{
  "best_metric": 0.5302808284759521,
  "best_model_checkpoint": "Human_action_classifier/checkpoint-4300",
  "epoch": 7.0,
  "eval_steps": 100,
  "global_step": 4410,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "grad_norm": 1.9168592691421509,
      "learning_rate": 0.00019954648526077098,
      "loss": 2.6612,
      "step": 10
    },
    {
      "epoch": 0.03,
      "grad_norm": 2.498575210571289,
      "learning_rate": 0.00019909297052154195,
      "loss": 2.4529,
      "step": 20
    },
    {
      "epoch": 0.05,
      "grad_norm": 2.4278714656829834,
      "learning_rate": 0.00019863945578231293,
      "loss": 2.3062,
      "step": 30
    },
    {
      "epoch": 0.06,
      "grad_norm": 2.5514345169067383,
      "learning_rate": 0.0001981859410430839,
      "loss": 2.169,
      "step": 40
    },
    {
      "epoch": 0.08,
      "grad_norm": 2.1051814556121826,
      "learning_rate": 0.0001977324263038549,
      "loss": 1.9571,
      "step": 50
    },
    {
      "epoch": 0.1,
      "grad_norm": 2.5904359817504883,
      "learning_rate": 0.00019727891156462587,
      "loss": 1.7926,
      "step": 60
    },
    {
      "epoch": 0.11,
      "grad_norm": 2.9325740337371826,
      "learning_rate": 0.00019682539682539682,
      "loss": 1.6761,
      "step": 70
    },
    {
      "epoch": 0.13,
      "grad_norm": 3.636162519454956,
      "learning_rate": 0.00019637188208616781,
      "loss": 1.6388,
      "step": 80
    },
    {
      "epoch": 0.14,
      "grad_norm": 3.7232675552368164,
      "learning_rate": 0.0001959183673469388,
      "loss": 1.527,
      "step": 90
    },
    {
      "epoch": 0.16,
      "grad_norm": 3.467459201812744,
      "learning_rate": 0.00019546485260770976,
      "loss": 1.4545,
      "step": 100
    },
    {
      "epoch": 0.16,
      "eval_accuracy": 0.6706349206349206,
      "eval_loss": 1.3145380020141602,
      "eval_runtime": 93.6176,
      "eval_samples_per_second": 26.918,
      "eval_steps_per_second": 3.365,
      "step": 100
    },
    {
      "epoch": 0.17,
      "grad_norm": 2.4409241676330566,
      "learning_rate": 0.00019501133786848073,
      "loss": 1.331,
      "step": 110
    },
    {
      "epoch": 0.19,
      "grad_norm": 3.841000556945801,
      "learning_rate": 0.0001945578231292517,
      "loss": 1.4273,
      "step": 120
    },
    {
      "epoch": 0.21,
      "grad_norm": 4.918443202972412,
      "learning_rate": 0.0001941043083900227,
      "loss": 1.4078,
      "step": 130
    },
    {
      "epoch": 0.22,
      "grad_norm": 3.317399740219116,
      "learning_rate": 0.00019365079365079365,
      "loss": 1.3905,
      "step": 140
    },
    {
      "epoch": 0.24,
      "grad_norm": 5.5322418212890625,
      "learning_rate": 0.00019319727891156462,
      "loss": 1.5453,
      "step": 150
    },
    {
      "epoch": 0.25,
      "grad_norm": 4.116984844207764,
      "learning_rate": 0.00019274376417233562,
      "loss": 1.394,
      "step": 160
    },
    {
      "epoch": 0.27,
      "grad_norm": 3.141216278076172,
      "learning_rate": 0.0001922902494331066,
      "loss": 1.2375,
      "step": 170
    },
    {
      "epoch": 0.29,
      "grad_norm": 6.414323329925537,
      "learning_rate": 0.00019183673469387756,
      "loss": 1.3073,
      "step": 180
    },
    {
      "epoch": 0.3,
      "grad_norm": 5.794140338897705,
      "learning_rate": 0.00019138321995464854,
      "loss": 1.3405,
      "step": 190
    },
    {
      "epoch": 0.32,
      "grad_norm": 4.384269714355469,
      "learning_rate": 0.0001909297052154195,
      "loss": 1.2568,
      "step": 200
    },
    {
      "epoch": 0.32,
      "eval_accuracy": 0.7178571428571429,
      "eval_loss": 1.0386934280395508,
      "eval_runtime": 94.8513,
      "eval_samples_per_second": 26.568,
      "eval_steps_per_second": 3.321,
      "step": 200
    },
    {
      "epoch": 0.33,
      "grad_norm": 2.881287097930908,
      "learning_rate": 0.00019047619047619048,
      "loss": 1.2329,
      "step": 210
    },
    {
      "epoch": 0.35,
      "grad_norm": 4.77938985824585,
      "learning_rate": 0.00019002267573696145,
      "loss": 1.1439,
      "step": 220
    },
    {
      "epoch": 0.37,
      "grad_norm": 5.865462303161621,
      "learning_rate": 0.00018956916099773243,
      "loss": 1.2314,
      "step": 230
    },
    {
      "epoch": 0.38,
      "grad_norm": 3.464250326156616,
      "learning_rate": 0.00018911564625850343,
      "loss": 1.329,
      "step": 240
    },
    {
      "epoch": 0.4,
      "grad_norm": 3.2501161098480225,
      "learning_rate": 0.0001886621315192744,
      "loss": 1.3277,
      "step": 250
    },
    {
      "epoch": 0.41,
      "grad_norm": 5.26331901550293,
      "learning_rate": 0.00018820861678004534,
      "loss": 1.2876,
      "step": 260
    },
    {
      "epoch": 0.43,
      "grad_norm": 2.62630295753479,
      "learning_rate": 0.00018775510204081634,
      "loss": 1.1428,
      "step": 270
    },
    {
      "epoch": 0.44,
      "grad_norm": 6.121326923370361,
      "learning_rate": 0.00018730158730158731,
      "loss": 1.2277,
      "step": 280
    },
    {
      "epoch": 0.46,
      "grad_norm": 5.415524005889893,
      "learning_rate": 0.0001868480725623583,
      "loss": 1.1641,
      "step": 290
    },
    {
      "epoch": 0.48,
      "grad_norm": 7.849498271942139,
      "learning_rate": 0.00018639455782312926,
      "loss": 1.3145,
      "step": 300
    },
    {
      "epoch": 0.48,
      "eval_accuracy": 0.7134920634920635,
      "eval_loss": 1.0026524066925049,
      "eval_runtime": 89.5917,
      "eval_samples_per_second": 28.128,
      "eval_steps_per_second": 3.516,
      "step": 300
    },
    {
      "epoch": 0.49,
      "grad_norm": 3.9149835109710693,
      "learning_rate": 0.00018594104308390023,
      "loss": 1.1359,
      "step": 310
    },
    {
      "epoch": 0.51,
      "grad_norm": 4.712550163269043,
      "learning_rate": 0.0001854875283446712,
      "loss": 1.1196,
      "step": 320
    },
    {
      "epoch": 0.52,
