{
  "best_metric": 0.583153247833252,
  "best_model_checkpoint": "./vit-base-brain-tumor-detection/checkpoint-1000",
  "epoch": 20.0,
  "eval_steps": 100,
  "global_step": 5000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.04,
      "grad_norm": 1.653470516204834,
      "learning_rate": 0.0001996,
      "loss": 1.3897,
      "step": 10
    },
    {
      "epoch": 0.08,
      "grad_norm": 0.5039440989494324,
      "learning_rate": 0.00019920000000000002,
      "loss": 1.4048,
      "step": 20
    },
    {
      "epoch": 0.12,
      "grad_norm": 0.8407747745513916,
      "learning_rate": 0.0001988,
      "loss": 1.3912,
      "step": 30
    },
    {
      "epoch": 0.16,
      "grad_norm": 0.898512601852417,
      "learning_rate": 0.0001984,
      "loss": 1.3661,
      "step": 40
    },
    {
      "epoch": 0.2,
      "grad_norm": 2.638078212738037,
      "learning_rate": 0.00019800000000000002,
      "loss": 1.2842,
      "step": 50
    },
    {
      "epoch": 0.24,
      "grad_norm": 1.7601313591003418,
      "learning_rate": 0.0001976,
      "loss": 1.239,
      "step": 60
    },
    {
      "epoch": 0.28,
      "grad_norm": 5.79738187789917,
      "learning_rate": 0.0001972,
      "loss": 1.1859,
      "step": 70
    },
    {
      "epoch": 0.32,
      "grad_norm": 3.172907829284668,
      "learning_rate": 0.0001968,
      "loss": 1.1265,
      "step": 80
    },
    {
      "epoch": 0.36,
      "grad_norm": 3.0865306854248047,
      "learning_rate": 0.0001964,
      "loss": 1.1506,
      "step": 90
    },
    {
      "epoch": 0.4,
      "grad_norm": 3.093891143798828,
      "learning_rate": 0.000196,
      "loss": 0.9535,
      "step": 100
    },
    {
      "epoch": 0.4,
      "eval_accuracy": 0.618,
      "eval_loss": 0.8965551853179932,
      "eval_runtime": 12.9233,
      "eval_samples_per_second": 77.38,
      "eval_steps_per_second": 9.672,
      "step": 100
    },
    {
      "epoch": 0.44,
      "grad_norm": 1.729243516921997,
      "learning_rate": 0.0001956,
      "loss": 0.9068,
      "step": 110
    },
    {
      "epoch": 0.48,
      "grad_norm": 4.636998653411865,
      "learning_rate": 0.0001952,
      "loss": 0.8465,
      "step": 120
    },
    {
      "epoch": 0.52,
      "grad_norm": 1.6259952783584595,
      "learning_rate": 0.0001948,
      "loss": 1.0864,
      "step": 130
    },
    {
      "epoch": 0.56,
      "grad_norm": 2.269352436065674,
      "learning_rate": 0.0001944,
      "loss": 1.0165,
      "step": 140
    },
    {
      "epoch": 0.6,
      "grad_norm": 2.516747236251831,
      "learning_rate": 0.000194,
      "loss": 0.8877,
      "step": 150
    },
    {
      "epoch": 0.64,
      "grad_norm": 3.543081760406494,
      "learning_rate": 0.00019360000000000002,
      "loss": 0.9845,
      "step": 160
    },
    {
      "epoch": 0.68,
      "grad_norm": 2.1874799728393555,
      "learning_rate": 0.0001932,
      "loss": 0.8626,
      "step": 170
    },
    {
      "epoch": 0.72,
      "grad_norm": 2.4686241149902344,
      "learning_rate": 0.0001928,
      "loss": 0.8459,
      "step": 180
    },
    {
      "epoch": 0.76,
      "grad_norm": 3.529902696609497,
      "learning_rate": 0.00019240000000000001,
      "loss": 1.0014,
      "step": 190
    },
    {
      "epoch": 0.8,
      "grad_norm": 6.457658767700195,
      "learning_rate": 0.000192,
      "loss": 0.862,
      "step": 200
    },
    {
      "epoch": 0.8,
      "eval_accuracy": 0.561,
      "eval_loss": 1.1149431467056274,
      "eval_runtime": 5.617,
      "eval_samples_per_second": 178.029,
      "eval_steps_per_second": 22.254,
      "step": 200
    },
    {
      "epoch": 0.84,
      "grad_norm": 3.746286153793335,
      "learning_rate": 0.0001916,
      "loss": 0.9211,
      "step": 210
    },
    {
      "epoch": 0.88,
      "grad_norm": 0.9981279373168945,
      "learning_rate": 0.0001912,
      "loss": 0.7935,
      "step": 220
    },
    {
      "epoch": 0.92,
      "grad_norm": 4.057267189025879,
      "learning_rate": 0.0001908,
      "loss": 0.9092,
      "step": 230
    },
    {
      "epoch": 0.96,
      "grad_norm": 1.723970890045166,
      "learning_rate": 0.0001904,
      "loss": 0.8029,
      "step": 240
    },
    {
      "epoch": 1.0,
      "grad_norm": 2.485327959060669,
      "learning_rate": 0.00019,
      "loss": 0.8081,
      "step": 250
    },
    {
      "epoch": 1.04,
      "grad_norm": 1.9523849487304688,
      "learning_rate": 0.0001896,
      "loss": 0.6713,
      "step": 260
    },
    {
      "epoch": 1.08,
      "grad_norm": 1.719793438911438,
      "learning_rate": 0.0001892,
      "loss": 0.6921,
      "step": 270
    },
    {
      "epoch": 1.12,
      "grad_norm": 1.283629059791565,
      "learning_rate": 0.0001888,
      "loss": 0.6989,
      "step": 280
    },
    {
      "epoch": 1.16,
      "grad_norm": 14.220324516296387,
      "learning_rate": 0.0001884,
      "loss": 0.9012,
      "step": 290
    },
    {
      "epoch": 1.2,
      "grad_norm": 1.621273398399353,
      "learning_rate": 0.000188,
      "loss": 0.7373,
      "step": 300
    },
    {
      "epoch": 1.2,
      "eval_accuracy": 0.605,
      "eval_loss": 0.8542758822441101,
      "eval_runtime": 5.4409,
      "eval_samples_per_second": 183.793,
      "eval_steps_per_second": 22.974,
      "step": 300
    },
    {
      "epoch": 1.24,
      "grad_norm": 1.4123197793960571,
      "learning_rate": 0.0001876,
      "loss": 0.8087,
      "step": 310
    },
    {
      "epoch": 1.28,
      "grad_norm": 4.211037635803223,
      "learning_rate": 0.00018720000000000002,
      "loss": 0.7055,
      "step": 320
    },
    {
      "epoch": 1.32,
      "grad_norm": 1.4883663654327393,
      "learning_rate": 0.00018680000000000001,
      "loss": 0.7988,
      "step": 330
    },
    {
      "epoch": 1.3599999999999999,
      "grad_norm": 1.515805959701538,
      "learning_rate": 0.00018640000000000003,
      "loss": 0.8664,
      "step": 340
    },
    {
      "epoch": 1.4,
      "grad_norm": 1.3907535076141357,
      "learning_rate": 0.00018600000000000002,
      "loss": 0.6773,
      "step": 350
    },
    {
      "epoch": 1.44,
      "grad_norm": 1.5013153553009033,
      "learning_rate": 0.0001856,
      "loss": 0.6876,
      "step": 360
    },
    {
      "epoch": 1.48,
      "grad_norm": 2.7530360221862793,
      "learning_rate": 0.00018520000000000003,
      "loss": 0.5832,
      "step": 370
    },
    {
      "epoch": 1.52,
      "grad_norm": 2.470029354095459,
      "learning_rate": 0.00018480000000000002,
      "loss": 0.7209,
      "step": 380
    },
    {
      "epoch": 1.56,
      "grad_norm": 1.3378897905349731,
      "learning_rate": 0.0001844,
      "loss": 0.6202,
      "step": 390
    },
    {
      "epoch": 1.6,
      "grad_norm": 2.541882038116455,
      "learning_rate": 0.00018400000000000003,
      "loss": 0.6476,
      "step": 400
    },
    {
      "epoch": 1.6,
      "eval_accuracy": 0.666,
      "eval_loss": 0.7306695580482483,
      "eval_runtime": 6.1267,
      "eval_samples_per_second": 163.221,
      "eval_steps_per_second": 20.403,
      "step": 400
    },
    {
      "epoch": 1.6400000000000001,
      "grad_norm": 2.5715789794921875,
      "learning_rate": 0.00018360000000000002,
      "loss": 0.5758,
      "step": 410
    },
    {
      "epoch": 1.6800000000000002,
      "grad_norm": 1.6180702447891235,
      "learning_rate": 0.0001832,
      "loss": 0.7008,
      "step": 420
    },
    {
      "epoch": 1.72,
      "grad_norm": 2.0645174980163574,
      "learning_rate": 0.00018280000000000003,
      "loss": 0.7472,
      "step": 430
    },
    {
      "epoch": 1.76,
      "grad_norm": 6.950054168701172,
      "learning_rate": 0.00018240000000000002,
      "loss": 0.9427,
      "step": 440
    },
    {
      "epoch": 1.8,
      "grad_norm": 2.622056722640991,
      "learning_rate": 0.000182,
      "loss": 0.6746,
      "step": 450
    },
    {
      "epoch": 1.8399999999999999,
      "grad_norm": 1.489776611328125,
      "learning_rate": 0.00018160000000000002,
      "loss": 0.7327,
      "step": 460
    },
    {
      "epoch": 1.88,
      "grad_norm": 1.0776207447052002,
      "learning_rate": 0.0001812,
      "loss": 0.5842,
      "step": 470
    },
    {
      "epoch": 1.92,
      "grad_norm": 2.286405086517334,
      "learning_rate": 0.0001808,
      "loss": 0.6962,
      "step": 480
    },
    {
      "epoch": 1.96,
      "grad_norm": 2.593034267425537,
      "learning_rate": 0.00018040000000000002,
      "loss": 0.6214,
      "step": 490
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.7781627178192139,
      "learning_rate": 0.00018,
      "loss": 0.6712,
      "step": 500
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.694,
      "eval_loss": 0.6953811049461365,
      "eval_runtime": 5.961,
      "eval_samples_per_second": 167.757,
      "eval_steps_per_second": 20.97,
      "step": 500
    },
    {
      "epoch": 2.04,
      "grad_norm": 1.331834316253662,
      "learning_rate": 0.0001796,
      "loss": 0.3866,
      "step": 510
    },
    {
      "epoch": 2.08,
      "grad_norm": 1.5159146785736084,
      "learning_rate": 0.00017920000000000002,
      "loss": 0.5219,
      "step": 520
    },
    {
      "epoch": 2.12,
      "grad_norm": 3.525988817214966,
      "learning_rate": 0.0001788,
      "loss": 0.4692,
      "step": 530
    },
    {
      "epoch": 2.16,
      "grad_norm": 6.470268249511719,
      "learning_rate": 0.0001784,
      "loss": 0.4269,
      "step": 540
    },
    {
      "epoch": 2.2,
      "grad_norm": 1.4871269464492798,
      "learning_rate": 0.00017800000000000002,
      "loss": 0.3844,
      "step": 550
    },
    {
      "epoch": 2.24,
      "grad_norm": 3.404496669769287,
      "learning_rate": 0.0001776,
      "loss": 0.6704,
      "step": 560
    },
    {
      "epoch": 2.2800000000000002,
      "grad_norm": 1.3112822771072388,
      "learning_rate": 0.0001772,
      "loss": 0.6746,
      "step": 570
    },
    {
      "epoch": 2.32,
      "grad_norm": 5.380651473999023,
      "learning_rate": 0.00017680000000000001,
      "loss": 0.7249,
      "step": 580
    },
    {
      "epoch": 2.36,
      "grad_norm": 1.006377935409546,
      "learning_rate": 0.0001764,
      "loss": 0.4615,
      "step": 590
    },
    {
      "epoch": 2.4,
      "grad_norm": 2.6458024978637695,
      "learning_rate": 0.00017600000000000002,
      "loss": 0.4892,
      "step": 600
    },
    {
      "epoch": 2.4,
      "eval_accuracy": 0.707,
      "eval_loss": 0.6391066908836365,
      "eval_runtime": 5.4367,
      "eval_samples_per_second": 183.934,
      "eval_steps_per_second": 22.992,
      "step": 600
    },
    {
      "epoch": 2.44,
      "grad_norm": 3.180305004119873,
      "learning_rate": 0.0001756,
      "loss": 0.5289,
      "step": 610
    },
    {
      "epoch": 2.48,
      "grad_norm": 3.7770426273345947,
      "learning_rate": 0.0001752,
      "loss": 0.4695,
      "step": 620
    },
    {
      "epoch": 2.52,
      "grad_norm": 3.895836591720581,
      "learning_rate": 0.00017480000000000002,
      "loss": 0.4387,
      "step": 630
    },
