eval_runtime": 1330.3439,
+      "eval_samples_per_second": 375.375,
+      "eval_steps_per_second": 1.303,
+      "step": 90
+    },
+    {
+      "epoch": 0.08500045535958228,
+      "grad_norm": 36661.265625,
+      "learning_rate": 4.1666666666666665e-05,
+      "loss": 5.1301,
+      "step": 100
+    },
+    {
+      "epoch": 0.09350050089554052,
+      "grad_norm": 28157.064453125,
+      "learning_rate": 4.5833333333333334e-05,
+      "loss": 5.0237,
+      "step": 110
+    },
+    {
+      "epoch": 0.10200054643149874,
+      "grad_norm": 31035.537109375,
+      "learning_rate": 5e-05,
+      "loss": 4.9291,
+      "step": 120
+    },
+    {
+      "epoch": 0.11050059196745697,
+      "grad_norm": 41497.1640625,
+      "learning_rate": 5.416666666666667e-05,
+      "loss": 4.8501,
+      "step": 130
+    },
+    {
+      "epoch": 0.11900063750341519,
+      "grad_norm": 44447.87890625,
+      "learning_rate": 5.833333333333333e-05,
+      "loss": 4.7938,
+      "step": 140
+    },
+    {
+      "epoch": 0.12750068303937342,
+      "grad_norm": 36388.1015625,
+      "learning_rate": 6.25e-05,
+      "loss": 4.7491,
+      "step": 150
+    },
+    {
+      "epoch": 0.13600072857533166,
+      "grad_norm": 24360.1875,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 4.7147,
+      "step": 160
+    },
+    {
+      "epoch": 0.1445007741112899,
+      "grad_norm": 19699.2734375,
+      "learning_rate": 7.083333333333334e-05,
+      "loss": 4.6829,
+      "step": 170
+    },
+    {
+      "epoch": 0.1530008196472481,
+      "grad_norm": 11622.908203125,
+      "learning_rate": 7.5e-05,
+      "loss": 4.6545,
+      "step": 180
+    },
+    {
+      "epoch": 0.1530008196472481,
+      "eval_accuracy": 0.284967956348924,
+      "eval_loss": 4.628236293792725,
+      "eval_runtime": 1276.531,
+      "eval_samples_per_second": 391.199,
+      "eval_steps_per_second": 1.358,
+      "step": 180
+    },
+    {
+      "epoch": 0.16150086518320633,
+      "grad_norm": 16312.888671875,
+      "learning_rate": 7.916666666666666e-05,
+      "loss": 4.6377,
+      "step": 190
+    },
+    {
+      "epoch": 0.17000091071916457,
+      "grad_norm": 37823.2890625,
+      "learning_rate": 8.333333333333333e-05,
+      "loss": 4.6244,
+      "step": 200
+    },
+    {
+      "epoch": 0.1785009562551228,
+      "grad_norm": 18776.1796875,
+      "learning_rate": 8.75e-05,
+      "loss": 4.6141,
+      "step": 210
+    },
+    {
+      "epoch": 0.18700100179108103,
+      "grad_norm": 34368.3359375,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 4.6043,
+      "step": 220
+    },
+    {
+      "epoch": 0.19550104732703924,
+      "grad_norm": 38324.95703125,
+      "learning_rate": 9.583333333333334e-05,
+      "loss": 4.5989,
+      "step": 230
+    },
+    {
+      "epoch": 0.20400109286299747,
+      "grad_norm": 35916.953125,
+      "learning_rate": 0.0001,
+      "loss": 4.5924,
+      "step": 240
+    },
+    {
+      "epoch": 0.2125011383989557,
+      "grad_norm": 34779.6328125,
+      "learning_rate": 0.00010416666666666667,
+      "loss": 4.5855,
+      "step": 250
+    },
+    {
+      "epoch": 0.22100118393491394,
+      "grad_norm": 38605.4921875,
+      "learning_rate": 0.00010833333333333334,
+      "loss": 4.5771,
+      "step": 260
+    },
+    {
+      "epoch": 0.22950122947087218,
+      "grad_norm": 17166.369140625,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 4.5704,
+      "step": 270
+    },
+    {
+      "epoch": 0.22950122947087218,
+      "eval_accuracy": 0.2895338322091062,
+      "eval_loss": 4.5623779296875,
+      "eval_runtime": 1273.631,
+      "eval_samples_per_second": 392.09,
+      "eval_steps_per_second": 1.361,
+      "step": 270
+    },
+    {
+      "epoch": 0.23800127500683038,
+      "grad_norm": 42357.046875,
+      "learning_rate": 0.00011666666666666667,
+      "loss": 4.5682,
+      "step": 280
+    },
+    {
+      "epoch": 0.24650132054278862,
+      "grad_norm": 20787.40625,
+      "learning_rate": 0.00012083333333333333,
+      "loss": 4.5646,
+      "step": 290
+    },
+    {
+      "epoch": 0.25500136607874685,
+      "grad_norm": 49082.19921875,
+      "learning_rate": 0.000125,
+      "loss": 4.559,
+      "step": 300
+    },
+    {
+      "epoch": 0.2635014116147051,
+      "grad_norm": 18796.533203125,
+      "learning_rate": 0.00012916666666666667,
+      "loss": 4.5534,
+      "step": 310
+    },
+    {
+      "epoch": 0.2720014571506633,
+      "grad_norm": 45862.125,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 4.5513,
+      "step": 320
+    },
+    {
+      "epoch": 0.28050150268662155,
+      "grad_norm": 27371.748046875,
+      "learning_rate": 0.0001375,
+      "loss": 4.5489,
+      "step": 330
+    },
+    {
+      "epoch": 0.2890015482225798,
+      "grad_norm": 26154.533203125,
+      "learning_rate": 0.00014166666666666668,
+      "loss": 4.5471,
+      "step": 340
+    },
+    {
+      "epoch": 0.297501593758538,
+      "grad_norm": 46019.3671875,
+      "learning_rate": 0.00014583333333333335,
+      "loss": 4.5453,
+      "step": 350
+    },
+    {
+      "epoch": 0.3060016392944962,
+      "grad_norm": 54414.8828125,
+      "learning_rate": 0.00015,
+      "loss": 4.5435,
+      "step": 360
+    },
+    {
+      "epoch": 0.3060016392944962,
+      "eval_accuracy": 0.28996644531895915,
+      "eval_loss": 4.538370609283447,
+      "eval_runtime": 1274.2161,
+      "eval_samples_per_second": 391.91,
+      "eval_steps_per_second": 1.361,
+      "step": 360
+    },
+    {
+      "epoch": 0.31450168483045443,
+      "grad_norm": 61712.6640625,
+      "learning_rate": 0.00015416666666666668,
+      "loss": 4.5387,
+      "step": 370
+    },
+    {
+      "epoch": 0.32300173036641266,
+      "grad_norm": 35200.015625,
+      "learning_rate": 0.00015833333333333332,
+      "loss": 4.537,
+      "step": 380
+    },
+    {
+      "epoch": 0.3315017759023709,
+      "grad_norm": 40758.90234375,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 4.5345,
+      "step": 390
+    },
+    {
+      "epoch": 0.34000182143832913,
+      "grad_norm": 38277.3515625,
+      "learning_rate": 0.00016666666666666666,
+      "loss": 4.5351,
+      "step": 400
+    },
+    {
+      "epoch": 0.34850186697428737,
+      "grad_norm": 28720.861328125,
+      "learning_rate": 0.00017083333333333333,
+      "loss": 4.534,
+      "step": 410
+    },
+    {
+      "epoch": 0.3570019125102456,
+      "grad_norm": 21332.2109375,
+      "learning_rate": 0.000175,
+      "loss": 4.5329,
+      "step": 420
+    },
+    {
+      "epoch": 0.36550195804620383,
+      "grad_norm": 46977.23828125,
+      "learning_rate": 0.00017916666666666667,
+      "loss": 4.5282,
+      "step": 430
+    },
+    {
+      "epoch": 0.37400200358216207,
+      "grad_norm": 54403.328125,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 4.5273,
+      "step": 440
+    },
+    {
+      "epoch": 0.3825020491181203,
+      "grad_norm": 33038.54296875,
+      "learning_rate": 0.0001875,
+      "loss": 4.5272,
+      "step": 450
+    },
+    {
+      "epoch": 0.3825020491181203,
+      "eval_accuracy": 0.2900167539077763,
+      "eval_loss": 4.523510932922363,
+      "eval_runtime": 1270.6276,
+      "eval_samples_per_second": 393.017,
+      "eval_steps_per_second": 1.365,
+      "step": 450
+    },
+    {
+      "epoch": 0.3910020946540785,
+      "grad_norm": 38754.59765625,
+      "learning_rate": 0.00019166666666666667,
+      "loss": 4.5216,
+      "step": 460
+    },
+    {
+      "epoch": 0.3995021401900367,
+      "grad_norm": 36281.0078125,
+      "learning_rate": 0.00019583333333333334,
+      "loss": 4.5237,
+      "step": 470
+    },
+    {
+      "epoch": 0.40800218572599495,
+      "grad_norm": 35634.79296875,
+      "learning_rate": 0.0002,
+      "loss": 4.5186,
+      "step": 480
+    },
+    {
+      "epoch": 0.4165022312619532,
+      "grad_norm": 15257.716796875,
+      "learning_rate": 0.00020416666666666668,
+      "loss": 4.5173,
+      "step": 490
+    },
+    {
+      "epoch": 0.4250022767979114,
+      "grad_norm": 36190.26171875,
+      "learning_rate": 0.00020833333333333335,
+      "loss": 4.5174,
+      "step": 500
+    },
+    {
+      "epoch": 0.43350232233386965,
+      "grad_norm": 57666.21875,
+      "learning_rate": 0.0002125,
+      "loss": 4.5144,
+      "step": 510
+    },
+    {
+      "epoch": 0.4420023678698279,
+      "grad_norm": 33545.0078125,
+      "learning_rate": 0.00021666666666666668,
+      "loss": 4.5131,
+      "step": 520
+    },
+    {
+      "epoch": 0.4505024134057861,
+      "grad_norm": 75054.1328125,
+      "learning_rate": 0.00022083333333333333,
+      "loss": 4.5116,
+      "step": 530
+    },
+    {
+      "epoch": 0.45900245894174435,
+      "grad_norm": 27162.548828125,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 4.5104,
+      "step": 540
+    },
+    {
+      "epoch": 0.45900245894174435,
+      "eval_accuracy": 0.2903866574093662,
+      "eval_loss": 4.506768703460693,
+      "eval_runtime": 1274.9615,
+      "eval_samples_per_second": 391.681,
+      "eval_steps_per_second": 1.36,
+      "step": 540
+    },
+    {
+      "epoch": 0.4675025044777026,
+      "grad_norm": 31226.064453125,
+      "learning_rate": 0.00022916666666666666,
+      "loss": 4.5064,
+      "step": 550
+    },
+    {
+      "epoch": 0.47600255001366076,
+      "grad_norm": 34113.46484375,
+      "learning_rate": 0.00023333333333333333,
+      "loss": 4.5085,
+      "step": 560
+    },
+    {
+      "epoch": 0.484502595549619,
+      "grad_norm": 26731.498046875,
+      "learning_rate": 0.0002375,
+      "loss": 4.5079,
+      "step": 570
+    },
+    {
+      "epoch": 0.49300264108557723,
+      "grad_norm": 18876.388671875,
+      "learning_rate": 0.00024166666666666667,
+      "loss": 4.5041,
+      "step": 580
+    },
+    {
+      "epoch": 0.5015026866215355,
+      "grad_norm": 30968.685546875,
+      "learning_rate": 0.0002458333333333333,
+      "loss": 4.5044,
+      "step": 590
+    },
+    {
+      "epoch": 0.5100027321574937,
+      "grad_norm": 33946.8125,
+      "learning_rate": 0.00025,
+      "loss": 4.5036,
+      "step": 600
+    },
+    {
+      "epoch": 0.5185027776934519,
+      "grad_norm": 27673.900390625,
+      "learning_rate": 0.00025416666666666665,
+      "loss": 4.5007,
+      "step": 610
+    },
+    {
+      "epoch": 0.5270028232294102,
+      "grad_norm": 17412.22265625,
+      "learning_rate": 0.00025833333333333334,
+      "loss": 4.5003,
+      "step": 620
+    },
+    {
+      "epoch": 0.5355028687653683,
+      "grad_norm": 38210.4140625,
+      "learning_rate": 0.00026250000000000004,
+      "loss": 4.4977,
+      "step": 630
+    },
+    {
+      "epoch": 0.5355028687653683,
+      "eval_accuracy": 0.2904019042861853,
+      "eval_loss": 4.49587869644165,
+      "eval_runtime": 1274.4439,
+      "eval_samples_per_second": 391.84,
+      "eval_steps_per_second": 1.361,
+      "step": 630
+    },
+    {
+      "epoch": 0.5440029143013266,
+      "grad_norm": 22160.921875,
+      "learning_rate": 0.0002666666666666667,
+      "loss": 4.498,
+      "step": 640
+    },
+    {
+      "epoch": 0.5525029598372848,
+      "grad_norm": 40694.95703125,
+      "learning_rate": 0.0002708333333333333,
+      "loss": 4.4947,
+      "step": 650
+    },
+    {
+      "epoch": 0.5610030053732431,
+      "grad_norm": 22046.97265625,
+      "learning_rate": 0.000275,
+      "loss": 4.495,
+      "step": 660
+    },
+    {
+      "epoch": 0.5695030509092013,
+      "grad_norm": 45871.296875,
+      "learning_rate": 0.00027916666666666666,
+      "loss": 4.4946,
+      "step": 670
+    },
+    {
+      "epoch": 0.5780030964451596,
+      "grad_norm": 32230.837890625,
+      "learning_rate": 0.00028333333333333335,
+      "loss": 4.4941,
+      "step": 680
+    },
+    {
+      "epoch": 0.5865031419811177,
+      "grad_norm": 31339.076171875,
+      "learning_rate": 0.0002875,
+      "loss": 4.4921,
+      "step": 690
+    },
+    {
+      "epoch": 0.595003187517076,
+      "grad_norm": 21844.26171875,
+      "learning_rate": 0.0002916666666666667,
+      "loss": 4.4912,
+      "step": 700
+    },
+    {
+      "epoch": 0.6035032330530342,
+      "grad_norm": 20966.12109375,
+      "learning_rate": 0.00029583333333333333,
+      "loss": 4.489,
+      "step": 710
+    },
+    {
+      "epoch": 0.6120032785889924,
+      "grad_norm": 19211.65625,
+      "learning_rate": 0.0003,
+      "loss": 4.4903,
+      "step": 720
+    },
+    {
+      "epoch": 0.6120032785889924,
+      "eval_accuracy": 0.2904662633238704,
+      "eval_loss": 4.487661361694336,
+      "eval_runtime": 1273.4031,
+      "eval_samples_per_second": 392.16,
+      "eval_steps_per_second": 1.362,
+      "step": 720
+    },
+    {
+      "epoch": 0.6205033241249507,
+      "grad_norm": 47674.4140625,
+      "learning_rate": 0.00030416666666666667,
+      "loss": 4.4886,
+      "step": 730
+    },
+    {
+      "epoch": 0.6290033696609089,
+      "grad_norm": 44354.7890625,
+      "learning_rate": 0.00030833333333333337,
+      "loss": 4.4865,
+      "step": 740
+    },
+    {
+      "epoch": 0.6375034151968672,
+      "grad_norm": 27939.216796875,
+      "learning_rate": 0.0003125,
+      "loss": 4.4868,
+      "step": 750
+    },
+    {
+      "epoch": 0.6460034607328253,
+      "grad_norm": 29380.662109375,
+      "learning_rate": 0.00031666666666666665,
+      "loss": 4.4858,
+      "step": 760
+    },
+    {
+      "epoch": 0.6545035062687836,
+      "grad_norm": 38158.73046875,
+      "learning_rate": 0.00032083333333333334,
+      "loss": 4.4834,
+      "step": 770
+    },
+    {
+      "epoch": 0.6630035518047418,
+      "grad_norm": 40741.23828125,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 4.4826,
+      "step": 780
+    },
+    {
+      "epoch": 0.6715035973407001,
+      "grad_norm": 25295.03515625,
+      "learning_rate": 0.0003291666666666667,
+      "loss": 4.4845,
+      "step": 790
+    },
+    {
+      "epoch": 0.6800036428766583,
+      "grad_norm": 49988.22265625,
+      "learning_rate": 0.0003333333333333333,
+      "loss": 4.4814,
+      "step": 800
+    },
+    {
+      "epoch": 0.6885036884126164,
+      "grad_norm": 32892.44921875,
+      "learning_rate": 0.0003375,
+      "loss": 4.4826,
+      "step": 810
+    },
+    {
+      "epoch": 0.6885036884126164,
+      "eval_accuracy": 0.29046815360567907,
+      "eval_loss": 4.479104518890381,
+      "eval_runtime": 1276.8746,
+      "eval_samples_per_second": 391.094,
+      "eval_steps_per_second": 1.358,
+      "step": 810
+    },
+    {
+      "epoch": 0.6970037339485747,
+      "grad_norm": 32854.1015625,
+      "learning_rate": 0.00034166666666666666,
+      "loss": 4.4804,
+      "step": 820
+    },
+    {
+      "epoch": 0.7055037794845329,
+      "grad_norm": 31755.115234375,
+      "learning_rate": 0.00034583333333333335,
+      "loss": 4.4803,
+      "step": 830
+    },
+    {
+      "epoch": 0.7140038250204912,
+      "grad_norm": 26255.55078125,
+      "learning_rate": 0.00035,
+      "loss": 4.4767,
+      "step": 840
+    },
+    {
+      "epoch": 0.7225038705564494,
+      "grad_norm": 23372.48046875,
+      "learning_rate": 0.0003541666666666667,
+      "loss": 4.477,
+      "step": 850
+    },
+    {
+      "epoch": 0.7310039160924077,
+      "grad_norm": 37866.93359375,
+      "learning_rate": 0.00035833333333333333,
+      "loss": 4.4765,
+      "step": 860
+    },
+    {
+      "epoch": 0.7395039616283658,
+      "grad_norm": 26253.822265625,
+      "learning_rate": 0.0003625,
+      "loss": 4.4738,
+      "step": 870
+    },
+    {
+      "epoch": 0.7480040071643241,
+      "grad_norm": 25544.201171875,
+      "learning_rate": 0.00036666666666666667,
+      "loss": 4.4745,
+      "step": 880
+    },
+    {
+      "epoch": 0.7565040527002823,
+      "grad_norm": 25811.166015625,
+      "learning_rate": 0.00037083333333333337,
+      "loss": 4.4734,
+      "step": 890
+    },
+    {
+      "epoch": 0.7650040982362406,
+      "grad_norm": 27087.75390625,
+      "learning_rate": 0.000375,
+      "loss": 4.4737,
+      "step": 900
+    },
+    {
+      "epoch": 0.7650040982362406,
+      "eval_accuracy": 0.2906699089310573,
+      "eval_loss": 4.468409538269043,
+      "eval_runtime": 1272.8084,
+      "eval_samples_per_second": 392.343,
+      "eval_steps_per_second": 1.362,
+      "step": 900
+    },
+    {
+      "epoch": 0.7735041437721988,
+      "grad_norm": 34065.42578125,
+      "learning_rate": 0.00037916666666666665,
+      "loss": 4.4712,
+      "step": 910
+    },
+    {
+      "epoch": 0.782004189308157,
+      "grad_norm": 20391.37109375,
+      "learning_rate": 0.00038333333333333334,
+      "loss": 4.4627,
+      "step": 920
+    },
+    {
+      "epoch": 0.7905042348441152,
+      "grad_norm": 22009.44140625,
+      "learning_rate": 0.00038750000000000004,
+      "loss": 4.4501,
+      "step": 930
+    },
+    {
+      "epoch": 0.7990042803800734,
+      "grad_norm": 37206.39453125,
+      "learning_rate": 0.0003916666666666667,
+      "loss": 4.4388,
+      "step": 940
+    },
+    {
+      "epoch": 0.8075043259160317,
+      "grad_norm": 51573.41015625,
+      "learning_rate": 0.0003958333333333333,
+      "loss": 4.4256,
+      "step": 950
+    },
+    {
+      "epoch": 0.8160043714519899,
+      "grad_norm": 47200.921875,
+      "learning_rate": 0.0004,
+      "loss": 4.4164,
+      "step": 960
+    },
+    {
+      "epoch": 0.8245044169879482,
+      "grad_norm": 34512.5234375,
+      "learning_rate": 0.00040416666666666666,
+      "loss": 4.4093,
+      "step": 970
+    },
+    {
+      "epoch": 0.8330044625239064,
+      "grad_norm": 38512.90234375,
+      "learning_rate": 0.00040833333333333336,
+      "loss": 4.4011,
+      "step": 980
+    },
+    {
+      "epoch": 0.8415045080598647,
+      "grad_norm": 44749.2109375,
+      "learning_rate": 0.0004125,
+      "loss": 4.3852,
+      "step": 990
+    },
+    {
+      "epoch": 0.8415045080598647,
+      "eval_accuracy": 0.2906419834101783,
+      "eval_loss": 4.365363597869873,
+      "eval_runtime": 1274.1773,
+      "eval_samples_per_second": 391.922,
+      "eval_steps_per_second": 1.361,
+      "step": 990
+    },
+    {
+      "epoch": 0.8500045535958228,
+      "grad_norm": 43667.578125,
+      "learning_rate": 0.0004166666666666667,
+      "loss": 4.3701,
+      "step": 1000
+    },
+    {
+      "epoch": 0.858504599131781,
+      "grad_norm": 55209.21875,
+      "learning_rate": 0.00042083333333333333,
+      "loss": 4.3609,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8670046446677393,
+      "grad_norm": 60726.68359375,
+      "learning_rate": 0.000425,
+      "loss": 4.3476,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8755046902036975,
+      "grad_norm": 49279.09765625,
+      "learning_rate": 0.00042916666666666667,
+      "loss": 4.3382,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8840047357396558,
+      "grad_norm": 44572.08203125,
+      "learning_rate": 0.00043333333333333337,
+      "loss": 4.3271,
+      "step": 1040
+    },
+    {
+      "epoch": 0.892504781275614,
+      "grad_norm": 70982.9140625,
+      "learning_rate": 0.0004375,
+      "loss": 4.3161,
+      "step": 1050
+    },
+    {
+      "epoch": 0.9010048268115722,
+      "grad_norm": 49678.1796875,
+      "learning_rate": 0.00044166666666666665,
+      "loss": 4.3032,
+      "step": 1060
+    },
+    {
+      "epoch": 0.9095048723475304,
+      "grad_norm": 42499.19140625,
+      "learning_rate": 0.00044583333333333335,
+      "loss": 4.2959,
+      "step": 1070
+    },
+    {
+      "epoch": 0.9180049178834887,
+      "grad_norm": 61131.90625,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 4.2901,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9180049178834887,
+      "eval_accuracy": 0.2915224046693052,
+      "eval_loss": 4.254197597503662,
+      "eval_runtime": 1275.3488,
+      "eval_samples_per_second": 391.562,
+      "eval_steps_per_second": 1.36,
+      "step": 1080
+    },
+    {
+      "epoch": 0.9265049634194469,
+      "grad_norm": 61474.8046875,
+      "learning_rate": 0.0004541666666666667,
+      "loss": 4.2782,
+      "step": 1090
+    },
+    {
+      "epoch": 0.9350050089554052,
+      "grad_norm": 81287.625,
+      "learning_rate": 0.0004583333333333333,
+      "loss": 4.2578,
+      "step": 1100
+    },
+    {
+      "epoch": 0.9435050544913633,
+      "grad_norm": 71441.078125,
+      "learning_rate": 0.0004625,
+      "loss": 4.2455,
+      "step": 1110
+    },
+    {
+      "epoch": 0.9520051000273215,
+      "grad_norm": 52111.4921875,
+      "learning_rate": 0.00046666666666666666,
+      "loss": 4.237,
+      "step": 1120
+    },
+    {
+      "epoch": 0.9605051455632798,
+      "grad_norm": 79908.65625,
+      "learning_rate": 0.00047083333333333336,
+      "loss": 4.2271,
+      "step": 1130
+    },
+    {
+      "epoch": 0.969005191099238,
+      "grad_norm": 57955.83203125,
+      "learning_rate": 0.000475,
+      "loss": 4.2194,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9775052366351963,
+      "grad_norm": 52858.40234375,
+      "learning_rate": 0.0004791666666666667,
+      "loss": 4.2118,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9860052821711545,
+      "grad_norm": 78292.59375,
+      "learning_rate": 0.00048333333333333334,
+      "loss": 4.2114,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9945053277071128,
+      "grad_norm": 69718.5859375,
+      "learning_rate": 0.0004875,
+      "loss": 4.1945,
+      "step": 1170
+    },
+    {
+      "epoch": 0.9945053277071128,
+      "eval_accuracy": 0.29390537675129147,
+      "eval_loss": 4.130845069885254,
+      "eval_runtime": 1275.9587,
+      "eval_samples_per_second": 391.375,
+      "eval_steps_per_second": 1.359,
+      "step": 1170
+    },
+    {
+      "epoch": 1.003005373243071,
+      "grad_norm": 75537.1875,
+      "learning_rate": 0.0004916666666666666,
+      "loss": 4.1849,
+      "step": 1180
+    },
+    {
+      "epoch": 1.011505418779029,
+      "grad_norm": 81501.7734375,
+      "learning_rate": 0.0004958333333333334,
+      "loss": 4.1814,
+      "step": 1190
+    },
+    {
+      "epoch": 1.0200054643149874,
+      "grad_norm": 148886.71875,
+      "learning_rate": 0.0005,
+      "loss": 4.1642,
+      "step": 1200
+    },
+    {
+      "epoch": 1.0285055098509457,
+      "grad_norm": 78217.7109375,
+      "learning_rate": 0.0005041666666666667,
+      "loss": 4.1625,
+      "step": 1210
+    },
+    {
+      "epoch": 1.0370055553869038,
+      "grad_norm": 74189.0234375,
+      "learning_rate": 0.0005083333333333333,
+      "loss": 4.141,
+      "step": 1220
+    },
+    {
+      "epoch": 1.045505600922862,
+      "grad_norm": 105369.5546875,
+      "learning_rate": 0.0005124999999999999,
+      "loss": 4.1294,
+      "step": 1230
+    },
+    {
+      "epoch": 1.0540056464588203,
+      "grad_norm": 105365.2109375,
+      "learning_rate": 0.0005166666666666667,
+      "loss": 4.123,
+      "step": 1240
+    },
+    {
+      "epoch": 1.0625056919947786,
+      "grad_norm": 81112.6328125,
+      "learning_rate": 0.0005208333333333334,
+      "loss": 4.1089,
+      "step": 1250
+    },
+    {
+      "epoch": 1.0710057375307367,
+      "grad_norm": 115335.5859375,
+      "learning_rate": 0.0005250000000000001,
+      "loss": 4.1027,
+      "step": 1260
+    },
+    {
+      "epoch": 1.0710057375307367,
+      "eval_accuracy": 0.3011365978433179,
+      "eval_loss": 4.004420757293701,
+      "eval_runtime": 1276.3722,
+      "eval_samples_per_second": 391.248,
+      "eval_steps_per_second": 1.359,
+      "step": 1260
+    },
+    {
+      "epoch": 1.079505783066695,
+      "grad_norm": 108373.8359375,
+      "learning_rate": 0.0005291666666666667,
+      "loss": 4.0735,
+      "step": 1270
+    },
+    {
+      "epoch": 1.0880058286026533,
+      "grad_norm": 112014.7890625,
+      "learning_rate": 0.0005333333333333334,
+      "loss": 4.0363,
+      "step": 1280
+    },
+    {
+      "epoch": 1.0965058741386116,
+      "grad_norm": 153617.40625,
+      "learning_rate": 0.0005375,
+      "loss": 4.0039,
+      "step": 1290
+    },
+    {
+      "epoch": 1.1050059196745696,
+      "grad_norm": 124114.390625,
+      "learning_rate": 0.0005416666666666666,
+      "loss": 3.9605,
+      "step": 1300
+    },
+    {
+      "epoch": 1.113505965210528,
+      "grad_norm": 146727.78125,
+      "learning_rate": 0.0005458333333333333,
+      "loss": 3.8962,
+      "step": 1310
+    },
+    {
+      "epoch": 1.1220060107464862,
+      "grad_norm": 178636.21875,
+      "learning_rate": 0.00055,
+      "loss": 3.8446,
+      "step": 1320
+    },
+    {
+      "epoch": 1.1305060562824445,
+      "grad_norm": 177699.4375,
+      "learning_rate": 0.0005541666666666667,
+      "loss": 3.8045,
+      "step": 1330
+    },
+    {
+      "epoch": 1.1390061018184026,
+      "grad_norm": 121470.7890625,
+      "learning_rate": 0.0005583333333333333,
+      "loss": 3.7662,
+      "step": 1340
+    },
+    {
+      "epoch": 1.1475061473543609,
+      "grad_norm": 185655.078125,
+      "learning_rate": 0.0005625000000000001,
+      "loss": 3.7529,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1475061473543609,
+      "eval_accuracy": 0.34186952633889467,
+      "eval_loss": 3.526193380355835,
+      "eval_runtime": 1274.9146,
+      "eval_samples_per_second": 391.695,
+      "eval_steps_per_second": 1.36,
+      "step": 1350
+    },
+    {
+      "epoch": 1.1560061928903191,
+      "grad_norm": 194355.203125,
+      "learning_rate": 0.0005666666666666667,
+      "loss": 3.6921,
+      "step": 1360
+    },
+    {
+      "epoch": 1.1645062384262772,
+      "grad_norm": 102708.9375,
+      "learning_rate": 0.0005708333333333333,
+      "loss": 3.6505,
+      "step": 1370
+    },
+    {
+      "epoch": 1.1730062839622355,
+      "grad_norm": 153301.15625,
+      "learning_rate": 0.000575,
+      "loss": 3.6163,
+      "step": 1380
+    },
+    {
+      "epoch": 1.1815063294981938,
+      "grad_norm": 158013.796875,
+      "learning_rate": 0.0005791666666666667,
+      "loss": 3.5602,
+      "step": 1390
+    },
+    {
+      "epoch": 1.190006375034152,
+      "grad_norm": 113413.78125,
+      "learning_rate": 0.0005833333333333334,
+      "loss": 3.5185,
+      "step": 1400
+    },
+    {
+      "epoch": 1.1985064205701101,
+      "grad_norm": 157948.4375,
+      "learning_rate": 0.0005875,
+      "loss": 3.4583,
+      "step": 1410
+    },
+    {
+      "epoch": 1.2070064661060684,
+      "grad_norm": 157980.484375,
+      "learning_rate": 0.0005916666666666667,
+      "loss": 3.3973,
+      "step": 1420
+    },
+    {
+      "epoch": 1.2155065116420267,
+      "grad_norm": 100785.5234375,
+      "learning_rate": 0.0005958333333333333,
+      "loss": 3.3448,
+      "step": 1430
+    },
+    {
+      "epoch": 1.2240065571779848,
+      "grad_norm": 146644.875,
+      "learning_rate": 0.0006,
+      "loss": 3.3135,
+      "step": 1440
+    },
+    {
+      "epoch": 1.2240065571779848,
+      "eval_accuracy": 0.40612111173593607,
+      "eval_loss": 3.1175479888916016,
+      "eval_runtime": 1275.6913,
+      "eval_samples_per_second": 391.457,
+      "eval_steps_per_second": 1.359,
+      "step": 1440
+    },
+    {
+      "epoch": 1.232506602713943,
+      "grad_norm": 169930.046875,
+      "learning_rate": 0.0006041666666666666,
+      "loss": 3.31,
+      "step": 1450
+    },
+    {
+      "epoch": 1.2410066482499014,
+      "grad_norm": 166900.796875,
+      "learning_rate": 0.0006083333333333333,
+      "loss": 3.2813,
+      "step": 1460
+    },
+    {
+      "epoch": 1.2495066937858597,
+      "grad_norm": 142726.859375,
+      "learning_rate": 0.0006125000000000001,
+      "loss": 3.2236,
+      "step": 1470
+    },
+    {
+      "epoch": 1.2580067393218177,
+      "grad_norm": 113215.296875,
+      "learning_rate": 0.0006166666666666667,
+      "loss": 3.1767,
+      "step": 1480
+    },
+    {
+      "epoch": 1.266506784857776,
+      "grad_norm": 150677.703125,
+      "learning_rate": 0.0006208333333333334,
+      "loss": 3.1548,
+      "step": 1490
+    },
+    {
+      "epoch": 1.2750068303937343,
+      "grad_norm": 134445.515625,
+      "learning_rate": 0.000625,
+      "loss": 3.0924,
+      "step": 1500
+    },
+    {
+      "epoch": 1.2835068759296924,
+      "grad_norm": 116583.6796875,
+      "learning_rate": 0.0006291666666666667,
+      "loss": 3.023,
+      "step": 1510
+    },
+    {
+      "epoch": 1.2920069214656507,
+      "grad_norm": 99715.9765625,
+      "learning_rate": 0.0006333333333333333,
+      "loss": 2.9354,
+      "step": 1520
+    },
+    {
+      "epoch": 1.300506967001609,
+      "grad_norm": 97103.8125,
+      "learning_rate": 0.0006374999999999999,
+      "loss": 2.8965,
+      "step": 1530
+    },
+    {
+      "epoch": 1.300506967001609,
+      "eval_accuracy": 0.46788244925851397,
+      "eval_loss": 2.698025941848755,
+      "eval_runtime": 1277.5104,
+      "eval_samples_per_second": 390.899,
+      "eval_steps_per_second": 1.357,
+      "step": 1530
+    },
+    {
+      "epoch": 1.3090070125375672,
+      "grad_norm": 87332.8515625,
+      "learning_rate": 0.0006416666666666667,
+      "loss": 2.8339,
+      "step": 1540
+    },
+    {
+      "epoch": 1.3175070580735255,
+      "grad_norm": 89923.7890625,
+      "learning_rate": 0.0006458333333333334,
+      "loss": 2.7669,
+      "step": 1550
+    },
+    {
+      "epoch": 1.3260071036094836,
+      "grad_norm": 75226.8515625,
+      "learning_rate": 0.0006500000000000001,
+      "loss": 2.7303,
+      "step": 1560
+    },
+    {
+      "epoch": 1.3345071491454419,
+      "grad_norm": 66864.1484375,
+      "learning_rate": 0.0006541666666666667,
+      "loss": 2.6759,
+      "step": 1570
+    },
+    {
+      "epoch": 1.3430071946814,
+      "grad_norm": 61924.27734375,
+      "learning_rate": 0.0006583333333333334,
+      "loss": 2.6349,
+      "step": 1580
+    },
+    {
+      "epoch": 1.3515072402173582,
+      "grad_norm": 54619.734375,
+      "learning_rate": 0.0006625,
+      "loss": 2.5945,
+      "step": 1590
+    },
+    {
+      "epoch": 1.3600072857533165,
+      "grad_norm": 74138.375,
+      "learning_rate": 0.0006666666666666666,
+      "loss": 2.5633,
+      "step": 1600
+    },
+    {
+      "epoch": 1.3685073312892748,
+      "grad_norm": 58250.90625,
+      "learning_rate": 0.0006708333333333333,
+      "loss": 2.5349,
+      "step": 1610
+    },
+    {
+      "epoch": 1.377007376825233,
+      "grad_norm": 76097.328125,
+      "learning_rate": 0.000675,
+      "loss": 2.4952,
+      "step": 1620
+    },
+    {
+      "epoch": 1.377007376825233,
+      "eval_accuracy": 0.5196907758785413,
+      "eval_loss": 2.3480491638183594,
+      "eval_runtime": 1280.8752,
+      "eval_samples_per_second": 389.872,
+      "eval_steps_per_second": 1.354,
+      "step": 1620
+    },
+    {
+      "epoch": 1.3855074223611912,
+      "grad_norm": 77774.96875,
+      "learning_rate": 0.0006791666666666667,
+      "loss": 2.4686,
+      "step": 1630
+    },
+    {
+      "epoch": 1.3940074678971495,
+      "grad_norm": 65427.421875,
+      "learning_rate": 0.0006833333333333333,
+      "loss": 2.4615,
+      "step": 1640
+    },
+    {
+      "epoch": 1.4025075134331078,
+      "grad_norm": 56960.09765625,
+      "learning_rate": 0.0006875,
+      "loss": 2.4228,
+      "step": 1650
+    },
+    {
+      "epoch": 1.4110075589690658,
+      "grad_norm": 74808.4296875,
+      "learning_rate": 0.0006916666666666667,
+      "loss": 2.391,
+      "step": 1660
+    },
+    {
+      "epoch": 1.4195076045050241,