      "grad_norm": 4.665726184844971,
      "learning_rate": 0.0001850340136054422,
      "loss": 1.1178,
      "step": 330
    },
    {
      "epoch": 0.54,
      "grad_norm": 4.421199798583984,
      "learning_rate": 0.00018458049886621315,
      "loss": 1.093,
      "step": 340
    },
    {
      "epoch": 0.56,
      "grad_norm": 5.4558305740356445,
      "learning_rate": 0.00018412698412698412,
      "loss": 1.0638,
      "step": 350
    },
    {
      "epoch": 0.57,
      "grad_norm": 6.7150559425354,
      "learning_rate": 0.00018367346938775512,
      "loss": 1.21,
      "step": 360
    },
    {
      "epoch": 0.59,
      "grad_norm": 4.524730205535889,
      "learning_rate": 0.0001832199546485261,
      "loss": 1.199,
      "step": 370
    },
    {
      "epoch": 0.6,
      "grad_norm": 5.457512855529785,
      "learning_rate": 0.00018276643990929706,
      "loss": 1.202,
      "step": 380
    },
    {
      "epoch": 0.62,
      "grad_norm": 5.54859733581543,
      "learning_rate": 0.00018231292517006804,
      "loss": 1.1671,
      "step": 390
    },
    {
      "epoch": 0.63,
      "grad_norm": 3.217099905014038,
      "learning_rate": 0.000181859410430839,
      "loss": 1.0866,
      "step": 400
    },
    {
      "epoch": 0.63,
      "eval_accuracy": 0.7376984126984127,
      "eval_loss": 0.8882665038108826,
      "eval_runtime": 94.504,
      "eval_samples_per_second": 26.666,
      "eval_steps_per_second": 3.333,
      "step": 400
    },
    {
      "epoch": 0.65,
      "grad_norm": 5.335367202758789,
      "learning_rate": 0.00018140589569161,
      "loss": 1.0176,
      "step": 410
    },
    {
      "epoch": 0.67,
      "grad_norm": 7.274291515350342,
      "learning_rate": 0.00018099773242630387,
      "loss": 1.2467,
      "step": 420
    },
    {
      "epoch": 0.68,
      "grad_norm": 5.159976959228516,
      "learning_rate": 0.00018054421768707484,
      "loss": 1.127,
      "step": 430
    },
    {
      "epoch": 0.7,
      "grad_norm": 5.351009845733643,
      "learning_rate": 0.0001800907029478458,
      "loss": 1.2382,
      "step": 440
    },
    {
      "epoch": 0.71,
      "grad_norm": 4.435301303863525,
      "learning_rate": 0.00017963718820861678,
      "loss": 1.061,
      "step": 450
    },
    {
      "epoch": 0.73,
      "grad_norm": 4.027517795562744,
      "learning_rate": 0.00017918367346938776,
      "loss": 1.0432,
      "step": 460
    },
    {
      "epoch": 0.75,
      "grad_norm": 4.309961795806885,
      "learning_rate": 0.00017873015873015876,
      "loss": 0.9949,
      "step": 470
    },
    {
      "epoch": 0.76,
      "grad_norm": 4.835757732391357,
      "learning_rate": 0.00017827664399092973,
      "loss": 1.1899,
      "step": 480
    },
    {
      "epoch": 0.78,
      "grad_norm": 5.26571798324585,
      "learning_rate": 0.00017782312925170067,
      "loss": 1.1235,
      "step": 490
    },
    {
      "epoch": 0.79,
      "grad_norm": 3.8406684398651123,
      "learning_rate": 0.00017736961451247167,
      "loss": 1.0036,
      "step": 500
    },
    {
      "epoch": 0.79,
      "eval_accuracy": 0.7321428571428571,
      "eval_loss": 0.897292971611023,
      "eval_runtime": 96.2225,
      "eval_samples_per_second": 26.189,
      "eval_steps_per_second": 3.274,
      "step": 500
    },
    {
      "epoch": 0.81,
      "grad_norm": 4.993399143218994,
      "learning_rate": 0.00017691609977324264,
      "loss": 0.8682,
      "step": 510
    },
    {
      "epoch": 0.83,
      "grad_norm": 4.4026875495910645,
      "learning_rate": 0.00017646258503401362,
      "loss": 1.0603,
      "step": 520
    },
    {
      "epoch": 0.84,
      "grad_norm": 4.832537651062012,
      "learning_rate": 0.0001760090702947846,
      "loss": 1.034,
      "step": 530
    },
    {
      "epoch": 0.86,
      "grad_norm": 5.253382682800293,
      "learning_rate": 0.00017555555555555556,
      "loss": 0.9938,
      "step": 540
    },
    {
      "epoch": 0.87,
      "grad_norm": 3.681997060775757,
      "learning_rate": 0.00017510204081632653,
      "loss": 1.0082,
      "step": 550
    },
    {
      "epoch": 0.89,
      "grad_norm": 6.324045181274414,
      "learning_rate": 0.0001746485260770975,
      "loss": 0.9629,
      "step": 560
    },
    {
      "epoch": 0.9,
      "grad_norm": 5.0340352058410645,
      "learning_rate": 0.00017419501133786848,
      "loss": 1.1184,
      "step": 570
    },
    {
      "epoch": 0.92,
      "grad_norm": 3.532378673553467,
      "learning_rate": 0.00017374149659863948,
      "loss": 0.9167,
      "step": 580
    },
    {
      "epoch": 0.94,
      "grad_norm": 5.029895305633545,
      "learning_rate": 0.00017328798185941045,
      "loss": 1.1213,
      "step": 590
    },
    {
      "epoch": 0.95,
      "grad_norm": 4.585740566253662,
      "learning_rate": 0.00017283446712018142,
      "loss": 1.1811,
      "step": 600
    },
    {
      "epoch": 0.95,
      "eval_accuracy": 0.7571428571428571,
      "eval_loss": 0.8048315644264221,
      "eval_runtime": 95.9399,
      "eval_samples_per_second": 26.266,
      "eval_steps_per_second": 3.283,
      "step": 600
    },
    {
      "epoch": 0.97,
      "grad_norm": 2.9866397380828857,
      "learning_rate": 0.0001723809523809524,
      "loss": 1.2417,
      "step": 610
    },
    {
      "epoch": 0.98,
      "grad_norm": 4.54673433303833,
      "learning_rate": 0.00017192743764172337,
      "loss": 1.1588,
      "step": 620
    },
    {
      "epoch": 1.0,
      "grad_norm": 4.411011695861816,
      "learning_rate": 0.00017147392290249434,
      "loss": 0.9512,
      "step": 630
    },
    {
      "epoch": 1.02,
      "grad_norm": 4.393170356750488,
      "learning_rate": 0.0001710204081632653,
      "loss": 0.9679,
      "step": 640
    },
    {
      "epoch": 1.03,
      "grad_norm": 5.261455535888672,
      "learning_rate": 0.00017056689342403628,
      "loss": 0.9089,
      "step": 650
    },
    {
      "epoch": 1.05,
      "grad_norm": 4.93878173828125,
      "learning_rate": 0.00017011337868480726,