    {
      "epoch": 2.56,
      "grad_norm": 3.006532907485962,
      "learning_rate": 0.0001744,
      "loss": 0.5095,
      "step": 640
    },
    {
      "epoch": 2.6,
      "grad_norm": 4.034090995788574,
      "learning_rate": 0.000174,
      "loss": 0.5297,
      "step": 650
    },
    {
      "epoch": 2.64,
      "grad_norm": 2.2521889209747314,
      "learning_rate": 0.00017360000000000002,
      "loss": 0.4649,
      "step": 660
    },
    {
      "epoch": 2.68,
      "grad_norm": 2.8466689586639404,
      "learning_rate": 0.0001732,
      "loss": 0.4819,
      "step": 670
    },
    {
      "epoch": 2.7199999999999998,
      "grad_norm": 8.159728050231934,
      "learning_rate": 0.0001728,
      "loss": 0.5005,
      "step": 680
    },
    {
      "epoch": 2.76,
      "grad_norm": 2.4270811080932617,
      "learning_rate": 0.00017240000000000002,
      "loss": 0.4146,
      "step": 690
    },
    {
      "epoch": 2.8,
      "grad_norm": 4.790659427642822,
      "learning_rate": 0.000172,
      "loss": 0.5801,
      "step": 700
    },
    {
      "epoch": 2.8,
      "eval_accuracy": 0.708,
      "eval_loss": 0.6246920824050903,
      "eval_runtime": 5.4112,
      "eval_samples_per_second": 184.802,
      "eval_steps_per_second": 23.1,
      "step": 700
    },
    {
      "epoch": 2.84,
      "grad_norm": 2.981046676635742,
      "learning_rate": 0.0001716,
      "loss": 0.5311,
      "step": 710
    },
    {
      "epoch": 2.88,
      "grad_norm": 3.6522932052612305,
      "learning_rate": 0.00017120000000000001,
      "loss": 0.4836,
      "step": 720
    },
    {
      "epoch": 2.92,
      "grad_norm": 2.448866367340088,
      "learning_rate": 0.0001708,
      "loss": 0.6052,
      "step": 730
    },
    {
      "epoch": 2.96,
      "grad_norm": 3.3822741508483887,
      "learning_rate": 0.0001704,
      "loss": 0.4695,
      "step": 740
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.214433431625366,
      "learning_rate": 0.00017,
      "loss": 0.6105,
      "step": 750
    },
    {
      "epoch": 3.04,
      "grad_norm": 1.5671595335006714,
      "learning_rate": 0.0001696,
      "loss": 0.4189,
      "step": 760
    },
    {
      "epoch": 3.08,
      "grad_norm": 3.648240327835083,
      "learning_rate": 0.0001692,
      "loss": 0.2745,
      "step": 770
    },
    {
      "epoch": 3.12,
      "grad_norm": 1.588549017906189,
      "learning_rate": 0.0001688,
      "loss": 0.2882,
      "step": 780
    },
    {
      "epoch": 3.16,
      "grad_norm": 2.041724920272827,
      "learning_rate": 0.0001684,
      "loss": 0.2642,
      "step": 790
    },
    {
      "epoch": 3.2,
      "grad_norm": 4.308396816253662,
      "learning_rate": 0.000168,
      "loss": 0.3505,
      "step": 800
    },
    {
      "epoch": 3.2,
      "eval_accuracy": 0.778,
      "eval_loss": 0.6055798530578613,
      "eval_runtime": 6.0689,
      "eval_samples_per_second": 164.773,
      "eval_steps_per_second": 20.597,
      "step": 800
    },
    {
      "epoch": 3.24,
      "grad_norm": 2.717337131500244,
      "learning_rate": 0.0001676,
      "loss": 0.2626,
      "step": 810
    },
    {
      "epoch": 3.2800000000000002,
      "grad_norm": 2.7859904766082764,
      "learning_rate": 0.0001672,
      "loss": 0.3826,
      "step": 820
    },
    {
      "epoch": 3.32,
      "grad_norm": 3.864297389984131,
      "learning_rate": 0.0001668,
      "loss": 0.4313,
      "step": 830
    },
    {
      "epoch": 3.36,
      "grad_norm": 3.782465934753418,
      "learning_rate": 0.0001664,
      "loss": 0.3361,
      "step": 840
    },
    {
      "epoch": 3.4,
      "grad_norm": 3.1425442695617676,
      "learning_rate": 0.000166,
      "loss": 0.2759,
      "step": 850
    },
    {
      "epoch": 3.44,
      "grad_norm": 3.1139609813690186,
      "learning_rate": 0.0001656,
      "loss": 0.3949,
      "step": 860
    },
    {
      "epoch": 3.48,
      "grad_norm": 4.945987224578857,
      "learning_rate": 0.0001652,
      "loss": 0.3701,
      "step": 870
    },
    {
      "epoch": 3.52,
      "grad_norm": 1.7310290336608887,
      "learning_rate": 0.0001648,
      "loss": 0.2879,
      "step": 880
    },
    {
      "epoch": 3.56,
      "grad_norm": 5.650954246520996,
      "learning_rate": 0.0001644,
      "loss": 0.4443,
      "step": 890
    },
    {
      "epoch": 3.6,
      "grad_norm": 2.8908536434173584,
      "learning_rate": 0.000164,
      "loss": 0.3503,
      "step": 900
    },
    {
      "epoch": 3.6,
      "eval_accuracy": 0.743,
      "eval_loss": 0.6263792514801025,
      "eval_runtime": 5.7324,
      "eval_samples_per_second": 174.449,
      "eval_steps_per_second": 21.806,
      "step": 900
    },
    {
      "epoch": 3.64,
      "grad_norm": 1.3399735689163208,
      "learning_rate": 0.0001636,
      "loss": 0.3594,
      "step": 910
    },
    {
      "epoch": 3.68,
      "grad_norm": 1.3512450456619263,
      "learning_rate": 0.0001632,
      "loss": 0.31,
      "step": 920
    },
    {
      "epoch": 3.7199999999999998,
      "grad_norm": 2.5052506923675537,
      "learning_rate": 0.0001628,
      "loss": 0.3234,
      "step": 930
    },
    {
      "epoch": 3.76,
      "grad_norm": 7.230278968811035,
      "learning_rate": 0.00016240000000000002,
      "loss": 0.3559,
      "step": 940
    },
    {
      "epoch": 3.8,
      "grad_norm": 1.464324712753296,
      "learning_rate": 0.000162,
      "loss": 0.2914,
      "step": 950
    },
    {
      "epoch": 3.84,
      "grad_norm": 3.140822649002075,
      "learning_rate": 0.00016160000000000002,
      "loss": 0.2292,
      "step": 960
    },
    {
      "epoch": 3.88,
      "grad_norm": 7.812990665435791,
      "learning_rate": 0.00016120000000000002,
      "loss": 0.401,
      "step": 970
    },
    {
      "epoch": 3.92,
      "grad_norm": 7.805127143859863,
      "learning_rate": 0.0001608,
      "loss": 0.4106,
      "step": 980
    },
    {
      "epoch": 3.96,
      "grad_norm": 5.645721435546875,
      "learning_rate": 0.00016040000000000002,
      "loss": 0.3844,
      "step": 990
    },
    {
      "epoch": 4.0,
      "grad_norm": 2.4538416862487793,
      "learning_rate": 0.00016,
      "loss": 0.3416,
      "step": 1000
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.785,
      "eval_loss": 0.583153247833252,
      "eval_runtime": 5.857,
      "eval_samples_per_second": 170.736,
      "eval_steps_per_second": 21.342,
      "step": 1000
    },
    {
      "epoch": 4.04,
      "grad_norm": 3.2518820762634277,
      "learning_rate": 0.0001596,
      "loss": 0.1471,
      "step": 1010
    },
    {
      "epoch": 4.08,
      "grad_norm": 5.489699840545654,
      "learning_rate": 0.00015920000000000002,
      "loss": 0.1807,
      "step": 1020
    },
    {
      "epoch": 4.12,
      "grad_norm": 0.6528148651123047,
      "learning_rate": 0.0001588,
      "loss": 0.189,
      "step": 1030
    },
    {
      "epoch": 4.16,
      "grad_norm": 5.080958843231201,
      "learning_rate": 0.00015840000000000003,
      "loss": 0.246,
      "step": 1040
    },
    {
      "epoch": 4.2,
      "grad_norm": 2.6048026084899902,
      "learning_rate": 0.00015800000000000002,
      "loss": 0.1937,
      "step": 1050
    },
    {
      "epoch": 4.24,
      "grad_norm": 2.226956844329834,
      "learning_rate": 0.0001576,
      "loss": 0.1781,
      "step": 1060
    },
    {
      "epoch": 4.28,
      "grad_norm": 1.7226955890655518,
      "learning_rate": 0.00015720000000000003,
      "loss": 0.2304,
      "step": 1070
    },
    {
      "epoch": 4.32,
      "grad_norm": 4.231192111968994,
      "learning_rate": 0.00015680000000000002,
      "loss": 0.1828,
      "step": 1080
    },
    {
      "epoch": 4.36,
      "grad_norm": 2.015580177307129,
      "learning_rate": 0.0001564,
      "loss": 0.2052,
      "step": 1090
    },
    {
      "epoch": 4.4,
      "grad_norm": 1.2273974418640137,
      "learning_rate": 0.00015600000000000002,
      "loss": 0.1427,
      "step": 1100
    },
    {
      "epoch": 4.4,
      "eval_accuracy": 0.769,
      "eval_loss": 0.7296926975250244,
      "eval_runtime": 5.2856,
      "eval_samples_per_second": 189.193,
      "eval_steps_per_second": 23.649,
      "step": 1100
    },
    {
      "epoch": 4.44,
      "grad_norm": 12.377334594726562,
      "learning_rate": 0.00015560000000000001,
      "loss": 0.2476,
      "step": 1110
    },
    {
      "epoch": 4.48,
      "grad_norm": 5.310695171356201,
      "learning_rate": 0.0001552,
      "loss": 0.2956,
      "step": 1120
    },
    {
      "epoch": 4.52,
      "grad_norm": 1.6223011016845703,
      "learning_rate": 0.00015480000000000002,
      "loss": 0.3322,
      "step": 1130
    },
    {
      "epoch": 4.5600000000000005,
      "grad_norm": 2.9505221843719482,
      "learning_rate": 0.0001544,
      "loss": 0.2848,
      "step": 1140
    },
    {
      "epoch": 4.6,
      "grad_norm": 8.578746795654297,
      "learning_rate": 0.000154,
      "loss": 0.1866,
      "step": 1150
    },
    {
      "epoch": 4.64,
      "grad_norm": 3.874375820159912,
      "learning_rate": 0.00015360000000000002,
      "loss": 0.2218,
      "step": 1160
    },
    {
      "epoch": 4.68,
      "grad_norm": 2.2798233032226562,
      "learning_rate": 0.0001532,
      "loss": 0.2767,
      "step": 1170
    },
    {
      "epoch": 4.72,
      "grad_norm": 1.274636149406433,
      "learning_rate": 0.0001528,
      "loss": 0.2295,
      "step": 1180
    },
    {
      "epoch": 4.76,
      "grad_norm": 4.265562534332275,
      "learning_rate": 0.00015240000000000002,
      "loss": 0.2062,
      "step": 1190
    },
    {
      "epoch": 4.8,
      "grad_norm": 3.9114277362823486,
      "learning_rate": 0.000152,
      "loss": 0.1982,
      "step": 1200
    },
    {
      "epoch": 4.8,
      "eval_accuracy": 0.73,
      "eval_loss": 0.7761247754096985,
      "eval_runtime": 5.3281,
      "eval_samples_per_second": 187.684,
      "eval_steps_per_second": 23.46,
      "step": 1200
    },
    {
      "epoch": 4.84,
      "grad_norm": 3.885077953338623,
      "learning_rate": 0.0001516,
      "loss": 0.2723,
      "step": 1210
    },
    {
      "epoch": 4.88,
      "grad_norm": 3.1466147899627686,
      "learning_rate": 0.00015120000000000002,
      "loss": 0.1816,
      "step": 1220
    },
    {
      "epoch": 4.92,
      "grad_norm": 1.5894757509231567,
      "learning_rate": 0.0001508,
      "loss": 0.2388,
      "step": 1230
    },
    {
      "epoch": 4.96,
      "grad_norm": 2.17746901512146,
      "learning_rate": 0.0001504,
      "loss": 0.2346,
      "step": 1240
    },
    {
      "epoch": 5.0,
      "grad_norm": 4.807671546936035,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.1019,
      "step": 1250
    },
    {
      "epoch": 5.04,
      "grad_norm": 0.5939441919326782,
      "learning_rate": 0.0001496,
      "loss": 0.2222,
      "step": 1260
    },
    {
      "epoch": 5.08,
      "grad_norm": 6.009125232696533,
      "learning_rate": 0.0001492,
      "loss": 0.1772,
      "step": 1270
    },
    {
      "epoch": 5.12,
      "grad_norm": 0.48813578486442566,