+      "grad_norm": 47526.44140625,
+      "learning_rate": 0.0006958333333333334,
+      "loss": 2.3693,
+      "step": 1670
+    },
+    {
+      "epoch": 1.4280076500409824,
+      "grad_norm": 54673.7578125,
+      "learning_rate": 0.0007,
+      "loss": 2.3505,
+      "step": 1680
+    },
+    {
+      "epoch": 1.4365076955769407,
+      "grad_norm": 33447.546875,
+      "learning_rate": 0.0007041666666666667,
+      "loss": 2.3205,
+      "step": 1690
+    },
+    {
+      "epoch": 1.4450077411128988,
+      "grad_norm": 56293.46484375,
+      "learning_rate": 0.0007083333333333334,
+      "loss": 2.318,
+      "step": 1700
+    },
+    {
+      "epoch": 1.453507786648857,
+      "grad_norm": 41077.09375,
+      "learning_rate": 0.0007125,
+      "loss": 2.2838,
+      "step": 1710
+    },
+    {
+      "epoch": 1.453507786648857,
+      "eval_accuracy": 0.5489981207105539,
+      "eval_loss": 2.1566824913024902,
+      "eval_runtime": 1279.9042,
+      "eval_samples_per_second": 390.168,
+      "eval_steps_per_second": 1.355,
+      "step": 1710
+    },
+    {
+      "epoch": 1.4620078321848153,
+      "grad_norm": 51759.4296875,
+      "learning_rate": 0.0007166666666666667,
+      "loss": 2.264,
+      "step": 1720
+    },
+    {
+      "epoch": 1.4705078777207734,
+      "grad_norm": 50884.80078125,
+      "learning_rate": 0.0007208333333333333,
+      "loss": 2.2408,
+      "step": 1730
+    },
+    {
+      "epoch": 1.4790079232567317,
+      "grad_norm": 38072.73046875,
+      "learning_rate": 0.000725,
+      "loss": 2.2221,
+      "step": 1740
+    },
+    {
+      "epoch": 1.48750796879269,
+      "grad_norm": 45036.5,
+      "learning_rate": 0.0007291666666666666,
+      "loss": 2.2046,
+      "step": 1750
+    },
+    {
+      "epoch": 1.4960080143286483,
+      "grad_norm": 30498.197265625,
+      "learning_rate": 0.0007333333333333333,
+      "loss": 2.1925,
+      "step": 1760
+    },
+    {
+      "epoch": 1.5045080598646066,
+      "grad_norm": 52624.26953125,
+      "learning_rate": 0.0007375000000000001,
+      "loss": 2.1845,
+      "step": 1770
+    },
+    {
+      "epoch": 1.5130081054005646,
+      "grad_norm": 39472.8125,
+      "learning_rate": 0.0007416666666666667,
+      "loss": 2.1579,
+      "step": 1780
+    },
+    {
+      "epoch": 1.521508150936523,
+      "grad_norm": 43868.4375,
+      "learning_rate": 0.0007458333333333334,
+      "loss": 2.1517,
+      "step": 1790
+    },
+    {
+      "epoch": 1.530008196472481,
+      "grad_norm": 39694.20703125,
+      "learning_rate": 0.00075,
+      "loss": 2.1324,
+      "step": 1800
+    },
+    {
+      "epoch": 1.530008196472481,
+      "eval_accuracy": 0.571443558749772,
+      "eval_loss": 2.0166752338409424,
+      "eval_runtime": 1279.8418,
+      "eval_samples_per_second": 390.187,
+      "eval_steps_per_second": 1.355,
+      "step": 1800
+    },
+    {
+      "epoch": 1.5385082420084393,
+      "grad_norm": 41399.078125,
+      "learning_rate": 0.0007541666666666667,
+      "loss": 2.1129,
+      "step": 1810
+    },
+    {
+      "epoch": 1.5470082875443976,
+      "grad_norm": 39988.51171875,
+      "learning_rate": 0.0007583333333333333,
+      "loss": 2.0979,
+      "step": 1820
+    },
+    {
+      "epoch": 1.5555083330803559,
+      "grad_norm": 40106.96875,
+      "learning_rate": 0.0007624999999999999,
+      "loss": 2.0861,
+      "step": 1830
+    },
+    {
+      "epoch": 1.5640083786163141,
+      "grad_norm": 37416.859375,
+      "learning_rate": 0.0007666666666666667,
+      "loss": 2.0691,
+      "step": 1840
+    },
+    {
+      "epoch": 1.5725084241522722,
+      "grad_norm": 45897.46484375,
+      "learning_rate": 0.0007708333333333334,
+      "loss": 2.0541,
+      "step": 1850
+    },
+    {
+      "epoch": 1.5810084696882305,
+      "grad_norm": 43155.91796875,
+      "learning_rate": 0.0007750000000000001,
+      "loss": 2.0409,
+      "step": 1860
+    },
+    {
+      "epoch": 1.5895085152241886,
+      "grad_norm": 44230.0703125,
+      "learning_rate": 0.0007791666666666667,
+      "loss": 2.026,
+      "step": 1870
+    },
+    {
+      "epoch": 1.5980085607601469,
+      "grad_norm": 41950.99609375,
+      "learning_rate": 0.0007833333333333334,
+      "loss": 2.007,
+      "step": 1880
+    },
+    {
+      "epoch": 1.6065086062961051,
+      "grad_norm": 50098.26171875,
+      "learning_rate": 0.0007875,
+      "loss": 2.004,
+      "step": 1890
+    },
+    {
+      "epoch": 1.6065086062961051,
+      "eval_accuracy": 0.5934511080462407,
+      "eval_loss": 1.89019775390625,
+      "eval_runtime": 1279.3367,
+      "eval_samples_per_second": 390.341,
+      "eval_steps_per_second": 1.355,
+      "step": 1890
+    },
+    {
+      "epoch": 1.6150086518320634,
+      "grad_norm": 38339.48046875,
+      "learning_rate": 0.0007916666666666666,
+      "loss": 1.992,
+      "step": 1900
+    },
+    {
+      "epoch": 1.6235086973680217,
+      "grad_norm": 45457.52734375,
+      "learning_rate": 0.0007958333333333333,
+      "loss": 1.9782,
+      "step": 1910
+    },
+    {
+      "epoch": 1.6320087429039798,
+      "grad_norm": 44519.359375,
+      "learning_rate": 0.0008,
+      "loss": 1.9713,
+      "step": 1920
+    },
+    {
+      "epoch": 1.640508788439938,
+      "grad_norm": 32601.9921875,
+      "learning_rate": 0.0008041666666666667,
+      "loss": 1.9564,
+      "step": 1930
+    },
+    {
+      "epoch": 1.6490088339758961,
+      "grad_norm": 38710.42578125,
+      "learning_rate": 0.0008083333333333333,
+      "loss": 1.9461,
+      "step": 1940
+    },
+    {
+      "epoch": 1.6575088795118544,
+      "grad_norm": 36938.29296875,
+      "learning_rate": 0.0008125000000000001,
+      "loss": 1.9386,
+      "step": 1950
+    },
+    {
+      "epoch": 1.6660089250478127,
+      "grad_norm": 39862.9609375,
+      "learning_rate": 0.0008166666666666667,
+      "loss": 1.9229,
+      "step": 1960
+    },
+    {
+      "epoch": 1.674508970583771,
+      "grad_norm": 29285.794921875,
+      "learning_rate": 0.0008208333333333334,
+      "loss": 1.9096,
+      "step": 1970
+    },
+    {
+      "epoch": 1.6830090161197293,
+      "grad_norm": 40694.77734375,
+      "learning_rate": 0.000825,
+      "loss": 1.8992,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6830090161197293,
+      "eval_accuracy": 0.6102293333175925,
+      "eval_loss": 1.7952998876571655,
+      "eval_runtime": 1281.484,
+      "eval_samples_per_second": 389.687,
+      "eval_steps_per_second": 1.353,
+      "step": 1980
+    },
+    {
+      "epoch": 1.6915090616556876,
+      "grad_norm": 52670.69140625,
+      "learning_rate": 0.0008291666666666667,
+      "loss": 1.8874,
+      "step": 1990
+    },
+    {
+      "epoch": 1.7000091071916457,
+      "grad_norm": 41263.34375,
+      "learning_rate": 0.0008333333333333334,
+      "loss": 1.8832,
+      "step": 2000
+    },
+    {
+      "epoch": 1.708509152727604,
+      "grad_norm": 48928.69921875,
+      "learning_rate": 0.0008375,
+      "loss": 1.8709,
+      "step": 2010
+    },
+    {
+      "epoch": 1.717009198263562,
+      "grad_norm": 34242.77734375,
+      "learning_rate": 0.0008416666666666667,
+      "loss": 1.8672,
+      "step": 2020
+    },
+    {
+      "epoch": 1.7255092437995203,
+      "grad_norm": 43822.64453125,
+      "learning_rate": 0.0008458333333333333,
+      "loss": 1.8542,
+      "step": 2030
+    },
+    {
+      "epoch": 1.7340092893354786,
+      "grad_norm": 30140.544921875,
+      "learning_rate": 0.00085,
+      "loss": 1.8418,
+      "step": 2040
+    },
+    {
+      "epoch": 1.7425093348714369,
+      "grad_norm": 32354.177734375,
+      "learning_rate": 0.0008541666666666666,
+      "loss": 1.833,
+      "step": 2050
+    },
+    {
+      "epoch": 1.7510093804073952,
+      "grad_norm": 37106.171875,
+      "learning_rate": 0.0008583333333333333,
+      "loss": 1.8285,
+      "step": 2060
+    },
+    {
+      "epoch": 1.7595094259433532,
+      "grad_norm": 35038.6875,
+      "learning_rate": 0.0008625000000000001,
+      "loss": 1.815,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7595094259433532,
+      "eval_accuracy": 0.6255571282197189,
+      "eval_loss": 1.7101428508758545,
+      "eval_runtime": 1279.0851,
+      "eval_samples_per_second": 390.418,
+      "eval_steps_per_second": 1.356,
+      "step": 2070
+    },
+    {
+      "epoch": 1.7680094714793115,
+      "grad_norm": 33263.3046875,
+      "learning_rate": 0.0008666666666666667,
+      "loss": 1.8021,
+      "step": 2080
+    },
+    {
+      "epoch": 1.7765095170152696,
+      "grad_norm": 29019.279296875,
+      "learning_rate": 0.0008708333333333334,
+      "loss": 1.7913,
+      "step": 2090
+    },
+    {
+      "epoch": 1.785009562551228,
+      "grad_norm": 40720.78125,
+      "learning_rate": 0.000875,
+      "loss": 1.7823,
+      "step": 2100
+    },
+    {
+      "epoch": 1.7935096080871862,
+      "grad_norm": 40485.59765625,
+      "learning_rate": 0.0008791666666666667,
+      "loss": 1.7738,
+      "step": 2110
+    },
+    {
+      "epoch": 1.8020096536231445,
+      "grad_norm": 34836.8984375,
+      "learning_rate": 0.0008833333333333333,
+      "loss": 1.7628,
+      "step": 2120
+    },
+    {
+      "epoch": 1.8105096991591028,
+      "grad_norm": 37340.25,
+      "learning_rate": 0.0008874999999999999,
+      "loss": 1.7564,
+      "step": 2130
+    },
+    {
+      "epoch": 1.8190097446950608,
+      "grad_norm": 34712.98828125,
+      "learning_rate": 0.0008916666666666667,
+      "loss": 1.7454,
+      "step": 2140
+    },
+    {
+      "epoch": 1.8275097902310191,
+      "grad_norm": 36124.44921875,
+      "learning_rate": 0.0008958333333333334,
+      "loss": 1.7375,
+      "step": 2150
+    },
+    {
+      "epoch": 1.8360098357669772,
+      "grad_norm": 26800.90234375,
+      "learning_rate": 0.0009000000000000001,
+      "loss": 1.724,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8360098357669772,
+      "eval_accuracy": 0.6417359150220479,
+      "eval_loss": 1.619744062423706,
+      "eval_runtime": 1279.7377,
+      "eval_samples_per_second": 390.219,
+      "eval_steps_per_second": 1.355,
+      "step": 2160
+    },
+    {
+      "epoch": 1.8445098813029355,
+      "grad_norm": 29157.935546875,
+      "learning_rate": 0.0009041666666666667,
+      "loss": 1.714,
+      "step": 2170
+    },
+    {
+      "epoch": 1.8530099268388938,
+      "grad_norm": 32622.201171875,
+      "learning_rate": 0.0009083333333333334,
+      "loss": 1.6972,
+      "step": 2180
+    },
+    {
+      "epoch": 1.861509972374852,
+      "grad_norm": 27361.466796875,
+      "learning_rate": 0.0009125,
+      "loss": 1.6807,
+      "step": 2190
+    },
+    {
+      "epoch": 1.8700100179108103,
+      "grad_norm": 30800.376953125,
+      "learning_rate": 0.0009166666666666666,
+      "loss": 1.6644,
+      "step": 2200
+    },
+    {
+      "epoch": 1.8785100634467686,
+      "grad_norm": 28547.041015625,
+      "learning_rate": 0.0009208333333333333,
+      "loss": 1.6546,
+      "step": 2210
+    },
+    {
+      "epoch": 1.8870101089827267,
+      "grad_norm": 32673.380859375,
+      "learning_rate": 0.000925,
+      "loss": 1.6407,
+      "step": 2220
+    },
+    {
+      "epoch": 1.8955101545186848,
+      "grad_norm": 38690.15625,
+      "learning_rate": 0.0009291666666666667,
+      "loss": 1.6327,
+      "step": 2230
+    },
+    {
+      "epoch": 1.904010200054643,
+      "grad_norm": 24017.017578125,
+      "learning_rate": 0.0009333333333333333,
+      "loss": 1.6185,
+      "step": 2240
+    },
+    {
+      "epoch": 1.9125102455906013,
+      "grad_norm": 33731.96484375,
+      "learning_rate": 0.0009375,
+      "loss": 1.6115,
+      "step": 2250
+    },
+    {
+      "epoch": 1.9125102455906013,
+      "eval_accuracy": 0.6670528637301408,
+      "eval_loss": 1.5023800134658813,
+      "eval_runtime": 1278.3406,
+      "eval_samples_per_second": 390.646,
+      "eval_steps_per_second": 1.356,
+      "step": 2250
+    },
+    {
+      "epoch": 1.9210102911265596,
+      "grad_norm": 35314.43359375,
+      "learning_rate": 0.0009416666666666667,
+      "loss": 1.5993,
+      "step": 2260
+    },
+    {
+      "epoch": 1.929510336662518,
+      "grad_norm": 36152.0078125,
+      "learning_rate": 0.0009458333333333334,
+      "loss": 1.5944,
+      "step": 2270
+    },
+    {
+      "epoch": 1.9380103821984762,
+      "grad_norm": 31576.48046875,
+      "learning_rate": 0.00095,
+      "loss": 1.5867,
+      "step": 2280
+    },
+    {
+      "epoch": 1.9465104277344343,
+      "grad_norm": 31250.708984375,
+      "learning_rate": 0.0009541666666666667,
+      "loss": 1.5779,
+      "step": 2290
+    },
+    {
+      "epoch": 1.9550104732703926,
+      "grad_norm": 39243.34375,
+      "learning_rate": 0.0009583333333333334,
+      "loss": 1.5693,
+      "step": 2300
+    },
+    {
+      "epoch": 1.9635105188063506,
+      "grad_norm": 26391.3046875,
+      "learning_rate": 0.0009625,
+      "loss": 1.5671,
+      "step": 2310
+    },
+    {
+      "epoch": 1.972010564342309,
+      "grad_norm": 34276.82421875,
+      "learning_rate": 0.0009666666666666667,
+      "loss": 1.56,
+      "step": 2320
+    },
+    {
+      "epoch": 1.9805106098782672,
+      "grad_norm": 28257.998046875,
+      "learning_rate": 0.0009708333333333333,
+      "loss": 1.5482,
+      "step": 2330
+    },
+    {
+      "epoch": 1.9890106554142255,
+      "grad_norm": 38320.5234375,
+      "learning_rate": 0.000975,
+      "loss": 1.5435,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9890106554142255,
+      "eval_accuracy": 0.679645168221919,
+      "eval_loss": 1.4393101930618286,
+      "eval_runtime": 1277.7056,
+      "eval_samples_per_second": 390.84,
+      "eval_steps_per_second": 1.357,
+      "step": 2340
+    },
+    {
+      "epoch": 1.9975107009501838,
+      "grad_norm": 30497.1171875,
+      "learning_rate": 0.0009791666666666666,
+      "loss": 1.5364,
+      "step": 2350
+    },
+    {
+      "epoch": 2.006010746486142,
+      "grad_norm": 30113.25,
+      "learning_rate": 0.0009833333333333332,
+      "loss": 1.5333,
+      "step": 2360
+    },
+    {
+      "epoch": 2.0145107920221,
+      "grad_norm": 23521.29296875,
+      "learning_rate": 0.0009875,
+      "loss": 1.528,
+      "step": 2370
+    },
+    {
+      "epoch": 2.023010837558058,
+      "grad_norm": 33135.26953125,
+      "learning_rate": 0.0009916666666666667,
+      "loss": 1.5211,
+      "step": 2380
+    },
+    {
+      "epoch": 2.0315108830940165,
+      "grad_norm": 22258.33984375,
+      "learning_rate": 0.0009958333333333334,
+      "loss": 1.5173,
+      "step": 2390
+    },
+    {
+      "epoch": 2.040010928629975,
+      "grad_norm": 30941.345703125,
+      "learning_rate": 0.001,
+      "loss": 1.5097,
+      "step": 2400
+    },
+    {
+      "epoch": 2.048510974165933,
+      "grad_norm": 27459.111328125,
+      "learning_rate": 0.000999009900990099,
+      "loss": 1.5022,
+      "step": 2410
+    },
+    {
+      "epoch": 2.0570110197018914,
+      "grad_norm": 23237.234375,
+      "learning_rate": 0.0009980198019801981,
+      "loss": 1.4956,
+      "step": 2420
+    },
+    {
+      "epoch": 2.0655110652378497,
+      "grad_norm": 24305.986328125,
+      "learning_rate": 0.000997029702970297,
+      "loss": 1.4921,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0655110652378497,
+      "eval_accuracy": 0.6886431014072092,
+      "eval_loss": 1.3913538455963135,
+      "eval_runtime": 1278.454,
+      "eval_samples_per_second": 390.611,
+      "eval_steps_per_second": 1.356,
+      "step": 2430
+    },
+    {
+      "epoch": 2.0740111107738075,
+      "grad_norm": 26497.21484375,
+      "learning_rate": 0.000996039603960396,
+      "loss": 1.4843,
+      "step": 2440
+    },
+    {
+      "epoch": 2.082511156309766,
+      "grad_norm": 24107.4375,
+      "learning_rate": 0.000995049504950495,
+      "loss": 1.4824,
+      "step": 2450
+    },
+    {
+      "epoch": 2.091011201845724,
+      "grad_norm": 26006.498046875,
+      "learning_rate": 0.0009940594059405941,
+      "loss": 1.4763,
+      "step": 2460
+    },
+    {
+      "epoch": 2.0995112473816824,
+      "grad_norm": 27216.076171875,
+      "learning_rate": 0.0009930693069306932,
+      "loss": 1.4704,
+      "step": 2470
+    },
+    {
+      "epoch": 2.1080112929176407,
+      "grad_norm": 21669.0546875,
+      "learning_rate": 0.000992079207920792,
+      "loss": 1.4636,
+      "step": 2480
+    },
+    {
+      "epoch": 2.116511338453599,
+      "grad_norm": 23455.484375,
+      "learning_rate": 0.000991089108910891,
+      "loss": 1.4566,
+      "step": 2490
+    },
+    {
+      "epoch": 2.1250113839895572,
+      "grad_norm": 20500.41015625,
+      "learning_rate": 0.0009900990099009901,
+      "loss": 1.4554,
+      "step": 2500
+    },
+    {
+      "epoch": 2.133511429525515,
+      "grad_norm": 26141.59765625,
+      "learning_rate": 0.0009891089108910892,
+      "loss": 1.4484,
+      "step": 2510
+    },
+    {
+      "epoch": 2.1420114750614734,
+      "grad_norm": 23275.765625,
+      "learning_rate": 0.0009881188118811882,
+      "loss": 1.4427,
+      "step": 2520
+    },
+    {
+      "epoch": 2.1420114750614734,
+      "eval_accuracy": 0.6980444623562028,
+      "eval_loss": 1.3418840169906616,
+      "eval_runtime": 1280.6296,
+      "eval_samples_per_second": 389.947,
+      "eval_steps_per_second": 1.354,
+      "step": 2520
+    },
+    {
+      "epoch": 2.1505115205974317,
+      "grad_norm": 26020.708984375,
+      "learning_rate": 0.000987128712871287,
+      "loss": 1.4374,
+      "step": 2530
+    },
+    {
+      "epoch": 2.15901156613339,
+      "grad_norm": 20408.48828125,
+      "learning_rate": 0.000986138613861386,
+      "loss": 1.4302,
+      "step": 2540
+    },
+    {
+      "epoch": 2.1675116116693482,
+      "grad_norm": 23640.2265625,
+      "learning_rate": 0.0009851485148514852,
+      "loss": 1.4277,
+      "step": 2550
+    },
+    {
+      "epoch": 2.1760116572053065,
+      "grad_norm": 22099.396484375,
+      "learning_rate": 0.0009841584158415842,
+      "loss": 1.4217,
+      "step": 2560
+    },
+    {
+      "epoch": 2.184511702741265,
+      "grad_norm": 23317.130859375,
+      "learning_rate": 0.0009831683168316833,
+      "loss": 1.4188,
+      "step": 2570
+    },
+    {
+      "epoch": 2.193011748277223,
+      "grad_norm": 24106.27734375,
+      "learning_rate": 0.000982178217821782,
+      "loss": 1.4147,
+      "step": 2580
+    },
+    {
+      "epoch": 2.201511793813181,
+      "grad_norm": 18410.43359375,
+      "learning_rate": 0.0009811881188118811,
+      "loss": 1.4107,
+      "step": 2590
+    },
+    {
+      "epoch": 2.2100118393491393,
+      "grad_norm": 22841.390625,
+      "learning_rate": 0.0009801980198019802,
+      "loss": 1.4075,
+      "step": 2600
+    },
+    {
+      "epoch": 2.2185118848850975,
+      "grad_norm": 24554.787109375,
+      "learning_rate": 0.0009792079207920793,
+      "loss": 1.404,
+      "step": 2610
+    },
+    {
+      "epoch": 2.2185118848850975,
+      "eval_accuracy": 0.7051629840505044,
+      "eval_loss": 1.3090746402740479,
+      "eval_runtime": 1279.6345,
+      "eval_samples_per_second": 390.25,
+      "eval_steps_per_second": 1.355,
+      "step": 2610
+    },
+    {
+      "epoch": 2.227011930421056,
+      "grad_norm": 25836.71484375,
+      "learning_rate": 0.0009782178217821783,
+      "loss": 1.4039,
+      "step": 2620
+    },
+    {
+      "epoch": 2.235511975957014,
+      "grad_norm": 21653.619140625,
+      "learning_rate": 0.0009772277227722771,
+      "loss": 1.3946,
+      "step": 2630
+    },
+    {
+      "epoch": 2.2440120214929724,
+      "grad_norm": 23728.3046875,
+      "learning_rate": 0.0009762376237623762,
+      "loss": 1.3929,
+      "step": 2640
+    },
+    {
+      "epoch": 2.2525120670289307,
+      "grad_norm": 21733.001953125,
+      "learning_rate": 0.0009752475247524752,
+      "loss": 1.3884,
+      "step": 2650
+    },
+    {
+      "epoch": 2.261012112564889,
+      "grad_norm": 21566.19921875,
+      "learning_rate": 0.0009742574257425743,
+      "loss": 1.3825,
+      "step": 2660
+    },
+    {
+      "epoch": 2.269512158100847,
+      "grad_norm": 19529.41015625,
+      "learning_rate": 0.0009732673267326732,
+      "loss": 1.3786,
+      "step": 2670
+    },
+    {
+      "epoch": 2.278012203636805,
+      "grad_norm": 20555.123046875,
+      "learning_rate": 0.0009722772277227723,
+      "loss": 1.3755,
+      "step": 2680
+    },
+    {
+      "epoch": 2.2865122491727634,
+      "grad_norm": 19776.90234375,
+      "learning_rate": 0.0009712871287128712,
+      "loss": 1.3706,
+      "step": 2690
+    },
+    {
+      "epoch": 2.2950122947087217,
+      "grad_norm": 18725.46875,
+      "learning_rate": 0.0009702970297029703,
+      "loss": 1.3703,
+      "step": 2700
+    },
+    {
+      "epoch": 2.2950122947087217,
+      "eval_accuracy": 0.7118258434856092,
+      "eval_loss": 1.2756892442703247,
+      "eval_runtime": 1278.3225,
+      "eval_samples_per_second": 390.651,
+      "eval_steps_per_second": 1.356,
+      "step": 2700
+    },
+    {
+      "epoch": 2.30351234024468,
+      "grad_norm": 21781.615234375,
+      "learning_rate": 0.0009693069306930693,
+      "loss": 1.369,
+      "step": 2710
+    },
+    {
+      "epoch": 2.3120123857806383,
+      "grad_norm": 21352.599609375,
+      "learning_rate": 0.0009683168316831683,
+      "loss": 1.3609,
+      "step": 2720
+    },
+    {
+      "epoch": 2.3205124313165966,
+      "grad_norm": 19621.919921875,
+      "learning_rate": 0.0009673267326732673,
+      "loss": 1.3579,
+      "step": 2730
+    },
+    {
+      "epoch": 2.3290124768525544,
+      "grad_norm": 15898.189453125,
+      "learning_rate": 0.0009663366336633663,
+      "loss": 1.3585,
+      "step": 2740
+    },
+    {
+      "epoch": 2.3375125223885127,
+      "grad_norm": 24640.2734375,
+      "learning_rate": 0.0009653465346534653,
+      "loss": 1.3521,
+      "step": 2750
+    },
+    {
+      "epoch": 2.346012567924471,
+      "grad_norm": 21775.046875,
+      "learning_rate": 0.0009643564356435644,
+      "loss": 1.3496,
+      "step": 2760
+    },
+    {
+      "epoch": 2.3545126134604293,
+      "grad_norm": 20199.751953125,
+      "learning_rate": 0.0009633663366336633,
+      "loss": 1.3477,
+      "step": 2770
+    },
+    {
+      "epoch": 2.3630126589963876,
+      "grad_norm": 23324.25,
+      "learning_rate": 0.0009623762376237624,
+      "loss": 1.344,
+      "step": 2780
+    },
+    {
+      "epoch": 2.371512704532346,
+      "grad_norm": 23789.677734375,
+      "learning_rate": 0.0009613861386138613,
+      "loss": 1.3423,
+      "step": 2790
+    },
+    {
+      "epoch": 2.371512704532346,
+      "eval_accuracy": 0.7178257874607675,
+      "eval_loss": 1.2434872388839722,
+      "eval_runtime": 1276.5226,
+      "eval_samples_per_second": 391.202,
+      "eval_steps_per_second": 1.358,
+      "step": 2790
+    },
+    {
+      "epoch": 2.380012750068304,
+      "grad_norm": 22512.796875,
+      "learning_rate": 0.0009603960396039604,
+      "loss": 1.3394,
+      "step": 2800
+    },
+    {
+      "epoch": 2.388512795604262,
+      "grad_norm": 24620.826171875,
+      "learning_rate": 0.0009594059405940594,
+      "loss": 1.3331,
+      "step": 2810
+    },
+    {
+      "epoch": 2.3970128411402203,
+      "grad_norm": 20756.77734375,
+      "learning_rate": 0.0009584158415841584,
+      "loss": 1.3293,
+      "step": 2820
+    },
+    {
+      "epoch": 2.4055128866761786,
+      "grad_norm": 17418.865234375,
+      "learning_rate": 0.0009574257425742574,
+      "loss": 1.3268,
+      "step": 2830
+    },
+    {
+      "epoch": 2.414012932212137,
+      "grad_norm": 19701.8984375,
+      "learning_rate": 0.0009564356435643564,
+      "loss": 1.3256,
+      "step": 2840
+    },
+    {
+      "epoch": 2.422512977748095,
+      "grad_norm": 22757.66015625,
+      "learning_rate": 0.0009554455445544554,
+      "loss": 1.3229,
+      "step": 2850
+    },
+    {
+      "epoch": 2.4310130232840534,
+      "grad_norm": 24491.03515625,
+      "learning_rate": 0.0009544554455445545,
+      "loss": 1.3187,
+      "step": 2860
+    },
+    {
+      "epoch": 2.4395130688200117,
+      "grad_norm": 19789.939453125,
+      "learning_rate": 0.0009534653465346534,
+      "loss": 1.3176,
+      "step": 2870
+    },
+    {
+      "epoch": 2.4480131143559696,
+      "grad_norm": 19331.78515625,
+      "learning_rate": 0.0009524752475247525,
+      "loss": 1.3165,
+      "step": 2880
+    },
+    {
+      "epoch": 2.4480131143559696,
+      "eval_accuracy": 0.7224858086201374,
+      "eval_loss": 1.2204521894454956,
+      "eval_runtime": 1278.1811,
+      "eval_samples_per_second": 390.694,
+      "eval_steps_per_second": 1.357,
+      "step": 2880
+    },
+    {
+      "epoch": 2.456513159891928,
+      "grad_norm": 19115.720703125,
+      "learning_rate": 0.0009514851485148514,
+      "loss": 1.3118,
+      "step": 2890
+    },
+    {
+      "epoch": 2.465013205427886,
+      "grad_norm": 18051.21875,
+      "learning_rate": 0.0009504950495049505,
+      "loss": 1.3087,
+      "step": 2900
+    },
+    {
+      "epoch": 2.4735132509638444,
+      "grad_norm": 17369.791015625,
+      "learning_rate": 0.0009495049504950495,
+      "loss": 1.3088,
+      "step": 2910
+    },
+    {
+      "epoch": 2.4820132964998027,
+      "grad_norm": 19926.080078125,
+      "learning_rate": 0.0009485148514851485,
+      "loss": 1.3067,
+      "step": 2920
+    },
+    {
+      "epoch": 2.490513342035761,
+      "grad_norm": 21108.0078125,
+      "learning_rate": 0.0009475247524752475,
+      "loss": 1.3034,
+      "step": 2930
+    },
+    {
+      "epoch": 2.4990133875717193,
+      "grad_norm": 20779.73828125,
+      "learning_rate": 0.0009465346534653465,
+      "loss": 1.2954,
+      "step": 2940
+    },
+    {
+      "epoch": 2.507513433107677,
+      "grad_norm": 19840.251953125,
+      "learning_rate": 0.0009455445544554455,
+      "loss": 1.2957,
+      "step": 2950
+    },
+    {
+      "epoch": 2.5160134786436354,
+      "grad_norm": 21993.7734375,
+      "learning_rate": 0.0009445544554455446,
+      "loss": 1.2948,
+      "step": 2960
+    },
+    {
+      "epoch": 2.5245135241795937,
+      "grad_norm": 23669.6171875,
+      "learning_rate": 0.0009435643564356435,
+      "loss": 1.2932,
+      "step": 2970
+    },
+    {
+      "epoch": 2.5245135241795937,
+      "eval_accuracy": 0.7274937619747395,
+      "eval_loss": 1.196637511253357,
+      "eval_runtime": 1278.0042,
+      "eval_samples_per_second": 390.748,
+      "eval_steps_per_second": 1.357,
+      "step": 2970
+    },
+    {
+      "epoch": 2.533013569715552,
+      "grad_norm": 16925.79296875,
+      "learning_rate": 0.0009425742574257426,
+      "loss": 1.2884,
+      "step": 2980
+    },
+    {
+      "epoch": 2.5415136152515103,
+      "grad_norm": 16753.69921875,
+      "learning_rate": 0.0009415841584158415,
+      "loss": 1.2862,
+      "step": 2990
+    },
+    {
+      "epoch": 2.5500136607874686,
+      "grad_norm": 19538.154296875,
+      "learning_rate": 0.0009405940594059406,
+      "loss": 1.2843,
+      "step": 3000
+    },
+    {
+      "epoch": 2.558513706323427,
+      "grad_norm": 18874.037109375,
+      "learning_rate": 0.0009396039603960396,
+      "loss": 1.2821,
+      "step": 3010
+    },
+    {
+      "epoch": 2.5670137518593847,
+      "grad_norm": 19626.3671875,
+      "learning_rate": 0.0009386138613861386,
+      "loss": 1.2793,
+      "step": 3020
+    },
+    {
+      "epoch": 2.575513797395343,
+      "grad_norm": 21052.349609375,
+      "learning_rate": 0.0009376237623762376,
+      "loss": 1.2789,
+      "step": 3030
+    },
+    {
+      "epoch": 2.5840138429313013,
+      "grad_norm": 19245.396484375,
+      "learning_rate": 0.0009366336633663367,
+      "loss": 1.2744,
+      "step": 3040
+    },
+    {
+      "epoch": 2.5925138884672596,
+      "grad_norm": 16579.640625,
+      "learning_rate": 0.0009356435643564357,
+      "loss": 1.2726,
+      "step": 3050
+    },
+    {
+      "epoch": 2.601013934003218,
+      "grad_norm": 17490.4609375,
+      "learning_rate": 0.0009346534653465348,
+      "loss": 1.2726,
+      "step": 3060
+    },
+    {
+      "epoch": 2.601013934003218,
+      "eval_accuracy": 0.7312536295087113,
+      "eval_loss": 1.1788941621780396,
+      "eval_runtime": 1282.3919,
+      "eval_samples_per_second": 389.411,
+      "eval_steps_per_second": 1.352,
+      "step": 3060
+    },
+    {
+      "epoch": 2.609513979539176,
+      "grad_norm": 21261.8125,
+      "learning_rate": 0.0009336633663366337,
+      "loss": 1.2683,
+      "step": 3070
+    },
+    {
+      "epoch": 2.6180140250751345,
+      "grad_norm": 21293.064453125,
+      "learning_rate": 0.0009326732673267328,
+      "loss": 1.2695,
+      "step": 3080
+    },
+    {
+      "epoch": 2.6265140706110923,
+      "grad_norm": 18312.087890625,
+      "learning_rate": 0.0009316831683168317,
+      "loss": 1.2651,
+      "step": 3090
+    },
+    {
+      "epoch": 2.635014116147051,
+      "grad_norm": 18225.01953125,
+      "learning_rate": 0.0009306930693069308,
+      "loss": 1.2608,
+      "step": 3100
+    },
+    {
+      "epoch": 2.643514161683009,
+      "grad_norm": 19320.2890625,
+      "learning_rate": 0.0009297029702970298,
+      "loss": 1.2628,
+      "step": 3110
+    },
+    {
+      "epoch": 2.652014207218967,
+      "grad_norm": 19210.060546875,
+      "learning_rate": 0.0009287128712871288,
+      "loss": 1.2602,
+      "step": 3120
+    },
+    {
+      "epoch": 2.6605142527549255,
+      "grad_norm": 20046.904296875,
+      "learning_rate": 0.0009277227722772278,
+      "loss": 1.2569,
+      "step": 3130
+    },
+    {
+      "epoch": 2.6690142982908838,
+      "grad_norm": 20291.744140625,
+      "learning_rate": 0.0009267326732673268,
+      "loss": 1.2546,
+      "step": 3140
+    },
+    {
+      "epoch": 2.677514343826842,
+      "grad_norm": 17438.1796875,
+      "learning_rate": 0.0009257425742574258,
+      "loss": 1.2553,
+      "step": 3150
+    },
+    {
+      "epoch": 2.677514343826842,
+      "eval_accuracy": 0.7340778616085779,
+      "eval_loss": 1.1635637283325195,
+      "eval_runtime": 1283.8261,
+      "eval_samples_per_second": 388.976,
+      "eval_steps_per_second": 1.351,
+      "step": 3150
+    },
+    {
+      "epoch": 2.6860143893628,
+      "grad_norm": 16703.447265625,
+      "learning_rate": 0.0009247524752475249,
+      "loss": 1.2526,
+      "step": 3160
+    },
+    {
+      "epoch": 2.6945144348987586,
+      "grad_norm": 16225.796875,
+      "learning_rate": 0.0009237623762376238,
+      "loss": 1.252,
+      "step": 3170
+    },
+    {
+      "epoch": 2.7030144804347165,
+      "grad_norm": 19865.849609375,
+      "learning_rate": 0.0009227722772277229,
+      "loss": 1.2472,
+      "step": 3180
+    },
+    {
+      "epoch": 2.7115145259706748,
+      "grad_norm": 18682.638671875,
+      "learning_rate": 0.0009217821782178218,
+      "loss": 1.2485,
+      "step": 3190
+    },
+    {
+      "epoch": 2.720014571506633,
+      "grad_norm": 24501.96875,
+      "learning_rate": 0.0009207920792079209,
+      "loss": 1.2447,
+      "step": 3200
+    },
+    {
+      "epoch": 2.7285146170425914,
+      "grad_norm": 16843.01953125,
+      "learning_rate": 0.0009198019801980199,
+      "loss": 1.2431,
+      "step": 3210
+    },
+    {
+      "epoch": 2.7370146625785496,
+      "grad_norm": 19249.625,
+      "learning_rate": 0.0009188118811881188,
+      "loss": 1.2404,
+      "step": 3220
+    },
+    {