      "loss": 0.8341,
      "step": 660
    },
    {
      "epoch": 1.06,
      "grad_norm": 4.5593156814575195,
      "learning_rate": 0.00016965986394557825,
      "loss": 0.9952,
      "step": 670
    },
    {
      "epoch": 1.08,
      "grad_norm": 3.9355390071868896,
      "learning_rate": 0.0001692063492063492,
      "loss": 0.9956,
      "step": 680
    },
    {
      "epoch": 1.1,
      "grad_norm": 4.096402645111084,
      "learning_rate": 0.00016875283446712017,
      "loss": 0.9566,
      "step": 690
    },
    {
      "epoch": 1.11,
      "grad_norm": 5.853695869445801,
      "learning_rate": 0.00016829931972789117,
      "loss": 0.9242,
      "step": 700
    },
    {
      "epoch": 1.11,
      "eval_accuracy": 0.7273809523809524,
      "eval_loss": 0.9095195531845093,
      "eval_runtime": 92.7769,
      "eval_samples_per_second": 27.162,
      "eval_steps_per_second": 3.395,
      "step": 700
    },
    {
      "epoch": 1.13,
      "grad_norm": 5.18225622177124,
      "learning_rate": 0.00016784580498866214,
      "loss": 1.1815,
      "step": 710
    },
    {
      "epoch": 1.14,
      "grad_norm": 6.094394683837891,
      "learning_rate": 0.00016739229024943312,
      "loss": 1.0088,
      "step": 720
    },
    {
      "epoch": 1.16,
      "grad_norm": 3.388333320617676,
      "learning_rate": 0.0001669387755102041,
      "loss": 1.0219,
      "step": 730
    },
    {
      "epoch": 1.17,
      "grad_norm": 2.702335834503174,
      "learning_rate": 0.00016648526077097506,
      "loss": 0.995,
      "step": 740
    },
    {
      "epoch": 1.19,
      "grad_norm": 6.5921735763549805,
      "learning_rate": 0.00016603174603174606,
      "loss": 1.0336,
      "step": 750
    },
    {
      "epoch": 1.21,
      "grad_norm": 7.6583781242370605,
      "learning_rate": 0.000165578231292517,
      "loss": 0.8889,
      "step": 760
    },
    {
      "epoch": 1.22,
      "grad_norm": 5.093008518218994,
      "learning_rate": 0.00016512471655328798,
      "loss": 0.8492,
      "step": 770
    },
    {
      "epoch": 1.24,
      "grad_norm": 4.372059345245361,
      "learning_rate": 0.00016467120181405898,
      "loss": 0.8115,
      "step": 780
    },
    {
      "epoch": 1.25,
      "grad_norm": 4.1783857345581055,
      "learning_rate": 0.00016421768707482995,
      "loss": 0.834,
      "step": 790
    },
    {
      "epoch": 1.27,
      "grad_norm": 4.295966148376465,
      "learning_rate": 0.0001638095238095238,
      "loss": 0.9477,
      "step": 800
    },
    {
      "epoch": 1.27,
      "eval_accuracy": 0.7619047619047619,
      "eval_loss": 0.8036767244338989,
      "eval_runtime": 89.9677,
      "eval_samples_per_second": 28.01,
      "eval_steps_per_second": 3.501,
      "step": 800
    },
    {
      "epoch": 1.29,
      "grad_norm": 4.976074695587158,
      "learning_rate": 0.0001633560090702948,
      "loss": 0.8778,
      "step": 810
    },
    {
      "epoch": 1.3,
      "grad_norm": 3.34804105758667,
      "learning_rate": 0.00016290249433106578,
      "loss": 0.9856,
      "step": 820
    },
    {
      "epoch": 1.32,
      "grad_norm": 6.416792869567871,
      "learning_rate": 0.00016244897959183672,
      "loss": 0.9829,
      "step": 830
    },
    {
      "epoch": 1.33,
      "grad_norm": 5.707892894744873,
      "learning_rate": 0.00016199546485260772,
      "loss": 0.9397,
      "step": 840
    },
    {
      "epoch": 1.35,
      "grad_norm": 4.620942115783691,
      "learning_rate": 0.0001615419501133787,
      "loss": 1.078,
      "step": 850
    },
    {
      "epoch": 1.37,
      "grad_norm": 4.452626705169678,
      "learning_rate": 0.00016108843537414967,
      "loss": 0.9531,
      "step": 860
    },
    {
      "epoch": 1.38,
      "grad_norm": 5.697545528411865,
      "learning_rate": 0.00016063492063492064,
      "loss": 1.0438,
      "step": 870
    },
    {
      "epoch": 1.4,
      "grad_norm": 4.365375518798828,
      "learning_rate": 0.0001601814058956916,
      "loss": 0.9024,
      "step": 880
    },
    {
      "epoch": 1.41,
      "grad_norm": 5.501079082489014,
      "learning_rate": 0.00015972789115646259,
      "loss": 0.9091,
      "step": 890
    },
    {
      "epoch": 1.43,
      "grad_norm": 4.409488201141357,
      "learning_rate": 0.00015927437641723358,
      "loss": 0.8634,
      "step": 900
    },
    {
      "epoch": 1.43,
      "eval_accuracy": 0.7642857142857142,
      "eval_loss": 0.7937940359115601,
      "eval_runtime": 92.0164,
      "eval_samples_per_second": 27.386,
      "eval_steps_per_second": 3.423,
      "step": 900
    },
    {
      "epoch": 1.44,
      "grad_norm": 4.543234825134277,
      "learning_rate": 0.00015882086167800453,
      "loss": 0.7896,
      "step": 910
    },
    {
      "epoch": 1.46,
      "grad_norm": 3.8954508304595947,
      "learning_rate": 0.0001583673469387755,
      "loss": 0.9597,
      "step": 920
    },
    {
      "epoch": 1.48,
      "grad_norm": 7.744637489318848,
      "learning_rate": 0.0001579138321995465,
      "loss": 0.9585,
      "step": 930
    },
    {
      "epoch": 1.49,
      "grad_norm": 3.619973659515381,
      "learning_rate": 0.00015746031746031747,
      "loss": 0.826,
      "step": 940
    },
    {
      "epoch": 1.51,
      "grad_norm": 10.431225776672363,
      "learning_rate": 0.00015700680272108845,
      "loss": 0.9651,
      "step": 950
    },
    {
      "epoch": 1.52,
      "grad_norm": 4.897637367248535,
      "learning_rate": 0.00015655328798185942,
      "loss": 0.9136,
      "step": 960
    },
    {
      "epoch": 1.54,
      "grad_norm": 2.429625988006592,
      "learning_rate": 0.0001560997732426304,
      "loss": 0.9676,
      "step": 970
    },
    {
      "epoch": 1.56,
      "grad_norm": 2.8244125843048096,
      "learning_rate": 0.00015564625850340136,
      "loss": 0.8796,
      "step": 980
    },
    {
      "epoch": 1.57,
      "grad_norm": 3.9079954624176025,
      "learning_rate": 0.00015519274376417234,
      "loss": 0.7595,
      "step": 990