      "learning_rate": 0.0001488,
      "loss": 0.1592,
      "step": 1280
    },
    {
      "epoch": 5.16,
      "grad_norm": 2.24546480178833,
      "learning_rate": 0.0001484,
      "loss": 0.2064,
      "step": 1290
    },
    {
      "epoch": 5.2,
      "grad_norm": 1.3820618391036987,
      "learning_rate": 0.000148,
      "loss": 0.193,
      "step": 1300
    },
    {
      "epoch": 5.2,
      "eval_accuracy": 0.741,
      "eval_loss": 0.8467380404472351,
      "eval_runtime": 6.1473,
      "eval_samples_per_second": 162.672,
      "eval_steps_per_second": 20.334,
      "step": 1300
    },
    {
      "epoch": 5.24,
      "grad_norm": 1.7147108316421509,
      "learning_rate": 0.0001476,
      "loss": 0.1436,
      "step": 1310
    },
    {
      "epoch": 5.28,
      "grad_norm": 0.31574681401252747,
      "learning_rate": 0.0001472,
      "loss": 0.1127,
      "step": 1320
    },
    {
      "epoch": 5.32,
      "grad_norm": 0.05109202861785889,
      "learning_rate": 0.00014680000000000002,
      "loss": 0.0844,
      "step": 1330
    },
    {
      "epoch": 5.36,
      "grad_norm": 3.080343246459961,
      "learning_rate": 0.0001464,
      "loss": 0.1685,
      "step": 1340
    },
    {
      "epoch": 5.4,
      "grad_norm": 0.8458659648895264,
      "learning_rate": 0.000146,
      "loss": 0.0856,
      "step": 1350
    },
    {
      "epoch": 5.44,
      "grad_norm": 2.8861546516418457,
      "learning_rate": 0.00014560000000000002,
      "loss": 0.1465,
      "step": 1360
    },
    {
      "epoch": 5.48,
      "grad_norm": 0.9648165702819824,
      "learning_rate": 0.0001452,
      "loss": 0.1534,
      "step": 1370
    },
    {
      "epoch": 5.52,
      "grad_norm": 0.8869705200195312,
      "learning_rate": 0.0001448,
      "loss": 0.1499,
      "step": 1380
    },
    {
      "epoch": 5.5600000000000005,
      "grad_norm": 5.056468963623047,
      "learning_rate": 0.0001444,
      "loss": 0.3252,
      "step": 1390
    },
    {
      "epoch": 5.6,
      "grad_norm": 4.4490485191345215,
      "learning_rate": 0.000144,
      "loss": 0.1831,
      "step": 1400
    },
    {
      "epoch": 5.6,
      "eval_accuracy": 0.774,
      "eval_loss": 0.6975213885307312,
      "eval_runtime": 5.5981,
      "eval_samples_per_second": 178.633,
      "eval_steps_per_second": 22.329,
      "step": 1400
    },
    {
      "epoch": 5.64,
      "grad_norm": 5.352523326873779,
      "learning_rate": 0.0001436,
      "loss": 0.1022,
      "step": 1410
    },
    {
      "epoch": 5.68,
      "grad_norm": 0.33864977955818176,
      "learning_rate": 0.0001432,
      "loss": 0.2627,
      "step": 1420
    },
    {
      "epoch": 5.72,
      "grad_norm": 3.7317004203796387,
      "learning_rate": 0.0001428,
      "loss": 0.1334,
      "step": 1430
    },
    {
      "epoch": 5.76,
      "grad_norm": 3.7961854934692383,
      "learning_rate": 0.0001424,
      "loss": 0.1658,
      "step": 1440
    },
    {
      "epoch": 5.8,
      "grad_norm": 0.13848626613616943,
      "learning_rate": 0.000142,
      "loss": 0.069,
      "step": 1450
    },
    {
      "epoch": 5.84,
      "grad_norm": 4.271240234375,
      "learning_rate": 0.0001416,
      "loss": 0.1401,
      "step": 1460
    },
    {
      "epoch": 5.88,
      "grad_norm": 3.202373504638672,
      "learning_rate": 0.0001412,
      "loss": 0.1468,
      "step": 1470
    },
    {
      "epoch": 5.92,
      "grad_norm": 0.39670008420944214,
      "learning_rate": 0.0001408,
      "loss": 0.0742,
      "step": 1480
    },
    {
      "epoch": 5.96,
      "grad_norm": 3.4956166744232178,
      "learning_rate": 0.0001404,
      "loss": 0.0929,
      "step": 1490
    },
    {
      "epoch": 6.0,
      "grad_norm": 3.5953352451324463,
      "learning_rate": 0.00014,
      "loss": 0.2612,
      "step": 1500
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.775,
      "eval_loss": 0.8718728423118591,
      "eval_runtime": 5.287,
      "eval_samples_per_second": 189.142,
      "eval_steps_per_second": 23.643,
      "step": 1500
    },
    {
      "epoch": 6.04,
      "grad_norm": 0.334338515996933,
      "learning_rate": 0.0001396,
      "loss": 0.0812,
      "step": 1510
    },
    {
      "epoch": 6.08,
      "grad_norm": 4.4882612228393555,
      "learning_rate": 0.0001392,
      "loss": 0.0565,
      "step": 1520
    },
    {
      "epoch": 6.12,
      "grad_norm": 6.175742149353027,
      "learning_rate": 0.00013879999999999999,
      "loss": 0.0966,
      "step": 1530
    },
    {
      "epoch": 6.16,
      "grad_norm": 3.152432680130005,
      "learning_rate": 0.0001384,
      "loss": 0.1777,
      "step": 1540
    },
    {
      "epoch": 6.2,
      "grad_norm": 0.0891793891787529,
      "learning_rate": 0.000138,
      "loss": 0.0406,
      "step": 1550
    },
    {
      "epoch": 6.24,
      "grad_norm": 11.810345649719238,
      "learning_rate": 0.00013759999999999998,
      "loss": 0.0862,
      "step": 1560
    },
    {
      "epoch": 6.28,
      "grad_norm": 0.12118436396121979,
      "learning_rate": 0.00013720000000000003,
      "loss": 0.0366,
      "step": 1570
    },
    {
      "epoch": 6.32,
      "grad_norm": 12.970296859741211,
      "learning_rate": 0.00013680000000000002,
      "loss": 0.0596,
      "step": 1580
    },
    {
      "epoch": 6.36,
      "grad_norm": 0.05544751510024071,
      "learning_rate": 0.0001364,
      "loss": 0.1192,
      "step": 1590
    },
    {
      "epoch": 6.4,
      "grad_norm": 0.16358701884746552,
      "learning_rate": 0.00013600000000000003,
      "loss": 0.102,
      "step": 1600
    },
    {
      "epoch": 6.4,
      "eval_accuracy": 0.788,
      "eval_loss": 0.9045199155807495,
      "eval_runtime": 5.1633,
      "eval_samples_per_second": 193.676,
      "eval_steps_per_second": 24.21,
      "step": 1600
    },
    {
      "epoch": 6.44,
      "grad_norm": 2.016242742538452,
      "learning_rate": 0.00013560000000000002,
      "loss": 0.0633,
      "step": 1610
    },
    {
      "epoch": 6.48,
      "grad_norm": 0.19605639576911926,
      "learning_rate": 0.0001352,
      "loss": 0.0223,
      "step": 1620
    },
    {
      "epoch": 6.52,
      "grad_norm": 0.13813719153404236,
      "learning_rate": 0.00013480000000000002,
      "loss": 0.1615,
      "step": 1630
    },
    {
      "epoch": 6.5600000000000005,
      "grad_norm": 0.18716655671596527,
      "learning_rate": 0.00013440000000000001,
      "loss": 0.0248,
      "step": 1640
    },
    {
      "epoch": 6.6,
      "grad_norm": 0.14301587641239166,
      "learning_rate": 0.000134,
      "loss": 0.0507,
      "step": 1650
    },
    {
      "epoch": 6.64,
      "grad_norm": 0.10028518736362457,
      "learning_rate": 0.00013360000000000002,
      "loss": 0.0334,
      "step": 1660
    },
    {
      "epoch": 6.68,
      "grad_norm": 6.466274261474609,
      "learning_rate": 0.0001332,
      "loss": 0.0881,
      "step": 1670
    },
    {
      "epoch": 6.72,
      "grad_norm": 0.15147142112255096,
      "learning_rate": 0.0001328,
      "loss": 0.1162,
      "step": 1680
    },
    {
      "epoch": 6.76,
      "grad_norm": 0.06184665858745575,
      "learning_rate": 0.00013240000000000002,
      "loss": 0.085,
      "step": 1690
    },
    {
      "epoch": 6.8,
      "grad_norm": 9.578435897827148,
      "learning_rate": 0.000132,
      "loss": 0.1029,
      "step": 1700
    },
    {
      "epoch": 6.8,
      "eval_accuracy": 0.783,
      "eval_loss": 0.9655129313468933,
      "eval_runtime": 5.9259,
      "eval_samples_per_second": 168.75,
      "eval_steps_per_second": 21.094,
      "step": 1700
    },
    {
      "epoch": 6.84,
      "grad_norm": 9.377839088439941,
      "learning_rate": 0.0001316,
      "loss": 0.0559,
      "step": 1710
    },
    {
      "epoch": 6.88,
      "grad_norm": 9.196260452270508,
      "learning_rate": 0.00013120000000000002,
      "loss": 0.0919,
      "step": 1720
    },
    {
      "epoch": 6.92,
      "grad_norm": 0.6224649548530579,
      "learning_rate": 0.0001308,
      "loss": 0.0328,
      "step": 1730
    },
    {
      "epoch": 6.96,
      "grad_norm": 0.8349862098693848,
      "learning_rate": 0.0001304,
      "loss": 0.0837,
      "step": 1740
    },
    {
      "epoch": 7.0,
      "grad_norm": 7.744389057159424,
      "learning_rate": 0.00013000000000000002,
      "loss": 0.1493,
      "step": 1750
    },
    {
      "epoch": 7.04,
      "grad_norm": 19.332578659057617,
      "learning_rate": 0.0001296,
      "loss": 0.1517,
      "step": 1760
    },
    {
      "epoch": 7.08,
      "grad_norm": 9.477415084838867,
      "learning_rate": 0.00012920000000000002,
      "loss": 0.0612,
      "step": 1770
    },
    {
      "epoch": 7.12,
      "grad_norm": 0.04479631036520004,
      "learning_rate": 0.00012880000000000001,
      "loss": 0.0629,
      "step": 1780
    },
    {
      "epoch": 7.16,
      "grad_norm": 0.03189521282911301,
      "learning_rate": 0.0001284,
      "loss": 0.0544,
      "step": 1790
    },
    {
      "epoch": 7.2,
      "grad_norm": 0.04392462968826294,
      "learning_rate": 0.00012800000000000002,
      "loss": 0.0735,
      "step": 1800
    },
    {
      "epoch": 7.2,
      "eval_accuracy": 0.78,
      "eval_loss": 0.9905818700790405,
      "eval_runtime": 5.871,
      "eval_samples_per_second": 170.328,
      "eval_steps_per_second": 21.291,
      "step": 1800
    },
    {
      "epoch": 7.24,
      "grad_norm": 0.06359164416790009,
      "learning_rate": 0.0001276,
      "loss": 0.1629,
      "step": 1810
    },
    {
      "epoch": 7.28,
      "grad_norm": 1.7004547119140625,
      "learning_rate": 0.0001272,
      "loss": 0.1058,
      "step": 1820
    },
    {
      "epoch": 7.32,
      "grad_norm": 0.6779342293739319,
      "learning_rate": 0.00012680000000000002,
      "loss": 0.0404,
      "step": 1830
    },
    {
      "epoch": 7.36,
      "grad_norm": 0.050379179418087006,
      "learning_rate": 0.0001264,
      "loss": 0.0565,
      "step": 1840
    },
    {
      "epoch": 7.4,
      "grad_norm": 0.23603555560112,
      "learning_rate": 0.000126,
      "loss": 0.0316,
      "step": 1850
    },
    {
      "epoch": 7.44,
      "grad_norm": 0.030859237536787987,
      "learning_rate": 0.00012560000000000002,
      "loss": 0.0781,
      "step": 1860
    },
    {
      "epoch": 7.48,
      "grad_norm": 0.14111267030239105,
      "learning_rate": 0.0001252,
      "loss": 0.1127,
      "step": 1870
    },
    {
      "epoch": 7.52,
      "grad_norm": 5.0431084632873535,
      "learning_rate": 0.0001248,
      "loss": 0.0226,
      "step": 1880
    },
    {
      "epoch": 7.5600000000000005,
      "grad_norm": 0.1641249656677246,
      "learning_rate": 0.00012440000000000002,
      "loss": 0.0875,
      "step": 1890
    },
    {
      "epoch": 7.6,
      "grad_norm": 1.8398263454437256,
      "learning_rate": 0.000124,
      "loss": 0.0715,
      "step": 1900
    },
    {
      "epoch": 7.6,
      "eval_accuracy": 0.787,
      "eval_loss": 0.8893071413040161,
      "eval_runtime": 5.6004,
      "eval_samples_per_second": 178.559,
      "eval_steps_per_second": 22.32,