+      "epoch": 2.7455147081145075,
+      "grad_norm": 16381.576171875,
+      "learning_rate": 0.0009178217821782179,
+      "loss": 1.2402,
+      "step": 3230
+    },
+    {
+      "epoch": 2.754014753650466,
+      "grad_norm": 17400.1015625,
+      "learning_rate": 0.0009168316831683168,
+      "loss": 1.2352,
+      "step": 3240
+    },
+    {
+      "epoch": 2.754014753650466,
+      "eval_accuracy": 0.738359822040415,
+      "eval_loss": 1.1412700414657593,
+      "eval_runtime": 1278.9235,
+      "eval_samples_per_second": 390.467,
+      "eval_steps_per_second": 1.356,
+      "step": 3240
+    },
+    {
+      "epoch": 2.762514799186424,
+      "grad_norm": 20167.6328125,
+      "learning_rate": 0.0009158415841584159,
+      "loss": 1.2357,
+      "step": 3250
+    },
+    {
+      "epoch": 2.7710148447223824,
+      "grad_norm": 17094.001953125,
+      "learning_rate": 0.000914851485148515,
+      "loss": 1.2327,
+      "step": 3260
+    },
+    {
+      "epoch": 2.7795148902583406,
+      "grad_norm": 19824.177734375,
+      "learning_rate": 0.0009138613861386139,
+      "loss": 1.2323,
+      "step": 3270
+    },
+    {
+      "epoch": 2.788014935794299,
+      "grad_norm": 20459.54296875,
+      "learning_rate": 0.0009128712871287129,
+      "loss": 1.2317,
+      "step": 3280
+    },
+    {
+      "epoch": 2.796514981330257,
+      "grad_norm": 19844.548828125,
+      "learning_rate": 0.0009118811881188119,
+      "loss": 1.2303,
+      "step": 3290
+    },
+    {
+      "epoch": 2.8050150268662155,
+      "grad_norm": 15563.7314453125,
+      "learning_rate": 0.0009108910891089109,
+      "loss": 1.2254,
+      "step": 3300
+    },
+    {
+      "epoch": 2.813515072402174,
+      "grad_norm": 19343.908203125,
+      "learning_rate": 0.00090990099009901,
+      "loss": 1.2253,
+      "step": 3310
+    },
+    {
+      "epoch": 2.8220151179381316,
+      "grad_norm": 15785.529296875,
+      "learning_rate": 0.0009089108910891089,
+      "loss": 1.2263,
+      "step": 3320
+    },
+    {
+      "epoch": 2.83051516347409,
+      "grad_norm": 21025.06640625,
+      "learning_rate": 0.000907920792079208,
+      "loss": 1.224,
+      "step": 3330
+    },
+    {
+      "epoch": 2.83051516347409,
+      "eval_accuracy": 0.7409290871538374,
+      "eval_loss": 1.1298062801361084,
+      "eval_runtime": 1281.4127,
+      "eval_samples_per_second": 389.709,
+      "eval_steps_per_second": 1.353,
+      "step": 3330
+    },
+    {
+      "epoch": 2.8390152090100482,
+      "grad_norm": 16538.94140625,
+      "learning_rate": 0.0009069306930693069,
+      "loss": 1.223,
+      "step": 3340
+    },
+    {
+      "epoch": 2.8475152545460065,
+      "grad_norm": 18982.095703125,
+      "learning_rate": 0.000905940594059406,
+      "loss": 1.2217,
+      "step": 3350
+    },
+    {
+      "epoch": 2.856015300081965,
+      "grad_norm": 18279.283203125,
+      "learning_rate": 0.000904950495049505,
+      "loss": 1.2178,
+      "step": 3360
+    },
+    {
+      "epoch": 2.864515345617923,
+      "grad_norm": 19073.97265625,
+      "learning_rate": 0.000903960396039604,
+      "loss": 1.2142,
+      "step": 3370
+    },
+    {
+      "epoch": 2.8730153911538814,
+      "grad_norm": 18313.646484375,
+      "learning_rate": 0.000902970297029703,
+      "loss": 1.2164,
+      "step": 3380
+    },
+    {
+      "epoch": 2.8815154366898392,
+      "grad_norm": 18153.08203125,
+      "learning_rate": 0.000901980198019802,
+      "loss": 1.2132,
+      "step": 3390
+    },
+    {
+      "epoch": 2.8900154822257975,
+      "grad_norm": 17545.95703125,
+      "learning_rate": 0.000900990099009901,
+      "loss": 1.211,
+      "step": 3400
+    },
+    {
+      "epoch": 2.898515527761756,
+      "grad_norm": 17355.42578125,
+      "learning_rate": 0.0009000000000000001,
+      "loss": 1.2137,
+      "step": 3410
+    },
+    {
+      "epoch": 2.907015573297714,
+      "grad_norm": 16659.796875,
+      "learning_rate": 0.000899009900990099,
+      "loss": 1.2094,
+      "step": 3420
+    },
+    {
+      "epoch": 2.907015573297714,
+      "eval_accuracy": 0.7434043563853827,
+      "eval_loss": 1.1177165508270264,
+      "eval_runtime": 1280.5435,
+      "eval_samples_per_second": 389.973,
+      "eval_steps_per_second": 1.354,
+      "step": 3420
+    },
+    {
+      "epoch": 2.9155156188336724,
+      "grad_norm": 20371.392578125,
+      "learning_rate": 0.0008980198019801981,
+      "loss": 1.2082,
+      "step": 3430
+    },
+    {
+      "epoch": 2.9240156643696307,
+      "grad_norm": 20884.041015625,
+      "learning_rate": 0.000897029702970297,
+      "loss": 1.2049,
+      "step": 3440
+    },
+    {
+      "epoch": 2.932515709905589,
+      "grad_norm": 18348.6328125,
+      "learning_rate": 0.0008960396039603961,
+      "loss": 1.2043,
+      "step": 3450
+    },
+    {
+      "epoch": 2.941015755441547,
+      "grad_norm": 17201.962890625,
+      "learning_rate": 0.0008950495049504951,
+      "loss": 1.2023,
+      "step": 3460
+    },
+    {
+      "epoch": 2.949515800977505,
+      "grad_norm": 18133.63671875,
+      "learning_rate": 0.0008940594059405941,
+      "loss": 1.2026,
+      "step": 3470
+    },
+    {
+      "epoch": 2.9580158465134634,
+      "grad_norm": 16873.494140625,
+      "learning_rate": 0.0008930693069306931,
+      "loss": 1.1983,
+      "step": 3480
+    },
+    {
+      "epoch": 2.9665158920494217,
+      "grad_norm": 18968.74609375,
+      "learning_rate": 0.0008920792079207921,
+      "loss": 1.2007,
+      "step": 3490
+    },
+    {
+      "epoch": 2.97501593758538,
+      "grad_norm": 16066.7578125,
+      "learning_rate": 0.0008910891089108911,
+      "loss": 1.1992,
+      "step": 3500
+    },
+    {
+      "epoch": 2.9835159831213383,
+      "grad_norm": 18147.33203125,
+      "learning_rate": 0.0008900990099009902,
+      "loss": 1.1955,
+      "step": 3510
+    },
+    {
+      "epoch": 2.9835159831213383,
+      "eval_accuracy": 0.7461886735137164,
+      "eval_loss": 1.101768970489502,
+      "eval_runtime": 1279.5411,
+      "eval_samples_per_second": 390.279,
+      "eval_steps_per_second": 1.355,
+      "step": 3510
+    },
+    {
+      "epoch": 2.9920160286572965,
+      "grad_norm": 16867.4296875,
+      "learning_rate": 0.0008891089108910891,
+      "loss": 1.1933,
+      "step": 3520
+    },
+    {
+      "epoch": 3.0005160741932544,
+      "grad_norm": 17814.7109375,
+      "learning_rate": 0.0008881188118811882,
+      "loss": 1.1949,
+      "step": 3530
+    },
+    {
+      "epoch": 3.0090161197292127,
+      "grad_norm": 17680.919921875,
+      "learning_rate": 0.0008871287128712871,
+      "loss": 1.1922,
+      "step": 3540
+    },
+    {
+      "epoch": 3.017516165265171,
+      "grad_norm": 20543.033203125,
+      "learning_rate": 0.0008861386138613862,
+      "loss": 1.1924,
+      "step": 3550
+    },
+    {
+      "epoch": 3.0260162108011293,
+      "grad_norm": 22293.501953125,
+      "learning_rate": 0.0008851485148514852,
+      "loss": 1.1918,
+      "step": 3560
+    },
+    {
+      "epoch": 3.0345162563370875,
+      "grad_norm": 18199.220703125,
+      "learning_rate": 0.0008841584158415842,
+      "loss": 1.1886,
+      "step": 3570
+    },
+    {
+      "epoch": 3.043016301873046,
+      "grad_norm": 17805.478515625,
+      "learning_rate": 0.0008831683168316832,
+      "loss": 1.1878,
+      "step": 3580
+    },
+    {
+      "epoch": 3.051516347409004,
+      "grad_norm": 17259.236328125,
+      "learning_rate": 0.0008821782178217822,
+      "loss": 1.1862,
+      "step": 3590
+    },
+    {
+      "epoch": 3.0600163929449624,
+      "grad_norm": 15020.2236328125,
+      "learning_rate": 0.0008811881188118812,
+      "loss": 1.1859,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0600163929449624,
+      "eval_accuracy": 0.7483353625113318,
+      "eval_loss": 1.0916943550109863,
+      "eval_runtime": 1279.6266,
+      "eval_samples_per_second": 390.253,
+      "eval_steps_per_second": 1.355,
+      "step": 3600
+    },
+    {
+      "epoch": 3.0685164384809203,
+      "grad_norm": 17711.4765625,
+      "learning_rate": 0.0008801980198019803,
+      "loss": 1.1849,
+      "step": 3610
+    },
+    {
+      "epoch": 3.0770164840168786,
+      "grad_norm": 21087.078125,
+      "learning_rate": 0.0008792079207920792,
+      "loss": 1.1835,
+      "step": 3620
+    },
+    {
+      "epoch": 3.085516529552837,
+      "grad_norm": 15549.55859375,
+      "learning_rate": 0.0008782178217821783,
+      "loss": 1.1821,
+      "step": 3630
+    },
+    {
+      "epoch": 3.094016575088795,
+      "grad_norm": 17686.115234375,
+      "learning_rate": 0.0008772277227722772,
+      "loss": 1.1793,
+      "step": 3640
+    },
+    {
+      "epoch": 3.1025166206247534,
+      "grad_norm": 18682.033203125,
+      "learning_rate": 0.0008762376237623763,
+      "loss": 1.1787,
+      "step": 3650
+    },
+    {
+      "epoch": 3.1110166661607117,
+      "grad_norm": 18502.626953125,
+      "learning_rate": 0.0008752475247524753,
+      "loss": 1.1776,
+      "step": 3660
+    },
+    {
+      "epoch": 3.11951671169667,
+      "grad_norm": 14834.3359375,
+      "learning_rate": 0.0008742574257425743,
+      "loss": 1.1772,
+      "step": 3670
+    },
+    {
+      "epoch": 3.128016757232628,
+      "grad_norm": 17454.373046875,
+      "learning_rate": 0.0008732673267326733,
+      "loss": 1.1766,
+      "step": 3680
+    },
+    {
+      "epoch": 3.136516802768586,
+      "grad_norm": 17697.75,
+      "learning_rate": 0.0008722772277227722,
+      "loss": 1.1764,
+      "step": 3690
+    },
+    {
+      "epoch": 3.136516802768586,
+      "eval_accuracy": 0.7508950932738141,
+      "eval_loss": 1.0795581340789795,
+      "eval_runtime": 1281.2745,
+      "eval_samples_per_second": 389.751,
+      "eval_steps_per_second": 1.353,
+      "step": 3690
+    },
+    {
+      "epoch": 3.1450168483045444,
+      "grad_norm": 20430.919921875,
+      "learning_rate": 0.0008712871287128713,
+      "loss": 1.1738,
+      "step": 3700
+    },
+    {
+      "epoch": 3.1535168938405027,
+      "grad_norm": 16903.4453125,
+      "learning_rate": 0.0008702970297029704,
+      "loss": 1.1709,
+      "step": 3710
+    },
+    {
+      "epoch": 3.162016939376461,
+      "grad_norm": 17143.486328125,
+      "learning_rate": 0.0008693069306930693,
+      "loss": 1.1698,
+      "step": 3720
+    },
+    {
+      "epoch": 3.1705169849124193,
+      "grad_norm": 15930.9501953125,
+      "learning_rate": 0.0008683168316831684,
+      "loss": 1.1696,
+      "step": 3730
+    },
+    {
+      "epoch": 3.1790170304483776,
+      "grad_norm": 16211.982421875,
+      "learning_rate": 0.0008673267326732673,
+      "loss": 1.169,
+      "step": 3740
+    },
+    {
+      "epoch": 3.1875170759843354,
+      "grad_norm": 19266.4140625,
+      "learning_rate": 0.0008663366336633663,
+      "loss": 1.1673,
+      "step": 3750
+    },
+    {
+      "epoch": 3.1960171215202937,
+      "grad_norm": 18455.83203125,
+      "learning_rate": 0.0008653465346534654,
+      "loss": 1.1682,
+      "step": 3760
+    },
+    {
+      "epoch": 3.204517167056252,
+      "grad_norm": 17270.205078125,
+      "learning_rate": 0.0008643564356435643,
+      "loss": 1.1662,
+      "step": 3770
+    },
+    {
+      "epoch": 3.2130172125922103,
+      "grad_norm": 15437.96484375,
+      "learning_rate": 0.0008633663366336634,
+      "loss": 1.1654,
+      "step": 3780
+    },
+    {
+      "epoch": 3.2130172125922103,
+      "eval_accuracy": 0.7525794036416251,
+      "eval_loss": 1.0734323263168335,
+      "eval_runtime": 1278.3867,
+      "eval_samples_per_second": 390.631,
+      "eval_steps_per_second": 1.356,
+      "step": 3780
+    },
+    {
+      "epoch": 3.2215172581281686,
+      "grad_norm": 16193.529296875,
+      "learning_rate": 0.0008623762376237623,
+      "loss": 1.1628,
+      "step": 3790
+    },
+    {
+      "epoch": 3.230017303664127,
+      "grad_norm": 17204.466796875,
+      "learning_rate": 0.0008613861386138614,
+      "loss": 1.1631,
+      "step": 3800
+    },
+    {
+      "epoch": 3.238517349200085,
+      "grad_norm": 17406.2265625,
+      "learning_rate": 0.0008603960396039604,
+      "loss": 1.1621,
+      "step": 3810
+    },
+    {
+      "epoch": 3.2470173947360434,
+      "grad_norm": 17054.076171875,
+      "learning_rate": 0.0008594059405940594,
+      "loss": 1.1614,
+      "step": 3820
+    },
+    {
+      "epoch": 3.2555174402720013,
+      "grad_norm": 18476.68359375,
+      "learning_rate": 0.0008584158415841584,
+      "loss": 1.1606,
+      "step": 3830
+    },
+    {
+      "epoch": 3.2640174858079596,
+      "grad_norm": 19170.802734375,
+      "learning_rate": 0.0008574257425742574,
+      "loss": 1.1581,
+      "step": 3840
+    },
+    {
+      "epoch": 3.272517531343918,
+      "grad_norm": 17178.326171875,
+      "learning_rate": 0.0008564356435643564,
+      "loss": 1.1562,
+      "step": 3850
+    },
+    {
+      "epoch": 3.281017576879876,
+      "grad_norm": 17949.396484375,
+      "learning_rate": 0.0008554455445544555,
+      "loss": 1.1557,
+      "step": 3860
+    },
+    {
+      "epoch": 3.2895176224158345,
+      "grad_norm": 15490.787109375,
+      "learning_rate": 0.0008544554455445544,
+      "loss": 1.1571,
+      "step": 3870
+    },
+    {
+      "epoch": 3.2895176224158345,
+      "eval_accuracy": 0.754430037196134,
+      "eval_loss": 1.0631576776504517,
+      "eval_runtime": 1278.1739,
+      "eval_samples_per_second": 390.696,
+      "eval_steps_per_second": 1.357,
+      "step": 3870
+    },
+    {
+      "epoch": 3.2980176679517927,
+      "grad_norm": 17647.640625,
+      "learning_rate": 0.0008534653465346535,
+      "loss": 1.1552,
+      "step": 3880
+    },
+    {
+      "epoch": 3.306517713487751,
+      "grad_norm": 16109.29296875,
+      "learning_rate": 0.0008524752475247524,
+      "loss": 1.153,
+      "step": 3890
+    },
+    {
+      "epoch": 3.315017759023709,
+      "grad_norm": 19355.884765625,
+      "learning_rate": 0.0008514851485148515,
+      "loss": 1.1543,
+      "step": 3900
+    },
+    {
+      "epoch": 3.323517804559667,
+      "grad_norm": 15843.697265625,
+      "learning_rate": 0.0008504950495049505,
+      "loss": 1.1533,
+      "step": 3910
+    },
+    {
+      "epoch": 3.3320178500956255,
+      "grad_norm": 20818.416015625,
+      "learning_rate": 0.0008495049504950495,
+      "loss": 1.1497,
+      "step": 3920
+    },
+    {
+      "epoch": 3.3405178956315837,
+      "grad_norm": 17527.841796875,
+      "learning_rate": 0.0008485148514851485,
+      "loss": 1.1508,
+      "step": 3930
+    },
+    {
+      "epoch": 3.349017941167542,
+      "grad_norm": 20557.205078125,
+      "learning_rate": 0.0008475247524752475,
+      "loss": 1.1508,
+      "step": 3940
+    },
+    {
+      "epoch": 3.3575179867035003,
+      "grad_norm": 17984.611328125,
+      "learning_rate": 0.0008465346534653465,
+      "loss": 1.1462,
+      "step": 3950
+    },
+    {
+      "epoch": 3.3660180322394586,
+      "grad_norm": 15756.77734375,
+      "learning_rate": 0.0008455445544554456,
+      "loss": 1.1456,
+      "step": 3960
+    },
+    {
+      "epoch": 3.3660180322394586,
+      "eval_accuracy": 0.7564186600937407,
+      "eval_loss": 1.0549671649932861,
+      "eval_runtime": 1283.4177,
+      "eval_samples_per_second": 389.1,
+      "eval_steps_per_second": 1.351,
+      "step": 3960
+    },
+    {
+      "epoch": 3.3745180777754165,
+      "grad_norm": 15412.625,
+      "learning_rate": 0.0008445544554455445,
+      "loss": 1.1451,
+      "step": 3970
+    },
+    {
+      "epoch": 3.3830181233113747,
+      "grad_norm": 16771.29296875,
+      "learning_rate": 0.0008435643564356436,
+      "loss": 1.1465,
+      "step": 3980
+    },
+    {
+      "epoch": 3.391518168847333,
+      "grad_norm": 18703.853515625,
+      "learning_rate": 0.0008425742574257425,
+      "loss": 1.1454,
+      "step": 3990
+    },
+    {
+      "epoch": 3.4000182143832913,
+      "grad_norm": 15575.3798828125,
+      "learning_rate": 0.0008415841584158416,
+      "loss": 1.1453,
+      "step": 4000
+    },
+    {
+      "epoch": 3.4085182599192496,
+      "grad_norm": 18316.923828125,
+      "learning_rate": 0.0008405940594059406,
+      "loss": 1.1408,
+      "step": 4010
+    },
+    {
+      "epoch": 3.417018305455208,
+      "grad_norm": 18408.486328125,
+      "learning_rate": 0.0008396039603960396,
+      "loss": 1.1432,
+      "step": 4020
+    },
+    {
+      "epoch": 3.425518350991166,
+      "grad_norm": 18312.7734375,
+      "learning_rate": 0.0008386138613861386,
+      "loss": 1.1404,
+      "step": 4030
+    },
+    {
+      "epoch": 3.434018396527124,
+      "grad_norm": 15673.939453125,
+      "learning_rate": 0.0008376237623762376,
+      "loss": 1.1406,
+      "step": 4040
+    },
+    {
+      "epoch": 3.4425184420630823,
+      "grad_norm": 18445.818359375,
+      "learning_rate": 0.0008366336633663366,
+      "loss": 1.1391,
+      "step": 4050
+    },
+    {
+      "epoch": 3.4425184420630823,
+      "eval_accuracy": 0.7579978835836059,
+      "eval_loss": 1.0442180633544922,
+      "eval_runtime": 1278.59,
+      "eval_samples_per_second": 390.569,
+      "eval_steps_per_second": 1.356,
+      "step": 4050
+    },
+    {
+      "epoch": 3.4510184875990406,
+      "grad_norm": 21039.810546875,
+      "learning_rate": 0.0008356435643564357,
+      "loss": 1.1395,
+      "step": 4060
+    },
+    {
+      "epoch": 3.459518533134999,
+      "grad_norm": 15481.697265625,
+      "learning_rate": 0.0008346534653465346,
+      "loss": 1.1379,
+      "step": 4070
+    },
+    {
+      "epoch": 3.468018578670957,
+      "grad_norm": 17034.869140625,
+      "learning_rate": 0.0008336633663366337,
+      "loss": 1.1365,
+      "step": 4080
+    },
+    {
+      "epoch": 3.4765186242069155,
+      "grad_norm": 17020.0703125,
+      "learning_rate": 0.0008326732673267326,
+      "loss": 1.1331,
+      "step": 4090
+    },
+    {
+      "epoch": 3.4850186697428738,
+      "grad_norm": 19914.296875,
+      "learning_rate": 0.0008316831683168317,
+      "loss": 1.1352,
+      "step": 4100
+    },
+    {
+      "epoch": 3.4935187152788316,
+      "grad_norm": 19380.546875,
+      "learning_rate": 0.0008306930693069307,
+      "loss": 1.1334,
+      "step": 4110
+    },
+    {
+      "epoch": 3.5020187608147904,
+      "grad_norm": 16126.1982421875,
+      "learning_rate": 0.0008297029702970297,
+      "loss": 1.133,
+      "step": 4120
+    },
+    {
+      "epoch": 3.510518806350748,
+      "grad_norm": 16914.3046875,
+      "learning_rate": 0.0008287128712871287,
+      "loss": 1.1328,
+      "step": 4130
+    },
+    {
+      "epoch": 3.5190188518867065,
+      "grad_norm": 15453.5771484375,
+      "learning_rate": 0.0008277227722772277,
+      "loss": 1.1302,
+      "step": 4140
+    },
+    {
+      "epoch": 3.5190188518867065,
+      "eval_accuracy": 0.7592944501997413,
+      "eval_loss": 1.0387505292892456,
+      "eval_runtime": 1280.9395,
+      "eval_samples_per_second": 389.853,
+      "eval_steps_per_second": 1.354,
+      "step": 4140
+    },
+    {
+      "epoch": 3.527518897422665,
+      "grad_norm": 17636.76171875,
+      "learning_rate": 0.0008267326732673267,
+      "loss": 1.1313,
+      "step": 4150
+    },
+    {
+      "epoch": 3.536018942958623,
+      "grad_norm": 16709.236328125,
+      "learning_rate": 0.0008257425742574258,
+      "loss": 1.1307,
+      "step": 4160
+    },
+    {
+      "epoch": 3.5445189884945814,
+      "grad_norm": 18499.865234375,
+      "learning_rate": 0.0008247524752475247,
+      "loss": 1.1294,
+      "step": 4170
+    },
+    {
+      "epoch": 3.553019034030539,
+      "grad_norm": 16481.85546875,
+      "learning_rate": 0.0008237623762376238,
+      "loss": 1.1276,
+      "step": 4180
+    },
+    {
+      "epoch": 3.561519079566498,
+      "grad_norm": 16035.626953125,
+      "learning_rate": 0.0008227722772277227,
+      "loss": 1.1259,
+      "step": 4190
+    },
+    {
+      "epoch": 3.570019125102456,
+      "grad_norm": 17429.3046875,
+      "learning_rate": 0.0008217821782178218,
+      "loss": 1.1281,
+      "step": 4200
+    },
+    {
+      "epoch": 3.578519170638414,
+      "grad_norm": 17563.748046875,
+      "learning_rate": 0.0008207920792079208,
+      "loss": 1.1244,
+      "step": 4210
+    },
+    {
+      "epoch": 3.5870192161743724,
+      "grad_norm": 17140.78125,
+      "learning_rate": 0.0008198019801980197,
+      "loss": 1.1264,
+      "step": 4220
+    },
+    {
+      "epoch": 3.5955192617103306,
+      "grad_norm": 18145.185546875,
+      "learning_rate": 0.0008188118811881188,
+      "loss": 1.1229,
+      "step": 4230
+    },
+    {
+      "epoch": 3.5955192617103306,
+      "eval_accuracy": 0.7607957295247235,
+      "eval_loss": 1.0331238508224487,
+      "eval_runtime": 1281.4596,
+      "eval_samples_per_second": 389.695,
+      "eval_steps_per_second": 1.353,
+      "step": 4230
+    },
+    {
+      "epoch": 3.604019307246289,
+      "grad_norm": 14324.3125,
+      "learning_rate": 0.0008178217821782177,
+      "loss": 1.124,
+      "step": 4240
+    },
+    {
+      "epoch": 3.612519352782247,
+      "grad_norm": 19569.59375,
+      "learning_rate": 0.0008168316831683168,
+      "loss": 1.124,
+      "step": 4250
+    },
+    {
+      "epoch": 3.6210193983182055,
+      "grad_norm": 18639.96484375,
+      "learning_rate": 0.0008158415841584159,
+      "loss": 1.1221,
+      "step": 4260
+    },
+    {
+      "epoch": 3.6295194438541634,
+      "grad_norm": 17077.609375,
+      "learning_rate": 0.0008148514851485148,
+      "loss": 1.1177,
+      "step": 4270
+    },
+    {
+      "epoch": 3.6380194893901217,
+      "grad_norm": 19580.677734375,
+      "learning_rate": 0.0008138613861386138,
+      "loss": 1.1208,
+      "step": 4280
+    },
+    {
+      "epoch": 3.64651953492608,
+      "grad_norm": 14986.6796875,
+      "learning_rate": 0.0008128712871287128,
+      "loss": 1.1192,
+      "step": 4290
+    },
+    {
+      "epoch": 3.6550195804620382,
+      "grad_norm": 16046.3740234375,
+      "learning_rate": 0.000811881188118812,
+      "loss": 1.1193,
+      "step": 4300
+    },
+    {
+      "epoch": 3.6635196259979965,
+      "grad_norm": 14974.083984375,
+      "learning_rate": 0.000810891089108911,
+      "loss": 1.1162,
+      "step": 4310
+    },
+    {
+      "epoch": 3.672019671533955,
+      "grad_norm": 17986.744140625,
+      "learning_rate": 0.00080990099009901,
+      "loss": 1.1183,
+      "step": 4320
+    },
+    {
+      "epoch": 3.672019671533955,
+      "eval_accuracy": 0.7619108635382325,
+      "eval_loss": 1.0262423753738403,
+      "eval_runtime": 1281.4568,
+      "eval_samples_per_second": 389.696,
+      "eval_steps_per_second": 1.353,
+      "step": 4320
+    },
+    {
+      "epoch": 3.680519717069913,
+      "grad_norm": 19318.87109375,
+      "learning_rate": 0.000808910891089109,
+      "loss": 1.1153,
+      "step": 4330
+    },
+    {
+      "epoch": 3.689019762605871,
+      "grad_norm": 18132.083984375,
+      "learning_rate": 0.0008079207920792079,
+      "loss": 1.1155,
+      "step": 4340
+    },
+    {
+      "epoch": 3.6975198081418292,
+      "grad_norm": 15848.3271484375,
+      "learning_rate": 0.000806930693069307,
+      "loss": 1.1172,
+      "step": 4350
+    },
+    {
+      "epoch": 3.7060198536777875,
+      "grad_norm": 16213.193359375,
+      "learning_rate": 0.000805940594059406,
+      "loss": 1.1136,
+      "step": 4360
+    },
+    {
+      "epoch": 3.714519899213746,
+      "grad_norm": 18759.615234375,
+      "learning_rate": 0.000804950495049505,
+      "loss": 1.1129,
+      "step": 4370
+    },
+    {
+      "epoch": 3.723019944749704,
+      "grad_norm": 19659.138671875,
+      "learning_rate": 0.000803960396039604,
+      "loss": 1.115,
+      "step": 4380
+    },
+    {
+      "epoch": 3.7315199902856624,
+      "grad_norm": 19299.97265625,
+      "learning_rate": 0.000802970297029703,
+      "loss": 1.1124,
+      "step": 4390
+    },
+    {
+      "epoch": 3.7400200358216207,
+      "grad_norm": 15511.5537109375,
+      "learning_rate": 0.000801980198019802,
+      "loss": 1.1121,
+      "step": 4400
+    },
+    {
+      "epoch": 3.7485200813575785,
+      "grad_norm": 18630.70703125,
+      "learning_rate": 0.0008009900990099011,
+      "loss": 1.1106,
+      "step": 4410
+    },
+    {
+      "epoch": 3.7485200813575785,
+      "eval_accuracy": 0.763328105961178,
+      "eval_loss": 1.0201424360275269,
+      "eval_runtime": 1277.6344,
+      "eval_samples_per_second": 390.861,
+      "eval_steps_per_second": 1.357,
+      "step": 4410
+    },
+    {
+      "epoch": 3.757020126893537,
+      "grad_norm": 17692.544921875,
+      "learning_rate": 0.0008,
+      "loss": 1.1116,
+      "step": 4420
+    },
+    {
+      "epoch": 3.765520172429495,
+      "grad_norm": 14010.234375,
+      "learning_rate": 0.0007990099009900991,
+      "loss": 1.1089,
+      "step": 4430
+    },
+    {
+      "epoch": 3.7740202179654534,
+      "grad_norm": 18022.48828125,
+      "learning_rate": 0.000798019801980198,
+      "loss": 1.1082,
+      "step": 4440
+    },
+    {
+      "epoch": 3.7825202635014117,
+      "grad_norm": 16178.5849609375,
+      "learning_rate": 0.0007970297029702971,
+      "loss": 1.108,
+      "step": 4450
+    },
+    {
+      "epoch": 3.79102030903737,
+      "grad_norm": 16874.7734375,
+      "learning_rate": 0.0007960396039603961,
+      "loss": 1.1071,
+      "step": 4460
+    },
+    {
+      "epoch": 3.7995203545733283,
+      "grad_norm": 16226.3564453125,
+      "learning_rate": 0.0007950495049504951,
+      "loss": 1.1052,
+      "step": 4470
+    },
+    {
+      "epoch": 3.808020400109286,
+      "grad_norm": 21136.931640625,
+      "learning_rate": 0.0007940594059405941,
+      "loss": 1.1071,
+      "step": 4480
+    },
+    {
+      "epoch": 3.8165204456452444,
+      "grad_norm": 19177.98828125,
+      "learning_rate": 0.0007930693069306931,
+      "loss": 1.1048,
+      "step": 4490
+    },
+    {
+      "epoch": 3.8250204911812027,
+      "grad_norm": 16423.876953125,
+      "learning_rate": 0.0007920792079207921,
+      "loss": 1.1021,
+      "step": 4500
+    },
+    {
+      "epoch": 3.8250204911812027,
+      "eval_accuracy": 0.7649368216993604,
+      "eval_loss": 1.0112674236297607,
+      "eval_runtime": 1282.4987,
+      "eval_samples_per_second": 389.379,
+      "eval_steps_per_second": 1.352,
+      "step": 4500
+    },
+    {
+      "epoch": 3.833520536717161,
+      "grad_norm": 15636.1162109375,
+      "learning_rate": 0.0007910891089108912,
+      "loss": 1.1007,
+      "step": 4510
+    },
+    {
+      "epoch": 3.8420205822531193,
+      "grad_norm": 16542.201171875,
+      "learning_rate": 0.0007900990099009901,
+      "loss": 1.1032,
+      "step": 4520
+    },
+    {
+      "epoch": 3.8505206277890776,
+      "grad_norm": 17437.408203125,
+      "learning_rate": 0.0007891089108910892,
+      "loss": 1.1015,
+      "step": 4530
+    },
+    {
+      "epoch": 3.859020673325036,
+      "grad_norm": 19502.89453125,
+      "learning_rate": 0.0007881188118811881,
+      "loss": 1.1025,
+      "step": 4540
+    },
+    {
+      "epoch": 3.8675207188609937,
+      "grad_norm": 18670.546875,
+      "learning_rate": 0.0007871287128712872,
+      "loss": 1.0993,
+      "step": 4550
+    },
+    {
+      "epoch": 3.876020764396952,
+      "grad_norm": 18770.947265625,
+      "learning_rate": 0.0007861386138613862,
+      "loss": 1.1026,
+      "step": 4560
+    },
+    {
+      "epoch": 3.8845208099329103,
+      "grad_norm": 17557.958984375,
+      "learning_rate": 0.0007851485148514852,
+      "loss": 1.0983,
+      "step": 4570
+    },
+    {
+      "epoch": 3.8930208554688686,
+      "grad_norm": 18551.2265625,
+      "learning_rate": 0.0007841584158415842,
+      "loss": 1.1001,
+      "step": 4580
+    },
+    {
+      "epoch": 3.901520901004827,
+      "grad_norm": 17471.43359375,
+      "learning_rate": 0.0007831683168316832,
+      "loss": 1.097,
+      "step": 4590
+    },
+    {
+      "epoch": 3.901520901004827,
+      "eval_accuracy": 0.7659643920785019,
+      "eval_loss": 1.0095082521438599,
+      "eval_runtime": 1281.981,
+      "eval_samples_per_second": 389.536,
+      "eval_steps_per_second": 1.353,
+      "step": 4590
+    },
+    {
+      "epoch": 3.910020946540785,
+      "grad_norm": 17272.111328125,
+      "learning_rate": 0.0007821782178217822,
+      "loss": 1.0978,
+      "step": 4600
+    },
+    {
+      "epoch": 3.9185209920767434,
+      "grad_norm": 13404.1982421875,
+      "learning_rate": 0.0007811881188118813,
+      "loss": 1.0995,
+      "step": 4610
+    },
+    {
+      "epoch": 3.9270210376127013,
+      "grad_norm": 14501.0966796875,
+      "learning_rate": 0.0007801980198019802,
+      "loss": 1.0989,
+      "step": 4620
+    },
+    {
+      "epoch": 3.9355210831486596,
+      "grad_norm": 19314.373046875,
+      "learning_rate": 0.0007792079207920793,
+      "loss": 1.0949,
+      "step": 4630
+    },
+    {
+      "epoch": 3.944021128684618,
+      "grad_norm": 17886.853515625,
+      "learning_rate": 0.0007782178217821782,
+      "loss": 1.097,
+      "step": 4640
+    },
+    {
+      "epoch": 3.952521174220576,
+      "grad_norm": 16101.6513671875,
+      "learning_rate": 0.0007772277227722773,
+      "loss": 1.0922,
+      "step": 4650
+    },
+    {
+      "epoch": 3.9610212197565344,
+      "grad_norm": 16483.103515625,
+      "learning_rate": 0.0007762376237623763,
+      "loss": 1.094,
+      "step": 4660
+    },
+    {
+      "epoch": 3.9695212652924927,
+      "grad_norm": 17740.6640625,
+      "learning_rate": 0.0007752475247524753,
+      "loss": 1.0923,
+      "step": 4670
+    },
+    {
+      "epoch": 3.978021310828451,
+      "grad_norm": 18332.16015625,
+      "learning_rate": 0.0007742574257425743,
+      "loss": 1.0932,
+      "step": 4680
+    },
+    {
+      "epoch": 3.978021310828451,
+      "eval_accuracy": 0.7671802734913508,
+      "eval_loss": 1.0010571479797363,
+      "eval_runtime": 1282.4979,
+      "eval_samples_per_second": 389.379,
+      "eval_steps_per_second": 1.352,
+      "step": 4680
+    },
+    {
+      "epoch": 3.986521356364409,
+      "grad_norm": 15639.021484375,
+      "learning_rate": 0.0007732673267326733,
+      "loss": 1.0928,
+      "step": 4690
+    },
+    {
+      "epoch": 3.9950214019003676,
+      "grad_norm": 15639.115234375,
+      "learning_rate": 0.0007722772277227723,
+      "loss": 1.0929,
+      "step": 4700
+    },
+    {
+      "epoch": 4.003521447436325,
+      "grad_norm": 16150.44140625,
+      "learning_rate": 0.0007712871287128714,
+      "loss": 1.0907,
+      "step": 4710
+    },
+    {
+      "epoch": 4.012021492972284,
+      "grad_norm": 20656.818359375,
+      "learning_rate": 0.0007702970297029703,
+      "loss": 1.0894,
+      "step": 4720
+    },
+    {
+      "epoch": 4.020521538508242,
+      "grad_norm": 18745.37890625,
+      "learning_rate": 0.0007693069306930694,
+      "loss": 1.0919,
+      "step": 4730
+    },
+    {
+      "epoch": 4.0290215840442,
+      "grad_norm": 16387.470703125,
+      "learning_rate": 0.0007683168316831683,
+      "loss": 1.09,
+      "step": 4740
+    },
+    {
+      "epoch": 4.037521629580159,
+      "grad_norm": 18658.521484375,
+      "learning_rate": 0.0007673267326732674,
+      "loss": 1.0873,
+      "step": 4750
+    },
+    {
+      "epoch": 4.046021675116116,
+      "grad_norm": 19641.572265625,
+      "learning_rate": 0.0007663366336633664,
+      "loss": 1.0886,
+      "step": 4760
+    },
+    {