    },
    {
      "epoch": 1.59,
      "grad_norm": 8.262408256530762,
      "learning_rate": 0.0001547392290249433,
      "loss": 1.0098,
      "step": 1000
    },
    {
      "epoch": 1.59,
      "eval_accuracy": 0.7765873015873016,
      "eval_loss": 0.7327961325645447,
      "eval_runtime": 90.6084,
      "eval_samples_per_second": 27.812,
      "eval_steps_per_second": 3.477,
      "step": 1000
    },
    {
      "epoch": 1.6,
      "grad_norm": 5.290775299072266,
      "learning_rate": 0.0001542857142857143,
      "loss": 0.8282,
      "step": 1010
    },
    {
      "epoch": 1.62,
      "grad_norm": 5.40395450592041,
      "learning_rate": 0.00015383219954648528,
      "loss": 1.0232,
      "step": 1020
    },
    {
      "epoch": 1.63,
      "grad_norm": 6.36408805847168,
      "learning_rate": 0.00015337868480725622,
      "loss": 0.9572,
      "step": 1030
    },
    {
      "epoch": 1.65,
      "grad_norm": 3.8068315982818604,
      "learning_rate": 0.00015292517006802722,
      "loss": 0.8457,
      "step": 1040
    },
    {
      "epoch": 1.67,
      "grad_norm": 5.893327236175537,
      "learning_rate": 0.0001524716553287982,
      "loss": 0.915,
      "step": 1050
    },
    {
      "epoch": 1.68,
      "grad_norm": 4.604687213897705,
      "learning_rate": 0.00015201814058956917,
      "loss": 0.9143,
      "step": 1060
    },
    {
      "epoch": 1.7,
      "grad_norm": 4.973442077636719,
      "learning_rate": 0.00015156462585034014,
      "loss": 0.8211,
      "step": 1070
    },
    {
      "epoch": 1.71,
      "grad_norm": 4.463946342468262,
      "learning_rate": 0.0001511111111111111,
      "loss": 0.8649,
      "step": 1080
    },
    {
      "epoch": 1.73,
      "grad_norm": 4.718800067901611,
      "learning_rate": 0.0001506575963718821,
      "loss": 0.7986,
      "step": 1090
    },
    {
      "epoch": 1.75,
      "grad_norm": 4.523365020751953,
      "learning_rate": 0.00015020408163265306,
      "loss": 0.8176,
      "step": 1100
    },
    {
      "epoch": 1.75,
      "eval_accuracy": 0.7515873015873016,
      "eval_loss": 0.8064602017402649,
      "eval_runtime": 93.1844,
      "eval_samples_per_second": 27.043,
      "eval_steps_per_second": 3.38,
      "step": 1100
    },
    {
      "epoch": 1.76,
      "grad_norm": 5.417792797088623,
      "learning_rate": 0.00014975056689342403,
      "loss": 1.0616,
      "step": 1110
    },
    {
      "epoch": 1.78,
      "grad_norm": 6.2752275466918945,
      "learning_rate": 0.00014929705215419503,
      "loss": 0.9484,
      "step": 1120
    },
    {
      "epoch": 1.79,
      "grad_norm": 6.395713806152344,
      "learning_rate": 0.000148843537414966,
      "loss": 1.0357,
      "step": 1130
    },
    {
      "epoch": 1.81,
      "grad_norm": 4.707973003387451,
      "learning_rate": 0.00014839002267573697,
      "loss": 0.922,
      "step": 1140
    },
    {
      "epoch": 1.83,
      "grad_norm": 6.663784980773926,
      "learning_rate": 0.00014793650793650795,
      "loss": 1.141,
      "step": 1150
    },
    {
      "epoch": 1.84,
      "grad_norm": 4.041499137878418,
      "learning_rate": 0.00014748299319727892,
      "loss": 0.9529,
      "step": 1160
    },
    {
      "epoch": 1.86,
      "grad_norm": 5.212307453155518,
      "learning_rate": 0.0001470294784580499,
      "loss": 0.6781,
      "step": 1170
    },
    {
      "epoch": 1.87,
      "grad_norm": 5.991590976715088,
      "learning_rate": 0.00014657596371882086,
      "loss": 0.8701,
      "step": 1180
    },
    {
      "epoch": 1.89,
      "grad_norm": 8.68984317779541,
      "learning_rate": 0.00014612244897959183,
      "loss": 1.0802,
      "step": 1190
    },
    {
      "epoch": 1.9,
      "grad_norm": 6.660585403442383,
      "learning_rate": 0.0001456689342403628,
      "loss": 0.8072,
      "step": 1200
    },
    {
      "epoch": 1.9,
      "eval_accuracy": 0.7694444444444445,
      "eval_loss": 0.77680903673172,
      "eval_runtime": 91.7003,
      "eval_samples_per_second": 27.481,
      "eval_steps_per_second": 3.435,
      "step": 1200
    },
    {
      "epoch": 1.92,
      "grad_norm": 3.435088872909546,
      "learning_rate": 0.0001452154195011338,
      "loss": 1.0272,
      "step": 1210
    },
    {
      "epoch": 1.94,
      "grad_norm": 3.200514078140259,
      "learning_rate": 0.00014476190476190475,
      "loss": 1.0246,
      "step": 1220
    },
    {
      "epoch": 1.95,
      "grad_norm": 3.882340669631958,
      "learning_rate": 0.00014430839002267575,
      "loss": 0.7745,
      "step": 1230
    },
    {
      "epoch": 1.97,
      "grad_norm": 3.1602838039398193,
      "learning_rate": 0.00014385487528344672,
      "loss": 0.9499,
      "step": 1240
    },
    {
      "epoch": 1.98,
      "grad_norm": 2.896543502807617,
      "learning_rate": 0.0001434013605442177,
      "loss": 0.9155,
      "step": 1250
    },
    {
      "epoch": 2.0,
      "grad_norm": 4.671875476837158,
      "learning_rate": 0.00014294784580498867,
      "loss": 0.8183,
      "step": 1260
    },
    {
      "epoch": 2.02,
      "grad_norm": 6.149994373321533,
      "learning_rate": 0.00014249433106575964,
      "loss": 0.8043,
      "step": 1270
    },
    {
      "epoch": 2.03,
      "grad_norm": 4.373509407043457,
      "learning_rate": 0.0001420408163265306,
      "loss": 0.7205,
      "step": 1280
    },
    {
      "epoch": 2.05,
      "grad_norm": 5.70250940322876,
      "learning_rate": 0.0001415873015873016,
      "loss": 0.8109,
      "step": 1290
    },
    {
      "epoch": 2.06,
      "grad_norm": 4.3769683837890625,
      "learning_rate": 0.00014113378684807256,
      "loss": 0.7739,
      "step": 1300
    },
    {
      "epoch": 2.06,
      "eval_accuracy": 0.7726190476190476,
      "eval_loss": 0.7623938322067261,
      "eval_runtime": 108.4142,
      "eval_samples_per_second": 23.244,
      "eval_steps_per_second": 2.906,
      "step": 1300
    },
    {
      "epoch": 2.08,
      "grad_norm": 8.936156272888184,
      "learning_rate": 0.00014068027210884353,