      "step": 1900
    },
    {
      "epoch": 7.64,
      "grad_norm": 11.241652488708496,
      "learning_rate": 0.0001236,
      "loss": 0.0418,
      "step": 1910
    },
    {
      "epoch": 7.68,
      "grad_norm": 1.0991768836975098,
      "learning_rate": 0.0001232,
      "loss": 0.0755,
      "step": 1920
    },
    {
      "epoch": 7.72,
      "grad_norm": 19.279315948486328,
      "learning_rate": 0.0001228,
      "loss": 0.2096,
      "step": 1930
    },
    {
      "epoch": 7.76,
      "grad_norm": 0.7445815801620483,
      "learning_rate": 0.0001224,
      "loss": 0.0395,
      "step": 1940
    },
    {
      "epoch": 7.8,
      "grad_norm": 0.07692953944206238,
      "learning_rate": 0.000122,
      "loss": 0.0305,
      "step": 1950
    },
    {
      "epoch": 7.84,
      "grad_norm": 0.1866171956062317,
      "learning_rate": 0.0001216,
      "loss": 0.0379,
      "step": 1960
    },
    {
      "epoch": 7.88,
      "grad_norm": 11.213932037353516,
      "learning_rate": 0.0001212,
      "loss": 0.0625,
      "step": 1970
    },
    {
      "epoch": 7.92,
      "grad_norm": 0.762454092502594,
      "learning_rate": 0.0001208,
      "loss": 0.0336,
      "step": 1980
    },
    {
      "epoch": 7.96,
      "grad_norm": 9.390318870544434,
      "learning_rate": 0.0001204,
      "loss": 0.0813,
      "step": 1990
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.019455738365650177,
      "learning_rate": 0.00012,
      "loss": 0.1254,
      "step": 2000
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.761,
      "eval_loss": 1.1220637559890747,
      "eval_runtime": 5.5365,
      "eval_samples_per_second": 180.619,
      "eval_steps_per_second": 22.577,
      "step": 2000
    },
    {
      "epoch": 8.04,
      "grad_norm": 0.1315418779850006,
      "learning_rate": 0.00011960000000000001,
      "loss": 0.1103,
      "step": 2010
    },
    {
      "epoch": 8.08,
      "grad_norm": 0.4731353521347046,
      "learning_rate": 0.0001192,
      "loss": 0.0696,
      "step": 2020
    },
    {
      "epoch": 8.12,
      "grad_norm": 2.56032133102417,
      "learning_rate": 0.0001188,
      "loss": 0.0488,
      "step": 2030
    },
    {
      "epoch": 8.16,
      "grad_norm": 1.2172237634658813,
      "learning_rate": 0.0001184,
      "loss": 0.031,
      "step": 2040
    },
    {
      "epoch": 8.2,
      "grad_norm": 0.04160616546869278,
      "learning_rate": 0.000118,
      "loss": 0.0332,
      "step": 2050
    },
    {
      "epoch": 8.24,
      "grad_norm": 0.02498347871005535,
      "learning_rate": 0.0001176,
      "loss": 0.0046,
      "step": 2060
    },
    {
      "epoch": 8.28,
      "grad_norm": 0.023147333413362503,
      "learning_rate": 0.0001172,
      "loss": 0.0354,
      "step": 2070
    },
    {
      "epoch": 8.32,
      "grad_norm": 24.763408660888672,
      "learning_rate": 0.00011679999999999999,
      "loss": 0.0182,
      "step": 2080
    },
    {
      "epoch": 8.36,
      "grad_norm": 0.1869715005159378,
      "learning_rate": 0.0001164,
      "loss": 0.0198,
      "step": 2090
    },
    {
      "epoch": 8.4,
      "grad_norm": 0.013836881145834923,
      "learning_rate": 0.000116,
      "loss": 0.021,
      "step": 2100
    },
    {
      "epoch": 8.4,
      "eval_accuracy": 0.779,
      "eval_loss": 1.1648471355438232,
      "eval_runtime": 5.2413,
      "eval_samples_per_second": 190.794,
      "eval_steps_per_second": 23.849,
      "step": 2100
    },
    {
      "epoch": 8.44,
      "grad_norm": 0.022183042019605637,
      "learning_rate": 0.00011559999999999999,
      "loss": 0.0551,
      "step": 2110
    },
    {
      "epoch": 8.48,
      "grad_norm": 7.832488536834717,
      "learning_rate": 0.0001152,
      "loss": 0.1164,
      "step": 2120
    },
    {
      "epoch": 8.52,
      "grad_norm": 0.02375981956720352,
      "learning_rate": 0.0001148,
      "loss": 0.0164,
      "step": 2130
    },
    {
      "epoch": 8.56,
      "grad_norm": 0.018566014245152473,
      "learning_rate": 0.0001144,
      "loss": 0.0155,
      "step": 2140
    },
    {
      "epoch": 8.6,
      "grad_norm": 2.625162124633789,
      "learning_rate": 0.00011399999999999999,
      "loss": 0.0702,
      "step": 2150
    },
    {
      "epoch": 8.64,
      "grad_norm": 0.022067397832870483,
      "learning_rate": 0.0001136,
      "loss": 0.0441,
      "step": 2160
    },
    {
      "epoch": 8.68,
      "grad_norm": 0.030472321435809135,
      "learning_rate": 0.0001132,
      "loss": 0.012,
      "step": 2170
    },
    {
      "epoch": 8.72,
      "grad_norm": 0.957785964012146,
      "learning_rate": 0.00011279999999999999,
      "loss": 0.0151,
      "step": 2180
    },
    {
      "epoch": 8.76,
      "grad_norm": 3.766643524169922,
      "learning_rate": 0.00011240000000000002,
      "loss": 0.0853,
      "step": 2190
    },
    {
      "epoch": 8.8,
      "grad_norm": 0.32392239570617676,
      "learning_rate": 0.00011200000000000001,
      "loss": 0.0133,
      "step": 2200
    },
    {
      "epoch": 8.8,
      "eval_accuracy": 0.806,
      "eval_loss": 0.9857252240180969,
      "eval_runtime": 5.2301,
      "eval_samples_per_second": 191.201,
      "eval_steps_per_second": 23.9,
      "step": 2200
    },
    {
      "epoch": 8.84,
      "grad_norm": 0.021277396008372307,
      "learning_rate": 0.00011160000000000002,
      "loss": 0.0314,
      "step": 2210
    },
    {
      "epoch": 8.88,
      "grad_norm": 13.449085235595703,
      "learning_rate": 0.00011120000000000002,
      "loss": 0.0629,
      "step": 2220
    },
    {
      "epoch": 8.92,
      "grad_norm": 0.02039525657892227,
      "learning_rate": 0.00011080000000000001,
      "loss": 0.0358,
      "step": 2230
    },
    {
      "epoch": 8.96,
      "grad_norm": 0.10957217961549759,
      "learning_rate": 0.00011040000000000001,
      "loss": 0.0379,
      "step": 2240
    },
    {
      "epoch": 9.0,
      "grad_norm": 0.016647523269057274,
      "learning_rate": 0.00011000000000000002,
      "loss": 0.0125,
      "step": 2250
    },
    {
      "epoch": 9.04,
      "grad_norm": 13.669293403625488,
      "learning_rate": 0.00010960000000000001,
      "loss": 0.035,
      "step": 2260
    },
    {
      "epoch": 9.08,
      "grad_norm": 0.012543587014079094,
      "learning_rate": 0.00010920000000000001,
      "loss": 0.031,
      "step": 2270
    },
    {
      "epoch": 9.12,
      "grad_norm": 0.011763609014451504,
      "learning_rate": 0.00010880000000000002,
      "loss": 0.0299,
      "step": 2280
    },
    {
      "epoch": 9.16,
      "grad_norm": 0.0155358100309968,
      "learning_rate": 0.00010840000000000002,
      "loss": 0.0339,
      "step": 2290
    },
    {
      "epoch": 9.2,
      "grad_norm": 0.2667885720729828,
      "learning_rate": 0.00010800000000000001,
      "loss": 0.0086,
      "step": 2300
    },
    {
      "epoch": 9.2,
      "eval_accuracy": 0.799,
      "eval_loss": 1.036458134651184,
      "eval_runtime": 6.0166,
      "eval_samples_per_second": 166.207,
      "eval_steps_per_second": 20.776,
      "step": 2300
    },
    {
      "epoch": 9.24,
      "grad_norm": 0.008514340966939926,
      "learning_rate": 0.00010760000000000001,
      "loss": 0.0177,
      "step": 2310
    },
    {
      "epoch": 9.28,
      "grad_norm": 0.033006951212882996,
      "learning_rate": 0.00010720000000000002,
      "loss": 0.0041,
      "step": 2320
    },
    {
      "epoch": 9.32,
      "grad_norm": 17.495553970336914,
      "learning_rate": 0.00010680000000000001,
      "loss": 0.0506,
      "step": 2330
    },
    {
      "epoch": 9.36,
      "grad_norm": 29.127471923828125,
      "learning_rate": 0.00010640000000000001,
      "loss": 0.0216,
      "step": 2340
    },
    {
      "epoch": 9.4,
      "grad_norm": 0.03006252832710743,
      "learning_rate": 0.00010600000000000002,
      "loss": 0.0273,
      "step": 2350
    },
    {
      "epoch": 9.44,
      "grad_norm": 0.021872328594326973,
      "learning_rate": 0.0001056,
      "loss": 0.0027,
      "step": 2360
    },
    {
      "epoch": 9.48,
      "grad_norm": 0.01040344312787056,
      "learning_rate": 0.00010520000000000001,
      "loss": 0.018,
      "step": 2370
    },
    {
      "epoch": 9.52,
      "grad_norm": 9.743282318115234,
      "learning_rate": 0.00010480000000000001,
      "loss": 0.1394,
      "step": 2380
    },
    {
      "epoch": 9.56,
      "grad_norm": 0.012420209124684334,
      "learning_rate": 0.0001044,
      "loss": 0.0257,
      "step": 2390
    },
    {
      "epoch": 9.6,
      "grad_norm": 0.21533189713954926,
      "learning_rate": 0.00010400000000000001,
      "loss": 0.0223,
      "step": 2400
    },
    {
      "epoch": 9.6,
      "eval_accuracy": 0.812,
      "eval_loss": 0.982612669467926,
      "eval_runtime": 5.176,
      "eval_samples_per_second": 193.201,
      "eval_steps_per_second": 24.15,
      "step": 2400
    },
    {
      "epoch": 9.64,
      "grad_norm": 0.02012082375586033,
      "learning_rate": 0.00010360000000000001,
      "loss": 0.0371,
      "step": 2410
    },
    {
      "epoch": 9.68,
      "grad_norm": 0.03367308899760246,
      "learning_rate": 0.0001032,
      "loss": 0.0053,
      "step": 2420
    },
    {
      "epoch": 9.72,
      "grad_norm": 0.9374167919158936,
      "learning_rate": 0.0001028,
      "loss": 0.058,
      "step": 2430
    },
    {
      "epoch": 9.76,
      "grad_norm": 0.010282558389008045,
      "learning_rate": 0.00010240000000000001,
      "loss": 0.0438,
      "step": 2440
    },
    {
      "epoch": 9.8,
      "grad_norm": 0.009324116632342339,
      "learning_rate": 0.00010200000000000001,
      "loss": 0.0028,
      "step": 2450
    },
    {
      "epoch": 9.84,
      "grad_norm": 0.009123372845351696,
      "learning_rate": 0.0001016,
      "loss": 0.1,
      "step": 2460
    },
    {
      "epoch": 9.88,
      "grad_norm": 0.010366716422140598,
      "learning_rate": 0.00010120000000000001,
      "loss": 0.1172,
      "step": 2470
    },
    {
      "epoch": 9.92,
      "grad_norm": 3.760880708694458,
      "learning_rate": 0.00010080000000000001,
      "loss": 0.005,
      "step": 2480
    },
    {
      "epoch": 9.96,
      "grad_norm": 0.033989980816841125,
      "learning_rate": 0.0001004,
      "loss": 0.0441,
      "step": 2490
    },
    {
      "epoch": 10.0,
      "grad_norm": 0.008832584135234356,
      "learning_rate": 0.0001,
      "loss": 0.0023,
      "step": 2500
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.795,
      "eval_loss": 1.0696669816970825,
      "eval_runtime": 5.262,
      "eval_samples_per_second": 190.041,
      "eval_steps_per_second": 23.755,
      "step": 2500
    },
    {
      "epoch": 10.04,
      "grad_norm": 0.06487290561199188,
      "learning_rate": 9.960000000000001e-05,
      "loss": 0.0359,
      "step": 2510
    },
    {
      "epoch": 10.08,
      "grad_norm": 0.06183789297938347,
      "learning_rate": 9.92e-05,
      "loss": 0.0019,
      "step": 2520
    },
    {
      "epoch": 10.12,
      "grad_norm": 0.007254609372466803,
      "learning_rate": 9.88e-05,
      "loss": 0.0443,