+      "epoch": 4.054521720652075,
+      "grad_norm": 16653.2578125,
+      "learning_rate": 0.0007653465346534654,
+      "loss": 1.0873,
+      "step": 4770
+    },
+    {
+      "epoch": 4.054521720652075,
+      "eval_accuracy": 0.7683563302945906,
+      "eval_loss": 0.9970803260803223,
+      "eval_runtime": 1282.3032,
+      "eval_samples_per_second": 389.438,
+      "eval_steps_per_second": 1.352,
+      "step": 4770
+    },
+    {
+      "epoch": 4.063021766188033,
+      "grad_norm": 17606.6015625,
+      "learning_rate": 0.0007643564356435644,
+      "loss": 1.0889,
+      "step": 4780
+    },
+    {
+      "epoch": 4.071521811723992,
+      "grad_norm": 17704.697265625,
+      "learning_rate": 0.0007633663366336634,
+      "loss": 1.0869,
+      "step": 4790
+    },
+    {
+      "epoch": 4.08002185725995,
+      "grad_norm": 20424.306640625,
+      "learning_rate": 0.0007623762376237624,
+      "loss": 1.0861,
+      "step": 4800
+    },
+    {
+      "epoch": 4.088521902795907,
+      "grad_norm": 15488.392578125,
+      "learning_rate": 0.0007613861386138615,
+      "loss": 1.0858,
+      "step": 4810
+    },
+    {
+      "epoch": 4.097021948331866,
+      "grad_norm": 18205.36328125,
+      "learning_rate": 0.0007603960396039604,
+      "loss": 1.0847,
+      "step": 4820
+    },
+    {
+      "epoch": 4.105521993867824,
+      "grad_norm": 15557.935546875,
+      "learning_rate": 0.0007594059405940595,
+      "loss": 1.0845,
+      "step": 4830
+    },
+    {
+      "epoch": 4.114022039403783,
+      "grad_norm": 17741.306640625,
+      "learning_rate": 0.0007584158415841584,
+      "loss": 1.0837,
+      "step": 4840
+    },
+    {
+      "epoch": 4.122522084939741,
+      "grad_norm": 21305.34765625,
+      "learning_rate": 0.0007574257425742574,
+      "loss": 1.0841,
+      "step": 4850
+    },
+    {
+      "epoch": 4.131022130475699,
+      "grad_norm": 20442.0390625,
+      "learning_rate": 0.0007564356435643565,
+      "loss": 1.0813,
+      "step": 4860
+    },
+    {
+      "epoch": 4.131022130475699,
+      "eval_accuracy": 0.769627324769287,
+      "eval_loss": 0.9902046322822571,
+      "eval_runtime": 1278.8399,
+      "eval_samples_per_second": 390.493,
+      "eval_steps_per_second": 1.356,
+      "step": 4860
+    },
+    {
+      "epoch": 4.139522176011657,
+      "grad_norm": 19991.08203125,
+      "learning_rate": 0.0007554455445544554,
+      "loss": 1.0809,
+      "step": 4870
+    },
+    {
+      "epoch": 4.148022221547615,
+      "grad_norm": 21327.169921875,
+      "learning_rate": 0.0007544554455445545,
+      "loss": 1.0822,
+      "step": 4880
+    },
+    {
+      "epoch": 4.156522267083574,
+      "grad_norm": 20165.03515625,
+      "learning_rate": 0.0007534653465346534,
+      "loss": 1.0824,
+      "step": 4890
+    },
+    {
+      "epoch": 4.165022312619532,
+      "grad_norm": 15693.8251953125,
+      "learning_rate": 0.0007524752475247525,
+      "loss": 1.0786,
+      "step": 4900
+    },
+    {
+      "epoch": 4.17352235815549,
+      "grad_norm": 18861.962890625,
+      "learning_rate": 0.0007514851485148515,
+      "loss": 1.0802,
+      "step": 4910
+    },
+    {
+      "epoch": 4.182022403691448,
+      "grad_norm": 18349.630859375,
+      "learning_rate": 0.0007504950495049505,
+      "loss": 1.0811,
+      "step": 4920
+    },
+    {
+      "epoch": 4.190522449227407,
+      "grad_norm": 16894.869140625,
+      "learning_rate": 0.0007495049504950495,
+      "loss": 1.0784,
+      "step": 4930
+    },
+    {
+      "epoch": 4.199022494763365,
+      "grad_norm": 15264.0703125,
+      "learning_rate": 0.0007485148514851485,
+      "loss": 1.0779,
+      "step": 4940
+    },
+    {
+      "epoch": 4.207522540299323,
+      "grad_norm": 18500.328125,
+      "learning_rate": 0.0007475247524752475,
+      "loss": 1.0769,
+      "step": 4950
+    },
+    {
+      "epoch": 4.207522540299323,
+      "eval_accuracy": 0.7703979549467554,
+      "eval_loss": 0.9842762351036072,
+      "eval_runtime": 1282.9458,
+      "eval_samples_per_second": 389.243,
+      "eval_steps_per_second": 1.352,
+      "step": 4950
+    },
+    {
+      "epoch": 4.216022585835281,
+      "grad_norm": 16738.04296875,
+      "learning_rate": 0.0007465346534653466,
+      "loss": 1.0763,
+      "step": 4960
+    },
+    {
+      "epoch": 4.224522631371239,
+      "grad_norm": 15980.9921875,
+      "learning_rate": 0.0007455445544554455,
+      "loss": 1.0779,
+      "step": 4970
+    },
+    {
+      "epoch": 4.233022676907198,
+      "grad_norm": 17264.802734375,
+      "learning_rate": 0.0007445544554455446,
+      "loss": 1.0752,
+      "step": 4980
+    },
+    {
+      "epoch": 4.241522722443156,
+      "grad_norm": 19698.16796875,
+      "learning_rate": 0.0007435643564356435,
+      "loss": 1.0726,
+      "step": 4990
+    },
+    {
+      "epoch": 4.2500227679791145,
+      "grad_norm": 16336.4052734375,
+      "learning_rate": 0.0007425742574257426,
+      "loss": 1.0749,
+      "step": 5000
+    },
+    {
+      "epoch": 4.258522813515072,
+      "grad_norm": 17774.734375,
+      "learning_rate": 0.0007415841584158416,
+      "loss": 1.073,
+      "step": 5010
+    },
+    {
+      "epoch": 4.26702285905103,
+      "grad_norm": 19482.98046875,
+      "learning_rate": 0.0007405940594059406,
+      "loss": 1.0754,
+      "step": 5020
+    },
+    {
+      "epoch": 4.275522904586989,
+      "grad_norm": 17985.568359375,
+      "learning_rate": 0.0007396039603960396,
+      "loss": 1.072,
+      "step": 5030
+    },
+    {
+      "epoch": 4.284022950122947,
+      "grad_norm": 16432.27734375,
+      "learning_rate": 0.0007386138613861386,
+      "loss": 1.0737,
+      "step": 5040
+    },
+    {
+      "epoch": 4.284022950122947,
+      "eval_accuracy": 0.7713623846481911,
+      "eval_loss": 0.982131838798523,
+      "eval_runtime": 1279.4531,
+      "eval_samples_per_second": 390.306,
+      "eval_steps_per_second": 1.355,
+      "step": 5040
+    },
+    {
+      "epoch": 4.2925229956589055,
+      "grad_norm": 18672.43359375,
+      "learning_rate": 0.0007376237623762376,
+      "loss": 1.0719,
+      "step": 5050
+    },
+    {
+      "epoch": 4.301023041194863,
+      "grad_norm": 15502.23828125,
+      "learning_rate": 0.0007366336633663367,
+      "loss": 1.0719,
+      "step": 5060
+    },
+    {
+      "epoch": 4.309523086730822,
+      "grad_norm": 17056.31640625,
+      "learning_rate": 0.0007356435643564356,
+      "loss": 1.0708,
+      "step": 5070
+    },
+    {
+      "epoch": 4.31802313226678,
+      "grad_norm": 17404.642578125,
+      "learning_rate": 0.0007346534653465347,
+      "loss": 1.0711,
+      "step": 5080
+    },
+    {
+      "epoch": 4.326523177802738,
+      "grad_norm": 16851.5390625,
+      "learning_rate": 0.0007336633663366336,
+      "loss": 1.0711,
+      "step": 5090
+    },
+    {
+      "epoch": 4.3350232233386965,
+      "grad_norm": 16218.2109375,
+      "learning_rate": 0.0007326732673267327,
+      "loss": 1.0705,
+      "step": 5100
+    },
+    {
+      "epoch": 4.343523268874654,
+      "grad_norm": 15998.1015625,
+      "learning_rate": 0.0007316831683168317,
+      "loss": 1.0706,
+      "step": 5110
+    },
+    {
+      "epoch": 4.352023314410613,
+      "grad_norm": 17217.677734375,
+      "learning_rate": 0.0007306930693069307,
+      "loss": 1.0688,
+      "step": 5120
+    },
+    {
+      "epoch": 4.360523359946571,
+      "grad_norm": 20126.892578125,
+      "learning_rate": 0.0007297029702970297,
+      "loss": 1.0687,
+      "step": 5130
+    },
+    {
+      "epoch": 4.360523359946571,
+      "eval_accuracy": 0.7720324660360576,
+      "eval_loss": 0.9775763154029846,
+      "eval_runtime": 1279.545,
+      "eval_samples_per_second": 390.278,
+      "eval_steps_per_second": 1.355,
+      "step": 5130
+    },
+    {
+      "epoch": 4.36902340548253,
+      "grad_norm": 18316.3203125,
+      "learning_rate": 0.0007287128712871287,
+      "loss": 1.0677,
+      "step": 5140
+    },
+    {
+      "epoch": 4.3775234510184875,
+      "grad_norm": 19130.66796875,
+      "learning_rate": 0.0007277227722772277,
+      "loss": 1.0673,
+      "step": 5150
+    },
+    {
+      "epoch": 4.386023496554446,
+      "grad_norm": 16458.451171875,
+      "learning_rate": 0.0007267326732673268,
+      "loss": 1.0662,
+      "step": 5160
+    },
+    {
+      "epoch": 4.394523542090404,
+      "grad_norm": 15268.9111328125,
+      "learning_rate": 0.0007257425742574257,
+      "loss": 1.0674,
+      "step": 5170
+    },
+    {
+      "epoch": 4.403023587626362,
+      "grad_norm": 15841.1025390625,
+      "learning_rate": 0.0007247524752475248,
+      "loss": 1.0661,
+      "step": 5180
+    },
+    {
+      "epoch": 4.411523633162321,
+      "grad_norm": 15076.5078125,
+      "learning_rate": 0.0007237623762376237,
+      "loss": 1.0656,
+      "step": 5190
+    },
+    {
+      "epoch": 4.4200236786982785,
+      "grad_norm": 17972.560546875,
+      "learning_rate": 0.0007227722772277228,
+      "loss": 1.0653,
+      "step": 5200
+    },
+    {
+      "epoch": 4.428523724234237,
+      "grad_norm": 17051.736328125,
+      "learning_rate": 0.0007217821782178218,
+      "loss": 1.0642,
+      "step": 5210
+    },
+    {
+      "epoch": 4.437023769770195,
+      "grad_norm": 18644.025390625,
+      "learning_rate": 0.0007207920792079208,
+      "loss": 1.0635,
+      "step": 5220
+    },
+    {
+      "epoch": 4.437023769770195,
+      "eval_accuracy": 0.7729909297402872,
+      "eval_loss": 0.9726957678794861,
+      "eval_runtime": 1279.9366,
+      "eval_samples_per_second": 390.158,
+      "eval_steps_per_second": 1.355,
+      "step": 5220
+    },
+    {
+      "epoch": 4.445523815306154,
+      "grad_norm": 17667.255859375,
+      "learning_rate": 0.0007198019801980198,
+      "loss": 1.0627,
+      "step": 5230
+    },
+    {
+      "epoch": 4.454023860842112,
+      "grad_norm": 16890.02734375,
+      "learning_rate": 0.0007188118811881188,
+      "loss": 1.0621,
+      "step": 5240
+    },
+    {
+      "epoch": 4.4625239063780695,
+      "grad_norm": 17795.29296875,
+      "learning_rate": 0.0007178217821782178,
+      "loss": 1.0609,
+      "step": 5250
+    },
+    {
+      "epoch": 4.471023951914028,
+      "grad_norm": 17139.46875,
+      "learning_rate": 0.0007168316831683169,
+      "loss": 1.0626,
+      "step": 5260
+    },
+    {
+      "epoch": 4.479523997449986,
+      "grad_norm": 16425.12109375,
+      "learning_rate": 0.0007158415841584158,
+      "loss": 1.0611,
+      "step": 5270
+    },
+    {
+      "epoch": 4.488024042985945,
+      "grad_norm": 20427.578125,
+      "learning_rate": 0.0007148514851485149,
+      "loss": 1.0575,
+      "step": 5280
+    },
+    {
+      "epoch": 4.496524088521903,
+      "grad_norm": 21385.232421875,
+      "learning_rate": 0.0007138613861386138,
+      "loss": 1.0609,
+      "step": 5290
+    },
+    {
+      "epoch": 4.505024134057861,
+      "grad_norm": 16601.791015625,
+      "learning_rate": 0.0007128712871287129,
+      "loss": 1.0594,
+      "step": 5300
+    },
+    {
+      "epoch": 4.513524179593819,
+      "grad_norm": 17113.41015625,
+      "learning_rate": 0.0007118811881188119,
+      "loss": 1.0583,
+      "step": 5310
+    },
+    {
+      "epoch": 4.513524179593819,
+      "eval_accuracy": 0.7740536654886605,
+      "eval_loss": 0.9671040773391724,
+      "eval_runtime": 1277.3232,
+      "eval_samples_per_second": 390.957,
+      "eval_steps_per_second": 1.358,
+      "step": 5310
+    },
+    {
+      "epoch": 4.522024225129778,
+      "grad_norm": 18315.3984375,
+      "learning_rate": 0.0007108910891089109,
+      "loss": 1.0581,
+      "step": 5320
+    },
+    {
+      "epoch": 4.530524270665736,
+      "grad_norm": 17246.2890625,
+      "learning_rate": 0.0007099009900990099,
+      "loss": 1.0594,
+      "step": 5330
+    },
+    {
+      "epoch": 4.539024316201694,
+      "grad_norm": 19144.822265625,
+      "learning_rate": 0.0007089108910891088,
+      "loss": 1.0577,
+      "step": 5340
+    },
+    {
+      "epoch": 4.547524361737652,
+      "grad_norm": 17691.115234375,
+      "learning_rate": 0.0007079207920792079,
+      "loss": 1.0581,
+      "step": 5350
+    },
+    {
+      "epoch": 4.55602440727361,
+      "grad_norm": 17498.75,
+      "learning_rate": 0.000706930693069307,
+      "loss": 1.0567,
+      "step": 5360
+    },
+    {
+      "epoch": 4.564524452809569,
+      "grad_norm": 19553.53515625,
+      "learning_rate": 0.0007059405940594059,
+      "loss": 1.0565,
+      "step": 5370
+    },
+    {
+      "epoch": 4.573024498345527,
+      "grad_norm": 17747.8359375,
+      "learning_rate": 0.000704950495049505,
+      "loss": 1.0603,
+      "step": 5380
+    },
+    {
+      "epoch": 4.581524543881486,
+      "grad_norm": 17742.814453125,
+      "learning_rate": 0.0007039603960396039,
+      "loss": 1.0567,
+      "step": 5390
+    },
+    {
+      "epoch": 4.590024589417443,
+      "grad_norm": 18410.59765625,
+      "learning_rate": 0.0007029702970297029,
+      "loss": 1.0561,
+      "step": 5400
+    },
+    {
+      "epoch": 4.590024589417443,
+      "eval_accuracy": 0.7741942335122652,
+      "eval_loss": 0.968053936958313,
+      "eval_runtime": 1278.1291,
+      "eval_samples_per_second": 390.71,
+      "eval_steps_per_second": 1.357,
+      "step": 5400
+    },
+    {
+      "epoch": 4.598524634953401,
+      "grad_norm": 14357.056640625,
+      "learning_rate": 0.000701980198019802,
+      "loss": 1.0544,
+      "step": 5410
+    },
+    {
+      "epoch": 4.60702468048936,
+      "grad_norm": 18278.533203125,
+      "learning_rate": 0.0007009900990099009,
+      "loss": 1.0531,
+      "step": 5420
+    },
+    {
+      "epoch": 4.615524726025318,
+      "grad_norm": 18012.599609375,
+      "learning_rate": 0.0007,
+      "loss": 1.0567,
+      "step": 5430
+    },
+    {
+      "epoch": 4.624024771561277,
+      "grad_norm": 18881.583984375,
+      "learning_rate": 0.0006990099009900989,
+      "loss": 1.0548,
+      "step": 5440
+    },
+    {
+      "epoch": 4.632524817097234,
+      "grad_norm": 19899.90234375,
+      "learning_rate": 0.000698019801980198,
+      "loss": 1.0541,
+      "step": 5450
+    },
+    {
+      "epoch": 4.641024862633193,
+      "grad_norm": 15934.2138671875,
+      "learning_rate": 0.000697029702970297,
+      "loss": 1.0536,
+      "step": 5460
+    },
+    {
+      "epoch": 4.649524908169151,
+      "grad_norm": 15930.8544921875,
+      "learning_rate": 0.000696039603960396,
+      "loss": 1.0529,
+      "step": 5470
+    },
+    {
+      "epoch": 4.658024953705109,
+      "grad_norm": 18346.009765625,
+      "learning_rate": 0.000695049504950495,
+      "loss": 1.0507,
+      "step": 5480
+    },
+    {
+      "epoch": 4.666524999241068,
+      "grad_norm": 17350.013671875,
+      "learning_rate": 0.000694059405940594,
+      "loss": 1.0528,
+      "step": 5490
+    },
+    {
+      "epoch": 4.666524999241068,
+      "eval_accuracy": 0.7752448834285515,
+      "eval_loss": 0.9624823927879333,
+      "eval_runtime": 1279.2263,
+      "eval_samples_per_second": 390.375,
+      "eval_steps_per_second": 1.356,
+      "step": 5490
+    },
+    {
+      "epoch": 4.675025044777025,
+      "grad_norm": 17104.44921875,
+      "learning_rate": 0.000693069306930693,
+      "loss": 1.0502,
+      "step": 5500
+    },
+    {
+      "epoch": 4.683525090312984,
+      "grad_norm": 16279.3994140625,
+      "learning_rate": 0.0006920792079207921,
+      "loss": 1.0525,
+      "step": 5510
+    },
+    {
+      "epoch": 4.692025135848942,
+      "grad_norm": 17741.423828125,
+      "learning_rate": 0.000691089108910891,
+      "loss": 1.0495,
+      "step": 5520
+    },
+    {
+      "epoch": 4.700525181384901,
+      "grad_norm": 18454.111328125,
+      "learning_rate": 0.0006900990099009901,
+      "loss": 1.0501,
+      "step": 5530
+    },
+    {
+      "epoch": 4.709025226920859,
+      "grad_norm": 20272.701171875,
+      "learning_rate": 0.000689108910891089,
+      "loss": 1.0518,
+      "step": 5540
+    },
+    {
+      "epoch": 4.717525272456816,
+      "grad_norm": 17187.880859375,
+      "learning_rate": 0.0006881188118811881,
+      "loss": 1.0498,
+      "step": 5550
+    },
+    {
+      "epoch": 4.726025317992775,
+      "grad_norm": 16848.552734375,
+      "learning_rate": 0.0006871287128712872,
+      "loss": 1.0475,
+      "step": 5560
+    },
+    {
+      "epoch": 4.734525363528733,
+      "grad_norm": 18908.748046875,
+      "learning_rate": 0.0006861386138613862,
+      "loss": 1.0477,
+      "step": 5570
+    },
+    {
+      "epoch": 4.743025409064692,
+      "grad_norm": 17731.822265625,
+      "learning_rate": 0.0006851485148514852,
+      "loss": 1.0496,
+      "step": 5580
+    },
+    {
+      "epoch": 4.743025409064692,
+      "eval_accuracy": 0.7761857833832465,
+      "eval_loss": 0.9594100117683411,
+      "eval_runtime": 1281.0251,
+      "eval_samples_per_second": 389.827,
+      "eval_steps_per_second": 1.354,
+      "step": 5580
+    },
+    {
+      "epoch": 4.75152545460065,
+      "grad_norm": 16594.486328125,
+      "learning_rate": 0.0006841584158415842,
+      "loss": 1.0479,
+      "step": 5590
+    },
+    {
+      "epoch": 4.760025500136608,
+      "grad_norm": 17834.193359375,
+      "learning_rate": 0.0006831683168316832,
+      "loss": 1.0465,
+      "step": 5600
+    },
+    {
+      "epoch": 4.768525545672566,
+      "grad_norm": 16431.404296875,
+      "learning_rate": 0.0006821782178217823,
+      "loss": 1.047,
+      "step": 5610
+    },
+    {
+      "epoch": 4.777025591208524,
+      "grad_norm": 18339.9921875,
+      "learning_rate": 0.0006811881188118812,
+      "loss": 1.0456,
+      "step": 5620
+    },
+    {
+      "epoch": 4.785525636744483,
+      "grad_norm": 15955.2890625,
+      "learning_rate": 0.0006801980198019803,
+      "loss": 1.0436,
+      "step": 5630
+    },
+    {
+      "epoch": 4.794025682280441,
+      "grad_norm": 18920.619140625,
+      "learning_rate": 0.0006792079207920792,
+      "loss": 1.0458,
+      "step": 5640
+    },
+    {
+      "epoch": 4.802525727816399,
+      "grad_norm": 16657.296875,
+      "learning_rate": 0.0006782178217821783,
+      "loss": 1.0452,
+      "step": 5650
+    },
+    {
+      "epoch": 4.811025773352357,
+      "grad_norm": 19700.88671875,
+      "learning_rate": 0.0006772277227722773,
+      "loss": 1.0473,
+      "step": 5660
+    },
+    {
+      "epoch": 4.819525818888316,
+      "grad_norm": 19387.7578125,
+      "learning_rate": 0.0006762376237623763,
+      "loss": 1.0458,
+      "step": 5670
+    },
+    {
+      "epoch": 4.819525818888316,
+      "eval_accuracy": 0.7771768462948865,
+      "eval_loss": 0.9521399736404419,
+      "eval_runtime": 1279.986,
+      "eval_samples_per_second": 390.143,
+      "eval_steps_per_second": 1.355,
+      "step": 5670
+    },
+    {
+      "epoch": 4.828025864424274,
+      "grad_norm": 18805.466796875,
+      "learning_rate": 0.0006752475247524753,
+      "loss": 1.0425,
+      "step": 5680
+    },
+    {
+      "epoch": 4.836525909960232,
+      "grad_norm": 17308.564453125,
+      "learning_rate": 0.0006742574257425743,
+      "loss": 1.043,
+      "step": 5690
+    },
+    {
+      "epoch": 4.84502595549619,
+      "grad_norm": 19598.796875,
+      "learning_rate": 0.0006732673267326733,
+      "loss": 1.0425,
+      "step": 5700
+    },
+    {
+      "epoch": 4.853526001032148,
+      "grad_norm": 17387.70703125,
+      "learning_rate": 0.0006722772277227724,
+      "loss": 1.0435,
+      "step": 5710
+    },
+    {
+      "epoch": 4.862026046568107,
+      "grad_norm": 18542.671875,
+      "learning_rate": 0.0006712871287128713,
+      "loss": 1.0429,
+      "step": 5720
+    },
+    {
+      "epoch": 4.870526092104065,
+      "grad_norm": 15447.505859375,
+      "learning_rate": 0.0006702970297029704,
+      "loss": 1.0428,
+      "step": 5730
+    },
+    {
+      "epoch": 4.8790261376400235,
+      "grad_norm": 19042.740234375,
+      "learning_rate": 0.0006693069306930693,
+      "loss": 1.0428,
+      "step": 5740
+    },
+    {
+      "epoch": 4.887526183175981,
+      "grad_norm": 19183.861328125,
+      "learning_rate": 0.0006683168316831684,
+      "loss": 1.042,
+      "step": 5750
+    },
+    {
+      "epoch": 4.896026228711939,
+      "grad_norm": 16694.583984375,
+      "learning_rate": 0.0006673267326732674,
+      "loss": 1.0407,
+      "step": 5760
+    },
+    {
+      "epoch": 4.896026228711939,
+      "eval_accuracy": 0.7776950393076092,
+      "eval_loss": 0.950414776802063,
+      "eval_runtime": 1280.3997,
+      "eval_samples_per_second": 390.017,
+      "eval_steps_per_second": 1.354,
+      "step": 5760
+    },
+    {
+      "epoch": 4.904526274247898,
+      "grad_norm": 16852.962890625,
+      "learning_rate": 0.0006663366336633664,
+      "loss": 1.0404,
+      "step": 5770
+    },
+    {
+      "epoch": 4.913026319783856,
+      "grad_norm": 20231.42578125,
+      "learning_rate": 0.0006653465346534654,
+      "loss": 1.0399,
+      "step": 5780
+    },
+    {
+      "epoch": 4.9215263653198145,
+      "grad_norm": 15834.216796875,
+      "learning_rate": 0.0006643564356435644,
+      "loss": 1.042,
+      "step": 5790
+    },
+    {
+      "epoch": 4.930026410855772,
+      "grad_norm": 15273.1630859375,
+      "learning_rate": 0.0006633663366336634,
+      "loss": 1.0397,
+      "step": 5800
+    },
+    {
+      "epoch": 4.938526456391731,
+      "grad_norm": 20834.515625,
+      "learning_rate": 0.0006623762376237625,
+      "loss": 1.0386,
+      "step": 5810
+    },
+    {
+      "epoch": 4.947026501927689,
+      "grad_norm": 21407.017578125,
+      "learning_rate": 0.0006613861386138614,
+      "loss": 1.0389,
+      "step": 5820
+    },
+    {
+      "epoch": 4.955526547463647,
+      "grad_norm": 19523.822265625,
+      "learning_rate": 0.0006603960396039605,
+      "loss": 1.038,
+      "step": 5830
+    },
+    {
+      "epoch": 4.9640265929996055,
+      "grad_norm": 18045.078125,
+      "learning_rate": 0.0006594059405940594,
+      "loss": 1.0359,
+      "step": 5840
+    },
+    {
+      "epoch": 4.972526638535563,
+      "grad_norm": 16661.830078125,
+      "learning_rate": 0.0006584158415841585,
+      "loss": 1.037,
+      "step": 5850
+    },
+    {
+      "epoch": 4.972526638535563,
+      "eval_accuracy": 0.7784280857816438,
+      "eval_loss": 0.9486609101295471,
+      "eval_runtime": 1279.799,
+      "eval_samples_per_second": 390.2,
+      "eval_steps_per_second": 1.355,
+      "step": 5850
+    },
+    {
+      "epoch": 4.981026684071522,
+      "grad_norm": 19275.583984375,
+      "learning_rate": 0.0006574257425742575,
+      "loss": 1.0376,
+      "step": 5860
+    },
+    {
+      "epoch": 4.98952672960748,
+      "grad_norm": 20361.98828125,
+      "learning_rate": 0.0006564356435643565,
+      "loss": 1.0386,
+      "step": 5870
+    },
+    {
+      "epoch": 4.998026775143439,
+      "grad_norm": 18276.357421875,
+      "learning_rate": 0.0006554455445544555,
+      "loss": 1.0364,
+      "step": 5880
+    },
+    {
+      "epoch": 5.0065268206793965,
+      "grad_norm": 18657.6484375,
+      "learning_rate": 0.0006544554455445545,
+      "loss": 1.0364,
+      "step": 5890
+    },
+    {
+      "epoch": 5.015026866215354,
+      "grad_norm": 17588.00390625,
+      "learning_rate": 0.0006534653465346535,
+      "loss": 1.0346,
+      "step": 5900
+    },
+    {
+      "epoch": 5.023526911751313,
+      "grad_norm": 20156.283203125,
+      "learning_rate": 0.0006524752475247526,
+      "loss": 1.0341,
+      "step": 5910
+    },
+    {
+      "epoch": 5.032026957287271,
+      "grad_norm": 18811.03125,
+      "learning_rate": 0.0006514851485148515,
+      "loss": 1.034,
+      "step": 5920
+    },
+    {
+      "epoch": 5.04052700282323,
+      "grad_norm": 18864.615234375,
+      "learning_rate": 0.0006504950495049506,
+      "loss": 1.0356,
+      "step": 5930
+    },
+    {
+      "epoch": 5.0490270483591875,
+      "grad_norm": 18199.67578125,
+      "learning_rate": 0.0006495049504950495,
+      "loss": 1.0339,
+      "step": 5940
+    },
+    {
+      "epoch": 5.0490270483591875,
+      "eval_accuracy": 0.7792297436274894,
+      "eval_loss": 0.9439004063606262,
+      "eval_runtime": 1279.6853,
+      "eval_samples_per_second": 390.235,
+      "eval_steps_per_second": 1.355,
+      "step": 5940
+    },
+    {
+      "epoch": 5.057527093895146,
+      "grad_norm": 19274.73828125,
+      "learning_rate": 0.0006485148514851485,
+      "loss": 1.0334,
+      "step": 5950
+    },
+    {
+      "epoch": 5.066027139431104,
+      "grad_norm": 20778.62109375,
+      "learning_rate": 0.0006475247524752476,
+      "loss": 1.0337,
+      "step": 5960
+    },
+    {
+      "epoch": 5.074527184967062,
+      "grad_norm": 19509.306640625,
+      "learning_rate": 0.0006465346534653465,
+      "loss": 1.0355,
+      "step": 5970
+    },
+    {
+      "epoch": 5.083027230503021,
+      "grad_norm": 18669.24609375,
+      "learning_rate": 0.0006455445544554456,
+      "loss": 1.0329,
+      "step": 5980
+    },
+    {
+      "epoch": 5.0915272760389785,
+      "grad_norm": 15668.9755859375,
+      "learning_rate": 0.0006445544554455445,
+      "loss": 1.0294,
+      "step": 5990
+    },
+    {
+      "epoch": 5.100027321574937,
+      "grad_norm": 18858.783203125,
+      "learning_rate": 0.0006435643564356436,
+      "loss": 1.0324,
+      "step": 6000
+    },
+    {
+      "epoch": 5.108527367110895,
+      "grad_norm": 16453.763671875,
+      "learning_rate": 0.0006425742574257426,
+      "loss": 1.033,
+      "step": 6010
+    },
+    {
+      "epoch": 5.117027412646854,
+      "grad_norm": 19821.6875,
+      "learning_rate": 0.0006415841584158416,
+      "loss": 1.0324,
+      "step": 6020
+    },
+    {
+      "epoch": 5.125527458182812,
+      "grad_norm": 18548.5078125,
+      "learning_rate": 0.0006405940594059406,
+      "loss": 1.0284,
+      "step": 6030
+    },
+    {
+      "epoch": 5.125527458182812,
+      "eval_accuracy": 0.7796817210719912,
+      "eval_loss": 0.9404940605163574,
+      "eval_runtime": 1281.8344,
+      "eval_samples_per_second": 389.581,
+      "eval_steps_per_second": 1.353,
+      "step": 6030
+    },
+    {
+      "epoch": 5.1340275037187695,
+      "grad_norm": 18221.74609375,
+      "learning_rate": 0.0006396039603960396,
+      "loss": 1.0309,
+      "step": 6040
+    },
+    {
+      "epoch": 5.142527549254728,
+      "grad_norm": 19257.5078125,
+      "learning_rate": 0.0006386138613861386,
+      "loss": 1.0316,
+      "step": 6050
+    },
+    {
+      "epoch": 5.151027594790686,
+      "grad_norm": 20660.0859375,
+      "learning_rate": 0.0006376237623762377,
+      "loss": 1.0281,
+      "step": 6060
+    },
+    {
+      "epoch": 5.159527640326645,
+      "grad_norm": 17646.05859375,
+      "learning_rate": 0.0006366336633663366,
+      "loss": 1.0298,
+      "step": 6070
+    },
+    {
+      "epoch": 5.168027685862603,
+      "grad_norm": 15755.4951171875,
+      "learning_rate": 0.0006356435643564357,
+      "loss": 1.0296,
+      "step": 6080
+    },
+    {
+      "epoch": 5.176527731398561,
+      "grad_norm": 19384.6484375,
+      "learning_rate": 0.0006346534653465346,
+      "loss": 1.0277,
+      "step": 6090
+    },
+    {
+      "epoch": 5.185027776934519,
+      "grad_norm": 16928.974609375,
+      "learning_rate": 0.0006336633663366337,
+      "loss": 1.0295,
+      "step": 6100
+    },
+    {
+      "epoch": 5.193527822470478,
+      "grad_norm": 15206.5625,
+      "learning_rate": 0.0006326732673267327,
+      "loss": 1.0288,
+      "step": 6110
+    },
+    {
+      "epoch": 5.202027868006436,
+      "grad_norm": 18075.7421875,
+      "learning_rate": 0.0006316831683168317,
+      "loss": 1.0288,
+      "step": 6120
+    },
+    {
+      "epoch": 5.202027868006436,
+      "eval_accuracy": 0.7805738833087461,
+      "eval_loss": 0.9371596574783325,
+      "eval_runtime": 1279.2071,
+      "eval_samples_per_second": 390.381,
+      "eval_steps_per_second": 1.356,
+      "step": 6120
+    },
+    {
+      "epoch": 5.210527913542394,
+      "grad_norm": 17331.333984375,
+      "learning_rate": 0.0006306930693069307,
+      "loss": 1.0245,
+      "step": 6130
+    },
+    {
+      "epoch": 5.219027959078352,
+      "grad_norm": 19757.05078125,
+      "learning_rate": 0.0006297029702970297,
+      "loss": 1.0266,
+      "step": 6140
+    },
+    {
+      "epoch": 5.22752800461431,
+      "grad_norm": 16021.1396484375,
+      "learning_rate": 0.0006287128712871287,
+      "loss": 1.0256,
+      "step": 6150
+    },
+    {
+      "epoch": 5.236028050150269,
+      "grad_norm": 16731.166015625,
+      "learning_rate": 0.0006277227722772278,
+      "loss": 1.0257,
+      "step": 6160
+    },
+    {
+      "epoch": 5.244528095686227,
+      "grad_norm": 17451.5390625,
+      "learning_rate": 0.0006267326732673267,
+      "loss": 1.0271,
+      "step": 6170
+    },
+    {
+      "epoch": 5.2530281412221855,
+      "grad_norm": 19212.619140625,
+      "learning_rate": 0.0006257425742574258,
+      "loss": 1.0255,
+      "step": 6180
+    },
+    {
+      "epoch": 5.261528186758143,
+      "grad_norm": 17397.1015625,
+      "learning_rate": 0.0006247524752475247,
+      "loss": 1.026,
+      "step": 6190
+    },
+    {
+      "epoch": 5.270028232294101,
+      "grad_norm": 18388.25,
+      "learning_rate": 0.0006237623762376238,
+      "loss": 1.0253,
+      "step": 6200
+    },
+    {
+      "epoch": 5.27852827783006,
+      "grad_norm": 18145.59765625,
+      "learning_rate": 0.0006227722772277228,
+      "loss": 1.0251,
+      "step": 6210
+    },
+    {
+      "epoch": 5.27852827783006,
+      "eval_accuracy": 0.7809789625861284,
+      "eval_loss": 0.9343125820159912,
+      "eval_runtime": 1280.9852,
+      "eval_samples_per_second": 389.839,
+      "eval_steps_per_second": 1.354,
+      "step": 6210
+    },
+    {
+      "epoch": 5.287028323366018,
+      "grad_norm": 18143.001953125,
+      "learning_rate": 0.0006217821782178218,
+      "loss": 1.0251,
+      "step": 6220
+    },
+    {
+      "epoch": 5.2955283689019765,
+      "grad_norm": 23030.259765625,
+      "learning_rate": 0.0006207920792079208,
+      "loss": 1.0248,
+      "step": 6230
+    },
+    {
+      "epoch": 5.304028414437934,
+      "grad_norm": 20499.400390625,
+      "learning_rate": 0.0006198019801980198,
+      "loss": 1.0239,
+      "step": 6240
+    },
+    {
+      "epoch": 5.312528459973893,
+      "grad_norm": 16773.736328125,
+      "learning_rate": 0.0006188118811881188,
+      "loss": 1.0251,
+      "step": 6250
+    },
+    {
+      "epoch": 5.321028505509851,
+      "grad_norm": 15036.5712890625,
+      "learning_rate": 0.0006178217821782179,
+      "loss": 1.0248,
+      "step": 6260
+    },
+    {
+      "epoch": 5.329528551045809,
+      "grad_norm": 18476.0703125,
+      "learning_rate": 0.0006168316831683168,
+      "loss": 1.0241,
+      "step": 6270
+    },
+    {
+      "epoch": 5.3380285965817675,
+      "grad_norm": 18741.55859375,
+      "learning_rate": 0.0006158415841584159,
+      "loss": 1.0238,
+      "step": 6280
+    },
+    {
+      "epoch": 5.346528642117725,
+      "grad_norm": 17602.275390625,
+      "learning_rate": 0.0006148514851485148,
+      "loss": 1.0222,
+      "step": 6290
+    },
+    {
+      "epoch": 5.355028687653684,
+      "grad_norm": 15750.1806640625,
+      "learning_rate": 0.0006138613861386139,
+      "loss": 1.0218,
+      "step": 6300
+    },
+    {
+      "epoch": 5.355028687653684,
+      "eval_accuracy": 0.7811827125583685,
+      "eval_loss": 0.9342101216316223,