      "loss": 0.7361,
      "step": 1310
    },
    {
      "epoch": 2.1,
      "grad_norm": 6.855538845062256,
      "learning_rate": 0.00014022675736961453,
      "loss": 0.7778,
      "step": 1320
    },
    {
      "epoch": 2.11,
      "grad_norm": 2.5195322036743164,
      "learning_rate": 0.0001397732426303855,
      "loss": 0.7778,
      "step": 1330
    },
    {
      "epoch": 2.13,
      "grad_norm": 4.916295051574707,
      "learning_rate": 0.00013931972789115645,
      "loss": 0.8522,
      "step": 1340
    },
    {
      "epoch": 2.14,
      "grad_norm": 5.055403232574463,
      "learning_rate": 0.00013886621315192745,
      "loss": 0.6851,
      "step": 1350
    },
    {
      "epoch": 2.16,
      "grad_norm": 5.334274768829346,
      "learning_rate": 0.00013841269841269842,
      "loss": 0.9657,
      "step": 1360
    },
    {
      "epoch": 2.17,
      "grad_norm": 6.083943843841553,
      "learning_rate": 0.00013795918367346942,
      "loss": 0.6776,
      "step": 1370
    },
    {
      "epoch": 2.19,
      "grad_norm": 4.362452983856201,
      "learning_rate": 0.00013750566893424036,
      "loss": 0.7703,
      "step": 1380
    },
    {
      "epoch": 2.21,
      "grad_norm": 6.978400707244873,
      "learning_rate": 0.00013705215419501133,
      "loss": 0.7911,
      "step": 1390
    },
    {
      "epoch": 2.22,
      "grad_norm": 4.561004638671875,
      "learning_rate": 0.00013659863945578233,
      "loss": 0.6851,
      "step": 1400
    },
    {
      "epoch": 2.22,
      "eval_accuracy": 0.794047619047619,
      "eval_loss": 0.668690025806427,
      "eval_runtime": 91.4622,
      "eval_samples_per_second": 27.552,
      "eval_steps_per_second": 3.444,
      "step": 1400
    },
    {
      "epoch": 2.24,
      "grad_norm": 4.068347930908203,
      "learning_rate": 0.0001361451247165533,
      "loss": 0.6705,
      "step": 1410
    },
    {
      "epoch": 2.25,
      "grad_norm": 4.628905773162842,
      "learning_rate": 0.00013569160997732425,
      "loss": 0.8955,
      "step": 1420
    },
    {
      "epoch": 2.27,
      "grad_norm": 6.214323997497559,
      "learning_rate": 0.00013523809523809525,
      "loss": 0.785,
      "step": 1430
    },
    {
      "epoch": 2.29,
      "grad_norm": 3.4356954097747803,
      "learning_rate": 0.00013478458049886622,
      "loss": 0.6415,
      "step": 1440
    },
    {
      "epoch": 2.3,
      "grad_norm": 6.76524019241333,
      "learning_rate": 0.0001343310657596372,
      "loss": 0.8914,
      "step": 1450
    },
    {
      "epoch": 2.32,
      "grad_norm": 4.742693901062012,
      "learning_rate": 0.00013387755102040817,
      "loss": 0.8758,
      "step": 1460
    },
    {
      "epoch": 2.33,
      "grad_norm": 3.940936326980591,
      "learning_rate": 0.00013342403628117914,
      "loss": 0.6984,
      "step": 1470
    },
    {
      "epoch": 2.35,
      "grad_norm": 7.005272388458252,
      "learning_rate": 0.0001329705215419501,
      "loss": 0.833,
      "step": 1480
    },
    {
      "epoch": 2.37,
      "grad_norm": 5.6659722328186035,
      "learning_rate": 0.0001325170068027211,
      "loss": 0.7352,
      "step": 1490
    },
    {
      "epoch": 2.38,
      "grad_norm": 4.9742045402526855,
      "learning_rate": 0.00013206349206349206,
      "loss": 0.7496,
      "step": 1500
    },
    {
      "epoch": 2.38,
      "eval_accuracy": 0.7948412698412698,
      "eval_loss": 0.6806091070175171,
      "eval_runtime": 91.1192,
      "eval_samples_per_second": 27.656,
      "eval_steps_per_second": 3.457,
      "step": 1500
    },
    {
      "epoch": 2.4,
      "grad_norm": 5.499560832977295,
      "learning_rate": 0.00013160997732426303,
      "loss": 0.8274,
      "step": 1510
    },
    {
      "epoch": 2.41,
      "grad_norm": 4.391964435577393,
      "learning_rate": 0.00013115646258503403,
      "loss": 0.7892,
      "step": 1520
    },
    {
      "epoch": 2.43,
      "grad_norm": 4.317266464233398,
      "learning_rate": 0.000130702947845805,
      "loss": 0.7462,
      "step": 1530
    },
    {
      "epoch": 2.44,
      "grad_norm": 3.7989251613616943,
      "learning_rate": 0.00013024943310657597,
      "loss": 0.8322,
      "step": 1540
    },
    {
      "epoch": 2.46,
      "grad_norm": 4.737931251525879,
      "learning_rate": 0.00012979591836734695,
      "loss": 0.9,
      "step": 1550
    },
    {
      "epoch": 2.48,
      "grad_norm": 2.4748082160949707,
      "learning_rate": 0.00012934240362811792,
      "loss": 0.7613,
      "step": 1560
    },
    {
      "epoch": 2.49,
      "grad_norm": 5.137518882751465,
      "learning_rate": 0.00012888888888888892,
      "loss": 0.8596,
      "step": 1570
    },
    {
      "epoch": 2.51,
      "grad_norm": 3.9287266731262207,
      "learning_rate": 0.00012843537414965986,
      "loss": 0.9221,
      "step": 1580
    },
    {
      "epoch": 2.52,
      "grad_norm": 3.864816665649414,
      "learning_rate": 0.00012798185941043083,
      "loss": 0.6858,
      "step": 1590
    },
    {
      "epoch": 2.54,
      "grad_norm": 3.6453895568847656,
      "learning_rate": 0.00012752834467120183,
      "loss": 0.7352,
      "step": 1600
    },
    {
      "epoch": 2.54,
      "eval_accuracy": 0.7896825396825397,
      "eval_loss": 0.6942620277404785,
      "eval_runtime": 90.1988,
      "eval_samples_per_second": 27.938,
      "eval_steps_per_second": 3.492,
      "step": 1600
    },
    {
      "epoch": 2.56,
      "grad_norm": 6.48276424407959,
      "learning_rate": 0.0001270748299319728,
      "loss": 0.7989,
      "step": 1610
    },
    {
      "epoch": 2.57,
      "grad_norm": 3.8373072147369385,
      "learning_rate": 0.00012662131519274375,
      "loss": 0.7562,
      "step": 1620
    },
    {
      "epoch": 2.59,
      "grad_norm": 5.370635986328125,
      "learning_rate": 0.00012616780045351475,
      "loss": 0.8958,
      "step": 1630
    },
    {
      "epoch": 2.6,
      "grad_norm": 2.2488410472869873,
      "learning_rate": 0.00012571428571428572,