      "step": 2530
    },
    {
      "epoch": 10.16,
      "grad_norm": 27.718936920166016,
      "learning_rate": 9.84e-05,
      "loss": 0.0429,
      "step": 2540
    },
    {
      "epoch": 10.2,
      "grad_norm": 0.007617979310452938,
      "learning_rate": 9.8e-05,
      "loss": 0.0034,
      "step": 2550
    },
    {
      "epoch": 10.24,
      "grad_norm": 0.42735064029693604,
      "learning_rate": 9.76e-05,
      "loss": 0.028,
      "step": 2560
    },
    {
      "epoch": 10.28,
      "grad_norm": 0.00830338429659605,
      "learning_rate": 9.72e-05,
      "loss": 0.0049,
      "step": 2570
    },
    {
      "epoch": 10.32,
      "grad_norm": 0.007815328426659107,
      "learning_rate": 9.680000000000001e-05,
      "loss": 0.0015,
      "step": 2580
    },
    {
      "epoch": 10.36,
      "grad_norm": 0.023354820907115936,
      "learning_rate": 9.64e-05,
      "loss": 0.0017,
      "step": 2590
    },
    {
      "epoch": 10.4,
      "grad_norm": 0.02857878804206848,
      "learning_rate": 9.6e-05,
      "loss": 0.0021,
      "step": 2600
    },
    {
      "epoch": 10.4,
      "eval_accuracy": 0.815,
      "eval_loss": 1.0489550828933716,
      "eval_runtime": 5.2508,
      "eval_samples_per_second": 190.447,
      "eval_steps_per_second": 23.806,
      "step": 2600
    },
    {
      "epoch": 10.44,
      "grad_norm": 0.020630279555916786,
      "learning_rate": 9.56e-05,
      "loss": 0.0337,
      "step": 2610
    },
    {
      "epoch": 10.48,
      "grad_norm": 0.2516959607601166,
      "learning_rate": 9.52e-05,
      "loss": 0.0099,
      "step": 2620
    },
    {
      "epoch": 10.52,
      "grad_norm": 0.09867813438177109,
      "learning_rate": 9.48e-05,
      "loss": 0.0258,
      "step": 2630
    },
    {
      "epoch": 10.56,
      "grad_norm": 0.005374481901526451,
      "learning_rate": 9.44e-05,
      "loss": 0.0014,
      "step": 2640
    },
    {
      "epoch": 10.6,
      "grad_norm": 0.012418796308338642,
      "learning_rate": 9.4e-05,
      "loss": 0.0052,
      "step": 2650
    },
    {
      "epoch": 10.64,
      "grad_norm": 0.006738652009516954,
      "learning_rate": 9.360000000000001e-05,
      "loss": 0.0015,
      "step": 2660
    },
    {
      "epoch": 10.68,
      "grad_norm": 3.031865358352661,
      "learning_rate": 9.320000000000002e-05,
      "loss": 0.0023,
      "step": 2670
    },
    {
      "epoch": 10.72,
      "grad_norm": 0.007892146706581116,
      "learning_rate": 9.28e-05,
      "loss": 0.0014,
      "step": 2680
    },
    {
      "epoch": 10.76,
      "grad_norm": 0.008096442557871342,
      "learning_rate": 9.240000000000001e-05,
      "loss": 0.0321,
      "step": 2690
    },
    {
      "epoch": 10.8,
      "grad_norm": 10.601773262023926,
      "learning_rate": 9.200000000000001e-05,
      "loss": 0.0401,
      "step": 2700
    },
    {
      "epoch": 10.8,
      "eval_accuracy": 0.8,
      "eval_loss": 1.1593725681304932,
      "eval_runtime": 5.7948,
      "eval_samples_per_second": 172.569,
      "eval_steps_per_second": 21.571,
      "step": 2700
    },
    {
      "epoch": 10.84,
      "grad_norm": 0.00845256820321083,
      "learning_rate": 9.16e-05,
      "loss": 0.0016,
      "step": 2710
    },
    {
      "epoch": 10.88,
      "grad_norm": 0.016398636624217033,
      "learning_rate": 9.120000000000001e-05,
      "loss": 0.0108,
      "step": 2720
    },
    {
      "epoch": 10.92,
      "grad_norm": 0.015316460281610489,
      "learning_rate": 9.080000000000001e-05,
      "loss": 0.0394,
      "step": 2730
    },
    {
      "epoch": 10.96,
      "grad_norm": 0.35788559913635254,
      "learning_rate": 9.04e-05,
      "loss": 0.0017,
      "step": 2740
    },
    {
      "epoch": 11.0,
      "grad_norm": 0.011767162941396236,
      "learning_rate": 9e-05,
      "loss": 0.0047,
      "step": 2750
    },
    {
      "epoch": 11.04,
      "grad_norm": 0.007368568331003189,
      "learning_rate": 8.960000000000001e-05,
      "loss": 0.0012,
      "step": 2760
    },
    {
      "epoch": 11.08,
      "grad_norm": 0.011666523292660713,
      "learning_rate": 8.92e-05,
      "loss": 0.0014,
      "step": 2770
    },
    {
      "epoch": 11.12,
      "grad_norm": 0.00521031254902482,
      "learning_rate": 8.88e-05,
      "loss": 0.0012,
      "step": 2780
    },
    {
      "epoch": 11.16,
      "grad_norm": 0.025624191388487816,
      "learning_rate": 8.840000000000001e-05,
      "loss": 0.0014,
      "step": 2790
    },
    {
      "epoch": 11.2,
      "grad_norm": 0.008872034959495068,
      "learning_rate": 8.800000000000001e-05,
      "loss": 0.0012,
      "step": 2800
    },
    {
      "epoch": 11.2,
      "eval_accuracy": 0.817,
      "eval_loss": 1.0810606479644775,
      "eval_runtime": 5.7202,
      "eval_samples_per_second": 174.82,
      "eval_steps_per_second": 21.853,
      "step": 2800
    },
    {
      "epoch": 11.24,
      "grad_norm": 0.006208410952240229,
      "learning_rate": 8.76e-05,
      "loss": 0.0014,
      "step": 2810
    },
    {
      "epoch": 11.28,
      "grad_norm": 0.006170503329485655,
      "learning_rate": 8.72e-05,
      "loss": 0.0011,
      "step": 2820
    },
    {
      "epoch": 11.32,
      "grad_norm": 0.005829779896885157,
      "learning_rate": 8.680000000000001e-05,
      "loss": 0.0015,
      "step": 2830
    },
    {
      "epoch": 11.36,
      "grad_norm": 0.0047014919109642506,
      "learning_rate": 8.64e-05,
      "loss": 0.001,
      "step": 2840
    },
    {
      "epoch": 11.4,
      "grad_norm": 0.01678207516670227,
      "learning_rate": 8.6e-05,
      "loss": 0.0019,
      "step": 2850
    },
    {
      "epoch": 11.44,
      "grad_norm": 0.004566165152937174,
      "learning_rate": 8.560000000000001e-05,
      "loss": 0.0014,
      "step": 2860
    },
    {
      "epoch": 11.48,
      "grad_norm": 0.005536119919270277,
      "learning_rate": 8.52e-05,
      "loss": 0.001,
      "step": 2870
    },
    {
      "epoch": 11.52,
      "grad_norm": 0.005134043283760548,
      "learning_rate": 8.48e-05,
      "loss": 0.001,
      "step": 2880
    },
    {
      "epoch": 11.56,
      "grad_norm": 0.005173493642359972,
      "learning_rate": 8.44e-05,
      "loss": 0.001,
      "step": 2890
    },
    {
      "epoch": 11.6,
      "grad_norm": 0.0057559204287827015,
      "learning_rate": 8.4e-05,
      "loss": 0.0034,
      "step": 2900
    },
    {
      "epoch": 11.6,
      "eval_accuracy": 0.825,
      "eval_loss": 1.0955668687820435,
      "eval_runtime": 5.4405,
      "eval_samples_per_second": 183.805,
      "eval_steps_per_second": 22.976,
      "step": 2900
    },
    {
      "epoch": 11.64,
      "grad_norm": 0.004013891331851482,
      "learning_rate": 8.36e-05,
      "loss": 0.0009,
      "step": 2910
    },
    {
      "epoch": 11.68,
      "grad_norm": 4.541440486907959,
      "learning_rate": 8.32e-05,
      "loss": 0.0744,
      "step": 2920
    },
    {
      "epoch": 11.72,
      "grad_norm": 0.004702790640294552,
      "learning_rate": 8.28e-05,
      "loss": 0.001,
      "step": 2930
    },
    {
      "epoch": 11.76,
      "grad_norm": 0.004574855323880911,
      "learning_rate": 8.24e-05,
      "loss": 0.0028,
      "step": 2940
    },
    {
      "epoch": 11.8,
      "grad_norm": 0.006892206612974405,
      "learning_rate": 8.2e-05,
      "loss": 0.0031,
      "step": 2950
    },
    {
      "epoch": 11.84,
      "grad_norm": 0.28937533497810364,
      "learning_rate": 8.16e-05,
      "loss": 0.002,
      "step": 2960
    },
    {
      "epoch": 11.88,
      "grad_norm": 0.004892553202807903,
      "learning_rate": 8.120000000000001e-05,
      "loss": 0.0009,
      "step": 2970
    },
    {
      "epoch": 11.92,
      "grad_norm": 0.00817884597927332,
      "learning_rate": 8.080000000000001e-05,
      "loss": 0.0012,
      "step": 2980
    },
    {
      "epoch": 11.96,
      "grad_norm": 0.007845502346754074,
      "learning_rate": 8.04e-05,
      "loss": 0.0169,
      "step": 2990
    },
    {
      "epoch": 12.0,
      "grad_norm": 0.004683198872953653,
      "learning_rate": 8e-05,
      "loss": 0.0012,
      "step": 3000
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.808,
      "eval_loss": 1.200992226600647,
      "eval_runtime": 5.2017,
      "eval_samples_per_second": 192.245,
      "eval_steps_per_second": 24.031,
      "step": 3000
    },
    {
      "epoch": 12.04,
      "grad_norm": 0.004240355920046568,
      "learning_rate": 7.960000000000001e-05,
      "loss": 0.001,
      "step": 3010
    },
    {
      "epoch": 12.08,
      "grad_norm": 0.04097750037908554,
      "learning_rate": 7.920000000000001e-05,
      "loss": 0.001,
      "step": 3020
    },
    {
      "epoch": 12.12,
      "grad_norm": 0.036695607006549835,
      "learning_rate": 7.88e-05,
      "loss": 0.0009,
      "step": 3030
    },
    {
      "epoch": 12.16,
      "grad_norm": 0.003702098038047552,
      "learning_rate": 7.840000000000001e-05,
      "loss": 0.0009,
      "step": 3040
    },
    {
      "epoch": 12.2,
      "grad_norm": 10.573301315307617,
      "learning_rate": 7.800000000000001e-05,
      "loss": 0.0321,
      "step": 3050
    },
    {
      "epoch": 12.24,
      "grad_norm": 0.004409339744597673,
      "learning_rate": 7.76e-05,
      "loss": 0.0008,
      "step": 3060
    },
    {
      "epoch": 12.28,
      "grad_norm": 0.004956026095896959,
      "learning_rate": 7.72e-05,
      "loss": 0.0008,
      "step": 3070
    },
    {
      "epoch": 12.32,
      "grad_norm": 0.006153183523565531,
      "learning_rate": 7.680000000000001e-05,
      "loss": 0.0009,
      "step": 3080
    },
    {
      "epoch": 12.36,
      "grad_norm": 0.005817431956529617,
      "learning_rate": 7.64e-05,
      "loss": 0.0009,
      "step": 3090
    },
    {
      "epoch": 12.4,
      "grad_norm": 0.004047242924571037,
      "learning_rate": 7.6e-05,
      "loss": 0.0011,
      "step": 3100
    },
    {
      "epoch": 12.4,
      "eval_accuracy": 0.81,
      "eval_loss": 1.1712357997894287,
      "eval_runtime": 5.9454,
      "eval_samples_per_second": 168.198,
      "eval_steps_per_second": 21.025,
      "step": 3100
    },
    {
      "epoch": 12.44,
      "grad_norm": 0.05110397934913635,
      "learning_rate": 7.560000000000001e-05,
      "loss": 0.0009,
      "step": 3110
    },
    {
      "epoch": 12.48,
      "grad_norm": 0.005057331640273333,
      "learning_rate": 7.52e-05,
      "loss": 0.0008,
      "step": 3120
    },
    {
      "epoch": 12.52,
      "grad_norm": 0.00844608899205923,
      "learning_rate": 7.48e-05,
      "loss": 0.0008,
      "step": 3130
    },
    {
      "epoch": 12.56,
      "grad_norm": 0.004293734673410654,
      "learning_rate": 7.44e-05,
      "loss": 0.0011,
      "step": 3140
    },
    {
      "epoch": 12.6,
      "grad_norm": 0.004264953080564737,
      "learning_rate": 7.4e-05,
      "loss": 0.0008,
      "step": 3150
    },
    {
      "epoch": 12.64,
      "grad_norm": 0.021664146333932877,
      "learning_rate": 7.36e-05,
      "loss": 0.0432,
      "step": 3160
    },
    {
      "epoch": 12.68,
      "grad_norm": 0.004172308370471001,