+      "eval_runtime": 1278.6765,
+      "eval_samples_per_second": 390.543,
+      "eval_steps_per_second": 1.356,
+      "step": 6300
+    },
+    {
+      "epoch": 5.363528733189642,
+      "grad_norm": 15361.734375,
+      "learning_rate": 0.0006128712871287129,
+      "loss": 1.0212,
+      "step": 6310
+    },
+    {
+      "epoch": 5.372028778725601,
+      "grad_norm": 17438.470703125,
+      "learning_rate": 0.0006118811881188119,
+      "loss": 1.0217,
+      "step": 6320
+    },
+    {
+      "epoch": 5.3805288242615585,
+      "grad_norm": 16182.955078125,
+      "learning_rate": 0.0006108910891089109,
+      "loss": 1.0207,
+      "step": 6330
+    },
+    {
+      "epoch": 5.389028869797516,
+      "grad_norm": 15098.9970703125,
+      "learning_rate": 0.0006099009900990099,
+      "loss": 1.021,
+      "step": 6340
+    },
+    {
+      "epoch": 5.397528915333475,
+      "grad_norm": 16916.810546875,
+      "learning_rate": 0.0006089108910891089,
+      "loss": 1.0187,
+      "step": 6350
+    },
+    {
+      "epoch": 5.406028960869433,
+      "grad_norm": 15510.921875,
+      "learning_rate": 0.000607920792079208,
+      "loss": 1.0203,
+      "step": 6360
+    },
+    {
+      "epoch": 5.414529006405392,
+      "grad_norm": 18232.8671875,
+      "learning_rate": 0.0006069306930693069,
+      "loss": 1.0209,
+      "step": 6370
+    },
+    {
+      "epoch": 5.4230290519413495,
+      "grad_norm": 15943.5810546875,
+      "learning_rate": 0.000605940594059406,
+      "loss": 1.0189,
+      "step": 6380
+    },
+    {
+      "epoch": 5.431529097477308,
+      "grad_norm": 17515.39453125,
+      "learning_rate": 0.0006049504950495049,
+      "loss": 1.0185,
+      "step": 6390
+    },
+    {
+      "epoch": 5.431529097477308,
+      "eval_accuracy": 0.7821448982293225,
+      "eval_loss": 0.9290784597396851,
+      "eval_runtime": 1274.8156,
+      "eval_samples_per_second": 391.726,
+      "eval_steps_per_second": 1.36,
+      "step": 6390
+    },
+    {
+      "epoch": 5.4420023678698275,
+      "grad_norm": 17196.8828125,
+      "learning_rate": 0.000603960396039604,
+      "loss": 1.0209,
+      "step": 6400
+    },
+    {
+      "epoch": 5.450502413405786,
+      "grad_norm": 19976.57421875,
+      "learning_rate": 0.000602970297029703,
+      "loss": 1.0202,
+      "step": 6410
+    },
+    {
+      "epoch": 5.459002458941744,
+      "grad_norm": 16986.595703125,
+      "learning_rate": 0.000601980198019802,
+      "loss": 1.0194,
+      "step": 6420
+    },
+    {
+      "epoch": 5.467502504477703,
+      "grad_norm": 16322.2490234375,
+      "learning_rate": 0.000600990099009901,
+      "loss": 1.018,
+      "step": 6430
+    },
+    {
+      "epoch": 5.476002550013661,
+      "grad_norm": 16943.923828125,
+      "learning_rate": 0.0006,
+      "loss": 1.0173,
+      "step": 6440
+    },
+    {
+      "epoch": 5.484502595549619,
+      "grad_norm": 17515.416015625,
+      "learning_rate": 0.000599009900990099,
+      "loss": 1.0171,
+      "step": 6450
+    },
+    {
+      "epoch": 5.493002641085577,
+      "grad_norm": 20824.935546875,
+      "learning_rate": 0.000598019801980198,
+      "loss": 1.0167,
+      "step": 6460
+    },
+    {
+      "epoch": 5.501502686621535,
+      "grad_norm": 19812.29296875,
+      "learning_rate": 0.000597029702970297,
+      "loss": 1.017,
+      "step": 6470
+    },
+    {
+      "epoch": 5.510002732157494,
+      "grad_norm": 20733.58984375,
+      "learning_rate": 0.000596039603960396,
+      "loss": 1.0143,
+      "step": 6480
+    },
+    {
+      "epoch": 5.510002732157494,
+      "eval_accuracy": 0.7827784542349862,
+      "eval_loss": 0.926770031452179,
+      "eval_runtime": 1034.8114,
+      "eval_samples_per_second": 482.579,
+      "eval_steps_per_second": 0.821,
+      "step": 6480
+    },
+    {
+      "epoch": 5.518502777693452,
+      "grad_norm": 19389.359375,
+      "learning_rate": 0.000595049504950495,
+      "loss": 1.0156,
+      "step": 6490
+    },
+    {
+      "epoch": 5.52700282322941,
+      "grad_norm": 18111.1796875,
+      "learning_rate": 0.000594059405940594,
+      "loss": 1.015,
+      "step": 6500
+    },
+    {
+      "epoch": 5.535502868765368,
+      "grad_norm": 17627.548828125,
+      "learning_rate": 0.0005930693069306931,
+      "loss": 1.0149,
+      "step": 6510
+    },
+    {
+      "epoch": 5.544002914301327,
+      "grad_norm": 19444.2421875,
+      "learning_rate": 0.000592079207920792,
+      "loss": 1.0161,
+      "step": 6520
+    },
+    {
+      "epoch": 5.552502959837285,
+      "grad_norm": 20379.31640625,
+      "learning_rate": 0.0005910891089108911,
+      "loss": 1.0151,
+      "step": 6530
+    },
+    {
+      "epoch": 5.561003005373243,
+      "grad_norm": 17187.162109375,
+      "learning_rate": 0.00059009900990099,
+      "loss": 1.0164,
+      "step": 6540
+    },
+    {
+      "epoch": 5.569503050909201,
+      "grad_norm": 20134.74609375,
+      "learning_rate": 0.0005891089108910891,
+      "loss": 1.0125,
+      "step": 6550
+    },
+    {
+      "epoch": 5.578003096445159,
+      "grad_norm": 16345.7314453125,
+      "learning_rate": 0.0005881188118811881,
+      "loss": 1.0143,
+      "step": 6560
+    },
+    {
+      "epoch": 5.586503141981118,
+      "grad_norm": 16676.40625,
+      "learning_rate": 0.0005871287128712871,
+      "loss": 1.0146,
+      "step": 6570
+    },
+    {
+      "epoch": 5.586503141981118,
+      "eval_accuracy": 0.7830355090944983,
+      "eval_loss": 0.925011157989502,
+      "eval_runtime": 1032.366,
+      "eval_samples_per_second": 483.722,
+      "eval_steps_per_second": 0.823,
+      "step": 6570
+    },
+    {
+      "epoch": 5.595003187517076,
+      "grad_norm": 18004.888671875,
+      "learning_rate": 0.0005861386138613861,
+      "loss": 1.0134,
+      "step": 6580
+    },
+    {
+      "epoch": 5.6035032330530345,
+      "grad_norm": 17843.01171875,
+      "learning_rate": 0.0005851485148514851,
+      "loss": 1.0121,
+      "step": 6590
+    },
+    {
+      "epoch": 5.612003278588992,
+      "grad_norm": 19261.90625,
+      "learning_rate": 0.0005841584158415841,
+      "loss": 1.014,
+      "step": 6600
+    },
+    {
+      "epoch": 5.62050332412495,
+      "grad_norm": 19034.001953125,
+      "learning_rate": 0.0005831683168316832,
+      "loss": 1.014,
+      "step": 6610
+    },
+    {
+      "epoch": 5.629003369660909,
+      "grad_norm": 15280.501953125,
+      "learning_rate": 0.0005821782178217821,
+      "loss": 1.0127,
+      "step": 6620
+    },
+    {
+      "epoch": 5.637503415196867,
+      "grad_norm": 16361.3017578125,
+      "learning_rate": 0.0005811881188118812,
+      "loss": 1.0135,
+      "step": 6630
+    },
+    {
+      "epoch": 5.6460034607328256,
+      "grad_norm": 18691.2578125,
+      "learning_rate": 0.0005801980198019801,
+      "loss": 1.0116,
+      "step": 6640
+    },
+    {
+      "epoch": 5.654503506268783,
+      "grad_norm": 17126.857421875,
+      "learning_rate": 0.0005792079207920792,
+      "loss": 1.0123,
+      "step": 6650
+    },
+    {
+      "epoch": 5.663003551804742,
+      "grad_norm": 18155.701171875,
+      "learning_rate": 0.0005782178217821782,
+      "loss": 1.01,
+      "step": 6660
+    },
+    {
+      "epoch": 5.663003551804742,
+      "eval_accuracy": 0.7837371016336877,
+      "eval_loss": 0.9207693338394165,
+      "eval_runtime": 1032.8984,
+      "eval_samples_per_second": 483.473,
+      "eval_steps_per_second": 0.823,
+      "step": 6660
+    },
+    {
+      "epoch": 5.6715035973407,
+      "grad_norm": 19733.861328125,
+      "learning_rate": 0.0005772277227722772,
+      "loss": 1.0122,
+      "step": 6670
+    },
+    {
+      "epoch": 5.680003642876658,
+      "grad_norm": 16721.662109375,
+      "learning_rate": 0.0005762376237623762,
+      "loss": 1.013,
+      "step": 6680
+    },
+    {
+      "epoch": 5.6885036884126166,
+      "grad_norm": 19389.369140625,
+      "learning_rate": 0.0005752475247524752,
+      "loss": 1.012,
+      "step": 6690
+    },
+    {
+      "epoch": 5.697003733948574,
+      "grad_norm": 20522.015625,
+      "learning_rate": 0.0005742574257425742,
+      "loss": 1.0087,
+      "step": 6700
+    },
+    {
+      "epoch": 5.705503779484533,
+      "grad_norm": 19054.39453125,
+      "learning_rate": 0.0005732673267326733,
+      "loss": 1.0092,
+      "step": 6710
+    },
+    {
+      "epoch": 5.714003825020491,
+      "grad_norm": 16219.0595703125,
+      "learning_rate": 0.0005722772277227722,
+      "loss": 1.0105,
+      "step": 6720
+    },
+    {
+      "epoch": 5.72250387055645,
+      "grad_norm": 19937.716796875,
+      "learning_rate": 0.0005712871287128713,
+      "loss": 1.01,
+      "step": 6730
+    },
+    {
+      "epoch": 5.731003916092408,
+      "grad_norm": 18509.220703125,
+      "learning_rate": 0.0005702970297029702,
+      "loss": 1.0084,
+      "step": 6740
+    },
+    {
+      "epoch": 5.739503961628365,
+      "grad_norm": 18168.08203125,
+      "learning_rate": 0.0005693069306930693,
+      "loss": 1.0081,
+      "step": 6750
+    },
+    {
+      "epoch": 5.739503961628365,
+      "eval_accuracy": 0.7840970171619854,
+      "eval_loss": 0.9189032316207886,
+      "eval_runtime": 1032.6648,
+      "eval_samples_per_second": 483.582,
+      "eval_steps_per_second": 0.823,
+      "step": 6750
+    },
+    {
+      "epoch": 5.748004007164324,
+      "grad_norm": 17099.634765625,
+      "learning_rate": 0.0005683168316831683,
+      "loss": 1.0085,
+      "step": 6760
+    },
+    {
+      "epoch": 5.756504052700282,
+      "grad_norm": 19988.625,
+      "learning_rate": 0.0005673267326732673,
+      "loss": 1.0079,
+      "step": 6770
+    },
+    {
+      "epoch": 5.765004098236241,
+      "grad_norm": 17239.306640625,
+      "learning_rate": 0.0005663366336633663,
+      "loss": 1.0075,
+      "step": 6780
+    },
+    {
+      "epoch": 5.773504143772199,
+      "grad_norm": 16728.1171875,
+      "learning_rate": 0.0005653465346534653,
+      "loss": 1.0062,
+      "step": 6790
+    },
+    {
+      "epoch": 5.782004189308157,
+      "grad_norm": 15729.82421875,
+      "learning_rate": 0.0005643564356435643,
+      "loss": 1.0074,
+      "step": 6800
+    },
+    {
+      "epoch": 5.790504234844115,
+      "grad_norm": 19562.201171875,
+      "learning_rate": 0.0005633663366336634,
+      "loss": 1.0075,
+      "step": 6810
+    },
+    {
+      "epoch": 5.799004280380073,
+      "grad_norm": 20575.86328125,
+      "learning_rate": 0.0005623762376237624,
+      "loss": 1.0056,
+      "step": 6820
+    },
+    {
+      "epoch": 5.807504325916032,
+      "grad_norm": 19013.0625,
+      "learning_rate": 0.0005613861386138615,
+      "loss": 1.0058,
+      "step": 6830
+    },
+    {
+      "epoch": 5.81600437145199,
+      "grad_norm": 18319.556640625,
+      "learning_rate": 0.0005603960396039604,
+      "loss": 1.0055,
+      "step": 6840
+    },
+    {
+      "epoch": 5.81600437145199,
+      "eval_accuracy": 0.7846662939402871,
+      "eval_loss": 0.9171181321144104,
+      "eval_runtime": 1033.9438,
+      "eval_samples_per_second": 482.984,
+      "eval_steps_per_second": 0.822,
+      "step": 6840
+    },
+    {
+      "epoch": 5.824504416987948,
+      "grad_norm": 19040.82421875,
+      "learning_rate": 0.0005594059405940595,
+      "loss": 1.0077,
+      "step": 6850
+    },
+    {
+      "epoch": 5.833004462523906,
+      "grad_norm": 19041.265625,
+      "learning_rate": 0.0005584158415841585,
+      "loss": 1.0064,
+      "step": 6860
+    },
+    {
+      "epoch": 5.841504508059865,
+      "grad_norm": 15583.31640625,
+      "learning_rate": 0.0005574257425742575,
+      "loss": 1.0052,
+      "step": 6870
+    },
+    {
+      "epoch": 5.850004553595823,
+      "grad_norm": 15519.5576171875,
+      "learning_rate": 0.0005564356435643565,
+      "loss": 1.0066,
+      "step": 6880
+    },
+    {
+      "epoch": 5.858504599131781,
+      "grad_norm": 19509.908203125,
+      "learning_rate": 0.0005554455445544555,
+      "loss": 1.005,
+      "step": 6890
+    },
+    {
+      "epoch": 5.867004644667739,
+      "grad_norm": 16785.40625,
+      "learning_rate": 0.0005544554455445545,
+      "loss": 1.0045,
+      "step": 6900
+    },
+    {
+      "epoch": 5.875504690203697,
+      "grad_norm": 20995.373046875,
+      "learning_rate": 0.0005534653465346536,
+      "loss": 1.0028,
+      "step": 6910
+    },
+    {
+      "epoch": 5.884004735739656,
+      "grad_norm": 16980.791015625,
+      "learning_rate": 0.0005524752475247525,
+      "loss": 1.0058,
+      "step": 6920
+    },
+    {
+      "epoch": 5.892504781275614,
+      "grad_norm": 17058.552734375,
+      "learning_rate": 0.0005514851485148516,
+      "loss": 1.0032,
+      "step": 6930
+    },
+    {
+      "epoch": 5.892504781275614,
+      "eval_accuracy": 0.7852809138194954,
+      "eval_loss": 0.9138230085372925,
+      "eval_runtime": 1038.0576,
+      "eval_samples_per_second": 481.07,
+      "eval_steps_per_second": 0.819,
+      "step": 6930
+    },
+    {
+      "epoch": 5.9010048268115725,
+      "grad_norm": 15984.5615234375,
+      "learning_rate": 0.0005504950495049505,
+      "loss": 1.002,
+      "step": 6940
+    },
+    {
+      "epoch": 5.90950487234753,
+      "grad_norm": 18076.875,
+      "learning_rate": 0.0005495049504950496,
+      "loss": 1.0029,
+      "step": 6950
+    },
+    {
+      "epoch": 5.918004917883489,
+      "grad_norm": 17178.27734375,
+      "learning_rate": 0.0005485148514851486,
+      "loss": 1.0024,
+      "step": 6960
+    },
+    {
+      "epoch": 5.926504963419447,
+      "grad_norm": 17936.349609375,
+      "learning_rate": 0.0005475247524752476,
+      "loss": 0.9992,
+      "step": 6970
+    },
+    {
+      "epoch": 5.935005008955406,
+      "grad_norm": 16773.853515625,
+      "learning_rate": 0.0005465346534653466,
+      "loss": 1.002,
+      "step": 6980
+    },
+    {
+      "epoch": 5.9435050544913635,
+      "grad_norm": 18567.22265625,
+      "learning_rate": 0.0005455445544554456,
+      "loss": 1.0019,
+      "step": 6990
+    },
+    {
+      "epoch": 5.952005100027321,
+      "grad_norm": 20730.1484375,
+      "learning_rate": 0.0005445544554455446,
+      "loss": 1.0032,
+      "step": 7000
+    },
+    {
+      "epoch": 5.96050514556328,
+      "grad_norm": 18468.900390625,
+      "learning_rate": 0.0005435643564356437,
+      "loss": 1.0029,
+      "step": 7010
+    },
+    {
+      "epoch": 5.969005191099238,
+      "grad_norm": 21202.24609375,
+      "learning_rate": 0.0005425742574257426,
+      "loss": 1.0019,
+      "step": 7020
+    },
+    {
+      "epoch": 5.969005191099238,
+      "eval_accuracy": 0.7857404893569327,
+      "eval_loss": 0.9110856652259827,
+      "eval_runtime": 1036.1907,
+      "eval_samples_per_second": 481.936,
+      "eval_steps_per_second": 0.82,
+      "step": 7020
+    },
+    {
+      "epoch": 5.977505236635197,
+      "grad_norm": 17446.140625,
+      "learning_rate": 0.0005415841584158417,
+      "loss": 1.0013,
+      "step": 7030
+    },
+    {
+      "epoch": 5.9860052821711545,
+      "grad_norm": 20787.7109375,
+      "learning_rate": 0.0005405940594059406,
+      "loss": 1.0006,
+      "step": 7040
+    },
+    {
+      "epoch": 5.994505327707113,
+      "grad_norm": 15513.68359375,
+      "learning_rate": 0.0005396039603960396,
+      "loss": 1.001,
+      "step": 7050
+    },
+    {
+      "epoch": 6.003005373243071,
+      "grad_norm": 17996.607421875,
+      "learning_rate": 0.0005386138613861387,
+      "loss": 1.0003,
+      "step": 7060
+    },
+    {
+      "epoch": 6.011505418779029,
+      "grad_norm": 17095.095703125,
+      "learning_rate": 0.0005376237623762376,
+      "loss": 1.0007,
+      "step": 7070
+    },
+    {
+      "epoch": 6.020005464314988,
+      "grad_norm": 17025.11328125,
+      "learning_rate": 0.0005366336633663367,
+      "loss": 1.0003,
+      "step": 7080
+    },
+    {
+      "epoch": 6.0285055098509455,
+      "grad_norm": 15303.3544921875,
+      "learning_rate": 0.0005356435643564356,
+      "loss": 1.0,
+      "step": 7090
+    },
+    {
+      "epoch": 6.037005555386904,
+      "grad_norm": 18501.576171875,
+      "learning_rate": 0.0005346534653465347,
+      "loss": 0.9981,
+      "step": 7100
+    },
+    {
+      "epoch": 6.045505600922862,
+      "grad_norm": 19782.1171875,
+      "learning_rate": 0.0005336633663366337,
+      "loss": 0.9989,
+      "step": 7110
+    },
+    {
+      "epoch": 6.045505600922862,
+      "eval_accuracy": 0.7861505064886588,
+      "eval_loss": 0.9094851613044739,
+      "eval_runtime": 1034.5248,
+      "eval_samples_per_second": 482.712,
+      "eval_steps_per_second": 0.822,
+      "step": 7110
+    },
+    {
+      "epoch": 6.05400564645882,
+      "grad_norm": 19169.7890625,
+      "learning_rate": 0.0005326732673267327,
+      "loss": 0.9998,
+      "step": 7120
+    },
+    {
+      "epoch": 6.062505691994779,
+      "grad_norm": 19047.958984375,
+      "learning_rate": 0.0005316831683168317,
+      "loss": 0.9973,
+      "step": 7130
+    },
+    {
+      "epoch": 6.0710057375307365,
+      "grad_norm": 20607.556640625,
+      "learning_rate": 0.0005306930693069307,
+      "loss": 0.9962,
+      "step": 7140
+    },
+    {
+      "epoch": 6.079505783066695,
+      "grad_norm": 16297.8662109375,
+      "learning_rate": 0.0005297029702970297,
+      "loss": 0.9968,
+      "step": 7150
+    },
+    {
+      "epoch": 6.088005828602653,
+      "grad_norm": 16640.685546875,
+      "learning_rate": 0.0005287128712871288,
+      "loss": 0.9987,
+      "step": 7160
+    },
+    {
+      "epoch": 6.096505874138612,
+      "grad_norm": 16723.65625,
+      "learning_rate": 0.0005277227722772277,
+      "loss": 0.9981,
+      "step": 7170
+    },
+    {
+      "epoch": 6.10500591967457,
+      "grad_norm": 18703.06640625,
+      "learning_rate": 0.0005267326732673268,
+      "loss": 0.9959,
+      "step": 7180
+    },
+    {
+      "epoch": 6.113505965210528,
+      "grad_norm": 16973.158203125,
+      "learning_rate": 0.0005257425742574257,
+      "loss": 0.9988,
+      "step": 7190
+    },
+    {
+      "epoch": 6.122006010746486,
+      "grad_norm": 16989.619140625,
+      "learning_rate": 0.0005247524752475248,
+      "loss": 0.9955,
+      "step": 7200
+    },
+    {
+      "epoch": 6.122006010746486,
+      "eval_accuracy": 0.7866073008018492,
+      "eval_loss": 0.9084503650665283,
+      "eval_runtime": 1037.0895,
+      "eval_samples_per_second": 481.519,
+      "eval_steps_per_second": 0.82,
+      "step": 7200
+    },
+    {
+      "epoch": 6.130506056282444,
+      "grad_norm": 16469.03515625,
+      "learning_rate": 0.0005237623762376238,
+      "loss": 0.9962,
+      "step": 7210
+    },
+    {
+      "epoch": 6.139006101818403,
+      "grad_norm": 16646.052734375,
+      "learning_rate": 0.0005227722772277228,
+      "loss": 0.9978,
+      "step": 7220
+    },
+    {
+      "epoch": 6.147506147354361,
+      "grad_norm": 18823.22265625,
+      "learning_rate": 0.0005217821782178218,
+      "loss": 0.9967,
+      "step": 7230
+    },
+    {
+      "epoch": 6.156006192890319,
+      "grad_norm": 16628.587890625,
+      "learning_rate": 0.0005207920792079208,
+      "loss": 0.9968,
+      "step": 7240
+    },
+    {
+      "epoch": 6.164506238426277,
+      "grad_norm": 17544.67578125,
+      "learning_rate": 0.0005198019801980198,
+      "loss": 0.9958,
+      "step": 7250
+    },
+    {
+      "epoch": 6.173006283962236,
+      "grad_norm": 16480.125,
+      "learning_rate": 0.0005188118811881189,
+      "loss": 0.997,
+      "step": 7260
+    },
+    {
+      "epoch": 6.181506329498194,
+      "grad_norm": 18358.3984375,
+      "learning_rate": 0.0005178217821782178,
+      "loss": 0.9956,
+      "step": 7270
+    },
+    {
+      "epoch": 6.190006375034152,
+      "grad_norm": 19029.322265625,
+      "learning_rate": 0.0005168316831683169,
+      "loss": 0.9977,
+      "step": 7280
+    },
+    {
+      "epoch": 6.19850642057011,
+      "grad_norm": 16023.6572265625,
+      "learning_rate": 0.0005158415841584158,
+      "loss": 0.9944,
+      "step": 7290
+    },
+    {
+      "epoch": 6.19850642057011,
+      "eval_accuracy": 0.7871460102262176,
+      "eval_loss": 0.9043031930923462,
+      "eval_runtime": 1032.4584,
+      "eval_samples_per_second": 483.679,
+      "eval_steps_per_second": 0.823,
+      "step": 7290
+    },
+    {
+      "epoch": 6.207006466106068,
+      "grad_norm": 17711.34765625,
+      "learning_rate": 0.0005148514851485149,
+      "loss": 0.9931,
+      "step": 7300
+    },
+    {
+      "epoch": 6.215506511642027,
+      "grad_norm": 15834.974609375,
+      "learning_rate": 0.0005138613861386139,
+      "loss": 0.9935,
+      "step": 7310
+    },
+    {
+      "epoch": 6.224006557177985,
+      "grad_norm": 15997.443359375,
+      "learning_rate": 0.0005128712871287129,
+      "loss": 0.9943,
+      "step": 7320
+    },
+    {
+      "epoch": 6.2325066027139435,
+      "grad_norm": 19241.9921875,
+      "learning_rate": 0.0005118811881188119,
+      "loss": 0.9942,
+      "step": 7330
+    },
+    {
+      "epoch": 6.241006648249901,
+      "grad_norm": 14933.837890625,
+      "learning_rate": 0.0005108910891089109,
+      "loss": 0.9949,
+      "step": 7340
+    },
+    {
+      "epoch": 6.249506693785859,
+      "grad_norm": 16723.673828125,
+      "learning_rate": 0.0005099009900990099,
+      "loss": 0.9937,
+      "step": 7350
+    },
+    {
+      "epoch": 6.258006739321818,
+      "grad_norm": 19108.876953125,
+      "learning_rate": 0.000508910891089109,
+      "loss": 0.9915,
+      "step": 7360
+    },
+    {
+      "epoch": 6.266506784857776,
+      "grad_norm": 14807.8232421875,
+      "learning_rate": 0.0005079207920792079,
+      "loss": 0.9912,
+      "step": 7370
+    },
+    {
+      "epoch": 6.2750068303937345,
+      "grad_norm": 16829.048828125,
+      "learning_rate": 0.000506930693069307,
+      "loss": 0.9917,
+      "step": 7380
+    },
+    {
+      "epoch": 6.2750068303937345,
+      "eval_accuracy": 0.7875770735366631,
+      "eval_loss": 0.9019953608512878,
+      "eval_runtime": 1035.5292,
+      "eval_samples_per_second": 482.244,
+      "eval_steps_per_second": 0.821,
+      "step": 7380
+    },
+    {
+      "epoch": 6.283506875929692,
+      "grad_norm": 17801.87109375,
+      "learning_rate": 0.0005059405940594059,
+      "loss": 0.9928,
+      "step": 7390
+    },
+    {
+      "epoch": 6.292006921465651,
+      "grad_norm": 18446.126953125,
+      "learning_rate": 0.000504950495049505,
+      "loss": 0.9915,
+      "step": 7400
+    },
+    {
+      "epoch": 6.300506967001609,
+      "grad_norm": 16055.3349609375,
+      "learning_rate": 0.000503960396039604,
+      "loss": 0.9903,
+      "step": 7410
+    },
+    {
+      "epoch": 6.309007012537567,
+      "grad_norm": 16897.765625,
+      "learning_rate": 0.000502970297029703,
+      "loss": 0.9908,
+      "step": 7420
+    },
+    {
+      "epoch": 6.3175070580735255,
+      "grad_norm": 19176.193359375,
+      "learning_rate": 0.000501980198019802,
+      "loss": 0.9898,
+      "step": 7430
+    },
+    {
+      "epoch": 6.326007103609483,
+      "grad_norm": 14390.6298828125,
+      "learning_rate": 0.000500990099009901,
+      "loss": 0.9904,
+      "step": 7440
+    },
+    {
+      "epoch": 6.334507149145442,
+      "grad_norm": 21656.654296875,
+      "learning_rate": 0.0005,
+      "loss": 0.9916,
+      "step": 7450
+    },
+    {
+      "epoch": 6.3430071946814,
+      "grad_norm": 18097.005859375,
+      "learning_rate": 0.0004990099009900991,
+      "loss": 0.9902,
+      "step": 7460
+    },
+    {
+      "epoch": 6.351507240217359,
+      "grad_norm": 17372.796875,
+      "learning_rate": 0.000498019801980198,
+      "loss": 0.9896,
+      "step": 7470
+    },
+    {
+      "epoch": 6.351507240217359,
+      "eval_accuracy": 0.7878735288789341,
+      "eval_loss": 0.9011977314949036,
+      "eval_runtime": 1035.7365,
+      "eval_samples_per_second": 482.148,
+      "eval_steps_per_second": 0.821,
+      "step": 7470
+    },
+    {
+      "epoch": 6.3600072857533165,
+      "grad_norm": 16490.4375,
+      "learning_rate": 0.0004970297029702971,
+      "loss": 0.989,
+      "step": 7480
+    },
+    {
+      "epoch": 6.368507331289274,
+      "grad_norm": 18962.072265625,
+      "learning_rate": 0.000496039603960396,
+      "loss": 0.9887,
+      "step": 7490
+    },
+    {
+      "epoch": 6.377007376825233,
+      "grad_norm": 19760.28125,
+      "learning_rate": 0.0004950495049504951,
+      "loss": 0.9891,
+      "step": 7500
+    },
+    {
+      "epoch": 6.385507422361191,
+      "grad_norm": 16014.2646484375,
+      "learning_rate": 0.0004940594059405941,
+      "loss": 0.9921,
+      "step": 7510
+    },
+    {
+      "epoch": 6.39400746789715,
+      "grad_norm": 18045.798828125,
+      "learning_rate": 0.000493069306930693,
+      "loss": 0.9886,
+      "step": 7520
+    },
+    {
+      "epoch": 6.4025075134331075,
+      "grad_norm": 18082.34375,
+      "learning_rate": 0.0004920792079207921,
+      "loss": 0.987,
+      "step": 7530
+    },
+    {
+      "epoch": 6.411007558969066,
+      "grad_norm": 17991.447265625,
+      "learning_rate": 0.000491089108910891,
+      "loss": 0.99,
+      "step": 7540
+    },
+    {
+      "epoch": 6.419507604505024,
+      "grad_norm": 15685.39453125,
+      "learning_rate": 0.0004900990099009901,
+      "loss": 0.9884,
+      "step": 7550
+    },
+    {
+      "epoch": 6.428007650040982,
+      "grad_norm": 16992.146484375,
+      "learning_rate": 0.0004891089108910892,
+      "loss": 0.9878,
+      "step": 7560
+    },
+    {
+      "epoch": 6.428007650040982,
+      "eval_accuracy": 0.7882446558820121,
+      "eval_loss": 0.8991919755935669,
+      "eval_runtime": 1034.5241,
+      "eval_samples_per_second": 482.713,
+      "eval_steps_per_second": 0.822,
+      "step": 7560
+    },
+    {
+      "epoch": 6.436507695576941,
+      "grad_norm": 18808.25390625,
+      "learning_rate": 0.0004881188118811881,
+      "loss": 0.9872,
+      "step": 7570
+    },
+    {
+      "epoch": 6.4450077411128985,
+      "grad_norm": 18018.94921875,
+      "learning_rate": 0.00048712871287128715,
+      "loss": 0.9882,
+      "step": 7580
+    },
+    {
+      "epoch": 6.453507786648857,
+      "grad_norm": 16361.677734375,
+      "learning_rate": 0.00048613861386138615,
+      "loss": 0.9867,
+      "step": 7590
+    },
+    {
+      "epoch": 6.462007832184815,
+      "grad_norm": 16750.744140625,
+      "learning_rate": 0.00048514851485148515,
+      "loss": 0.9878,
+      "step": 7600
+    },
+    {
+      "epoch": 6.470507877720774,
+      "grad_norm": 19413.93359375,
+      "learning_rate": 0.00048415841584158414,
+      "loss": 0.9868,
+      "step": 7610
+    },
+    {
+      "epoch": 6.479007923256732,
+      "grad_norm": 17738.328125,
+      "learning_rate": 0.00048316831683168314,
+      "loss": 0.9869,
+      "step": 7620
+    },
+    {
+      "epoch": 6.4875079687926895,
+      "grad_norm": 16600.763671875,
+      "learning_rate": 0.0004821782178217822,
+      "loss": 0.9868,
+      "step": 7630
+    },
+    {
+      "epoch": 6.496008014328648,
+      "grad_norm": 19458.306640625,
+      "learning_rate": 0.0004811881188118812,
+      "loss": 0.9842,
+      "step": 7640
+    },
+    {
+      "epoch": 6.504508059864606,
+      "grad_norm": 15481.515625,
+      "learning_rate": 0.0004801980198019802,
+      "loss": 0.9886,
+      "step": 7650
+    },
+    {
+      "epoch": 6.504508059864606,
+      "eval_accuracy": 0.7887328723155168,
+      "eval_loss": 0.8970336318016052,
+      "eval_runtime": 1037.2132,
+      "eval_samples_per_second": 481.461,
+      "eval_steps_per_second": 0.82,
+      "step": 7650
+    },
+    {
+      "epoch": 6.513402750371877,
+      "grad_norm": 20331.080078125,
+      "learning_rate": 0.0004792079207920792,
+      "loss": 0.9857,
+      "step": 7660
+    },
+    {
+      "epoch": 6.5219027959078355,
+      "grad_norm": 17513.203125,
+      "learning_rate": 0.0004782178217821782,
+      "loss": 0.9846,
+      "step": 7670
+    },
+    {
+      "epoch": 6.530402841443793,
+      "grad_norm": 18388.54296875,
+      "learning_rate": 0.00047722772277227724,
+      "loss": 0.9853,
+      "step": 7680
+    },
+    {
+      "epoch": 6.538902886979752,
+      "grad_norm": 18303.26953125,
+      "learning_rate": 0.00047623762376237624,
+      "loss": 0.9853,
+      "step": 7690
+    },
+    {
+      "epoch": 6.54740293251571,
+      "grad_norm": 16617.505859375,
+      "learning_rate": 0.00047524752475247524,
+      "loss": 0.9859,
+      "step": 7700
+    },
+    {
+      "epoch": 6.555902978051668,
+      "grad_norm": 17269.935546875,
+      "learning_rate": 0.00047425742574257423,
+      "loss": 0.9855,
+      "step": 7710
+    },
+    {
+      "epoch": 6.5644030235876265,
+      "grad_norm": 16519.927734375,
+      "learning_rate": 0.00047326732673267323,
+      "loss": 0.9843,
+      "step": 7720
+    },
+    {
+      "epoch": 6.572903069123584,
+      "grad_norm": 15807.8505859375,
+      "learning_rate": 0.0004722772277227723,
+      "loss": 0.983,
+      "step": 7730
+    },
+    {
+      "epoch": 6.581403114659543,
+      "grad_norm": 15810.3408203125,
+      "learning_rate": 0.0004712871287128713,
+      "loss": 0.9861,
+      "step": 7740
+    },
+    {
+      "epoch": 6.581403114659543,
+      "eval_accuracy": 0.7891655969340131,
+      "eval_loss": 0.8951780200004578,
+      "eval_runtime": 1035.4986,
+      "eval_samples_per_second": 482.259,
+      "eval_steps_per_second": 0.821,
+      "step": 7740
+    },
+    {
+      "epoch": 6.589903160195501,
+      "grad_norm": 16766.341796875,
+      "learning_rate": 0.0004702970297029703,
+      "loss": 0.9851,
+      "step": 7750
+    },
+    {
+      "epoch": 6.59840320573146,
+      "grad_norm": 20359.9140625,
+      "learning_rate": 0.0004693069306930693,
+      "loss": 0.9844,
+      "step": 7760
+    },
+    {
+      "epoch": 6.6069032512674175,
+      "grad_norm": 16457.33984375,
+      "learning_rate": 0.00046831683168316833,
+      "loss": 0.9822,
+      "step": 7770
+    },
+    {
+      "epoch": 6.615403296803375,
+      "grad_norm": 18674.6015625,
+      "learning_rate": 0.0004673267326732674,
+      "loss": 0.983,
+      "step": 7780
+    },
+    {
+      "epoch": 6.623903342339334,
+      "grad_norm": 18476.642578125,
+      "learning_rate": 0.0004663366336633664,
+      "loss": 0.9826,
+      "step": 7790
+    },
+    {
+      "epoch": 6.632403387875292,
+      "grad_norm": 15693.6640625,
+      "learning_rate": 0.0004653465346534654,
+      "loss": 0.9806,
+      "step": 7800
+    },
+    {
+      "epoch": 6.640903433411251,
+      "grad_norm": 18678.359375,
+      "learning_rate": 0.0004643564356435644,
+      "loss": 0.9826,
+      "step": 7810
+    },
+    {
+      "epoch": 6.6494034789472085,
+      "grad_norm": 18860.5390625,
+      "learning_rate": 0.0004633663366336634,
+      "loss": 0.9843,
+      "step": 7820
+    },
+    {
+      "epoch": 6.657903524483167,
+      "grad_norm": 18948.015625,
+      "learning_rate": 0.00046237623762376243,
+      "loss": 0.9814,
+      "step": 7830
+    },
+    {
+      "epoch": 6.657903524483167,
+      "eval_accuracy": 0.789608308121163,
+      "eval_loss": 0.8939031362533569,
+      "eval_runtime": 1028.1352,
+      "eval_samples_per_second": 485.712,
+      "eval_steps_per_second": 0.827,
+      "step": 7830
+    },
+    {
+      "epoch": 6.666403570019125,
+      "grad_norm": 17197.728515625,
+      "learning_rate": 0.00046138613861386143,
+      "loss": 0.9816,