      "loss": 0.8911,
      "step": 1640
    },
    {
      "epoch": 2.62,
      "grad_norm": 4.261588096618652,
      "learning_rate": 0.0001252607709750567,
      "loss": 0.7508,
      "step": 1650
    },
    {
      "epoch": 2.63,
      "grad_norm": 4.3199286460876465,
      "learning_rate": 0.00012480725623582767,
      "loss": 0.6107,
      "step": 1660
    },
    {
      "epoch": 2.65,
      "grad_norm": 5.610477447509766,
      "learning_rate": 0.00012435374149659864,
      "loss": 0.5786,
      "step": 1670
    },
    {
      "epoch": 2.67,
      "grad_norm": 6.289084434509277,
      "learning_rate": 0.00012390022675736964,
      "loss": 0.5845,
      "step": 1680
    },
    {
      "epoch": 2.68,
      "grad_norm": 3.4954001903533936,
      "learning_rate": 0.0001234467120181406,
      "loss": 0.8207,
      "step": 1690
    },
    {
      "epoch": 2.7,
      "grad_norm": 3.2665460109710693,
      "learning_rate": 0.00012299319727891156,
      "loss": 0.7311,
      "step": 1700
    },
    {
      "epoch": 2.7,
      "eval_accuracy": 0.7714285714285715,
      "eval_loss": 0.7353097200393677,
      "eval_runtime": 91.0228,
      "eval_samples_per_second": 27.685,
      "eval_steps_per_second": 3.461,
      "step": 1700
    },
    {
      "epoch": 2.71,
      "grad_norm": 4.803609848022461,
      "learning_rate": 0.00012253968253968256,
      "loss": 0.7369,
      "step": 1710
    },
    {
      "epoch": 2.73,
      "grad_norm": 4.8724565505981445,
      "learning_rate": 0.00012208616780045353,
      "loss": 0.623,
      "step": 1720
    },
    {
      "epoch": 2.75,
      "grad_norm": 5.754215240478516,
      "learning_rate": 0.00012163265306122449,
      "loss": 0.6377,
      "step": 1730
    },
    {
      "epoch": 2.76,
      "grad_norm": 7.614988803863525,
      "learning_rate": 0.00012117913832199547,
      "loss": 0.8485,
      "step": 1740
    },
    {
      "epoch": 2.78,
      "grad_norm": 2.6383018493652344,
      "learning_rate": 0.00012072562358276644,
      "loss": 0.6942,
      "step": 1750
    },
    {
      "epoch": 2.79,
      "grad_norm": 5.747374057769775,
      "learning_rate": 0.00012027210884353742,
      "loss": 0.7562,
      "step": 1760
    },
    {
      "epoch": 2.81,
      "grad_norm": 5.084746360778809,
      "learning_rate": 0.0001198185941043084,
      "loss": 0.69,
      "step": 1770
    },
    {
      "epoch": 2.83,
      "grad_norm": 6.089473247528076,
      "learning_rate": 0.00011936507936507938,
      "loss": 0.9495,
      "step": 1780
    },
    {
      "epoch": 2.84,
      "grad_norm": 5.547908306121826,
      "learning_rate": 0.00011891156462585033,
      "loss": 0.6505,
      "step": 1790
    },
    {
      "epoch": 2.86,
      "grad_norm": 3.5899226665496826,
      "learning_rate": 0.00011845804988662132,
      "loss": 0.7181,
      "step": 1800
    },
    {
      "epoch": 2.86,
      "eval_accuracy": 0.792063492063492,
      "eval_loss": 0.6831231713294983,
      "eval_runtime": 91.1983,
      "eval_samples_per_second": 27.632,
      "eval_steps_per_second": 3.454,
      "step": 1800
    },
    {
      "epoch": 2.87,
      "grad_norm": 4.794887542724609,
      "learning_rate": 0.00011800453514739229,
      "loss": 0.6245,
      "step": 1810
    },
    {
      "epoch": 2.89,
      "grad_norm": 7.641270160675049,
      "learning_rate": 0.00011755102040816328,
      "loss": 0.8966,
      "step": 1820
    },
    {
      "epoch": 2.9,
      "grad_norm": 4.0006608963012695,
      "learning_rate": 0.00011709750566893425,
      "loss": 0.6106,
      "step": 1830
    },
    {
      "epoch": 2.92,
      "grad_norm": 5.983084201812744,
      "learning_rate": 0.00011664399092970522,
      "loss": 0.887,
      "step": 1840
    },
    {
      "epoch": 2.94,
      "grad_norm": 4.492617130279541,
      "learning_rate": 0.00011619047619047621,
      "loss": 0.7974,
      "step": 1850
    },
    {
      "epoch": 2.95,
      "grad_norm": 4.350939750671387,
      "learning_rate": 0.00011573696145124717,
      "loss": 0.8817,
      "step": 1860
    },
    {
      "epoch": 2.97,
      "grad_norm": 4.855531692504883,
      "learning_rate": 0.00011528344671201814,
      "loss": 0.9235,
      "step": 1870
    },
    {
      "epoch": 2.98,
      "grad_norm": 5.735949993133545,
      "learning_rate": 0.00011482993197278912,
      "loss": 0.6679,
      "step": 1880
    },
    {
      "epoch": 3.0,
      "grad_norm": 3.4668216705322266,
      "learning_rate": 0.0001143764172335601,
      "loss": 0.6783,
      "step": 1890
    },
    {
      "epoch": 3.02,
      "grad_norm": 4.096149921417236,
      "learning_rate": 0.00011392290249433107,
      "loss": 0.5986,
      "step": 1900
    },
    {
      "epoch": 3.02,
      "eval_accuracy": 0.7896825396825397,
      "eval_loss": 0.6930129528045654,
      "eval_runtime": 91.4004,
      "eval_samples_per_second": 27.571,
      "eval_steps_per_second": 3.446,
      "step": 1900
    },
    {
      "epoch": 3.03,
      "grad_norm": 3.9420833587646484,
      "learning_rate": 0.00011346938775510206,
      "loss": 0.5611,
      "step": 1910
    },
    {
      "epoch": 3.05,
      "grad_norm": 2.5048274993896484,
      "learning_rate": 0.00011301587301587301,
      "loss": 0.6619,
      "step": 1920
    },
    {
      "epoch": 3.06,
      "grad_norm": 4.899613380432129,
      "learning_rate": 0.00011256235827664399,
      "loss": 0.6889,
      "step": 1930
    },
    {
      "epoch": 3.08,
      "grad_norm": 4.282281398773193,
      "learning_rate": 0.00011210884353741497,
      "loss": 0.6182,
      "step": 1940
    },
    {
      "epoch": 3.1,
      "grad_norm": 3.1164541244506836,
      "learning_rate": 0.00011165532879818594,
      "loss": 0.6822,
      "step": 1950
    },
    {
      "epoch": 3.11,
      "grad_norm": 4.760287761688232,
      "learning_rate": 0.00011120181405895693,
      "loss": 0.6557,
      "step": 1960
    },
    {
      "epoch": 3.13,
      "grad_norm": 3.0668411254882812,
      "learning_rate": 0.0001107482993197279,
      "loss": 0.6537,