      "learning_rate": 7.32e-05,
      "loss": 0.0008,
      "step": 3170
    },
    {
      "epoch": 12.72,
      "grad_norm": 0.006536141969263554,
      "learning_rate": 7.280000000000001e-05,
      "loss": 0.0008,
      "step": 3180
    },
    {
      "epoch": 12.76,
      "grad_norm": 0.0048880744725465775,
      "learning_rate": 7.24e-05,
      "loss": 0.0008,
      "step": 3190
    },
    {
      "epoch": 12.8,
      "grad_norm": 0.004492601379752159,
      "learning_rate": 7.2e-05,
      "loss": 0.0092,
      "step": 3200
    },
    {
      "epoch": 12.8,
      "eval_accuracy": 0.813,
      "eval_loss": 1.1813948154449463,
      "eval_runtime": 5.2063,
      "eval_samples_per_second": 192.074,
      "eval_steps_per_second": 24.009,
      "step": 3200
    },
    {
      "epoch": 12.84,
      "grad_norm": 0.003316229674965143,
      "learning_rate": 7.16e-05,
      "loss": 0.0008,
      "step": 3210
    },
    {
      "epoch": 12.88,
      "grad_norm": 0.004760683514177799,
      "learning_rate": 7.12e-05,
      "loss": 0.0007,
      "step": 3220
    },
    {
      "epoch": 12.92,
      "grad_norm": 0.004011666867882013,
      "learning_rate": 7.08e-05,
      "loss": 0.0007,
      "step": 3230
    },
    {
      "epoch": 12.96,
      "grad_norm": 0.004005063325166702,
      "learning_rate": 7.04e-05,
      "loss": 0.0362,
      "step": 3240
    },
    {
      "epoch": 13.0,
      "grad_norm": 0.0050388905219733715,
      "learning_rate": 7e-05,
      "loss": 0.0008,
      "step": 3250
    },
    {
      "epoch": 13.04,
      "grad_norm": 0.01386964414268732,
      "learning_rate": 6.96e-05,
      "loss": 0.0009,
      "step": 3260
    },
    {
      "epoch": 13.08,
      "grad_norm": 0.004433898255228996,
      "learning_rate": 6.92e-05,
      "loss": 0.0008,
      "step": 3270
    },
    {
      "epoch": 13.12,
      "grad_norm": 0.004408930893987417,
      "learning_rate": 6.879999999999999e-05,
      "loss": 0.0008,
      "step": 3280
    },
    {
      "epoch": 13.16,
      "grad_norm": 0.004361864645034075,
      "learning_rate": 6.840000000000001e-05,
      "loss": 0.0007,
      "step": 3290
    },
    {
      "epoch": 13.2,
      "grad_norm": 0.010670163668692112,
      "learning_rate": 6.800000000000001e-05,
      "loss": 0.0007,
      "step": 3300
    },
    {
      "epoch": 13.2,
      "eval_accuracy": 0.818,
      "eval_loss": 1.1677281856536865,
      "eval_runtime": 5.3663,
      "eval_samples_per_second": 186.347,
      "eval_steps_per_second": 23.293,
      "step": 3300
    },
    {
      "epoch": 13.24,
      "grad_norm": 0.004130377899855375,
      "learning_rate": 6.76e-05,
      "loss": 0.0007,
      "step": 3310
    },
    {
      "epoch": 13.28,
      "grad_norm": 0.0031264375429600477,
      "learning_rate": 6.720000000000001e-05,
      "loss": 0.0007,
      "step": 3320
    },
    {
      "epoch": 13.32,
      "grad_norm": 0.004546809010207653,
      "learning_rate": 6.680000000000001e-05,
      "loss": 0.0007,
      "step": 3330
    },
    {
      "epoch": 13.36,
      "grad_norm": 0.0035698998253792524,
      "learning_rate": 6.64e-05,
      "loss": 0.0007,
      "step": 3340
    },
    {
      "epoch": 13.4,
      "grad_norm": 0.004544632509350777,
      "learning_rate": 6.6e-05,
      "loss": 0.0007,
      "step": 3350
    },
    {
      "epoch": 13.44,
      "grad_norm": 0.0035377303138375282,
      "learning_rate": 6.560000000000001e-05,
      "loss": 0.0006,
      "step": 3360
    },
    {
      "epoch": 13.48,
      "grad_norm": 0.003488726681098342,
      "learning_rate": 6.52e-05,
      "loss": 0.0007,
      "step": 3370
    },
    {
      "epoch": 13.52,
      "grad_norm": 0.016187245026230812,
      "learning_rate": 6.48e-05,
      "loss": 0.0007,
      "step": 3380
    },
    {
      "epoch": 13.56,
      "grad_norm": 0.0037824937608093023,
      "learning_rate": 6.440000000000001e-05,
      "loss": 0.0006,
      "step": 3390
    },
    {
      "epoch": 13.6,
      "grad_norm": 0.0033098619896918535,
      "learning_rate": 6.400000000000001e-05,
      "loss": 0.0007,
      "step": 3400
    },
    {
      "epoch": 13.6,
      "eval_accuracy": 0.818,
      "eval_loss": 1.172300934791565,
      "eval_runtime": 5.2344,
      "eval_samples_per_second": 191.045,
      "eval_steps_per_second": 23.881,
      "step": 3400
    },
    {
      "epoch": 13.64,
      "grad_norm": 0.0036126759368926287,
      "learning_rate": 6.36e-05,
      "loss": 0.0007,
      "step": 3410
    },
    {
      "epoch": 13.68,
      "grad_norm": 0.0037607806734740734,
      "learning_rate": 6.32e-05,
      "loss": 0.0006,
      "step": 3420
    },
    {
      "epoch": 13.72,
      "grad_norm": 0.002861644374206662,
      "learning_rate": 6.280000000000001e-05,
      "loss": 0.0006,
      "step": 3430
    },
    {
      "epoch": 13.76,
      "grad_norm": 0.0034740378614515066,
      "learning_rate": 6.24e-05,
      "loss": 0.0006,
      "step": 3440
    },
    {
      "epoch": 13.8,
      "grad_norm": 0.0031694734934717417,
      "learning_rate": 6.2e-05,
      "loss": 0.0006,
      "step": 3450
    },
    {
      "epoch": 13.84,
      "grad_norm": 0.0038870801217854023,
      "learning_rate": 6.16e-05,
      "loss": 0.0006,
      "step": 3460
    },
    {
      "epoch": 13.88,
      "grad_norm": 0.0030017481185495853,
      "learning_rate": 6.12e-05,
      "loss": 0.0006,
      "step": 3470
    },
    {
      "epoch": 13.92,
      "grad_norm": 0.004931191448122263,
      "learning_rate": 6.08e-05,
      "loss": 0.0006,
      "step": 3480
    },
    {
      "epoch": 13.96,
      "grad_norm": 0.0034845368936657906,
      "learning_rate": 6.04e-05,
      "loss": 0.0008,
      "step": 3490
    },
    {
      "epoch": 14.0,
      "grad_norm": 0.003087324323132634,
      "learning_rate": 6e-05,
      "loss": 0.0006,
      "step": 3500
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.821,
      "eval_loss": 1.1852285861968994,
      "eval_runtime": 5.8798,
      "eval_samples_per_second": 170.074,
      "eval_steps_per_second": 21.259,
      "step": 3500
    },
    {
      "epoch": 14.04,
      "grad_norm": 0.003496804041787982,
      "learning_rate": 5.96e-05,
      "loss": 0.0006,
      "step": 3510
    },
    {
      "epoch": 14.08,
      "grad_norm": 0.0027559709269553423,
      "learning_rate": 5.92e-05,
      "loss": 0.0006,
      "step": 3520
    },
    {
      "epoch": 14.12,
      "grad_norm": 0.0026562130078673363,
      "learning_rate": 5.88e-05,
      "loss": 0.0006,
      "step": 3530
    },
    {
      "epoch": 14.16,
      "grad_norm": 0.002727353246882558,
      "learning_rate": 5.8399999999999997e-05,
      "loss": 0.0006,
      "step": 3540
    },
    {
      "epoch": 14.2,
      "grad_norm": 0.02825564332306385,
      "learning_rate": 5.8e-05,
      "loss": 0.0006,
      "step": 3550
    },
    {
      "epoch": 14.24,
      "grad_norm": 0.0026983567513525486,
      "learning_rate": 5.76e-05,
      "loss": 0.0007,
      "step": 3560
    },
    {
      "epoch": 14.28,
      "grad_norm": 0.005061229225248098,
      "learning_rate": 5.72e-05,
      "loss": 0.0006,
      "step": 3570
    },
    {
      "epoch": 14.32,
      "grad_norm": 0.002966745523735881,
      "learning_rate": 5.68e-05,
      "loss": 0.0005,
      "step": 3580
    },
    {
      "epoch": 14.36,
      "grad_norm": 0.002947271103039384,
      "learning_rate": 5.6399999999999995e-05,
      "loss": 0.0006,
      "step": 3590
    },
    {
      "epoch": 14.4,
      "grad_norm": 0.0029176445677876472,
      "learning_rate": 5.6000000000000006e-05,
      "loss": 0.0005,
      "step": 3600
    },
    {
      "epoch": 14.4,
      "eval_accuracy": 0.82,
      "eval_loss": 1.1927635669708252,
      "eval_runtime": 5.2928,
      "eval_samples_per_second": 188.936,
      "eval_steps_per_second": 23.617,
      "step": 3600
    },
    {
      "epoch": 14.44,
      "grad_norm": 0.0033958940766751766,
      "learning_rate": 5.560000000000001e-05,
      "loss": 0.0005,
      "step": 3610
    },
    {
      "epoch": 14.48,
      "grad_norm": 0.003239579265937209,
      "learning_rate": 5.520000000000001e-05,
      "loss": 0.0006,
      "step": 3620
    },
    {
      "epoch": 14.52,
      "grad_norm": 0.002811953192576766,
      "learning_rate": 5.4800000000000004e-05,
      "loss": 0.0005,
      "step": 3630
    },
    {
      "epoch": 14.56,
      "grad_norm": 0.0023668641224503517,
      "learning_rate": 5.440000000000001e-05,
      "loss": 0.0005,
      "step": 3640
    },
    {
      "epoch": 14.6,
      "grad_norm": 0.0036032095085829496,
      "learning_rate": 5.4000000000000005e-05,
      "loss": 0.0005,
      "step": 3650
    },
    {
      "epoch": 14.64,
      "grad_norm": 0.0026333522982895374,
      "learning_rate": 5.360000000000001e-05,
      "loss": 0.0005,
      "step": 3660
    },
    {
      "epoch": 14.68,
      "grad_norm": 0.0027461478020995855,
      "learning_rate": 5.3200000000000006e-05,
      "loss": 0.0005,
      "step": 3670
    },
    {
      "epoch": 14.72,
      "grad_norm": 0.002880874089896679,
      "learning_rate": 5.28e-05,
      "loss": 0.0005,
      "step": 3680
    },
    {
      "epoch": 14.76,
      "grad_norm": 0.003313728841021657,
      "learning_rate": 5.2400000000000007e-05,
      "loss": 0.0005,
      "step": 3690
    },
    {
      "epoch": 14.8,
      "grad_norm": 0.002891551936045289,
      "learning_rate": 5.2000000000000004e-05,
      "loss": 0.0005,
      "step": 3700
    },
    {
      "epoch": 14.8,
      "eval_accuracy": 0.819,
      "eval_loss": 1.203009009361267,
      "eval_runtime": 5.2248,
      "eval_samples_per_second": 191.394,
      "eval_steps_per_second": 23.924,
      "step": 3700
    },
    {
      "epoch": 14.84,
      "grad_norm": 0.002585276961326599,
      "learning_rate": 5.16e-05,
      "loss": 0.0005,
      "step": 3710
    },
    {
      "epoch": 14.88,
      "grad_norm": 0.002555152401328087,
      "learning_rate": 5.1200000000000004e-05,
      "loss": 0.0005,
      "step": 3720
    },
    {
      "epoch": 14.92,
      "grad_norm": 0.0025339240673929453,
      "learning_rate": 5.08e-05,
      "loss": 0.0005,
      "step": 3730
    },
    {
      "epoch": 14.96,
      "grad_norm": 0.0028862394392490387,
      "learning_rate": 5.0400000000000005e-05,
      "loss": 0.0005,
      "step": 3740
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.0030768855940550566,
      "learning_rate": 5e-05,
      "loss": 0.0005,
      "step": 3750
    },
    {
      "epoch": 15.04,
      "grad_norm": 0.0024443198926746845,
      "learning_rate": 4.96e-05,
      "loss": 0.0005,
      "step": 3760
    },
    {
      "epoch": 15.08,
      "grad_norm": 0.0026144583243876696,
      "learning_rate": 4.92e-05,
      "loss": 0.0005,
      "step": 3770
    },
    {
      "epoch": 15.12,
      "grad_norm": 0.002447100356221199,
      "learning_rate": 4.88e-05,
      "loss": 0.0005,
      "step": 3780
    },
    {
      "epoch": 15.16,
      "grad_norm": 0.003170330310240388,
      "learning_rate": 4.8400000000000004e-05,
      "loss": 0.0005,
      "step": 3790