+      "step": 7840
+    },
+    {
+      "epoch": 6.674903615555083,
+      "grad_norm": 21051.08984375,
+      "learning_rate": 0.0004603960396039604,
+      "loss": 0.9812,
+      "step": 7850
+    },
+    {
+      "epoch": 6.683403661091042,
+      "grad_norm": 19255.662109375,
+      "learning_rate": 0.0004594059405940594,
+      "loss": 0.9836,
+      "step": 7860
+    },
+    {
+      "epoch": 6.6919037066269995,
+      "grad_norm": 16365.4599609375,
+      "learning_rate": 0.0004584158415841584,
+      "loss": 0.9824,
+      "step": 7870
+    },
+    {
+      "epoch": 6.700403752162958,
+      "grad_norm": 17249.7109375,
+      "learning_rate": 0.0004574257425742575,
+      "loss": 0.9793,
+      "step": 7880
+    },
+    {
+      "epoch": 6.708903797698916,
+      "grad_norm": 15945.6630859375,
+      "learning_rate": 0.00045643564356435647,
+      "loss": 0.9819,
+      "step": 7890
+    },
+    {
+      "epoch": 6.717403843234875,
+      "grad_norm": 16529.767578125,
+      "learning_rate": 0.00045544554455445547,
+      "loss": 0.9801,
+      "step": 7900
+    },
+    {
+      "epoch": 6.725903888770833,
+      "grad_norm": 16305.484375,
+      "learning_rate": 0.00045445544554455447,
+      "loss": 0.9823,
+      "step": 7910
+    },
+    {
+      "epoch": 6.7344039343067905,
+      "grad_norm": 17304.328125,
+      "learning_rate": 0.00045346534653465347,
+      "loss": 0.9782,
+      "step": 7920
+    },
+    {
+      "epoch": 6.7344039343067905,
+      "eval_accuracy": 0.7898104897932849,
+      "eval_loss": 0.8923720121383667,
+      "eval_runtime": 1032.3531,
+      "eval_samples_per_second": 483.728,
+      "eval_steps_per_second": 0.823,
+      "step": 7920
+    },
+    {
+      "epoch": 6.742903979842749,
+      "grad_norm": 16057.078125,
+      "learning_rate": 0.0004524752475247525,
+      "loss": 0.9804,
+      "step": 7930
+    },
+    {
+      "epoch": 6.751404025378707,
+      "grad_norm": 16539.8046875,
+      "learning_rate": 0.0004514851485148515,
+      "loss": 0.9816,
+      "step": 7940
+    },
+    {
+      "epoch": 6.759904070914666,
+      "grad_norm": 18756.751953125,
+      "learning_rate": 0.0004504950495049505,
+      "loss": 0.9803,
+      "step": 7950
+    },
+    {
+      "epoch": 6.768404116450624,
+      "grad_norm": 16140.8369140625,
+      "learning_rate": 0.0004495049504950495,
+      "loss": 0.9807,
+      "step": 7960
+    },
+    {
+      "epoch": 6.776904161986582,
+      "grad_norm": 16101.984375,
+      "learning_rate": 0.0004485148514851485,
+      "loss": 0.9795,
+      "step": 7970
+    },
+    {
+      "epoch": 6.78540420752254,
+      "grad_norm": 18084.240234375,
+      "learning_rate": 0.00044752475247524756,
+      "loss": 0.9798,
+      "step": 7980
+    },
+    {
+      "epoch": 6.793904253058498,
+      "grad_norm": 17187.412109375,
+      "learning_rate": 0.00044653465346534656,
+      "loss": 0.9801,
+      "step": 7990
+    },
+    {
+      "epoch": 6.802404298594457,
+      "grad_norm": 16965.23046875,
+      "learning_rate": 0.00044554455445544556,
+      "loss": 0.9784,
+      "step": 8000
+    },
+    {
+      "epoch": 6.810904344130415,
+      "grad_norm": 16516.47265625,
+      "learning_rate": 0.00044455445544554456,
+      "loss": 0.9786,
+      "step": 8010
+    },
+    {
+      "epoch": 6.810904344130415,
+      "eval_accuracy": 0.7902325926967286,
+      "eval_loss": 0.8902812004089355,
+      "eval_runtime": 1032.3456,
+      "eval_samples_per_second": 483.731,
+      "eval_steps_per_second": 0.823,
+      "step": 8010
+    },
+    {
+      "epoch": 6.819404389666373,
+      "grad_norm": 18656.958984375,
+      "learning_rate": 0.00044356435643564356,
+      "loss": 0.9794,
+      "step": 8020
+    },
+    {
+      "epoch": 6.827904435202331,
+      "grad_norm": 18381.744140625,
+      "learning_rate": 0.0004425742574257426,
+      "loss": 0.978,
+      "step": 8030
+    },
+    {
+      "epoch": 6.83640448073829,
+      "grad_norm": 16134.7392578125,
+      "learning_rate": 0.0004415841584158416,
+      "loss": 0.9783,
+      "step": 8040
+    },
+    {
+      "epoch": 6.844904526274248,
+      "grad_norm": 15535.76953125,
+      "learning_rate": 0.0004405940594059406,
+      "loss": 0.9771,
+      "step": 8050
+    },
+    {
+      "epoch": 6.853404571810206,
+      "grad_norm": 17276.91015625,
+      "learning_rate": 0.0004396039603960396,
+      "loss": 0.9778,
+      "step": 8060
+    },
+    {
+      "epoch": 6.861904617346164,
+      "grad_norm": 16477.451171875,
+      "learning_rate": 0.0004386138613861386,
+      "loss": 0.9762,
+      "step": 8070
+    },
+    {
+      "epoch": 6.870404662882122,
+      "grad_norm": 17235.421875,
+      "learning_rate": 0.00043762376237623765,
+      "loss": 0.9758,
+      "step": 8080
+    },
+    {
+      "epoch": 6.878904708418081,
+      "grad_norm": 23922.306640625,
+      "learning_rate": 0.00043663366336633665,
+      "loss": 0.9771,
+      "step": 8090
+    },
+    {
+      "epoch": 6.887404753954039,
+      "grad_norm": 20163.564453125,
+      "learning_rate": 0.00043564356435643565,
+      "loss": 0.9786,
+      "step": 8100
+    },
+    {
+      "epoch": 6.887404753954039,
+      "eval_accuracy": 0.7905720984224686,
+      "eval_loss": 0.8882827162742615,
+      "eval_runtime": 1035.2187,
+      "eval_samples_per_second": 482.389,
+      "eval_steps_per_second": 0.821,
+      "step": 8100
+    },
+    {
+      "epoch": 6.895904799489998,
+      "grad_norm": 22567.283203125,
+      "learning_rate": 0.00043465346534653465,
+      "loss": 0.9767,
+      "step": 8110
+    },
+    {
+      "epoch": 6.904404845025955,
+      "grad_norm": 18677.17578125,
+      "learning_rate": 0.00043366336633663365,
+      "loss": 0.9768,
+      "step": 8120
+    },
+    {
+      "epoch": 6.912904890561913,
+      "grad_norm": 18427.8046875,
+      "learning_rate": 0.0004326732673267327,
+      "loss": 0.9765,
+      "step": 8130
+    },
+    {
+      "epoch": 6.921404936097872,
+      "grad_norm": 16562.19140625,
+      "learning_rate": 0.0004316831683168317,
+      "loss": 0.9778,
+      "step": 8140
+    },
+    {
+      "epoch": 6.92990498163383,
+      "grad_norm": 18019.8046875,
+      "learning_rate": 0.0004306930693069307,
+      "loss": 0.9765,
+      "step": 8150
+    },
+    {
+      "epoch": 6.938405027169789,
+      "grad_norm": 15273.8798828125,
+      "learning_rate": 0.0004297029702970297,
+      "loss": 0.9756,
+      "step": 8160
+    },
+    {
+      "epoch": 6.946905072705746,
+      "grad_norm": 16736.18359375,
+      "learning_rate": 0.0004287128712871287,
+      "loss": 0.9739,
+      "step": 8170
+    },
+    {
+      "epoch": 6.955405118241705,
+      "grad_norm": 17612.673828125,
+      "learning_rate": 0.00042772277227722774,
+      "loss": 0.9772,
+      "step": 8180
+    },
+    {
+      "epoch": 6.963905163777663,
+      "grad_norm": 15665.3857421875,
+      "learning_rate": 0.00042673267326732674,
+      "loss": 0.9742,
+      "step": 8190
+    },
+    {
+      "epoch": 6.963905163777663,
+      "eval_accuracy": 0.7910029077910385,
+      "eval_loss": 0.8868052363395691,
+      "eval_runtime": 1033.6186,
+      "eval_samples_per_second": 483.136,
+      "eval_steps_per_second": 0.822,
+      "step": 8190
+    },
+    {
+      "epoch": 6.972405209313621,
+      "grad_norm": 16346.849609375,
+      "learning_rate": 0.00042574257425742574,
+      "loss": 0.9773,
+      "step": 8200
+    },
+    {
+      "epoch": 6.98090525484958,
+      "grad_norm": 15904.78515625,
+      "learning_rate": 0.00042475247524752474,
+      "loss": 0.9756,
+      "step": 8210
+    },
+    {
+      "epoch": 6.989405300385537,
+      "grad_norm": 16064.7041015625,
+      "learning_rate": 0.00042376237623762374,
+      "loss": 0.974,
+      "step": 8220
+    },
+    {
+      "epoch": 6.997905345921496,
+      "grad_norm": 17884.13671875,
+      "learning_rate": 0.0004227722772277228,
+      "loss": 0.9747,
+      "step": 8230
+    },
+    {
+      "epoch": 7.006405391457454,
+      "grad_norm": 14780.166015625,
+      "learning_rate": 0.0004217821782178218,
+      "loss": 0.9743,
+      "step": 8240
+    },
+    {
+      "epoch": 7.014905436993413,
+      "grad_norm": 17861.595703125,
+      "learning_rate": 0.0004207920792079208,
+      "loss": 0.9744,
+      "step": 8250
+    },
+    {
+      "epoch": 7.023405482529371,
+      "grad_norm": 16821.708984375,
+      "learning_rate": 0.0004198019801980198,
+      "loss": 0.9732,
+      "step": 8260
+    },
+    {
+      "epoch": 7.031905528065329,
+      "grad_norm": 18202.1796875,
+      "learning_rate": 0.0004188118811881188,
+      "loss": 0.9746,
+      "step": 8270
+    },
+    {
+      "epoch": 7.040405573601287,
+      "grad_norm": 15645.1865234375,
+      "learning_rate": 0.00041782178217821784,
+      "loss": 0.9735,
+      "step": 8280
+    },
+    {
+      "epoch": 7.040405573601287,
+      "eval_accuracy": 0.7912855873371129,
+      "eval_loss": 0.8865892887115479,
+      "eval_runtime": 1039.2279,
+      "eval_samples_per_second": 480.528,
+      "eval_steps_per_second": 0.818,
+      "step": 8280
+    },
+    {
+      "epoch": 7.048905619137245,
+      "grad_norm": 16949.755859375,
+      "learning_rate": 0.00041683168316831683,
+      "loss": 0.9736,
+      "step": 8290
+    },
+    {
+      "epoch": 7.057405664673204,
+      "grad_norm": 18461.337890625,
+      "learning_rate": 0.00041584158415841583,
+      "loss": 0.9728,
+      "step": 8300
+    },
+    {
+      "epoch": 7.065905710209162,
+      "grad_norm": 17668.466796875,
+      "learning_rate": 0.00041485148514851483,
+      "loss": 0.9721,
+      "step": 8310
+    },
+    {
+      "epoch": 7.07440575574512,
+      "grad_norm": 15126.234375,
+      "learning_rate": 0.00041386138613861383,
+      "loss": 0.9718,
+      "step": 8320
+    },
+    {
+      "epoch": 7.082905801281078,
+      "grad_norm": 16266.1083984375,
+      "learning_rate": 0.0004128712871287129,
+      "loss": 0.9733,
+      "step": 8330
+    },
+    {
+      "epoch": 7.091405846817037,
+      "grad_norm": 19422.59375,
+      "learning_rate": 0.0004118811881188119,
+      "loss": 0.9733,
+      "step": 8340
+    },
+    {
+      "epoch": 7.099905892352995,
+      "grad_norm": 17168.087890625,
+      "learning_rate": 0.0004108910891089109,
+      "loss": 0.9723,
+      "step": 8350
+    },
+    {
+      "epoch": 7.108405937888953,
+      "grad_norm": 16709.5625,
+      "learning_rate": 0.0004099009900990099,
+      "loss": 0.9712,
+      "step": 8360
+    },
+    {
+      "epoch": 7.116905983424911,
+      "grad_norm": 15946.3232421875,
+      "learning_rate": 0.0004089108910891089,
+      "loss": 0.9726,
+      "step": 8370
+    },
+    {
+      "epoch": 7.116905983424911,
+      "eval_accuracy": 0.7916597760806168,
+      "eval_loss": 0.8836008310317993,
+      "eval_runtime": 1033.0801,
+      "eval_samples_per_second": 483.388,
+      "eval_steps_per_second": 0.823,
+      "step": 8370
+    },
+    {
+      "epoch": 7.125406028960869,
+      "grad_norm": 13808.0322265625,
+      "learning_rate": 0.0004079207920792079,
+      "loss": 0.9693,
+      "step": 8380
+    },
+    {
+      "epoch": 7.133906074496828,
+      "grad_norm": 19899.408203125,
+      "learning_rate": 0.0004069306930693069,
+      "loss": 0.9695,
+      "step": 8390
+    },
+    {
+      "epoch": 7.142406120032786,
+      "grad_norm": 14987.80078125,
+      "learning_rate": 0.000405940594059406,
+      "loss": 0.9698,
+      "step": 8400
+    },
+    {
+      "epoch": 7.1509061655687445,
+      "grad_norm": 16658.63671875,
+      "learning_rate": 0.000404950495049505,
+      "loss": 0.9711,
+      "step": 8410
+    },
+    {
+      "epoch": 7.159406211104702,
+      "grad_norm": 16382.1494140625,
+      "learning_rate": 0.00040396039603960397,
+      "loss": 0.9717,
+      "step": 8420
+    },
+    {
+      "epoch": 7.16790625664066,
+      "grad_norm": 16640.423828125,
+      "learning_rate": 0.000402970297029703,
+      "loss": 0.9706,
+      "step": 8430
+    },
+    {
+      "epoch": 7.176406302176619,
+      "grad_norm": 17918.15625,
+      "learning_rate": 0.000401980198019802,
+      "loss": 0.9723,
+      "step": 8440
+    },
+    {
+      "epoch": 7.184906347712577,
+      "grad_norm": 16958.197265625,
+      "learning_rate": 0.000400990099009901,
+      "loss": 0.9704,
+      "step": 8450
+    },
+    {
+      "epoch": 7.1934063932485355,
+      "grad_norm": 15377.521484375,
+      "learning_rate": 0.0004,
+      "loss": 0.97,
+      "step": 8460
+    },
+    {
+      "epoch": 7.1934063932485355,
+      "eval_accuracy": 0.7920878335859806,
+      "eval_loss": 0.8810552954673767,
+      "eval_runtime": 1033.6339,
+      "eval_samples_per_second": 483.129,
+      "eval_steps_per_second": 0.822,
+      "step": 8460
+    },
+    {
+      "epoch": 7.201906438784493,
+      "grad_norm": 15135.03515625,
+      "learning_rate": 0.000399009900990099,
+      "loss": 0.9684,
+      "step": 8470
+    },
+    {
+      "epoch": 7.210406484320452,
+      "grad_norm": 16375.2431640625,
+      "learning_rate": 0.00039801980198019807,
+      "loss": 0.9691,
+      "step": 8480
+    },
+    {
+      "epoch": 7.21890652985641,
+      "grad_norm": 16684.763671875,
+      "learning_rate": 0.00039702970297029707,
+      "loss": 0.9678,
+      "step": 8490
+    },
+    {
+      "epoch": 7.227406575392368,
+      "grad_norm": 17551.935546875,
+      "learning_rate": 0.00039603960396039607,
+      "loss": 0.9688,
+      "step": 8500
+    },
+    {
+      "epoch": 7.2359066209283265,
+      "grad_norm": 17371.619140625,
+      "learning_rate": 0.00039504950495049506,
+      "loss": 0.9717,
+      "step": 8510
+    },
+    {
+      "epoch": 7.244406666464284,
+      "grad_norm": 20004.443359375,
+      "learning_rate": 0.00039405940594059406,
+      "loss": 0.9672,
+      "step": 8520
+    },
+    {
+      "epoch": 7.252906712000243,
+      "grad_norm": 18529.32421875,
+      "learning_rate": 0.0003930693069306931,
+      "loss": 0.9693,
+      "step": 8530
+    },
+    {
+      "epoch": 7.261406757536201,
+      "grad_norm": 16488.56640625,
+      "learning_rate": 0.0003920792079207921,
+      "loss": 0.9678,
+      "step": 8540
+    },
+    {
+      "epoch": 7.26990680307216,
+      "grad_norm": 22428.466796875,
+      "learning_rate": 0.0003910891089108911,
+      "loss": 0.9693,
+      "step": 8550
+    },
+    {
+      "epoch": 7.26990680307216,
+      "eval_accuracy": 0.792256339480986,
+      "eval_loss": 0.8810757398605347,
+      "eval_runtime": 1033.8252,
+      "eval_samples_per_second": 483.039,
+      "eval_steps_per_second": 0.822,
+      "step": 8550
+    },
+    {
+      "epoch": 7.2784068486081175,
+      "grad_norm": 17588.4453125,
+      "learning_rate": 0.0003900990099009901,
+      "loss": 0.9679,
+      "step": 8560
+    },
+    {
+      "epoch": 7.286906894144076,
+      "grad_norm": 18152.78515625,
+      "learning_rate": 0.0003891089108910891,
+      "loss": 0.9668,
+      "step": 8570
+    },
+    {
+      "epoch": 7.295406939680034,
+      "grad_norm": 14463.046875,
+      "learning_rate": 0.00038811881188118816,
+      "loss": 0.9679,
+      "step": 8580
+    },
+    {
+      "epoch": 7.303906985215992,
+      "grad_norm": 19075.265625,
+      "learning_rate": 0.00038712871287128716,
+      "loss": 0.9681,
+      "step": 8590
+    },
+    {
+      "epoch": 7.312407030751951,
+      "grad_norm": 21981.548828125,
+      "learning_rate": 0.00038613861386138616,
+      "loss": 0.9682,
+      "step": 8600
+    },
+    {
+      "epoch": 7.3209070762879085,
+      "grad_norm": 20439.08984375,
+      "learning_rate": 0.00038514851485148515,
+      "loss": 0.968,
+      "step": 8610
+    },
+    {
+      "epoch": 7.329407121823867,
+      "grad_norm": 17120.98046875,
+      "learning_rate": 0.00038415841584158415,
+      "loss": 0.9664,
+      "step": 8620
+    },
+    {
+      "epoch": 7.337907167359825,
+      "grad_norm": 18788.953125,
+      "learning_rate": 0.0003831683168316832,
+      "loss": 0.965,
+      "step": 8630
+    },
+    {
+      "epoch": 7.346407212895784,
+      "grad_norm": 14774.0830078125,
+      "learning_rate": 0.0003821782178217822,
+      "loss": 0.9666,
+      "step": 8640
+    },
+    {
+      "epoch": 7.346407212895784,
+      "eval_accuracy": 0.7926509573029413,
+      "eval_loss": 0.8783407807350159,
+      "eval_runtime": 1035.0192,
+      "eval_samples_per_second": 482.482,
+      "eval_steps_per_second": 0.821,
+      "step": 8640
+    },
+    {
+      "epoch": 7.354907258431742,
+      "grad_norm": 17107.38671875,
+      "learning_rate": 0.0003811881188118812,
+      "loss": 0.9666,
+      "step": 8650
+    },
+    {
+      "epoch": 7.3634073039676995,
+      "grad_norm": 15051.875,
+      "learning_rate": 0.0003801980198019802,
+      "loss": 0.9668,
+      "step": 8660
+    },
+    {
+      "epoch": 7.371907349503658,
+      "grad_norm": 16625.01171875,
+      "learning_rate": 0.0003792079207920792,
+      "loss": 0.9665,
+      "step": 8670
+    },
+    {
+      "epoch": 7.380407395039616,
+      "grad_norm": 15725.7158203125,
+      "learning_rate": 0.00037821782178217825,
+      "loss": 0.9665,
+      "step": 8680
+    },
+    {
+      "epoch": 7.388907440575575,
+      "grad_norm": 17703.29296875,
+      "learning_rate": 0.00037722772277227725,
+      "loss": 0.9653,
+      "step": 8690
+    },
+    {
+      "epoch": 7.397407486111533,
+      "grad_norm": 18342.701171875,
+      "learning_rate": 0.00037623762376237625,
+      "loss": 0.968,
+      "step": 8700
+    },
+    {
+      "epoch": 7.405907531647491,
+      "grad_norm": 16660.326171875,
+      "learning_rate": 0.00037524752475247524,
+      "loss": 0.9633,
+      "step": 8710
+    },
+    {
+      "epoch": 7.414407577183449,
+      "grad_norm": 14380.01171875,
+      "learning_rate": 0.00037425742574257424,
+      "loss": 0.9652,
+      "step": 8720
+    },
+    {
+      "epoch": 7.422907622719407,
+      "grad_norm": 15314.05859375,
+      "learning_rate": 0.0003732673267326733,
+      "loss": 0.9645,
+      "step": 8730
+    },
+    {
+      "epoch": 7.422907622719407,
+      "eval_accuracy": 0.7929085777897834,
+      "eval_loss": 0.8774629235267639,
+      "eval_runtime": 1036.1479,
+      "eval_samples_per_second": 481.956,
+      "eval_steps_per_second": 0.82,
+      "step": 8730
+    },
+    {
+      "epoch": 7.431407668255366,
+      "grad_norm": 15377.4306640625,
+      "learning_rate": 0.0003722772277227723,
+      "loss": 0.9661,
+      "step": 8740
+    },
+    {
+      "epoch": 7.439907713791324,
+      "grad_norm": 20652.443359375,
+      "learning_rate": 0.0003712871287128713,
+      "loss": 0.9638,
+      "step": 8750
+    },
+    {
+      "epoch": 7.448407759327282,
+      "grad_norm": 15570.021484375,
+      "learning_rate": 0.0003702970297029703,
+      "loss": 0.9647,
+      "step": 8760
+    },
+    {
+      "epoch": 7.45690780486324,
+      "grad_norm": 17774.888671875,
+      "learning_rate": 0.0003693069306930693,
+      "loss": 0.9651,
+      "step": 8770
+    },
+    {
+      "epoch": 7.465407850399199,
+      "grad_norm": 19272.056640625,
+      "learning_rate": 0.00036831683168316834,
+      "loss": 0.964,
+      "step": 8780
+    },
+    {
+      "epoch": 7.473907895935157,
+      "grad_norm": 17114.37109375,
+      "learning_rate": 0.00036732673267326734,
+      "loss": 0.963,
+      "step": 8790
+    },
+    {
+      "epoch": 7.482407941471115,
+      "grad_norm": 18956.96484375,
+      "learning_rate": 0.00036633663366336634,
+      "loss": 0.9648,
+      "step": 8800
+    },
+    {
+      "epoch": 7.490907987007073,
+      "grad_norm": 16027.392578125,
+      "learning_rate": 0.00036534653465346533,
+      "loss": 0.9646,
+      "step": 8810
+    },
+    {
+      "epoch": 7.499408032543031,
+      "grad_norm": 16447.48046875,
+      "learning_rate": 0.00036435643564356433,
+      "loss": 0.963,
+      "step": 8820
+    },
+    {
+      "epoch": 7.499408032543031,
+      "eval_accuracy": 0.7934318103664766,
+      "eval_loss": 0.8747227191925049,
+      "eval_runtime": 1031.384,
+      "eval_samples_per_second": 484.182,
+      "eval_steps_per_second": 0.824,
+      "step": 8820
+    },
+    {
+      "epoch": 7.50790807807899,
+      "grad_norm": 16898.626953125,
+      "learning_rate": 0.0003633663366336634,
+      "loss": 0.9651,
+      "step": 8830
+    },
+    {
+      "epoch": 7.516408123614948,
+      "grad_norm": 16217.8544921875,
+      "learning_rate": 0.0003623762376237624,
+      "loss": 0.9618,
+      "step": 8840
+    },
+    {
+      "epoch": 7.5249081691509065,
+      "grad_norm": 20356.615234375,
+      "learning_rate": 0.0003613861386138614,
+      "loss": 0.9601,
+      "step": 8850
+    },
+    {
+      "epoch": 7.533408214686864,
+      "grad_norm": 17453.779296875,
+      "learning_rate": 0.0003603960396039604,
+      "loss": 0.9627,
+      "step": 8860
+    },
+    {
+      "epoch": 7.541908260222822,
+      "grad_norm": 14681.1591796875,
+      "learning_rate": 0.0003594059405940594,
+      "loss": 0.9616,
+      "step": 8870
+    },
+    {
+      "epoch": 7.550408305758781,
+      "grad_norm": 16191.0478515625,
+      "learning_rate": 0.00035841584158415843,
+      "loss": 0.9623,
+      "step": 8880
+    },
+    {
+      "epoch": 7.558908351294739,
+      "grad_norm": 15307.25,
+      "learning_rate": 0.00035742574257425743,
+      "loss": 0.9604,
+      "step": 8890
+    },
+    {
+      "epoch": 7.5674083968306975,
+      "grad_norm": 18245.2265625,
+      "learning_rate": 0.0003564356435643564,
+      "loss": 0.9619,
+      "step": 8900
+    },
+    {
+      "epoch": 7.575908442366655,
+      "grad_norm": 16105.3232421875,
+      "learning_rate": 0.0003554455445544554,
+      "loss": 0.9643,
+      "step": 8910
+    },
+    {
+      "epoch": 7.575908442366655,
+      "eval_accuracy": 0.793696329436229,
+      "eval_loss": 0.8743059039115906,
+      "eval_runtime": 1035.7771,
+      "eval_samples_per_second": 482.129,
+      "eval_steps_per_second": 0.821,
+      "step": 8910
+    },
+    {
+      "epoch": 7.584408487902614,
+      "grad_norm": 16268.6376953125,
+      "learning_rate": 0.0003544554455445544,
+      "loss": 0.9632,
+      "step": 8920
+    },
+    {
+      "epoch": 7.592908533438572,
+      "grad_norm": 16207.1279296875,
+      "learning_rate": 0.0003534653465346535,
+      "loss": 0.9614,
+      "step": 8930
+    },
+    {
+      "epoch": 7.60140857897453,
+      "grad_norm": 15891.35546875,
+      "learning_rate": 0.0003524752475247525,
+      "loss": 0.9623,
+      "step": 8940
+    },
+    {
+      "epoch": 7.6099086245104886,
+      "grad_norm": 15914.4921875,
+      "learning_rate": 0.00035148514851485147,
+      "loss": 0.9609,
+      "step": 8950
+    },
+    {
+      "epoch": 7.618408670046446,
+      "grad_norm": 15182.478515625,
+      "learning_rate": 0.00035049504950495047,
+      "loss": 0.9601,
+      "step": 8960
+    },
+    {
+      "epoch": 7.626908715582405,
+      "grad_norm": 15736.513671875,
+      "learning_rate": 0.00034950495049504947,
+      "loss": 0.9613,
+      "step": 8970
+    },
+    {
+      "epoch": 7.635408761118363,
+      "grad_norm": 17880.693359375,
+      "learning_rate": 0.0003485148514851485,
+      "loss": 0.9606,
+      "step": 8980
+    },
+    {
+      "epoch": 7.643908806654322,
+      "grad_norm": 15555.0341796875,
+      "learning_rate": 0.0003475247524752475,
+      "loss": 0.9618,
+      "step": 8990
+    },
+    {
+      "epoch": 7.6524088521902796,
+      "grad_norm": 17536.287109375,
+      "learning_rate": 0.0003465346534653465,
+      "loss": 0.9606,
+      "step": 9000
+    },
+    {
+      "epoch": 7.6524088521902796,
+      "eval_accuracy": 0.7939732586317862,
+      "eval_loss": 0.8728435039520264,
+      "eval_runtime": 1035.2446,
+      "eval_samples_per_second": 482.377,
+      "eval_steps_per_second": 0.821,
+      "step": 9000
+    },
+    {
+      "epoch": 7.660908897726237,
+      "grad_norm": 16182.7158203125,
+      "learning_rate": 0.0003455445544554455,
+      "loss": 0.9605,
+      "step": 9010
+    },
+    {
+      "epoch": 7.669408943262196,
+      "grad_norm": 17641.724609375,
+      "learning_rate": 0.0003445544554455445,
+      "loss": 0.9611,
+      "step": 9020
+    },
+    {
+      "epoch": 7.677908988798154,
+      "grad_norm": 16540.5625,
+      "learning_rate": 0.0003435643564356436,
+      "loss": 0.9596,
+      "step": 9030
+    },
+    {
+      "epoch": 7.686409034334113,
+      "grad_norm": 15668.515625,
+      "learning_rate": 0.0003425742574257426,
+      "loss": 0.9592,
+      "step": 9040
+    },
+    {
+      "epoch": 7.694909079870071,
+      "grad_norm": 16943.318359375,
+      "learning_rate": 0.0003415841584158416,
+      "loss": 0.9619,
+      "step": 9050
+    },
+    {
+      "epoch": 7.703409125406029,
+      "grad_norm": 15736.9775390625,
+      "learning_rate": 0.0003405940594059406,
+      "loss": 0.959,
+      "step": 9060
+    },
+    {
+      "epoch": 7.711909170941987,
+      "grad_norm": 17073.3203125,
+      "learning_rate": 0.0003396039603960396,
+      "loss": 0.9609,
+      "step": 9070
+    },
+    {
+      "epoch": 7.720409216477945,
+      "grad_norm": 14390.5009765625,
+      "learning_rate": 0.00033861386138613867,
+      "loss": 0.9576,
+      "step": 9080
+    },
+    {
+      "epoch": 7.728909262013904,
+      "grad_norm": 17994.197265625,
+      "learning_rate": 0.00033762376237623766,
+      "loss": 0.9576,
+      "step": 9090
+    },
+    {
+      "epoch": 7.728909262013904,
+      "eval_accuracy": 0.7942283085944538,
+      "eval_loss": 0.8715931177139282,
+      "eval_runtime": 1034.9441,
+      "eval_samples_per_second": 482.517,
+      "eval_steps_per_second": 0.821,
+      "step": 9090
+    },
+    {
+      "epoch": 7.737409307549862,
+      "grad_norm": 14687.2646484375,
+      "learning_rate": 0.00033663366336633666,
+      "loss": 0.957,
+      "step": 9100
+    },
+    {
+      "epoch": 7.74590935308582,
+      "grad_norm": 16330.0478515625,
+      "learning_rate": 0.00033564356435643566,
+      "loss": 0.9595,
+      "step": 9110
+    },
+    {
+      "epoch": 7.754409398621778,
+      "grad_norm": 14873.998046875,
+      "learning_rate": 0.00033465346534653466,
+      "loss": 0.9596,
+      "step": 9120
+    },
+    {
+      "epoch": 7.762909444157737,
+      "grad_norm": 14278.1953125,
+      "learning_rate": 0.0003336633663366337,
+      "loss": 0.959,
+      "step": 9130
+    },
+    {
+      "epoch": 7.771409489693695,
+      "grad_norm": 17600.927734375,
+      "learning_rate": 0.0003326732673267327,
+      "loss": 0.9573,
+      "step": 9140
+    },
+    {
+      "epoch": 7.779909535229653,
+      "grad_norm": 15872.1806640625,
+      "learning_rate": 0.0003316831683168317,
+      "loss": 0.9573,
+      "step": 9150
+    },
+    {
+      "epoch": 7.788409580765611,
+      "grad_norm": 19368.9765625,
+      "learning_rate": 0.0003306930693069307,
+      "loss": 0.957,
+      "step": 9160
+    },
+    {
+      "epoch": 7.796909626301569,
+      "grad_norm": 19541.958984375,
+      "learning_rate": 0.0003297029702970297,
+      "loss": 0.9581,
+      "step": 9170
+    },
+    {
+      "epoch": 7.805409671837528,
+      "grad_norm": 16817.328125,
+      "learning_rate": 0.00032871287128712876,
+      "loss": 0.957,
+      "step": 9180
+    },
+    {
+      "epoch": 7.805409671837528,
+      "eval_accuracy": 0.7946937469682237,
+      "eval_loss": 0.8684272766113281,
+      "eval_runtime": 1033.8439,
+      "eval_samples_per_second": 483.03,
+      "eval_steps_per_second": 0.822,
+      "step": 9180
+    },
+    {
+      "epoch": 7.813909717373486,
+      "grad_norm": 18241.623046875,
+      "learning_rate": 0.00032772277227722775,
+      "loss": 0.9567,
+      "step": 9190
+    },
+    {
+      "epoch": 7.8224097629094445,
+      "grad_norm": 15321.79296875,
+      "learning_rate": 0.00032673267326732675,
+      "loss": 0.9574,
+      "step": 9200
+    },
+    {
+      "epoch": 7.830909808445402,
+      "grad_norm": 14790.1884765625,
+      "learning_rate": 0.00032574257425742575,
+      "loss": 0.9558,
+      "step": 9210
+    },
+    {
+      "epoch": 7.83940985398136,
+      "grad_norm": 13895.65234375,
+      "learning_rate": 0.00032475247524752475,
+      "loss": 0.9567,
+      "step": 9220
+    },
+    {
+      "epoch": 7.847909899517319,
+      "grad_norm": 17413.087890625,
+      "learning_rate": 0.0003237623762376238,
+      "loss": 0.9582,
+      "step": 9230
+    },
+    {
+      "epoch": 7.856409945053277,
+      "grad_norm": 15125.3447265625,
+      "learning_rate": 0.0003227722772277228,
+      "loss": 0.9555,
+      "step": 9240
+    },
+    {
+      "epoch": 7.8649099905892355,
+      "grad_norm": 15147.716796875,
+      "learning_rate": 0.0003217821782178218,
+      "loss": 0.9552,
+      "step": 9250
+    },
+    {
+      "epoch": 7.873410036125193,
+      "grad_norm": 16691.9765625,
+      "learning_rate": 0.0003207920792079208,
+      "loss": 0.9569,
+      "step": 9260
+    },
+    {
+      "epoch": 7.881910081661152,
+      "grad_norm": 15234.0234375,
+      "learning_rate": 0.0003198019801980198,
+      "loss": 0.957,
+      "step": 9270
+    },
+    {
+      "epoch": 7.881910081661152,
+      "eval_accuracy": 0.7946698774649099,
+      "eval_loss": 0.8689009547233582,
+      "eval_runtime": 1037.4044,
+      "eval_samples_per_second": 481.373,
+      "eval_steps_per_second": 0.819,
+      "step": 9270
+    },
+    {
+      "epoch": 7.89041012719711,
+      "grad_norm": 20217.794921875,
+      "learning_rate": 0.00031881188118811885,
+      "loss": 0.9558,
+      "step": 9280
+    },
+    {
+      "epoch": 7.898910172733068,
+      "grad_norm": 16890.482421875,
+      "learning_rate": 0.00031782178217821784,
+      "loss": 0.957,
+      "step": 9290
+    },
+    {
+      "epoch": 7.9074102182690265,
+      "grad_norm": 16310.05078125,
+      "learning_rate": 0.00031683168316831684,
+      "loss": 0.9557,
+      "step": 9300
+    },
+    {
+      "epoch": 7.915910263804984,
+      "grad_norm": 15330.0146484375,
+      "learning_rate": 0.00031584158415841584,
+      "loss": 0.9554,
+      "step": 9310
+    },
+    {
+      "epoch": 7.924410309340943,
+      "grad_norm": 14532.234375,
+      "learning_rate": 0.00031485148514851484,
+      "loss": 0.9542,
+      "step": 9320
+    },
+    {
+      "epoch": 7.932910354876901,
+      "grad_norm": 16035.0615234375,
+      "learning_rate": 0.0003138613861386139,
+      "loss": 0.9544,
+      "step": 9330
+    },
+    {
+      "epoch": 7.94141040041286,
+      "grad_norm": 16156.765625,
+      "learning_rate": 0.0003128712871287129,
+      "loss": 0.9549,
+      "step": 9340
+    },
+    {
+      "epoch": 7.9499104459488175,
+      "grad_norm": 16913.37109375,
+      "learning_rate": 0.0003118811881188119,
+      "loss": 0.9537,
+      "step": 9350
+    },
+    {
+      "epoch": 7.958410491484776,
+      "grad_norm": 16376.814453125,
+      "learning_rate": 0.0003108910891089109,
+      "loss": 0.9556,
+      "step": 9360
+    },
+    {
+      "epoch": 7.958410491484776,
+      "eval_accuracy": 0.7950512170789835,
+      "eval_loss": 0.8671656847000122,
+      "eval_runtime": 1035.1375,
+      "eval_samples_per_second": 482.427,
+      "eval_steps_per_second": 0.821,
+      "step": 9360
+    },
+    {
+      "epoch": 7.966910537020734,
+      "grad_norm": 14731.126953125,
+      "learning_rate": 0.0003099009900990099,
+      "loss": 0.9529,
+      "step": 9370
+    },
+    {
+      "epoch": 7.975410582556692,
+      "grad_norm": 16993.232421875,
+      "learning_rate": 0.00030891089108910894,
+      "loss": 0.9548,