      "step": 1970
    },
    {
      "epoch": 3.14,
      "grad_norm": 4.527184009552002,
      "learning_rate": 0.00011029478458049886,
      "loss": 0.5513,
      "step": 1980
    },
    {
      "epoch": 3.16,
      "grad_norm": 2.029935598373413,
      "learning_rate": 0.00010984126984126986,
      "loss": 0.5102,
      "step": 1990
    },
    {
      "epoch": 3.17,
      "grad_norm": 4.294469833374023,
      "learning_rate": 0.00010938775510204082,
      "loss": 0.5716,
      "step": 2000
    },
    {
      "epoch": 3.17,
      "eval_accuracy": 0.8047619047619048,
      "eval_loss": 0.6684977412223816,
      "eval_runtime": 90.6284,
      "eval_samples_per_second": 27.806,
      "eval_steps_per_second": 3.476,
      "step": 2000
    },
    {
      "epoch": 3.19,
      "grad_norm": 6.498531818389893,
      "learning_rate": 0.00010893424036281179,
      "loss": 0.6862,
      "step": 2010
    },
    {
      "epoch": 3.21,
      "grad_norm": 5.384933948516846,
      "learning_rate": 0.00010848072562358278,
      "loss": 0.4793,
      "step": 2020
    },
    {
      "epoch": 3.22,
      "grad_norm": 4.124419212341309,
      "learning_rate": 0.00010802721088435375,
      "loss": 0.7055,
      "step": 2030
    },
    {
      "epoch": 3.24,
      "grad_norm": 4.198364734649658,
      "learning_rate": 0.00010757369614512471,
      "loss": 0.7231,
      "step": 2040
    },
    {
      "epoch": 3.25,
      "grad_norm": 4.410750389099121,
      "learning_rate": 0.00010712018140589571,
      "loss": 0.751,
      "step": 2050
    },
    {
      "epoch": 3.27,
      "grad_norm": 6.240994930267334,
      "learning_rate": 0.00010666666666666667,
      "loss": 0.6299,
      "step": 2060
    },
    {
      "epoch": 3.29,
      "grad_norm": 4.845404148101807,
      "learning_rate": 0.00010621315192743764,
      "loss": 0.6496,
      "step": 2070
    },
    {
      "epoch": 3.3,
      "grad_norm": 4.447664260864258,
      "learning_rate": 0.00010575963718820862,
      "loss": 0.5909,
      "step": 2080
    },
    {
      "epoch": 3.32,
      "grad_norm": 4.779788494110107,
      "learning_rate": 0.0001053061224489796,
      "loss": 0.6505,
      "step": 2090
    },
    {
      "epoch": 3.33,
      "grad_norm": 3.946617841720581,
      "learning_rate": 0.00010485260770975056,
      "loss": 0.5218,
      "step": 2100
    },
    {
      "epoch": 3.33,
      "eval_accuracy": 0.7916666666666666,
      "eval_loss": 0.7152296900749207,
      "eval_runtime": 89.3624,
      "eval_samples_per_second": 28.2,
      "eval_steps_per_second": 3.525,
      "step": 2100
    },
    {
      "epoch": 3.35,
      "grad_norm": 4.13429069519043,
      "learning_rate": 0.00010439909297052155,
      "loss": 0.6193,
      "step": 2110
    },
    {
      "epoch": 3.37,
      "grad_norm": 5.503479480743408,
      "learning_rate": 0.00010394557823129251,
      "loss": 0.6206,
      "step": 2120
    },
    {
      "epoch": 3.38,
      "grad_norm": 6.610657215118408,
      "learning_rate": 0.00010349206349206351,
      "loss": 0.7244,
      "step": 2130
    },
    {
      "epoch": 3.4,
      "grad_norm": 3.592276096343994,
      "learning_rate": 0.00010303854875283447,
      "loss": 0.502,
      "step": 2140
    },
    {
      "epoch": 3.41,
      "grad_norm": 7.070047378540039,
      "learning_rate": 0.00010258503401360544,
      "loss": 0.8749,
      "step": 2150
    },
    {
      "epoch": 3.43,
      "grad_norm": 8.58852481842041,
      "learning_rate": 0.00010213151927437643,
      "loss": 0.5832,
      "step": 2160
    },
    {
      "epoch": 3.44,
      "grad_norm": 5.416906356811523,
      "learning_rate": 0.0001016780045351474,
      "loss": 0.5901,
      "step": 2170
    },
    {
      "epoch": 3.46,
      "grad_norm": 6.415400981903076,
      "learning_rate": 0.00010122448979591836,
      "loss": 0.4842,
      "step": 2180
    },
    {
      "epoch": 3.48,
      "grad_norm": 1.809209942817688,
      "learning_rate": 0.00010077097505668936,
      "loss": 0.6591,
      "step": 2190
    },
    {
      "epoch": 3.49,
      "grad_norm": 4.30819034576416,
      "learning_rate": 0.00010031746031746032,
      "loss": 0.8469,
      "step": 2200
    },
    {
      "epoch": 3.49,
      "eval_accuracy": 0.801984126984127,
      "eval_loss": 0.6404625177383423,
      "eval_runtime": 91.7544,
      "eval_samples_per_second": 27.465,
      "eval_steps_per_second": 3.433,
      "step": 2200
    },
    {
      "epoch": 3.51,
      "grad_norm": 1.2012885808944702,
      "learning_rate": 9.98639455782313e-05,
      "loss": 0.5602,
      "step": 2210
    },
    {
      "epoch": 3.52,
      "grad_norm": 6.166879653930664,
      "learning_rate": 9.941043083900228e-05,
      "loss": 0.693,
      "step": 2220
    },
    {
      "epoch": 3.54,
      "grad_norm": 2.257598876953125,
      "learning_rate": 9.895691609977325e-05,
      "loss": 0.6969,
      "step": 2230
    },
    {
      "epoch": 3.56,
      "grad_norm": 3.866694450378418,
      "learning_rate": 9.850340136054422e-05,
      "loss": 0.7577,
      "step": 2240
    },
    {
      "epoch": 3.57,
      "grad_norm": 4.493105888366699,
      "learning_rate": 9.804988662131521e-05,
      "loss": 0.7494,
      "step": 2250
    },
    {
      "epoch": 3.59,
      "grad_norm": 6.990530490875244,
      "learning_rate": 9.759637188208617e-05,
      "loss": 0.6278,
      "step": 2260
    },
    {
      "epoch": 3.6,
      "grad_norm": 3.6842026710510254,
      "learning_rate": 9.714285714285715e-05,
      "loss": 0.6444,
      "step": 2270
    },
    {
      "epoch": 3.62,
      "grad_norm": 3.766533851623535,
      "learning_rate": 9.668934240362812e-05,
      "loss": 0.6254,
      "step": 2280
    },
    {
      "epoch": 3.63,
      "grad_norm": 3.9097561836242676,
      "learning_rate": 9.62358276643991e-05,
      "loss": 0.6871,
      "step": 2290
    },
    {
      "epoch": 3.65,
      "grad_norm": 5.504273414611816,
      "learning_rate": 9.578231292517007e-05,
      "loss": 0.5783,
      "step": 2300