    },
    {
      "epoch": 15.2,
      "grad_norm": 0.0029565359000116587,
      "learning_rate": 4.8e-05,
      "loss": 0.0005,
      "step": 3800
    },
    {
      "epoch": 15.2,
      "eval_accuracy": 0.818,
      "eval_loss": 1.2092506885528564,
      "eval_runtime": 5.9208,
      "eval_samples_per_second": 168.895,
      "eval_steps_per_second": 21.112,
      "step": 3800
    },
    {
      "epoch": 15.24,
      "grad_norm": 0.0024090250954031944,
      "learning_rate": 4.76e-05,
      "loss": 0.0005,
      "step": 3810
    },
    {
      "epoch": 15.28,
      "grad_norm": 0.0021153378766030073,
      "learning_rate": 4.72e-05,
      "loss": 0.0005,
      "step": 3820
    },
    {
      "epoch": 15.32,
      "grad_norm": 0.0027159815654158592,
      "learning_rate": 4.6800000000000006e-05,
      "loss": 0.0005,
      "step": 3830
    },
    {
      "epoch": 15.36,
      "grad_norm": 0.0028497877065092325,
      "learning_rate": 4.64e-05,
      "loss": 0.0005,
      "step": 3840
    },
    {
      "epoch": 15.4,
      "grad_norm": 0.0024237327743321657,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.0005,
      "step": 3850
    },
    {
      "epoch": 15.44,
      "grad_norm": 0.0031945749651640654,
      "learning_rate": 4.5600000000000004e-05,
      "loss": 0.0005,
      "step": 3860
    },
    {
      "epoch": 15.48,
      "grad_norm": 0.0026720704045146704,
      "learning_rate": 4.52e-05,
      "loss": 0.0005,
      "step": 3870
    },
    {
      "epoch": 15.52,
      "grad_norm": 0.0030961642041802406,
      "learning_rate": 4.4800000000000005e-05,
      "loss": 0.0005,
      "step": 3880
    },
    {
      "epoch": 15.56,
      "grad_norm": 0.002183083677664399,
      "learning_rate": 4.44e-05,
      "loss": 0.0005,
      "step": 3890
    },
    {
      "epoch": 15.6,
      "grad_norm": 0.0029423057567328215,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.0005,
      "step": 3900
    },
    {
      "epoch": 15.6,
      "eval_accuracy": 0.818,
      "eval_loss": 1.2159565687179565,
      "eval_runtime": 5.3023,
      "eval_samples_per_second": 188.596,
      "eval_steps_per_second": 23.575,
      "step": 3900
    },
    {
      "epoch": 15.64,
      "grad_norm": 0.002488673897460103,
      "learning_rate": 4.36e-05,
      "loss": 0.0005,
      "step": 3910
    },
    {
      "epoch": 15.68,
      "grad_norm": 0.0022629755549132824,
      "learning_rate": 4.32e-05,
      "loss": 0.0005,
      "step": 3920
    },
    {
      "epoch": 15.72,
      "grad_norm": 0.0022780627477914095,
      "learning_rate": 4.2800000000000004e-05,
      "loss": 0.0004,
      "step": 3930
    },
    {
      "epoch": 15.76,
      "grad_norm": 0.002429110463708639,
      "learning_rate": 4.24e-05,
      "loss": 0.0005,
      "step": 3940
    },
    {
      "epoch": 15.8,
      "grad_norm": 0.003042061347514391,
      "learning_rate": 4.2e-05,
      "loss": 0.0004,
      "step": 3950
    },
    {
      "epoch": 15.84,
      "grad_norm": 0.0023479724768549204,
      "learning_rate": 4.16e-05,
      "loss": 0.0005,
      "step": 3960
    },
    {
      "epoch": 15.88,
      "grad_norm": 0.0032163311261683702,
      "learning_rate": 4.12e-05,
      "loss": 0.0005,
      "step": 3970
    },
    {
      "epoch": 15.92,
      "grad_norm": 0.0020656727720052004,
      "learning_rate": 4.08e-05,
      "loss": 0.0004,
      "step": 3980
    },
    {
      "epoch": 15.96,
      "grad_norm": 0.002548342105001211,
      "learning_rate": 4.0400000000000006e-05,
      "loss": 0.0005,
      "step": 3990
    },
    {
      "epoch": 16.0,
      "grad_norm": 0.001894648070447147,
      "learning_rate": 4e-05,
      "loss": 0.0004,
      "step": 4000
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.819,
      "eval_loss": 1.2232067584991455,
      "eval_runtime": 5.3459,
      "eval_samples_per_second": 187.058,
      "eval_steps_per_second": 23.382,
      "step": 4000
    },
    {
      "epoch": 16.04,
      "grad_norm": 0.002135420450940728,
      "learning_rate": 3.960000000000001e-05,
      "loss": 0.0004,
      "step": 4010
    },
    {
      "epoch": 16.08,
      "grad_norm": 0.002859125379472971,
      "learning_rate": 3.9200000000000004e-05,
      "loss": 0.0004,
      "step": 4020
    },
    {
      "epoch": 16.12,
      "grad_norm": 0.002459143288433552,
      "learning_rate": 3.88e-05,
      "loss": 0.0004,
      "step": 4030
    },
    {
      "epoch": 16.16,
      "grad_norm": 0.0026586554013192654,
      "learning_rate": 3.8400000000000005e-05,
      "loss": 0.0004,
      "step": 4040
    },
    {
      "epoch": 16.2,
      "grad_norm": 0.0031422963365912437,
      "learning_rate": 3.8e-05,
      "loss": 0.0004,
      "step": 4050
    },
    {
      "epoch": 16.24,
      "grad_norm": 0.0020834060851484537,
      "learning_rate": 3.76e-05,
      "loss": 0.0004,
      "step": 4060
    },
    {
      "epoch": 16.28,
      "grad_norm": 0.00235546356998384,
      "learning_rate": 3.72e-05,
      "loss": 0.0004,
      "step": 4070
    },
    {
      "epoch": 16.32,
      "grad_norm": 0.0019886652007699013,
      "learning_rate": 3.68e-05,
      "loss": 0.0004,
      "step": 4080
    },
    {
      "epoch": 16.36,
      "grad_norm": 0.0024375116918236017,
      "learning_rate": 3.6400000000000004e-05,
      "loss": 0.0004,
      "step": 4090
    },
    {
      "epoch": 16.4,
      "grad_norm": 0.002063624793663621,
      "learning_rate": 3.6e-05,
      "loss": 0.0004,
      "step": 4100
    },
    {
      "epoch": 16.4,
      "eval_accuracy": 0.819,
      "eval_loss": 1.2302378416061401,
      "eval_runtime": 5.3045,
      "eval_samples_per_second": 188.52,
      "eval_steps_per_second": 23.565,
      "step": 4100
    },
    {
      "epoch": 16.44,
      "grad_norm": 0.0025726554449647665,
      "learning_rate": 3.56e-05,
      "loss": 0.0004,
      "step": 4110
    },
    {
      "epoch": 16.48,
      "grad_norm": 0.0025369401555508375,
      "learning_rate": 3.52e-05,
      "loss": 0.0004,
      "step": 4120
    },
    {
      "epoch": 16.52,
      "grad_norm": 0.0020231506787240505,
      "learning_rate": 3.48e-05,
      "loss": 0.0004,
      "step": 4130
    },
    {
      "epoch": 16.56,
      "grad_norm": 0.0021831688936799765,
      "learning_rate": 3.4399999999999996e-05,
      "loss": 0.0004,
      "step": 4140
    },
    {
      "epoch": 16.6,
      "grad_norm": 0.0023912801407277584,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.0004,
      "step": 4150
    },
    {
      "epoch": 16.64,
      "grad_norm": 0.0023613148368895054,
      "learning_rate": 3.3600000000000004e-05,
      "loss": 0.0004,
      "step": 4160
    },
    {
      "epoch": 16.68,
      "grad_norm": 0.0024576077703386545,
      "learning_rate": 3.32e-05,
      "loss": 0.0004,
      "step": 4170
    },
    {
      "epoch": 16.72,
      "grad_norm": 0.002091441536322236,
      "learning_rate": 3.2800000000000004e-05,
      "loss": 0.0004,
      "step": 4180
    },
    {
      "epoch": 16.76,
      "grad_norm": 0.0024098132271319628,
      "learning_rate": 3.24e-05,
      "loss": 0.0004,
      "step": 4190
    },
    {
      "epoch": 16.8,
      "grad_norm": 0.0022078922484070063,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0004,
      "step": 4200
    },
    {
      "epoch": 16.8,
      "eval_accuracy": 0.819,
      "eval_loss": 1.2350249290466309,
      "eval_runtime": 5.4057,
      "eval_samples_per_second": 184.99,
      "eval_steps_per_second": 23.124,
      "step": 4200
    },
    {
      "epoch": 16.84,
      "grad_norm": 0.0018520988523960114,
      "learning_rate": 3.16e-05,
      "loss": 0.0004,
      "step": 4210
    },
    {
      "epoch": 16.88,
      "grad_norm": 0.00181812874507159,
      "learning_rate": 3.12e-05,
      "loss": 0.0004,
      "step": 4220
    },
    {
      "epoch": 16.92,
      "grad_norm": 0.0023901008535176516,
      "learning_rate": 3.08e-05,
      "loss": 0.0004,
      "step": 4230
    },
    {
      "epoch": 16.96,
      "grad_norm": 0.002264365553855896,
      "learning_rate": 3.04e-05,
      "loss": 0.0004,
      "step": 4240
    },
    {
      "epoch": 17.0,
      "grad_norm": 0.002284231362864375,
      "learning_rate": 3e-05,
      "loss": 0.0004,
      "step": 4250
    },
    {
      "epoch": 17.04,
      "grad_norm": 0.0025141683872789145,
      "learning_rate": 2.96e-05,
      "loss": 0.0004,
      "step": 4260
    },
    {
      "epoch": 17.08,
      "grad_norm": 0.0023226479534059763,
      "learning_rate": 2.9199999999999998e-05,
      "loss": 0.0004,
      "step": 4270
    },
    {
      "epoch": 17.12,
      "grad_norm": 0.002532349433749914,
      "learning_rate": 2.88e-05,
      "loss": 0.0004,
      "step": 4280
    },
    {
      "epoch": 17.16,
      "grad_norm": 0.0018973592668771744,
      "learning_rate": 2.84e-05,
      "loss": 0.0004,
      "step": 4290
    },
    {
      "epoch": 17.2,
      "grad_norm": 0.0026193023659288883,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.0004,
      "step": 4300
    },
    {
      "epoch": 17.2,
      "eval_accuracy": 0.82,
      "eval_loss": 1.2400437593460083,
      "eval_runtime": 5.314,
      "eval_samples_per_second": 188.181,
      "eval_steps_per_second": 23.523,
      "step": 4300
    },
    {
      "epoch": 17.24,
      "grad_norm": 0.001868457649834454,
      "learning_rate": 2.7600000000000003e-05,
      "loss": 0.0004,
      "step": 4310
    },
    {
      "epoch": 17.28,
      "grad_norm": 0.0027018135879188776,
      "learning_rate": 2.7200000000000004e-05,
      "loss": 0.0004,
      "step": 4320
    },
    {
      "epoch": 17.32,
      "grad_norm": 0.0023467419669032097,
      "learning_rate": 2.6800000000000004e-05,
      "loss": 0.0004,
      "step": 4330
    },
    {
      "epoch": 17.36,
      "grad_norm": 0.002205600030720234,
      "learning_rate": 2.64e-05,
      "loss": 0.0004,
      "step": 4340
    },
    {
      "epoch": 17.4,
      "grad_norm": 0.0021359582897275686,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.0004,
      "step": 4350
    },
    {
      "epoch": 17.44,
      "grad_norm": 0.0024898636620491743,
      "learning_rate": 2.5600000000000002e-05,
      "loss": 0.0004,
      "step": 4360
    },
    {
      "epoch": 17.48,
      "grad_norm": 0.0017528502503409982,
      "learning_rate": 2.5200000000000003e-05,
      "loss": 0.0004,
      "step": 4370
    },
    {
      "epoch": 17.52,
      "grad_norm": 0.0022504818625748158,
      "learning_rate": 2.48e-05,
      "loss": 0.0004,
      "step": 4380
    },
    {
      "epoch": 17.56,
      "grad_norm": 0.00209965487010777,
      "learning_rate": 2.44e-05,
      "loss": 0.0004,
      "step": 4390
    },
    {
      "epoch": 17.6,
      "grad_norm": 0.0020423857495188713,
      "learning_rate": 2.4e-05,
      "loss": 0.0004,
      "step": 4400
    },
    {
      "epoch": 17.6,
      "eval_accuracy": 0.821,
      "eval_loss": 1.2442196607589722,
      "eval_runtime": 5.4679,
      "eval_samples_per_second": 182.887,
      "eval_steps_per_second": 22.861,
      "step": 4400
    },
    {
      "epoch": 17.64,