+      "step": 9380
+    },
+    {
+      "epoch": 7.983910628092651,
+      "grad_norm": 16157.2880859375,
+      "learning_rate": 0.00030792079207920793,
+      "loss": 0.9543,
+      "step": 9390
+    },
+    {
+      "epoch": 7.9924106736286085,
+      "grad_norm": 17463.8984375,
+      "learning_rate": 0.00030693069306930693,
+      "loss": 0.9537,
+      "step": 9400
+    },
+    {
+      "epoch": 8.000910719164567,
+      "grad_norm": 18573.9765625,
+      "learning_rate": 0.00030594059405940593,
+      "loss": 0.9543,
+      "step": 9410
+    },
+    {
+      "epoch": 8.009410764700526,
+      "grad_norm": 18137.865234375,
+      "learning_rate": 0.00030495049504950493,
+      "loss": 0.9529,
+      "step": 9420
+    },
+    {
+      "epoch": 8.017910810236483,
+      "grad_norm": 17822.484375,
+      "learning_rate": 0.000303960396039604,
+      "loss": 0.9528,
+      "step": 9430
+    },
+    {
+      "epoch": 8.026410855772442,
+      "grad_norm": 16745.134765625,
+      "learning_rate": 0.000302970297029703,
+      "loss": 0.9529,
+      "step": 9440
+    },
+    {
+      "epoch": 8.0349109013084,
+      "grad_norm": 17063.685546875,
+      "learning_rate": 0.000301980198019802,
+      "loss": 0.9525,
+      "step": 9450
+    },
+    {
+      "epoch": 8.0349109013084,
+      "eval_accuracy": 0.7955066441462213,
+      "eval_loss": 0.8656001687049866,
+      "eval_runtime": 1034.9789,
+      "eval_samples_per_second": 482.501,
+      "eval_steps_per_second": 0.821,
+      "step": 9450
+    },
+    {
+      "epoch": 8.043410946844357,
+      "grad_norm": 14951.6552734375,
+      "learning_rate": 0.000300990099009901,
+      "loss": 0.9539,
+      "step": 9460
+    },
+    {
+      "epoch": 8.051910992380316,
+      "grad_norm": 18525.021484375,
+      "learning_rate": 0.0003,
+      "loss": 0.9529,
+      "step": 9470
+    },
+    {
+      "epoch": 8.060411037916275,
+      "grad_norm": 16683.45703125,
+      "learning_rate": 0.000299009900990099,
+      "loss": 0.9535,
+      "step": 9480
+    },
+    {
+      "epoch": 8.068911083452234,
+      "grad_norm": 17987.796875,
+      "learning_rate": 0.000298019801980198,
+      "loss": 0.9544,
+      "step": 9490
+    },
+    {
+      "epoch": 8.07741112898819,
+      "grad_norm": 14249.4453125,
+      "learning_rate": 0.000297029702970297,
+      "loss": 0.953,
+      "step": 9500
+    },
+    {
+      "epoch": 8.08591117452415,
+      "grad_norm": 15678.244140625,
+      "learning_rate": 0.000296039603960396,
+      "loss": 0.9517,
+      "step": 9510
+    },
+    {
+      "epoch": 8.094411220060108,
+      "grad_norm": 16945.619140625,
+      "learning_rate": 0.000295049504950495,
+      "loss": 0.9518,
+      "step": 9520
+    },
+    {
+      "epoch": 8.102911265596065,
+      "grad_norm": 16090.876953125,
+      "learning_rate": 0.00029405940594059407,
+      "loss": 0.9515,
+      "step": 9530
+    },
+    {
+      "epoch": 8.111411311132024,
+      "grad_norm": 15359.59375,
+      "learning_rate": 0.00029306930693069307,
+      "loss": 0.9509,
+      "step": 9540
+    },
+    {
+      "epoch": 8.111411311132024,
+      "eval_accuracy": 0.7956073067738585,
+      "eval_loss": 0.8639572858810425,
+      "eval_runtime": 1036.0688,
+      "eval_samples_per_second": 481.993,
+      "eval_steps_per_second": 0.82,
+      "step": 9540
+    },
+    {
+      "epoch": 8.119911356667982,
+      "grad_norm": 14432.5576171875,
+      "learning_rate": 0.00029207920792079207,
+      "loss": 0.9498,
+      "step": 9550
+    },
+    {
+      "epoch": 8.128411402203941,
+      "grad_norm": 15058.017578125,
+      "learning_rate": 0.00029108910891089107,
+      "loss": 0.9512,
+      "step": 9560
+    },
+    {
+      "epoch": 8.136911447739898,
+      "grad_norm": 14191.0791015625,
+      "learning_rate": 0.00029009900990099006,
+      "loss": 0.9517,
+      "step": 9570
+    },
+    {
+      "epoch": 8.145411493275857,
+      "grad_norm": 15950.939453125,
+      "learning_rate": 0.0002891089108910891,
+      "loss": 0.9506,
+      "step": 9580
+    },
+    {
+      "epoch": 8.153911538811816,
+      "grad_norm": 17082.619140625,
+      "learning_rate": 0.0002881188118811881,
+      "loss": 0.9513,
+      "step": 9590
+    },
+    {
+      "epoch": 8.162411584347772,
+      "grad_norm": 14874.2890625,
+      "learning_rate": 0.0002871287128712871,
+      "loss": 0.9501,
+      "step": 9600
+    },
+    {
+      "epoch": 8.170911629883731,
+      "grad_norm": 14615.6669921875,
+      "learning_rate": 0.0002861386138613861,
+      "loss": 0.951,
+      "step": 9610
+    },
+    {
+      "epoch": 8.17941167541969,
+      "grad_norm": 15949.583984375,
+      "learning_rate": 0.0002851485148514851,
+      "loss": 0.9511,
+      "step": 9620
+    },
+    {
+      "epoch": 8.187911720955649,
+      "grad_norm": 15501.064453125,
+      "learning_rate": 0.00028415841584158416,
+      "loss": 0.9503,
+      "step": 9630
+    },
+    {
+      "epoch": 8.187911720955649,
+      "eval_accuracy": 0.7959485569423415,
+      "eval_loss": 0.8628361821174622,
+      "eval_runtime": 1038.2269,
+      "eval_samples_per_second": 480.991,
+      "eval_steps_per_second": 0.819,
+      "step": 9630
+    },
+    {
+      "epoch": 8.196411766491606,
+      "grad_norm": 14815.34375,
+      "learning_rate": 0.00028316831683168316,
+      "loss": 0.9516,
+      "step": 9640
+    },
+    {
+      "epoch": 8.204911812027564,
+      "grad_norm": 19284.322265625,
+      "learning_rate": 0.00028217821782178216,
+      "loss": 0.9508,
+      "step": 9650
+    },
+    {
+      "epoch": 8.213411857563523,
+      "grad_norm": 15244.177734375,
+      "learning_rate": 0.0002811881188118812,
+      "loss": 0.9494,
+      "step": 9660
+    },
+    {
+      "epoch": 8.22191190309948,
+      "grad_norm": 14576.9482421875,
+      "learning_rate": 0.0002801980198019802,
+      "loss": 0.9485,
+      "step": 9670
+    },
+    {
+      "epoch": 8.230411948635439,
+      "grad_norm": 15386.0654296875,
+      "learning_rate": 0.00027920792079207926,
+      "loss": 0.9501,
+      "step": 9680
+    },
+    {
+      "epoch": 8.238911994171398,
+      "grad_norm": 15885.4248046875,
+      "learning_rate": 0.00027821782178217826,
+      "loss": 0.9493,
+      "step": 9690
+    },
+    {
+      "epoch": 8.247412039707356,
+      "grad_norm": 16190.021484375,
+      "learning_rate": 0.00027722772277227726,
+      "loss": 0.9498,
+      "step": 9700
+    },
+    {
+      "epoch": 8.255912085243313,
+      "grad_norm": 14406.8857421875,
+      "learning_rate": 0.00027623762376237626,
+      "loss": 0.9473,
+      "step": 9710
+    },
+    {
+      "epoch": 8.264412130779272,
+      "grad_norm": 14419.498046875,
+      "learning_rate": 0.00027524752475247525,
+      "loss": 0.9499,
+      "step": 9720
+    },
+    {
+      "epoch": 8.264412130779272,
+      "eval_accuracy": 0.7962577759360605,
+      "eval_loss": 0.8620018362998962,
+      "eval_runtime": 1037.7206,
+      "eval_samples_per_second": 481.226,
+      "eval_steps_per_second": 0.819,
+      "step": 9720
+    },
+    {
+      "epoch": 8.27291217631523,
+      "grad_norm": 16101.1533203125,
+      "learning_rate": 0.0002742574257425743,
+      "loss": 0.9502,
+      "step": 9730
+    },
+    {
+      "epoch": 8.281412221851188,
+      "grad_norm": 15671.9033203125,
+      "learning_rate": 0.0002732673267326733,
+      "loss": 0.9498,
+      "step": 9740
+    },
+    {
+      "epoch": 8.289912267387146,
+      "grad_norm": 15215.625,
+      "learning_rate": 0.0002722772277227723,
+      "loss": 0.9482,
+      "step": 9750
+    },
+    {
+      "epoch": 8.298412312923105,
+      "grad_norm": 16016.99609375,
+      "learning_rate": 0.0002712871287128713,
+      "loss": 0.9494,
+      "step": 9760
+    },
+    {
+      "epoch": 8.306912358459064,
+      "grad_norm": 17474.0859375,
+      "learning_rate": 0.0002702970297029703,
+      "loss": 0.9497,
+      "step": 9770
+    },
+    {
+      "epoch": 8.31541240399502,
+      "grad_norm": 13742.4052734375,
+      "learning_rate": 0.00026930693069306935,
+      "loss": 0.947,
+      "step": 9780
+    },
+    {
+      "epoch": 8.32391244953098,
+      "grad_norm": 15790.54296875,
+      "learning_rate": 0.00026831683168316835,
+      "loss": 0.9471,
+      "step": 9790
+    },
+    {
+      "epoch": 8.332412495066938,
+      "grad_norm": 16040.060546875,
+      "learning_rate": 0.00026732673267326735,
+      "loss": 0.9489,
+      "step": 9800
+    },
+    {
+      "epoch": 8.340912540602897,
+      "grad_norm": 14828.8095703125,
+      "learning_rate": 0.00026633663366336635,
+      "loss": 0.9482,
+      "step": 9810
+    },
+    {
+      "epoch": 8.340912540602897,
+      "eval_accuracy": 0.7962955020928217,
+      "eval_loss": 0.8608699440956116,
+      "eval_runtime": 1034.0211,
+      "eval_samples_per_second": 482.948,
+      "eval_steps_per_second": 0.822,
+      "step": 9810
+    },
+    {
+      "epoch": 8.349412586138854,
+      "grad_norm": 16481.876953125,
+      "learning_rate": 0.00026534653465346534,
+      "loss": 0.9478,
+      "step": 9820
+    },
+    {
+      "epoch": 8.357912631674813,
+      "grad_norm": 16586.9609375,
+      "learning_rate": 0.0002643564356435644,
+      "loss": 0.9491,
+      "step": 9830
+    },
+    {
+      "epoch": 8.366412677210771,
+      "grad_norm": 16209.767578125,
+      "learning_rate": 0.0002633663366336634,
+      "loss": 0.9484,
+      "step": 9840
+    },
+    {
+      "epoch": 8.374912722746728,
+      "grad_norm": 15675.52734375,
+      "learning_rate": 0.0002623762376237624,
+      "loss": 0.9483,
+      "step": 9850
+    },
+    {
+      "epoch": 8.383412768282687,
+      "grad_norm": 16780.919921875,
+      "learning_rate": 0.0002613861386138614,
+      "loss": 0.9466,
+      "step": 9860
+    },
+    {
+      "epoch": 8.391912813818646,
+      "grad_norm": 14785.990234375,
+      "learning_rate": 0.0002603960396039604,
+      "loss": 0.9459,
+      "step": 9870
+    },
+    {
+      "epoch": 8.400412859354603,
+      "grad_norm": 14482.4794921875,
+      "learning_rate": 0.00025940594059405944,
+      "loss": 0.9464,
+      "step": 9880
+    },
+    {
+      "epoch": 8.408912904890562,
+      "grad_norm": 17260.060546875,
+      "learning_rate": 0.00025841584158415844,
+      "loss": 0.947,
+      "step": 9890
+    },
+    {
+      "epoch": 8.41741295042652,
+      "grad_norm": 15805.9501953125,
+      "learning_rate": 0.00025742574257425744,
+      "loss": 0.9454,
+      "step": 9900
+    },
+    {
+      "epoch": 8.41741295042652,
+      "eval_accuracy": 0.7969292599781366,
+      "eval_loss": 0.8581969738006592,
+      "eval_runtime": 1037.2614,
+      "eval_samples_per_second": 481.439,
+      "eval_steps_per_second": 0.819,
+      "step": 9900
+    },
+    {
+      "epoch": 8.425912995962479,
+      "grad_norm": 19391.345703125,
+      "learning_rate": 0.00025643564356435644,
+      "loss": 0.9447,
+      "step": 9910
+    },
+    {
+      "epoch": 8.434413041498436,
+      "grad_norm": 15079.96875,
+      "learning_rate": 0.00025544554455445543,
+      "loss": 0.9472,
+      "step": 9920
+    },
+    {
+      "epoch": 8.442913087034395,
+      "grad_norm": 16290.3056640625,
+      "learning_rate": 0.0002544554455445545,
+      "loss": 0.9455,
+      "step": 9930
+    },
+    {
+      "epoch": 8.451413132570353,
+      "grad_norm": 14815.7783203125,
+      "learning_rate": 0.0002534653465346535,
+      "loss": 0.9453,
+      "step": 9940
+    },
+    {
+      "epoch": 8.459913178106312,
+      "grad_norm": 14300.2734375,
+      "learning_rate": 0.0002524752475247525,
+      "loss": 0.9448,
+      "step": 9950
+    },
+    {
+      "epoch": 8.46841322364227,
+      "grad_norm": 16484.212890625,
+      "learning_rate": 0.0002514851485148515,
+      "loss": 0.9471,
+      "step": 9960
+    },
+    {
+      "epoch": 8.476913269178228,
+      "grad_norm": 14792.8935546875,
+      "learning_rate": 0.0002504950495049505,
+      "loss": 0.9459,
+      "step": 9970
+    },
+    {
+      "epoch": 8.485413314714187,
+      "grad_norm": 15095.5869140625,
+      "learning_rate": 0.00024950495049504953,
+      "loss": 0.9439,
+      "step": 9980
+    },
+    {
+      "epoch": 8.493913360250144,
+      "grad_norm": 15246.3203125,
+      "learning_rate": 0.00024851485148514853,
+      "loss": 0.9459,
+      "step": 9990
+    },
+    {
+      "epoch": 8.493913360250144,
+      "eval_accuracy": 0.7969879777309228,
+      "eval_loss": 0.8577154278755188,
+      "eval_runtime": 1035.8486,
+      "eval_samples_per_second": 482.096,
+      "eval_steps_per_second": 0.821,
+      "step": 9990
+    },
+    {
+      "epoch": 8.502413405786102,
+      "grad_norm": 15397.66796875,
+      "learning_rate": 0.00024752475247524753,
+      "loss": 0.9443,
+      "step": 10000
+    },
+    {
+      "epoch": 8.510913451322061,
+      "grad_norm": 15283.72265625,
+      "learning_rate": 0.0002465346534653465,
+      "loss": 0.9446,
+      "step": 10010
+    },
+    {
+      "epoch": 8.519413496858018,
+      "grad_norm": 15016.5234375,
+      "learning_rate": 0.0002455445544554455,
+      "loss": 0.9446,
+      "step": 10020
+    },
+    {
+      "epoch": 8.527913542393977,
+      "grad_norm": 14347.6044921875,
+      "learning_rate": 0.0002445544554455446,
+      "loss": 0.945,
+      "step": 10030
+    },
+    {
+      "epoch": 8.536413587929935,
+      "grad_norm": 17782.130859375,
+      "learning_rate": 0.00024356435643564357,
+      "loss": 0.9445,
+      "step": 10040
+    },
+    {
+      "epoch": 8.544913633465894,
+      "grad_norm": 14884.66796875,
+      "learning_rate": 0.00024257425742574257,
+      "loss": 0.9429,
+      "step": 10050
+    },
+    {
+      "epoch": 8.553413679001851,
+      "grad_norm": 17597.05859375,
+      "learning_rate": 0.00024158415841584157,
+      "loss": 0.9443,
+      "step": 10060
+    },
+    {
+      "epoch": 8.56191372453781,
+      "grad_norm": 17000.009765625,
+      "learning_rate": 0.0002405940594059406,
+      "loss": 0.9449,
+      "step": 10070
+    },
+    {
+      "epoch": 8.570413770073769,
+      "grad_norm": 17303.015625,
+      "learning_rate": 0.0002396039603960396,
+      "loss": 0.9444,
+      "step": 10080
+    },
+    {
+      "epoch": 8.570413770073769,
+      "eval_accuracy": 0.7970984014477712,
+      "eval_loss": 0.8577408194541931,
+      "eval_runtime": 1033.5768,
+      "eval_samples_per_second": 483.155,
+      "eval_steps_per_second": 0.822,
+      "step": 10080
+    },
+    {
+      "epoch": 8.578913815609727,
+      "grad_norm": 14883.8759765625,
+      "learning_rate": 0.00023861386138613862,
+      "loss": 0.9429,
+      "step": 10090
+    },
+    {
+      "epoch": 8.587413861145684,
+      "grad_norm": 16271.083984375,
+      "learning_rate": 0.00023762376237623762,
+      "loss": 0.9441,
+      "step": 10100
+    },
+    {
+      "epoch": 8.595913906681643,
+      "grad_norm": 18421.546875,
+      "learning_rate": 0.00023663366336633662,
+      "loss": 0.9444,
+      "step": 10110
+    },
+    {
+      "epoch": 8.604413952217602,
+      "grad_norm": 19686.576171875,
+      "learning_rate": 0.00023564356435643564,
+      "loss": 0.9434,
+      "step": 10120
+    },
+    {
+      "epoch": 8.612913997753559,
+      "grad_norm": 14919.0126953125,
+      "learning_rate": 0.00023465346534653464,
+      "loss": 0.9427,
+      "step": 10130
+    },
+    {
+      "epoch": 8.621414043289517,
+      "grad_norm": 16443.49609375,
+      "learning_rate": 0.0002336633663366337,
+      "loss": 0.9412,
+      "step": 10140
+    },
+    {
+      "epoch": 8.629914088825476,
+      "grad_norm": 14300.4599609375,
+      "learning_rate": 0.0002326732673267327,
+      "loss": 0.9436,
+      "step": 10150
+    },
+    {
+      "epoch": 8.638414134361435,
+      "grad_norm": 17048.576171875,
+      "learning_rate": 0.0002316831683168317,
+      "loss": 0.9438,
+      "step": 10160
+    },
+    {
+      "epoch": 8.646914179897392,
+      "grad_norm": 14095.205078125,
+      "learning_rate": 0.00023069306930693071,
+      "loss": 0.9419,
+      "step": 10170
+    },
+    {
+      "epoch": 8.646914179897392,
+      "eval_accuracy": 0.797700358706027,
+      "eval_loss": 0.854520857334137,
+      "eval_runtime": 1038.4284,
+      "eval_samples_per_second": 480.898,
+      "eval_steps_per_second": 0.819,
+      "step": 10170
+    },
+    {
+      "epoch": 8.65541422543335,
+      "grad_norm": 16067.9052734375,
+      "learning_rate": 0.0002297029702970297,
+      "loss": 0.9409,
+      "step": 10180
+    },
+    {
+      "epoch": 8.66391427096931,
+      "grad_norm": 14993.923828125,
+      "learning_rate": 0.00022871287128712874,
+      "loss": 0.9421,
+      "step": 10190
+    },
+    {
+      "epoch": 8.672414316505266,
+      "grad_norm": 14386.7099609375,
+      "learning_rate": 0.00022772277227722774,
+      "loss": 0.9428,
+      "step": 10200
+    },
+    {
+      "epoch": 8.680914362041225,
+      "grad_norm": 15530.26953125,
+      "learning_rate": 0.00022673267326732673,
+      "loss": 0.9425,
+      "step": 10210
+    },
+    {
+      "epoch": 8.689414407577184,
+      "grad_norm": 15528.4541015625,
+      "learning_rate": 0.00022574257425742576,
+      "loss": 0.941,
+      "step": 10220
+    },
+    {
+      "epoch": 8.697914453113142,
+      "grad_norm": 14304.1318359375,
+      "learning_rate": 0.00022475247524752476,
+      "loss": 0.9402,
+      "step": 10230
+    },
+    {
+      "epoch": 8.7064144986491,
+      "grad_norm": 15193.4375,
+      "learning_rate": 0.00022376237623762378,
+      "loss": 0.9415,
+      "step": 10240
+    },
+    {
+      "epoch": 8.714914544185058,
+      "grad_norm": 14937.7109375,
+      "learning_rate": 0.00022277227722772278,
+      "loss": 0.9421,
+      "step": 10250
+    },
+    {
+      "epoch": 8.723414589721017,
+      "grad_norm": 15369.6669921875,
+      "learning_rate": 0.00022178217821782178,
+      "loss": 0.9415,
+      "step": 10260
+    },
+    {
+      "epoch": 8.723414589721017,
+      "eval_accuracy": 0.797684357783119,
+      "eval_loss": 0.8541524410247803,
+      "eval_runtime": 1037.1311,
+      "eval_samples_per_second": 481.499,
+      "eval_steps_per_second": 0.82,
+      "step": 10260
+    },
+    {
+      "epoch": 8.731914635256974,
+      "grad_norm": 17404.70703125,
+      "learning_rate": 0.0002207920792079208,
+      "loss": 0.9417,
+      "step": 10270
+    },
+    {
+      "epoch": 8.740414680792933,
+      "grad_norm": 16622.16796875,
+      "learning_rate": 0.0002198019801980198,
+      "loss": 0.9409,
+      "step": 10280
+    },
+    {
+      "epoch": 8.748914726328891,
+      "grad_norm": 13510.8125,
+      "learning_rate": 0.00021881188118811883,
+      "loss": 0.9409,
+      "step": 10290
+    },
+    {
+      "epoch": 8.75741477186485,
+      "grad_norm": 14584.3173828125,
+      "learning_rate": 0.00021782178217821783,
+      "loss": 0.9411,
+      "step": 10300
+    },
+    {
+      "epoch": 8.765914817400807,
+      "grad_norm": 13928.373046875,
+      "learning_rate": 0.00021683168316831682,
+      "loss": 0.9408,
+      "step": 10310
+    },
+    {
+      "epoch": 8.774414862936766,
+      "grad_norm": 13927.751953125,
+      "learning_rate": 0.00021584158415841585,
+      "loss": 0.9404,
+      "step": 10320
+    },
+    {
+      "epoch": 8.782914908472724,
+      "grad_norm": 15996.322265625,
+      "learning_rate": 0.00021485148514851485,
+      "loss": 0.9415,
+      "step": 10330
+    },
+    {
+      "epoch": 8.791414954008681,
+      "grad_norm": 14078.8876953125,
+      "learning_rate": 0.00021386138613861387,
+      "loss": 0.9392,
+      "step": 10340
+    },
+    {
+      "epoch": 8.79991499954464,
+      "grad_norm": 14735.3623046875,
+      "learning_rate": 0.00021287128712871287,
+      "loss": 0.9395,
+      "step": 10350
+    },
+    {
+      "epoch": 8.79991499954464,
+      "eval_accuracy": 0.7978062402609887,
+      "eval_loss": 0.8535209894180298,
+      "eval_runtime": 1036.7203,
+      "eval_samples_per_second": 481.69,
+      "eval_steps_per_second": 0.82,
+      "step": 10350
+    },
+    {
+      "epoch": 8.808415045080599,
+      "grad_norm": 14420.7861328125,
+      "learning_rate": 0.00021188118811881187,
+      "loss": 0.9412,
+      "step": 10360
+    },
+    {
+      "epoch": 8.816915090616558,
+      "grad_norm": 14367.3955078125,
+      "learning_rate": 0.0002108910891089109,
+      "loss": 0.9382,
+      "step": 10370
+    },
+    {
+      "epoch": 8.825415136152515,
+      "grad_norm": 14271.1162109375,
+      "learning_rate": 0.0002099009900990099,
+      "loss": 0.9406,
+      "step": 10380
+    },
+    {
+      "epoch": 8.833915181688473,
+      "grad_norm": 14588.7724609375,
+      "learning_rate": 0.00020891089108910892,
+      "loss": 0.9397,
+      "step": 10390
+    },
+    {
+      "epoch": 8.842415227224432,
+      "grad_norm": 13847.20703125,
+      "learning_rate": 0.00020792079207920792,
+      "loss": 0.9394,
+      "step": 10400
+    },
+    {
+      "epoch": 8.850915272760389,
+      "grad_norm": 14702.3779296875,
+      "learning_rate": 0.00020693069306930691,
+      "loss": 0.9386,
+      "step": 10410
+    },
+    {
+      "epoch": 8.859415318296348,
+      "grad_norm": 13664.92578125,
+      "learning_rate": 0.00020594059405940594,
+      "loss": 0.939,
+      "step": 10420
+    },
+    {
+      "epoch": 8.867915363832306,
+      "grad_norm": 14917.166015625,
+      "learning_rate": 0.00020495049504950494,
+      "loss": 0.9392,
+      "step": 10430
+    },
+    {
+      "epoch": 8.876415409368265,
+      "grad_norm": 15233.123046875,
+      "learning_rate": 0.00020396039603960396,
+      "loss": 0.9411,
+      "step": 10440
+    },
+    {
+      "epoch": 8.876415409368265,
+      "eval_accuracy": 0.7982283642493349,
+      "eval_loss": 0.8512039184570312,
+      "eval_runtime": 1037.2463,
+      "eval_samples_per_second": 481.446,
+      "eval_steps_per_second": 0.819,
+      "step": 10440
+    },
+    {
+      "epoch": 8.884915454904222,
+      "grad_norm": 14748.51953125,
+      "learning_rate": 0.000202970297029703,
+      "loss": 0.9379,
+      "step": 10450
+    },
+    {
+      "epoch": 8.893415500440181,
+      "grad_norm": 14725.625,
+      "learning_rate": 0.00020198019801980199,
+      "loss": 0.9394,
+      "step": 10460
+    },
+    {
+      "epoch": 8.90191554597614,
+      "grad_norm": 14719.560546875,
+      "learning_rate": 0.000200990099009901,
+      "loss": 0.9396,
+      "step": 10470
+    },
+    {
+      "epoch": 8.910415591512097,
+      "grad_norm": 15252.0546875,
+      "learning_rate": 0.0002,
+      "loss": 0.9386,
+      "step": 10480
+    },
+    {
+      "epoch": 8.918915637048055,
+      "grad_norm": 14756.244140625,
+      "learning_rate": 0.00019900990099009903,
+      "loss": 0.9378,
+      "step": 10490
+    },
+    {
+      "epoch": 8.927415682584014,
+      "grad_norm": 14705.61328125,
+      "learning_rate": 0.00019801980198019803,
+      "loss": 0.9377,
+      "step": 10500
+    },
+    {
+      "epoch": 8.935915728119973,
+      "grad_norm": 14071.0009765625,
+      "learning_rate": 0.00019702970297029703,
+      "loss": 0.9394,
+      "step": 10510
+    },
+    {
+      "epoch": 8.94441577365593,
+      "grad_norm": 13337.140625,
+      "learning_rate": 0.00019603960396039606,
+      "loss": 0.9384,
+      "step": 10520
+    },
+    {
+      "epoch": 8.952915819191889,
+      "grad_norm": 14109.1650390625,
+      "learning_rate": 0.00019504950495049505,
+      "loss": 0.9362,
+      "step": 10530
+    },
+    {
+      "epoch": 8.952915819191889,
+      "eval_accuracy": 0.7982979521046931,
+      "eval_loss": 0.8517733812332153,
+      "eval_runtime": 1038.3162,
+      "eval_samples_per_second": 480.95,
+      "eval_steps_per_second": 0.819,
+      "step": 10530
+    },
+    {
+      "epoch": 8.961415864727847,
+      "grad_norm": 17169.69140625,
+      "learning_rate": 0.00019405940594059408,
+      "loss": 0.9373,
+      "step": 10540
+    },
+    {
+      "epoch": 8.969915910263804,
+      "grad_norm": 14149.8935546875,
+      "learning_rate": 0.00019306930693069308,
+      "loss": 0.9365,
+      "step": 10550
+    },
+    {
+      "epoch": 8.978415955799763,
+      "grad_norm": 16067.255859375,
+      "learning_rate": 0.00019207920792079208,
+      "loss": 0.9371,
+      "step": 10560
+    },
+    {
+      "epoch": 8.986916001335722,
+      "grad_norm": 14375.869140625,
+      "learning_rate": 0.0001910891089108911,
+      "loss": 0.9375,
+      "step": 10570
+    },
+    {
+      "epoch": 8.99541604687168,
+      "grad_norm": 14484.7421875,
+      "learning_rate": 0.0001900990099009901,
+      "loss": 0.9372,
+      "step": 10580
+    },
+    {
+      "epoch": 9.003916092407637,
+      "grad_norm": 13759.2275390625,
+      "learning_rate": 0.00018910891089108913,
+      "loss": 0.9361,
+      "step": 10590
+    },
+    {
+      "epoch": 9.012416137943596,
+      "grad_norm": 14114.7119140625,
+      "learning_rate": 0.00018811881188118812,
+      "loss": 0.9369,
+      "step": 10600
+    },
+    {
+      "epoch": 9.020916183479555,
+      "grad_norm": 13548.9287109375,
+      "learning_rate": 0.00018712871287128712,
+      "loss": 0.9363,
+      "step": 10610
+    },
+    {
+      "epoch": 9.029416229015512,
+      "grad_norm": 13845.412109375,
+      "learning_rate": 0.00018613861386138615,
+      "loss": 0.9369,
+      "step": 10620
+    },
+    {
+      "epoch": 9.029416229015512,
+      "eval_accuracy": 0.7986381309887046,
+      "eval_loss": 0.8504059314727783,
+      "eval_runtime": 1037.8808,
+      "eval_samples_per_second": 481.152,
+      "eval_steps_per_second": 0.819,
+      "step": 10620
+    },
+    {
+      "epoch": 9.03791627455147,
+      "grad_norm": 14264.1455078125,
+      "learning_rate": 0.00018514851485148514,
+      "loss": 0.9371,
+      "step": 10630
+    },
+    {
+      "epoch": 9.04641632008743,
+      "grad_norm": 12720.0791015625,
+      "learning_rate": 0.00018415841584158417,
+      "loss": 0.9351,
+      "step": 10640
+    },
+    {
+      "epoch": 9.054916365623388,
+      "grad_norm": 14962.2216796875,
+      "learning_rate": 0.00018316831683168317,
+      "loss": 0.9347,
+      "step": 10650
+    },
+    {
+      "epoch": 9.063416411159345,
+      "grad_norm": 14343.6259765625,
+      "learning_rate": 0.00018217821782178217,
+      "loss": 0.9366,
+      "step": 10660
+    },
+    {
+      "epoch": 9.071916456695304,
+      "grad_norm": 15166.927734375,
+      "learning_rate": 0.0001811881188118812,
+      "loss": 0.9379,
+      "step": 10670
+    },
+    {
+      "epoch": 9.080416502231262,
+      "grad_norm": 13523.9873046875,
+      "learning_rate": 0.0001801980198019802,
+      "loss": 0.9369,
+      "step": 10680
+    },
+    {
+      "epoch": 9.08891654776722,
+      "grad_norm": 15057.4638671875,
+      "learning_rate": 0.00017920792079207922,
+      "loss": 0.9368,
+      "step": 10690
+    },
+    {
+      "epoch": 9.097416593303178,
+      "grad_norm": 13766.353515625,
+      "learning_rate": 0.0001782178217821782,
+      "loss": 0.935,
+      "step": 10700
+    },
+    {
+      "epoch": 9.105916638839137,
+      "grad_norm": 15783.3056640625,
+      "learning_rate": 0.0001772277227722772,
+      "loss": 0.9365,
+      "step": 10710
+    },
+    {
+      "epoch": 9.105916638839137,
+      "eval_accuracy": 0.7986280323340061,
+      "eval_loss": 0.8495949506759644,
+      "eval_runtime": 1036.1574,
+      "eval_samples_per_second": 481.952,
+      "eval_steps_per_second": 0.82,
+      "step": 10710
+    },
+    {
+      "epoch": 9.114416684375096,
+      "grad_norm": 14088.314453125,
+      "learning_rate": 0.00017623762376237624,
+      "loss": 0.9363,
+      "step": 10720
+    },
+    {
+      "epoch": 9.122916729911053,
+      "grad_norm": 14255.0244140625,
+      "learning_rate": 0.00017524752475247524,
+      "loss": 0.9351,
+      "step": 10730
+    },
+    {
+      "epoch": 9.131416775447011,
+      "grad_norm": 15012.841796875,
+      "learning_rate": 0.00017425742574257426,
+      "loss": 0.9341,
+      "step": 10740
+    },
+    {
+      "epoch": 9.13991682098297,
+      "grad_norm": 15262.3740234375,
+      "learning_rate": 0.00017326732673267326,
+      "loss": 0.9342,
+      "step": 10750
+    },
+    {
+      "epoch": 9.148416866518927,
+      "grad_norm": 14390.744140625,
+      "learning_rate": 0.00017227722772277226,
+      "loss": 0.9359,
+      "step": 10760
+    },
+    {
+      "epoch": 9.156916912054886,
+      "grad_norm": 14124.2373046875,
+      "learning_rate": 0.0001712871287128713,
+      "loss": 0.9353,
+      "step": 10770
+    },
+    {
+      "epoch": 9.165416957590844,
+      "grad_norm": 13361.82421875,
+      "learning_rate": 0.0001702970297029703,
+      "loss": 0.9352,
+      "step": 10780
+    },
+    {
+      "epoch": 9.173917003126803,
+      "grad_norm": 13731.3984375,
+      "learning_rate": 0.00016930693069306933,
+      "loss": 0.9345,
+      "step": 10790
+    },
+    {
+      "epoch": 9.18241704866276,
+      "grad_norm": 13207.9716796875,
+      "learning_rate": 0.00016831683168316833,
+      "loss": 0.9336,
+      "step": 10800
+    },
+    {
+      "epoch": 9.18241704866276,
+      "eval_accuracy": 0.7991505659503615,
+      "eval_loss": 0.8472453355789185,
+      "eval_runtime": 1035.1715,
+      "eval_samples_per_second": 482.411,
+      "eval_steps_per_second": 0.821,
+      "step": 10800
+    },
+    {
+      "epoch": 9.190917094198719,
+      "grad_norm": 14880.1533203125,
+      "learning_rate": 0.00016732673267326733,
+      "loss": 0.9333,
+      "step": 10810
+    },
+    {
+      "epoch": 9.199417139734678,
+      "grad_norm": 12751.9345703125,
+      "learning_rate": 0.00016633663366336635,
+      "loss": 0.935,
+      "step": 10820
+    },
+    {
+      "epoch": 9.207917185270635,
+      "grad_norm": 13587.4658203125,
+      "learning_rate": 0.00016534653465346535,
+      "loss": 0.9336,
+      "step": 10830
+    },
+    {
+      "epoch": 9.216417230806593,
+      "grad_norm": 13916.77734375,
+      "learning_rate": 0.00016435643564356438,
+      "loss": 0.9348,
+      "step": 10840
+    },
+    {
+      "epoch": 9.224917276342552,
+      "grad_norm": 14346.2119140625,
+      "learning_rate": 0.00016336633663366338,
+      "loss": 0.9328,
+      "step": 10850
+    },
+    {
+      "epoch": 9.23341732187851,
+      "grad_norm": 11983.17578125,
+      "learning_rate": 0.00016237623762376237,
+      "loss": 0.9341,
+      "step": 10860
+    },
+    {
+      "epoch": 9.241917367414468,
+      "grad_norm": 13424.4794921875,
+      "learning_rate": 0.0001613861386138614,
+      "loss": 0.9341,
+      "step": 10870
+    },
+    {
+      "epoch": 9.250417412950426,
+      "grad_norm": 13093.9267578125,
+      "learning_rate": 0.0001603960396039604,
+      "loss": 0.9336,
+      "step": 10880
+    },
+    {
+      "epoch": 9.258917458486385,
+      "grad_norm": 14450.0888671875,
+      "learning_rate": 0.00015940594059405942,
+      "loss": 0.9326,
+      "step": 10890
+    },
+    {
+      "epoch": 9.258917458486385,
+      "eval_accuracy": 0.7993350474335167,
+      "eval_loss": 0.846928596496582,
+      "eval_runtime": 1034.8058,
+      "eval_samples_per_second": 482.581,
+      "eval_steps_per_second": 0.821,
+      "step": 10890
+    },
+    {
+      "epoch": 9.267417504022344,
+      "grad_norm": 12639.7099609375,
+      "learning_rate": 0.00015841584158415842,
+      "loss": 0.9331,
+      "step": 10900
+    },
+    {
+      "epoch": 9.2759175495583,
+      "grad_norm": 14105.6357421875,
+      "learning_rate": 0.00015742574257425742,
+      "loss": 0.9323,
+      "step": 10910