      "grad_norm": 0.0022337392438203096,
      "learning_rate": 2.36e-05,
      "loss": 0.0004,
      "step": 4410
    },
    {
      "epoch": 17.68,
      "grad_norm": 0.0018173023127019405,
      "learning_rate": 2.32e-05,
      "loss": 0.0004,
      "step": 4420
    },
    {
      "epoch": 17.72,
      "grad_norm": 0.002295331796631217,
      "learning_rate": 2.2800000000000002e-05,
      "loss": 0.0004,
      "step": 4430
    },
    {
      "epoch": 17.76,
      "grad_norm": 0.001931387116201222,
      "learning_rate": 2.2400000000000002e-05,
      "loss": 0.0004,
      "step": 4440
    },
    {
      "epoch": 17.8,
      "grad_norm": 0.0019796870183199644,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.0004,
      "step": 4450
    },
    {
      "epoch": 17.84,
      "grad_norm": 0.0016401337925344706,
      "learning_rate": 2.16e-05,
      "loss": 0.0004,
      "step": 4460
    },
    {
      "epoch": 17.88,
      "grad_norm": 0.001987120369449258,
      "learning_rate": 2.12e-05,
      "loss": 0.0004,
      "step": 4470
    },
    {
      "epoch": 17.92,
      "grad_norm": 0.002337078796699643,
      "learning_rate": 2.08e-05,
      "loss": 0.0004,
      "step": 4480
    },
    {
      "epoch": 17.96,
      "grad_norm": 0.0024157376028597355,
      "learning_rate": 2.04e-05,
      "loss": 0.0004,
      "step": 4490
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.0021758731454610825,
      "learning_rate": 2e-05,
      "loss": 0.0004,
      "step": 4500
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.821,
      "eval_loss": 1.2483272552490234,
      "eval_runtime": 5.8485,
      "eval_samples_per_second": 170.983,
      "eval_steps_per_second": 21.373,
      "step": 4500
    },
    {
      "epoch": 18.04,
      "grad_norm": 0.002017402555793524,
      "learning_rate": 1.9600000000000002e-05,
      "loss": 0.0004,
      "step": 4510
    },
    {
      "epoch": 18.08,
      "grad_norm": 0.0020504500716924667,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 0.0004,
      "step": 4520
    },
    {
      "epoch": 18.12,
      "grad_norm": 0.0019253026694059372,
      "learning_rate": 1.88e-05,
      "loss": 0.0004,
      "step": 4530
    },
    {
      "epoch": 18.16,
      "grad_norm": 0.0020694080740213394,
      "learning_rate": 1.84e-05,
      "loss": 0.0004,
      "step": 4540
    },
    {
      "epoch": 18.2,
      "grad_norm": 0.00169233581982553,
      "learning_rate": 1.8e-05,
      "loss": 0.0004,
      "step": 4550
    },
    {
      "epoch": 18.24,
      "grad_norm": 0.0019015150610357523,
      "learning_rate": 1.76e-05,
      "loss": 0.0004,
      "step": 4560
    },
    {
      "epoch": 18.28,
      "grad_norm": 0.002108811168000102,
      "learning_rate": 1.7199999999999998e-05,
      "loss": 0.0004,
      "step": 4570
    },
    {
      "epoch": 18.32,
      "grad_norm": 0.0021513872779905796,
      "learning_rate": 1.6800000000000002e-05,
      "loss": 0.0004,
      "step": 4580
    },
    {
      "epoch": 18.36,
      "grad_norm": 0.0019856479484587908,
      "learning_rate": 1.6400000000000002e-05,
      "loss": 0.0004,
      "step": 4590
    },
    {
      "epoch": 18.4,
      "grad_norm": 0.0020635281689465046,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0004,
      "step": 4600
    },
    {
      "epoch": 18.4,
      "eval_accuracy": 0.821,
      "eval_loss": 1.25178062915802,
      "eval_runtime": 5.3158,
      "eval_samples_per_second": 188.119,
      "eval_steps_per_second": 23.515,
      "step": 4600
    },
    {
      "epoch": 18.44,
      "grad_norm": 0.0020957705564796925,
      "learning_rate": 1.56e-05,
      "loss": 0.0004,
      "step": 4610
    },
    {
      "epoch": 18.48,
      "grad_norm": 0.002100448589771986,
      "learning_rate": 1.52e-05,
      "loss": 0.0004,
      "step": 4620
    },
    {
      "epoch": 18.52,
      "grad_norm": 0.0017688070656731725,
      "learning_rate": 1.48e-05,
      "loss": 0.0004,
      "step": 4630
    },
    {
      "epoch": 18.56,
      "grad_norm": 0.0018529172521084547,
      "learning_rate": 1.44e-05,
      "loss": 0.0004,
      "step": 4640
    },
    {
      "epoch": 18.6,
      "grad_norm": 0.0016175085911527276,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.0004,
      "step": 4650
    },
    {
      "epoch": 18.64,
      "grad_norm": 0.001927876495756209,
      "learning_rate": 1.3600000000000002e-05,
      "loss": 0.0004,
      "step": 4660
    },
    {
      "epoch": 18.68,
      "grad_norm": 0.00210273964330554,
      "learning_rate": 1.32e-05,
      "loss": 0.0004,
      "step": 4670
    },
    {
      "epoch": 18.72,
      "grad_norm": 0.002222416689619422,
      "learning_rate": 1.2800000000000001e-05,
      "loss": 0.0004,
      "step": 4680
    },
    {
      "epoch": 18.76,
      "grad_norm": 0.0018316478235647082,
      "learning_rate": 1.24e-05,
      "loss": 0.0004,
      "step": 4690
    },
    {
      "epoch": 18.8,
      "grad_norm": 0.0019819007720798254,
      "learning_rate": 1.2e-05,
      "loss": 0.0004,
      "step": 4700
    },
    {
      "epoch": 18.8,
      "eval_accuracy": 0.821,
      "eval_loss": 1.254600167274475,
      "eval_runtime": 5.2417,
      "eval_samples_per_second": 190.777,
      "eval_steps_per_second": 23.847,
      "step": 4700
    },
    {
      "epoch": 18.84,
      "grad_norm": 0.0016140459338203073,
      "learning_rate": 1.16e-05,
      "loss": 0.0003,
      "step": 4710
    },
    {
      "epoch": 18.88,
      "grad_norm": 0.0018461669096723199,
      "learning_rate": 1.1200000000000001e-05,
      "loss": 0.0003,
      "step": 4720
    },
    {
      "epoch": 18.92,
      "grad_norm": 0.0018047132762148976,
      "learning_rate": 1.08e-05,
      "loss": 0.0003,
      "step": 4730
    },
    {
      "epoch": 18.96,
      "grad_norm": 0.0021749059669673443,
      "learning_rate": 1.04e-05,
      "loss": 0.0004,
      "step": 4740
    },
    {
      "epoch": 19.0,
      "grad_norm": 0.002118439180776477,
      "learning_rate": 1e-05,
      "loss": 0.0004,
      "step": 4750
    },
    {
      "epoch": 19.04,
      "grad_norm": 0.0019508595578372478,
      "learning_rate": 9.600000000000001e-06,
      "loss": 0.0004,
      "step": 4760
    },
    {
      "epoch": 19.08,
      "grad_norm": 0.002842509187757969,
      "learning_rate": 9.2e-06,
      "loss": 0.0004,
      "step": 4770
    },
    {
      "epoch": 19.12,
      "grad_norm": 0.0016302353469654918,
      "learning_rate": 8.8e-06,
      "loss": 0.0003,
      "step": 4780
    },
    {
      "epoch": 19.16,
      "grad_norm": 0.001962054753676057,
      "learning_rate": 8.400000000000001e-06,
      "loss": 0.0003,
      "step": 4790
    },
    {
      "epoch": 19.2,
      "grad_norm": 0.0021732356399297714,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0004,
      "step": 4800
    },
    {
      "epoch": 19.2,
      "eval_accuracy": 0.821,
      "eval_loss": 1.2561498880386353,
      "eval_runtime": 10.3614,
      "eval_samples_per_second": 96.512,
      "eval_steps_per_second": 12.064,
      "step": 4800
    },
    {
      "epoch": 19.24,
      "grad_norm": 0.0018920012516900897,
      "learning_rate": 7.6e-06,
      "loss": 0.0004,
      "step": 4810
    },
    {
      "epoch": 19.28,
      "grad_norm": 0.002033686963841319,
      "learning_rate": 7.2e-06,
      "loss": 0.0003,
      "step": 4820
    },
    {
      "epoch": 19.32,
      "grad_norm": 0.0019939634948968887,
      "learning_rate": 6.800000000000001e-06,
      "loss": 0.0004,
      "step": 4830
    },
    {
      "epoch": 19.36,
      "grad_norm": 0.0024624329525977373,
      "learning_rate": 6.4000000000000006e-06,
      "loss": 0.0004,
      "step": 4840
    },
    {
      "epoch": 19.4,
      "grad_norm": 0.002102550817653537,
      "learning_rate": 6e-06,
      "loss": 0.0004,
      "step": 4850
    },
    {
      "epoch": 19.44,
      "grad_norm": 0.0017401843797415495,
      "learning_rate": 5.600000000000001e-06,
      "loss": 0.0004,
      "step": 4860
    },
    {
      "epoch": 19.48,
      "grad_norm": 0.0016316601540893316,
      "learning_rate": 5.2e-06,
      "loss": 0.0003,
      "step": 4870
    },
    {
      "epoch": 19.52,
      "grad_norm": 0.0024664264637976885,
      "learning_rate": 4.800000000000001e-06,
      "loss": 0.0003,
      "step": 4880
    },
    {
      "epoch": 19.56,
      "grad_norm": 0.0017682453617453575,
      "learning_rate": 4.4e-06,
      "loss": 0.0004,
      "step": 4890
    },
    {
      "epoch": 19.6,
      "grad_norm": 0.0019737447146326303,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0004,
      "step": 4900
    },
    {
      "epoch": 19.6,
      "eval_accuracy": 0.82,
      "eval_loss": 1.2573806047439575,
      "eval_runtime": 5.3728,
      "eval_samples_per_second": 186.124,
      "eval_steps_per_second": 23.265,
      "step": 4900
    },
    {
      "epoch": 19.64,
      "grad_norm": 0.002168442588299513,
      "learning_rate": 3.6e-06,
      "loss": 0.0003,
      "step": 4910
    },
    {
      "epoch": 19.68,
      "grad_norm": 0.0016441222978755832,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 0.0003,
      "step": 4920
    },
    {
      "epoch": 19.72,
      "grad_norm": 0.0016924934461712837,
      "learning_rate": 2.8000000000000003e-06,
      "loss": 0.0003,
      "step": 4930
    },
    {
      "epoch": 19.76,
      "grad_norm": 0.0026684573385864496,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.0004,
      "step": 4940
    },
    {
      "epoch": 19.8,
      "grad_norm": 0.002037678612396121,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0004,
      "step": 4950
    },
    {
      "epoch": 19.84,
      "grad_norm": 0.0019053890137001872,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 0.0004,
      "step": 4960
    },
    {
      "epoch": 19.88,
      "grad_norm": 0.002319651423022151,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 0.0003,
      "step": 4970
    },
    {
      "epoch": 19.92,
      "grad_norm": 0.0018768333829939365,
      "learning_rate": 8.000000000000001e-07,
      "loss": 0.0004,
      "step": 4980
    },
    {
      "epoch": 19.96,
      "grad_norm": 0.0020151715725660324,
      "learning_rate": 4.0000000000000003e-07,
      "loss": 0.0004,
      "step": 4990
    },
    {
      "epoch": 20.0,
      "grad_norm": 0.0020873937755823135,
      "learning_rate": 0.0,
      "loss": 0.0004,
      "step": 5000
    },
    {
      "epoch": 20.0,
      "eval_accuracy": 0.82,
      "eval_loss": 1.257745385169983,
      "eval_runtime": 5.3256,
      "eval_samples_per_second": 187.773,
      "eval_steps_per_second": 23.472,
      "step": 5000
    },
    {
      "epoch": 20.0,
      "step": 5000,
      "total_flos": 6.19947029495808e+18,
      "train_loss": 0.16214322472065687,
      "train_runtime": 2178.5774,
      "train_samples_per_second": 36.721,
      "train_steps_per_second": 2.295
    }
  ],
  "logging_steps": 10,
  "max_steps": 5000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 100,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.19947029495808e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}