+    },
+    {
+      "epoch": 9.28441759509426,
+      "grad_norm": 13878.3876953125,
+      "learning_rate": 0.00015643564356435644,
+      "loss": 0.9319,
+      "step": 10920
+    },
+    {
+      "epoch": 9.292917640630218,
+      "grad_norm": 14202.0380859375,
+      "learning_rate": 0.00015544554455445544,
+      "loss": 0.9309,
+      "step": 10930
+    },
+    {
+      "epoch": 9.301417686166175,
+      "grad_norm": 12203.2275390625,
+      "learning_rate": 0.00015445544554455447,
+      "loss": 0.9323,
+      "step": 10940
+    },
+    {
+      "epoch": 9.309917731702134,
+      "grad_norm": 12799.16015625,
+      "learning_rate": 0.00015346534653465347,
+      "loss": 0.933,
+      "step": 10950
+    },
+    {
+      "epoch": 9.318417777238093,
+      "grad_norm": 14222.4921875,
+      "learning_rate": 0.00015247524752475246,
+      "loss": 0.9335,
+      "step": 10960
+    },
+    {
+      "epoch": 9.32691782277405,
+      "grad_norm": 13150.4560546875,
+      "learning_rate": 0.0001514851485148515,
+      "loss": 0.9334,
+      "step": 10970
+    },
+    {
+      "epoch": 9.335417868310008,
+      "grad_norm": 13035.662109375,
+      "learning_rate": 0.0001504950495049505,
+      "loss": 0.9336,
+      "step": 10980
+    },
+    {
+      "epoch": 9.335417868310008,
+      "eval_accuracy": 0.7995125837200675,
+      "eval_loss": 0.8457638621330261,
+      "eval_runtime": 1036.0457,
+      "eval_samples_per_second": 482.004,
+      "eval_steps_per_second": 0.82,
+      "step": 10980
+    },
+    {
+      "epoch": 9.343917913845967,
+      "grad_norm": 13063.5947265625,
+      "learning_rate": 0.0001495049504950495,
+      "loss": 0.9358,
+      "step": 10990
+    },
+    {
+      "epoch": 9.352417959381926,
+      "grad_norm": 13189.0205078125,
+      "learning_rate": 0.0001485148514851485,
+      "loss": 0.9314,
+      "step": 11000
+    },
+    {
+      "epoch": 9.360918004917883,
+      "grad_norm": 14116.0537109375,
+      "learning_rate": 0.0001475247524752475,
+      "loss": 0.9322,
+      "step": 11010
+    },
+    {
+      "epoch": 9.369418050453842,
+      "grad_norm": 13098.5810546875,
+      "learning_rate": 0.00014653465346534653,
+      "loss": 0.9306,
+      "step": 11020
+    },
+    {
+      "epoch": 9.3779180959898,
+      "grad_norm": 12630.921875,
+      "learning_rate": 0.00014554455445544553,
+      "loss": 0.9294,
+      "step": 11030
+    },
+    {
+      "epoch": 9.386418141525759,
+      "grad_norm": 12936.2509765625,
+      "learning_rate": 0.00014455445544554456,
+      "loss": 0.9323,
+      "step": 11040
+    },
+    {
+      "epoch": 9.394918187061716,
+      "grad_norm": 13526.9853515625,
+      "learning_rate": 0.00014356435643564356,
+      "loss": 0.9311,
+      "step": 11050
+    },
+    {
+      "epoch": 9.403418232597675,
+      "grad_norm": 13259.3603515625,
+      "learning_rate": 0.00014257425742574255,
+      "loss": 0.9332,
+      "step": 11060
+    },
+    {
+      "epoch": 9.411918278133633,
+      "grad_norm": 14287.5361328125,
+      "learning_rate": 0.00014158415841584158,
+      "loss": 0.9321,
+      "step": 11070
+    },
+    {
+      "epoch": 9.411918278133633,
+      "eval_accuracy": 0.7996804125615878,
+      "eval_loss": 0.8447943329811096,
+      "eval_runtime": 1035.7879,
+      "eval_samples_per_second": 482.124,
+      "eval_steps_per_second": 0.821,
+      "step": 11070
+    },
+    {
+      "epoch": 9.42041832366959,
+      "grad_norm": 12443.166015625,
+      "learning_rate": 0.0001405940594059406,
+      "loss": 0.9303,
+      "step": 11080
+    },
+    {
+      "epoch": 9.42891836920555,
+      "grad_norm": 13510.2119140625,
+      "learning_rate": 0.00013960396039603963,
+      "loss": 0.9298,
+      "step": 11090
+    },
+    {
+      "epoch": 9.437418414741508,
+      "grad_norm": 11732.4970703125,
+      "learning_rate": 0.00013861386138613863,
+      "loss": 0.9314,
+      "step": 11100
+    },
+    {
+      "epoch": 9.445918460277467,
+      "grad_norm": 13648.421875,
+      "learning_rate": 0.00013762376237623763,
+      "loss": 0.9318,
+      "step": 11110
+    },
+    {
+      "epoch": 9.454418505813424,
+      "grad_norm": 12602.56640625,
+      "learning_rate": 0.00013663366336633665,
+      "loss": 0.9318,
+      "step": 11120
+    },
+    {
+      "epoch": 9.462918551349382,
+      "grad_norm": 13816.3720703125,
+      "learning_rate": 0.00013564356435643565,
+      "loss": 0.9319,
+      "step": 11130
+    },
+    {
+      "epoch": 9.471418596885341,
+      "grad_norm": 14125.748046875,
+      "learning_rate": 0.00013465346534653468,
+      "loss": 0.9297,
+      "step": 11140
+    },
+    {
+      "epoch": 9.479918642421298,
+      "grad_norm": 13495.037109375,
+      "learning_rate": 0.00013366336633663367,
+      "loss": 0.9314,
+      "step": 11150
+    },
+    {
+      "epoch": 9.488418687957257,
+      "grad_norm": 12772.404296875,
+      "learning_rate": 0.00013267326732673267,
+      "loss": 0.9283,
+      "step": 11160
+    },
+    {
+      "epoch": 9.488418687957257,
+      "eval_accuracy": 0.799732345307766,
+      "eval_loss": 0.8448570370674133,
+      "eval_runtime": 1038.7893,
+      "eval_samples_per_second": 480.731,
+      "eval_steps_per_second": 0.818,
+      "step": 11160
+    },
+    {
+      "epoch": 9.496918733493215,
+      "grad_norm": 13424.62890625,
+      "learning_rate": 0.0001316831683168317,
+      "loss": 0.9297,
+      "step": 11170
+    },
+    {
+      "epoch": 9.505418779029174,
+      "grad_norm": 14624.0283203125,
+      "learning_rate": 0.0001306930693069307,
+      "loss": 0.9295,
+      "step": 11180
+    },
+    {
+      "epoch": 9.513918824565131,
+      "grad_norm": 12779.9169921875,
+      "learning_rate": 0.00012970297029702972,
+      "loss": 0.9322,
+      "step": 11190
+    },
+    {
+      "epoch": 9.52241887010109,
+      "grad_norm": 13789.7421875,
+      "learning_rate": 0.00012871287128712872,
+      "loss": 0.9312,
+      "step": 11200
+    },
+    {
+      "epoch": 9.530918915637049,
+      "grad_norm": 12679.0703125,
+      "learning_rate": 0.00012772277227722772,
+      "loss": 0.9289,
+      "step": 11210
+    },
+    {
+      "epoch": 9.539418961173006,
+      "grad_norm": 14479.2919921875,
+      "learning_rate": 0.00012673267326732674,
+      "loss": 0.931,
+      "step": 11220
+    },
+    {
+      "epoch": 9.547919006708964,
+      "grad_norm": 12816.5322265625,
+      "learning_rate": 0.00012574257425742574,
+      "loss": 0.9304,
+      "step": 11230
+    },
+    {
+      "epoch": 9.556419052244923,
+      "grad_norm": 13485.4990234375,
+      "learning_rate": 0.00012475247524752477,
+      "loss": 0.928,
+      "step": 11240
+    },
+    {
+      "epoch": 9.564919097780882,
+      "grad_norm": 13575.083984375,
+      "learning_rate": 0.00012376237623762376,
+      "loss": 0.929,
+      "step": 11250
+    },
+    {
+      "epoch": 9.564919097780882,
+      "eval_accuracy": 0.7999083549113648,
+      "eval_loss": 0.8442253470420837,
+      "eval_runtime": 1035.3058,
+      "eval_samples_per_second": 482.348,
+      "eval_steps_per_second": 0.821,
+      "step": 11250
+    },
+    {
+      "epoch": 9.573419143316839,
+      "grad_norm": 13030.95703125,
+      "learning_rate": 0.00012277227722772276,
+      "loss": 0.9303,
+      "step": 11260
+    },
+    {
+      "epoch": 9.581919188852797,
+      "grad_norm": 12883.271484375,
+      "learning_rate": 0.00012178217821782179,
+      "loss": 0.9304,
+      "step": 11270
+    },
+    {
+      "epoch": 9.590419234388756,
+      "grad_norm": 12350.5322265625,
+      "learning_rate": 0.00012079207920792079,
+      "loss": 0.9298,
+      "step": 11280
+    },
+    {
+      "epoch": 9.598919279924713,
+      "grad_norm": 12253.1220703125,
+      "learning_rate": 0.0001198019801980198,
+      "loss": 0.9301,
+      "step": 11290
+    },
+    {
+      "epoch": 9.607419325460672,
+      "grad_norm": 13372.912109375,
+      "learning_rate": 0.00011881188118811881,
+      "loss": 0.9298,
+      "step": 11300
+    },
+    {
+      "epoch": 9.61591937099663,
+      "grad_norm": 12150.8623046875,
+      "learning_rate": 0.00011782178217821782,
+      "loss": 0.9276,
+      "step": 11310
+    },
+    {
+      "epoch": 9.62441941653259,
+      "grad_norm": 12663.35546875,
+      "learning_rate": 0.00011683168316831685,
+      "loss": 0.9305,
+      "step": 11320
+    },
+    {
+      "epoch": 9.632919462068546,
+      "grad_norm": 12773.25,
+      "learning_rate": 0.00011584158415841584,
+      "loss": 0.929,
+      "step": 11330
+    },
+    {
+      "epoch": 9.641419507604505,
+      "grad_norm": 12251.90234375,
+      "learning_rate": 0.00011485148514851486,
+      "loss": 0.9282,
+      "step": 11340
+    },
+    {
+      "epoch": 9.641419507604505,
+      "eval_accuracy": 0.8002882889018181,
+      "eval_loss": 0.842271089553833,
+      "eval_runtime": 1038.9797,
+      "eval_samples_per_second": 480.643,
+      "eval_steps_per_second": 0.818,
+      "step": 11340
+    },
+    {
+      "epoch": 9.649919553140464,
+      "grad_norm": 12380.44140625,
+      "learning_rate": 0.00011386138613861387,
+      "loss": 0.9275,
+      "step": 11350
+    },
+    {
+      "epoch": 9.65841959867642,
+      "grad_norm": 12090.271484375,
+      "learning_rate": 0.00011287128712871288,
+      "loss": 0.9275,
+      "step": 11360
+    },
+    {
+      "epoch": 9.66691964421238,
+      "grad_norm": 13393.486328125,
+      "learning_rate": 0.00011188118811881189,
+      "loss": 0.9293,
+      "step": 11370
+    },
+    {
+      "epoch": 9.675419689748338,
+      "grad_norm": 11574.048828125,
+      "learning_rate": 0.00011089108910891089,
+      "loss": 0.9269,
+      "step": 11380
+    },
+    {
+      "epoch": 9.683919735284297,
+      "grad_norm": 13117.8916015625,
+      "learning_rate": 0.0001099009900990099,
+      "loss": 0.928,
+      "step": 11390
+    },
+    {
+      "epoch": 9.692419780820254,
+      "grad_norm": 13132.0810546875,
+      "learning_rate": 0.00010891089108910891,
+      "loss": 0.9272,
+      "step": 11400
+    },
+    {
+      "epoch": 9.700919826356213,
+      "grad_norm": 13526.349609375,
+      "learning_rate": 0.00010792079207920792,
+      "loss": 0.9271,
+      "step": 11410
+    },
+    {
+      "epoch": 9.709419871892171,
+      "grad_norm": 13662.8740234375,
+      "learning_rate": 0.00010693069306930694,
+      "loss": 0.9261,
+      "step": 11420
+    },
+    {
+      "epoch": 9.717919917428128,
+      "grad_norm": 12285.236328125,
+      "learning_rate": 0.00010594059405940593,
+      "loss": 0.9271,
+      "step": 11430
+    },
+    {
+      "epoch": 9.717919917428128,
+      "eval_accuracy": 0.8004033551877789,
+      "eval_loss": 0.8413528800010681,
+      "eval_runtime": 1034.3691,
+      "eval_samples_per_second": 482.785,
+      "eval_steps_per_second": 0.822,
+      "step": 11430
+    },
+    {
+      "epoch": 9.726419962964087,
+      "grad_norm": 12984.609375,
+      "learning_rate": 0.00010495049504950495,
+      "loss": 0.9272,
+      "step": 11440
+    },
+    {
+      "epoch": 9.734920008500046,
+      "grad_norm": 13035.4326171875,
+      "learning_rate": 0.00010396039603960396,
+      "loss": 0.927,
+      "step": 11450
+    },
+    {
+      "epoch": 9.743420054036005,
+      "grad_norm": 12569.7998046875,
+      "learning_rate": 0.00010297029702970297,
+      "loss": 0.9276,
+      "step": 11460
+    },
+    {
+      "epoch": 9.751920099571961,
+      "grad_norm": 13077.8876953125,
+      "learning_rate": 0.00010198019801980198,
+      "loss": 0.9271,
+      "step": 11470
+    },
+    {
+      "epoch": 9.76042014510792,
+      "grad_norm": 12659.830078125,
+      "learning_rate": 0.00010099009900990099,
+      "loss": 0.9276,
+      "step": 11480
+    },
+    {
+      "epoch": 9.768920190643879,
+      "grad_norm": 13212.763671875,
+      "learning_rate": 0.0001,
+      "loss": 0.9268,
+      "step": 11490
+    },
+    {
+      "epoch": 9.777420236179836,
+      "grad_norm": 12120.390625,
+      "learning_rate": 9.900990099009902e-05,
+      "loss": 0.9267,
+      "step": 11500
+    },
+    {
+      "epoch": 9.785920281715795,
+      "grad_norm": 12504.2646484375,
+      "learning_rate": 9.801980198019803e-05,
+      "loss": 0.9281,
+      "step": 11510
+    },
+    {
+      "epoch": 9.794420327251753,
+      "grad_norm": 12107.5166015625,
+      "learning_rate": 9.702970297029704e-05,
+      "loss": 0.9264,
+      "step": 11520
+    },
+    {
+      "epoch": 9.794420327251753,
+      "eval_accuracy": 0.800677157520577,
+      "eval_loss": 0.840539276599884,
+      "eval_runtime": 1035.1246,
+      "eval_samples_per_second": 482.433,
+      "eval_steps_per_second": 0.821,
+      "step": 11520
+    },
+    {
+      "epoch": 9.802920372787712,
+      "grad_norm": 13302.0888671875,
+      "learning_rate": 9.603960396039604e-05,
+      "loss": 0.9249,
+      "step": 11530
+    },
+    {
+      "epoch": 9.811420418323669,
+      "grad_norm": 11464.931640625,
+      "learning_rate": 9.504950495049505e-05,
+      "loss": 0.928,
+      "step": 11540
+    },
+    {
+      "epoch": 9.819920463859628,
+      "grad_norm": 11620.611328125,
+      "learning_rate": 9.405940594059406e-05,
+      "loss": 0.9269,
+      "step": 11550
+    },
+    {
+      "epoch": 9.828420509395587,
+      "grad_norm": 12684.673828125,
+      "learning_rate": 9.306930693069307e-05,
+      "loss": 0.9264,
+      "step": 11560
+    },
+    {
+      "epoch": 9.836920554931543,
+      "grad_norm": 12113.302734375,
+      "learning_rate": 9.207920792079209e-05,
+      "loss": 0.9239,
+      "step": 11570
+    },
+    {
+      "epoch": 9.845420600467502,
+      "grad_norm": 12095.48828125,
+      "learning_rate": 9.108910891089108e-05,
+      "loss": 0.9262,
+      "step": 11580
+    },
+    {
+      "epoch": 9.853920646003461,
+      "grad_norm": 12388.09765625,
+      "learning_rate": 9.00990099009901e-05,
+      "loss": 0.9241,
+      "step": 11590
+    },
+    {
+      "epoch": 9.86242069153942,
+      "grad_norm": 12502.1171875,
+      "learning_rate": 8.91089108910891e-05,
+      "loss": 0.9256,
+      "step": 11600
+    },
+    {
+      "epoch": 9.870920737075377,
+      "grad_norm": 12518.240234375,
+      "learning_rate": 8.811881188118812e-05,
+      "loss": 0.9268,
+      "step": 11610
+    },
+    {
+      "epoch": 9.870920737075377,
+      "eval_accuracy": 0.800816669053509,
+      "eval_loss": 0.8393772840499878,
+      "eval_runtime": 1035.7736,
+      "eval_samples_per_second": 482.13,
+      "eval_steps_per_second": 0.821,
+      "step": 11610
+    },
+    {
+      "epoch": 9.879420782611335,
+      "grad_norm": 11850.7548828125,
+      "learning_rate": 8.712871287128713e-05,
+      "loss": 0.9257,
+      "step": 11620
+    },
+    {
+      "epoch": 9.887920828147294,
+      "grad_norm": 12279.2197265625,
+      "learning_rate": 8.613861386138613e-05,
+      "loss": 0.9251,
+      "step": 11630
+    },
+    {
+      "epoch": 9.896420873683251,
+      "grad_norm": 12372.4765625,
+      "learning_rate": 8.514851485148515e-05,
+      "loss": 0.9241,
+      "step": 11640
+    },
+    {
+      "epoch": 9.90492091921921,
+      "grad_norm": 11712.88671875,
+      "learning_rate": 8.415841584158417e-05,
+      "loss": 0.9233,
+      "step": 11650
+    },
+    {
+      "epoch": 9.913420964755169,
+      "grad_norm": 12502.9453125,
+      "learning_rate": 8.316831683168318e-05,
+      "loss": 0.9252,
+      "step": 11660
+    },
+    {
+      "epoch": 9.921921010291127,
+      "grad_norm": 13177.66796875,
+      "learning_rate": 8.217821782178219e-05,
+      "loss": 0.9237,
+      "step": 11670
+    },
+    {
+      "epoch": 9.930421055827084,
+      "grad_norm": 12558.2802734375,
+      "learning_rate": 8.118811881188119e-05,
+      "loss": 0.9251,
+      "step": 11680
+    },
+    {
+      "epoch": 9.938921101363043,
+      "grad_norm": 11745.330078125,
+      "learning_rate": 8.01980198019802e-05,
+      "loss": 0.9264,
+      "step": 11690
+    },
+    {
+      "epoch": 9.947421146899002,
+      "grad_norm": 11350.08203125,
+      "learning_rate": 7.920792079207921e-05,
+      "loss": 0.924,
+      "step": 11700
+    },
+    {
+      "epoch": 9.947421146899002,
+      "eval_accuracy": 0.8009528506484064,
+      "eval_loss": 0.8385128974914551,
+      "eval_runtime": 1035.7833,
+      "eval_samples_per_second": 482.126,
+      "eval_steps_per_second": 0.821,
+      "step": 11700
+    },
+    {
+      "epoch": 9.955921192434959,
+      "grad_norm": 11514.369140625,
+      "learning_rate": 7.821782178217822e-05,
+      "loss": 0.9236,
+      "step": 11710
+    },
+    {
+      "epoch": 9.964421237970917,
+      "grad_norm": 11468.5146484375,
+      "learning_rate": 7.722772277227723e-05,
+      "loss": 0.9226,
+      "step": 11720
+    },
+    {
+      "epoch": 9.972921283506876,
+      "grad_norm": 11347.8369140625,
+      "learning_rate": 7.623762376237623e-05,
+      "loss": 0.9244,
+      "step": 11730
+    },
+    {
+      "epoch": 9.981421329042835,
+      "grad_norm": 11171.818359375,
+      "learning_rate": 7.524752475247524e-05,
+      "loss": 0.9247,
+      "step": 11740
+    },
+    {
+      "epoch": 9.989921374578792,
+      "grad_norm": 11165.7451171875,
+      "learning_rate": 7.425742574257426e-05,
+      "loss": 0.926,
+      "step": 11750
+    },
+    {
+      "epoch": 9.99842142011475,
+      "grad_norm": 11604.375,
+      "learning_rate": 7.326732673267327e-05,
+      "loss": 0.9237,
+      "step": 11760
+    },
+    {
+      "epoch": 10.00692146565071,
+      "grad_norm": 11761.7099609375,
+      "learning_rate": 7.227722772277228e-05,
+      "loss": 0.9245,
+      "step": 11770
+    },
+    {
+      "epoch": 10.015421511186666,
+      "grad_norm": 11363.318359375,
+      "learning_rate": 7.128712871287128e-05,
+      "loss": 0.9242,
+      "step": 11780
+    },
+    {
+      "epoch": 10.023921556722625,
+      "grad_norm": 11492.59765625,
+      "learning_rate": 7.02970297029703e-05,
+      "loss": 0.9243,
+      "step": 11790
+    },
+    {
+      "epoch": 10.023921556722625,
+      "eval_accuracy": 0.8013337325505296,
+      "eval_loss": 0.8371462225914001,
+      "eval_runtime": 1036.3502,
+      "eval_samples_per_second": 481.862,
+      "eval_steps_per_second": 0.82,
+      "step": 11790
+    },
+    {
+      "epoch": 10.032421602258584,
+      "grad_norm": 11367.05859375,
+      "learning_rate": 6.930693069306931e-05,
+      "loss": 0.9241,
+      "step": 11800
+    },
+    {
+      "epoch": 10.040921647794542,
+      "grad_norm": 11473.013671875,
+      "learning_rate": 6.831683168316833e-05,
+      "loss": 0.9223,
+      "step": 11810
+    },
+    {
+      "epoch": 10.0494216933305,
+      "grad_norm": 11349.96484375,
+      "learning_rate": 6.732673267326734e-05,
+      "loss": 0.9225,
+      "step": 11820
+    },
+    {
+      "epoch": 10.057921738866458,
+      "grad_norm": 11236.7431640625,
+      "learning_rate": 6.633663366336634e-05,
+      "loss": 0.9236,
+      "step": 11830
+    },
+    {
+      "epoch": 10.066421784402417,
+      "grad_norm": 11882.5078125,
+      "learning_rate": 6.534653465346535e-05,
+      "loss": 0.9227,
+      "step": 11840
+    },
+    {
+      "epoch": 10.074921829938374,
+      "grad_norm": 11027.5966796875,
+      "learning_rate": 6.435643564356436e-05,
+      "loss": 0.9212,
+      "step": 11850
+    },
+    {
+      "epoch": 10.083421875474333,
+      "grad_norm": 11009.87890625,
+      "learning_rate": 6.336633663366337e-05,
+      "loss": 0.9235,
+      "step": 11860
+    },
+    {
+      "epoch": 10.091921921010291,
+      "grad_norm": 10943.912109375,
+      "learning_rate": 6.237623762376238e-05,
+      "loss": 0.9236,
+      "step": 11870
+    },
+    {
+      "epoch": 10.10042196654625,
+      "grad_norm": 11583.521484375,
+      "learning_rate": 6.138613861386138e-05,
+      "loss": 0.9251,
+      "step": 11880
+    },
+    {
+      "epoch": 10.10042196654625,
+      "eval_accuracy": 0.8014076660668283,
+      "eval_loss": 0.8366426825523376,
+      "eval_runtime": 1035.5282,
+      "eval_samples_per_second": 482.245,
+      "eval_steps_per_second": 0.821,
+      "step": 11880
+    },
+    {
+      "epoch": 10.108922012082207,
+      "grad_norm": 10507.802734375,
+      "learning_rate": 6.039603960396039e-05,
+      "loss": 0.9218,
+      "step": 11890
+    },
+    {
+      "epoch": 10.117422057618166,
+      "grad_norm": 12043.7763671875,
+      "learning_rate": 5.9405940594059404e-05,
+      "loss": 0.9213,
+      "step": 11900
+    },
+    {
+      "epoch": 10.125922103154124,
+      "grad_norm": 10901.1728515625,
+      "learning_rate": 5.841584158415842e-05,
+      "loss": 0.92,
+      "step": 11910
+    },
+    {
+      "epoch": 10.134422148690081,
+      "grad_norm": 11496.796875,
+      "learning_rate": 5.742574257425743e-05,
+      "loss": 0.922,
+      "step": 11920
+    },
+    {
+      "epoch": 10.14292219422604,
+      "grad_norm": 11207.1455078125,
+      "learning_rate": 5.643564356435644e-05,
+      "loss": 0.9207,
+      "step": 11930
+    },
+    {
+      "epoch": 10.151422239761999,
+      "grad_norm": 10568.88671875,
+      "learning_rate": 5.5445544554455445e-05,
+      "loss": 0.923,
+      "step": 11940
+    },
+    {
+      "epoch": 10.159922285297958,
+      "grad_norm": 11236.5009765625,
+      "learning_rate": 5.4455445544554456e-05,
+      "loss": 0.9226,
+      "step": 11950
+    },
+    {
+      "epoch": 10.168422330833915,
+      "grad_norm": 10793.302734375,
+      "learning_rate": 5.346534653465347e-05,
+      "loss": 0.921,
+      "step": 11960
+    },
+    {
+      "epoch": 10.176922376369873,
+      "grad_norm": 10562.205078125,
+      "learning_rate": 5.247524752475247e-05,
+      "loss": 0.9211,
+      "step": 11970
+    },
+    {
+      "epoch": 10.176922376369873,
+      "eval_accuracy": 0.8014045174418934,
+      "eval_loss": 0.836881697177887,
+      "eval_runtime": 1035.287,
+      "eval_samples_per_second": 482.357,
+      "eval_steps_per_second": 0.821,
+      "step": 11970
+    },
+    {
+      "epoch": 10.185422421905832,
+      "grad_norm": 10615.4091796875,
+      "learning_rate": 5.1485148514851485e-05,
+      "loss": 0.9216,
+      "step": 11980
+    },
+    {
+      "epoch": 10.19392246744179,
+      "grad_norm": 11039.455078125,
+      "learning_rate": 5.0495049504950497e-05,
+      "loss": 0.9193,
+      "step": 11990
+    },
+    {
+      "epoch": 10.202422512977748,
+      "grad_norm": 11875.810546875,
+      "learning_rate": 4.950495049504951e-05,
+      "loss": 0.9221,
+      "step": 12000
+    },
+    {
+      "epoch": 10.210922558513706,
+      "grad_norm": 10764.8984375,
+      "learning_rate": 4.851485148514852e-05,
+      "loss": 0.922,
+      "step": 12010
+    },
+    {
+      "epoch": 10.219422604049665,
+      "grad_norm": 10610.0224609375,
+      "learning_rate": 4.7524752475247525e-05,
+      "loss": 0.9225,
+      "step": 12020
+    },
+    {
+      "epoch": 10.227922649585622,
+      "grad_norm": 10679.6396484375,
+      "learning_rate": 4.653465346534654e-05,
+      "loss": 0.921,
+      "step": 12030
+    },
+    {
+      "epoch": 10.23642269512158,
+      "grad_norm": 10098.7451171875,
+      "learning_rate": 4.554455445544554e-05,
+      "loss": 0.9227,
+      "step": 12040
+    },
+    {
+      "epoch": 10.24492274065754,
+      "grad_norm": 11032.216796875,
+      "learning_rate": 4.455445544554455e-05,
+      "loss": 0.9223,
+      "step": 12050
+    },
+    {
+      "epoch": 10.253422786193497,
+      "grad_norm": 11303.1201171875,
+      "learning_rate": 4.3564356435643565e-05,
+      "loss": 0.9223,
+      "step": 12060
+    },
+    {
+      "epoch": 10.253422786193497,
+      "eval_accuracy": 0.8015698536327949,
+      "eval_loss": 0.8357640504837036,
+      "eval_runtime": 1032.412,
+      "eval_samples_per_second": 483.7,
+      "eval_steps_per_second": 0.823,
+      "step": 12060
+    },
+    {
+      "epoch": 10.261922831729455,
+      "grad_norm": 10582.2373046875,
+      "learning_rate": 4.257425742574258e-05,
+      "loss": 0.9192,
+      "step": 12070
+    },
+    {
+      "epoch": 10.270422877265414,
+      "grad_norm": 10299.9345703125,
+      "learning_rate": 4.158415841584159e-05,
+      "loss": 0.9209,
+      "step": 12080
+    },
+    {
+      "epoch": 10.278922922801373,
+      "grad_norm": 10322.740234375,
+      "learning_rate": 4.0594059405940594e-05,
+      "loss": 0.922,
+      "step": 12090
+    },
+    {
+      "epoch": 10.28742296833733,
+      "grad_norm": 10159.0244140625,
+      "learning_rate": 3.9603960396039605e-05,
+      "loss": 0.9189,
+      "step": 12100
+    },
+    {
+      "epoch": 10.295923013873288,
+      "grad_norm": 10132.728515625,
+      "learning_rate": 3.861386138613862e-05,
+      "loss": 0.9209,
+      "step": 12110
+    },
+    {
+      "epoch": 10.304423059409247,
+      "grad_norm": 10718.8837890625,
+      "learning_rate": 3.762376237623762e-05,
+      "loss": 0.9208,
+      "step": 12120
+    },
+    {
+      "epoch": 10.312923104945206,
+      "grad_norm": 10195.5869140625,
+      "learning_rate": 3.6633663366336634e-05,
+      "loss": 0.9225,
+      "step": 12130
+    },
+    {
+      "epoch": 10.321423150481163,
+      "grad_norm": 10577.4375,
+      "learning_rate": 3.564356435643564e-05,
+      "loss": 0.9204,
+      "step": 12140
+    },
+    {
+      "epoch": 10.329923196017122,
+      "grad_norm": 10045.123046875,
+      "learning_rate": 3.465346534653466e-05,
+      "loss": 0.9182,
+      "step": 12150
+    },
+    {
+      "epoch": 10.329923196017122,
+      "eval_accuracy": 0.8016586056285943,
+      "eval_loss": 0.835513710975647,
+      "eval_runtime": 1039.9403,
+      "eval_samples_per_second": 480.199,
+      "eval_steps_per_second": 0.817,
+      "step": 12150
+    },
+    {
+      "epoch": 10.33842324155308,
+      "grad_norm": 10554.9501953125,
+      "learning_rate": 3.366336633663367e-05,
+      "loss": 0.9219,
+      "step": 12160
+    },
+    {
+      "epoch": 10.346923287089037,
+      "grad_norm": 10395.685546875,
+      "learning_rate": 3.2673267326732674e-05,
+      "loss": 0.9189,
+      "step": 12170
+    },
+    {
+      "epoch": 10.355423332624996,
+      "grad_norm": 10223.9580078125,
+      "learning_rate": 3.1683168316831686e-05,
+      "loss": 0.9205,
+      "step": 12180
+    },
+    {
+      "epoch": 10.363923378160955,
+      "grad_norm": 10570.7958984375,
+      "learning_rate": 3.069306930693069e-05,
+      "loss": 0.9207,
+      "step": 12190
+    },
+    {
+      "epoch": 10.372423423696914,
+      "grad_norm": 9892.7470703125,
+      "learning_rate": 2.9702970297029702e-05,
+      "loss": 0.918,
+      "step": 12200
+    },
+    {
+      "epoch": 10.38092346923287,
+      "grad_norm": 9921.88671875,
+      "learning_rate": 2.8712871287128714e-05,
+      "loss": 0.9196,
+      "step": 12210
+    },
+    {
+      "epoch": 10.38942351476883,
+      "grad_norm": 9997.732421875,
+      "learning_rate": 2.7722772277227722e-05,
+      "loss": 0.9217,
+      "step": 12220
+    },
+    {
+      "epoch": 10.397923560304788,
+      "grad_norm": 10292.541015625,
+      "learning_rate": 2.6732673267326734e-05,
+      "loss": 0.9195,
+      "step": 12230
+    },
+    {
+      "epoch": 10.406423605840745,
+      "grad_norm": 10017.322265625,
+      "learning_rate": 2.5742574257425742e-05,
+      "loss": 0.9215,
+      "step": 12240
+    },
+    {
+      "epoch": 10.406423605840745,
+      "eval_accuracy": 0.8019068191825514,
+      "eval_loss": 0.83404940366745,
+      "eval_runtime": 1036.3785,
+      "eval_samples_per_second": 481.849,
+      "eval_steps_per_second": 0.82,
+      "step": 12240
+    },
+    {
+      "epoch": 10.414923651376704,
+      "grad_norm": 10193.8349609375,
+      "learning_rate": 2.4752475247524754e-05,
+      "loss": 0.9206,
+      "step": 12250
+    },
+    {
+      "epoch": 10.423423696912662,
+      "grad_norm": 10791.517578125,
+      "learning_rate": 2.3762376237623762e-05,
+      "loss": 0.9192,
+      "step": 12260
+    },
+    {
+      "epoch": 10.431923742448621,
+      "grad_norm": 9776.1796875,
+      "learning_rate": 2.277227722772277e-05,
+      "loss": 0.9183,
+      "step": 12270
+    },
+    {
+      "epoch": 10.440423787984578,
+      "grad_norm": 9862.8271484375,
+      "learning_rate": 2.1782178217821783e-05,
+      "loss": 0.9194,
+      "step": 12280
+    },
+    {
+      "epoch": 10.448923833520537,
+      "grad_norm": 10056.2607421875,
+      "learning_rate": 2.0792079207920794e-05,
+      "loss": 0.9207,
+      "step": 12290
+    },
+    {
+      "epoch": 10.457423879056496,
+      "grad_norm": 9770.7578125,
+      "learning_rate": 1.9801980198019803e-05,
+      "loss": 0.9198,
+      "step": 12300
+    },
+    {
+      "epoch": 10.465923924592452,
+      "grad_norm": 10069.5634765625,
+      "learning_rate": 1.881188118811881e-05,
+      "loss": 0.9201,
+      "step": 12310
+    },
+    {
+      "epoch": 10.474423970128411,
+      "grad_norm": 9465.1181640625,
+      "learning_rate": 1.782178217821782e-05,
+      "loss": 0.9197,
+      "step": 12320
+    },
+    {
+      "epoch": 10.48292401566437,
+      "grad_norm": 9807.8330078125,
+      "learning_rate": 1.6831683168316834e-05,
+      "loss": 0.9179,
+      "step": 12330
+    },
+    {
+      "epoch": 10.48292401566437,
+      "eval_accuracy": 0.8019292754647317,
+      "eval_loss": 0.8340857625007629,
+      "eval_runtime": 1037.0091,
+      "eval_samples_per_second": 481.556,
+      "eval_steps_per_second": 0.82,
+      "step": 12330
+    },
+    {
+      "epoch": 10.491424061200329,
+      "grad_norm": 9650.10546875,
+      "learning_rate": 1.5841584158415843e-05,
+      "loss": 0.9205,
+      "step": 12340
+    },
+    {
+      "epoch": 10.499924106736286,
+      "grad_norm": 9682.900390625,
+      "learning_rate": 1.4851485148514851e-05,
+      "loss": 0.9195,
+      "step": 12350
+    },
+    {
+      "epoch": 10.508424152272244,
+      "grad_norm": 9976.9375,
+      "learning_rate": 1.3861386138613861e-05,
+      "loss": 0.9191,
+      "step": 12360
+    },
+    {
+      "epoch": 10.516924197808203,
+      "grad_norm": 9428.4052734375,
+      "learning_rate": 1.2871287128712871e-05,
+      "loss": 0.9189,
+      "step": 12370
+    },
+    {
+      "epoch": 10.52542424334416,
+      "grad_norm": 10025.7158203125,
+      "learning_rate": 1.1881188118811881e-05,
+      "loss": 0.918,
+      "step": 12380
+    },
+    {
+      "epoch": 10.533924288880119,
+      "grad_norm": 9750.935546875,
+      "learning_rate": 1.0891089108910891e-05,
+      "loss": 0.9187,
+      "step": 12390
+    },
+    {
+      "epoch": 10.542424334416078,
+      "grad_norm": 9010.52734375,
+      "learning_rate": 9.900990099009901e-06,
+      "loss": 0.9189,
+      "step": 12400
+    },
+    {
+      "epoch": 10.550924379952036,
+      "grad_norm": 9163.4169921875,
+      "learning_rate": 8.91089108910891e-06,
+      "loss": 0.9191,
+      "step": 12410
+    },
+    {
+      "epoch": 10.559424425487993,
+      "grad_norm": 9015.5771484375,
+      "learning_rate": 7.920792079207921e-06,
+      "loss": 0.9195,
+      "step": 12420
+    },
+    {
+      "epoch": 10.559424425487993,
+      "eval_accuracy": 0.802042445850802,
+      "eval_loss": 0.833830714225769,
+      "eval_runtime": 1033.7642,
+      "eval_samples_per_second": 483.068,
+      "eval_steps_per_second": 0.822,
+      "step": 12420
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 12500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 11,
+  "save_steps": 90,
+  "total_flos": 7.8411669992781e+18,
+  "train_batch_size": 288,
+  "trial_name": null,
+  "trial_params": null
+}