    },
    {
      "epoch": 3.65,
      "eval_accuracy": 0.7956349206349206,
      "eval_loss": 0.6727890968322754,
      "eval_runtime": 91.691,
      "eval_samples_per_second": 27.484,
      "eval_steps_per_second": 3.435,
      "step": 2300
    },
    {
      "epoch": 3.67,
      "grad_norm": 6.009402751922607,
      "learning_rate": 9.532879818594105e-05,
      "loss": 0.6296,
      "step": 2310
    },
    {
      "epoch": 3.68,
      "grad_norm": 3.723788022994995,
      "learning_rate": 9.487528344671203e-05,
      "loss": 0.61,
      "step": 2320
    },
    {
      "epoch": 3.7,
      "grad_norm": 1.6886146068572998,
      "learning_rate": 9.4421768707483e-05,
      "loss": 0.6042,
      "step": 2330
    },
    {
      "epoch": 3.71,
      "grad_norm": 3.2953665256500244,
      "learning_rate": 9.396825396825397e-05,
      "loss": 0.5417,
      "step": 2340
    },
    {
      "epoch": 3.73,
      "grad_norm": 5.710081100463867,
      "learning_rate": 9.351473922902494e-05,
      "loss": 0.4744,
      "step": 2350
    },
    {
      "epoch": 3.75,
      "grad_norm": 8.341416358947754,
      "learning_rate": 9.306122448979592e-05,
      "loss": 0.6468,
      "step": 2360
    },
    {
      "epoch": 3.76,
      "grad_norm": 5.668067455291748,
      "learning_rate": 9.26077097505669e-05,
      "loss": 0.7058,
      "step": 2370
    },
    {
      "epoch": 3.78,
      "grad_norm": 7.821765422821045,
      "learning_rate": 9.215419501133787e-05,
      "loss": 0.6014,
      "step": 2380
    },
    {
      "epoch": 3.79,
      "grad_norm": 3.075150728225708,
      "learning_rate": 9.170068027210885e-05,
      "loss": 0.5857,
      "step": 2390
    },
    {
      "epoch": 3.81,
      "grad_norm": 5.349681377410889,
      "learning_rate": 9.124716553287982e-05,
      "loss": 0.7202,
      "step": 2400
    },
    {
      "epoch": 3.81,
      "eval_accuracy": 0.8154761904761905,
      "eval_loss": 0.6007378697395325,
      "eval_runtime": 91.6882,
      "eval_samples_per_second": 27.484,
      "eval_steps_per_second": 3.436,
      "step": 2400
    },
    {
      "epoch": 3.83,
      "grad_norm": 6.9100117683410645,
      "learning_rate": 9.079365079365079e-05,
      "loss": 0.5335,
      "step": 2410
    },
    {
      "epoch": 3.84,
      "grad_norm": NaN,
      "learning_rate": 9.038548752834468e-05,
      "loss": 0.7311,
      "step": 2420
    },
    {
      "epoch": 3.86,
      "grad_norm": 3.347836494445801,
      "learning_rate": 8.993197278911565e-05,
      "loss": 0.5405,
      "step": 2430
    },
    {
      "epoch": 3.87,
      "grad_norm": 10.232234954833984,
      "learning_rate": 8.947845804988662e-05,
      "loss": 0.7061,
      "step": 2440
    },
    {
      "epoch": 3.89,
      "grad_norm": 5.486011981964111,
      "learning_rate": 8.902494331065761e-05,
      "loss": 0.6905,
      "step": 2450
    },
    {
      "epoch": 3.9,
      "grad_norm": 4.699936389923096,
      "learning_rate": 8.857142857142857e-05,
      "loss": 0.5979,
      "step": 2460
    },
    {
      "epoch": 3.92,
      "grad_norm": 7.7097296714782715,
      "learning_rate": 8.811791383219955e-05,
      "loss": 0.5904,
      "step": 2470
    },
    {
      "epoch": 3.94,
      "grad_norm": 3.3451244831085205,
      "learning_rate": 8.766439909297052e-05,
      "loss": 0.5359,
      "step": 2480
    },
    {
      "epoch": 3.95,
      "grad_norm": 7.01662540435791,
      "learning_rate": 8.72108843537415e-05,
      "loss": 0.728,
      "step": 2490
    },
    {
      "epoch": 3.97,
      "grad_norm": 4.137801170349121,
      "learning_rate": 8.675736961451247e-05,
      "loss": 0.5525,
      "step": 2500
    },
    {
      "epoch": 3.97,
      "eval_accuracy": 0.8055555555555556,
      "eval_loss": 0.6558998823165894,
      "eval_runtime": 94.8951,
      "eval_samples_per_second": 26.556,
      "eval_steps_per_second": 3.319,
      "step": 2500
    },
    {
      "epoch": 3.98,
      "grad_norm": 6.564231872558594,
      "learning_rate": 8.630385487528345e-05,
      "loss": 0.6735,
      "step": 2510
    },
    {
      "epoch": 4.0,
      "grad_norm": 5.252976894378662,
      "learning_rate": 8.585034013605443e-05,
      "loss": 0.6903,
      "step": 2520
    },
    {
      "epoch": 4.02,
      "grad_norm": 4.289461612701416,
      "learning_rate": 8.53968253968254e-05,
      "loss": 0.5104,
      "step": 2530
    },
    {
      "epoch": 4.03,
      "grad_norm": 4.131557941436768,
      "learning_rate": 8.494331065759637e-05,
      "loss": 0.5064,
      "step": 2540
    },
    {
      "epoch": 4.05,
      "grad_norm": 3.5607643127441406,
      "learning_rate": 8.448979591836736e-05,
      "loss": 0.4825,
      "step": 2550
    },
    {
      "epoch": 4.06,
      "grad_norm": 2.5051770210266113,
      "learning_rate": 8.403628117913832e-05,
      "loss": 0.5668,
      "step": 2560
    },
    {
      "epoch": 4.08,
      "grad_norm": 1.6936321258544922,
      "learning_rate": 8.35827664399093e-05,
      "loss": 0.4412,
      "step": 2570
    },
    {
      "epoch": 4.1,
      "grad_norm": 3.99070143699646,
      "learning_rate": 8.312925170068027e-05,
      "loss": 0.5046,
      "step": 2580
    },
    {
      "epoch": 4.11,
      "grad_norm": 4.099004745483398,
      "learning_rate": 8.267573696145126e-05,
      "loss": 0.5753,
      "step": 2590
    },
    {
      "epoch": 4.13,
      "grad_norm": 3.5789458751678467,
      "learning_rate": 8.222222222222222e-05,
      "loss": 0.519,
      "step": 2600
    },
    {
      "epoch": 4.13,
      "eval_accuracy": 0.8222222222222222,
      "eval_loss": 0.5868101716041565,
      "eval_runtime": 93.686,
      "eval_samples_per_second": 26.898,
      "eval_steps_per_second": 3.362,
      "step": 2600
    },
    {
      "epoch": 4.14,
      "grad_norm": 4.785660266876221,
      "learning_rate": 8.17687074829932e-05,
      "loss": 0.4747,
      "step": 2610
    },
    {
      "epoch": 4.16,
      "grad_norm": 3.662282705307007,
      "learning_rate": 8.131519274376418e-05,
      "loss": 0.4822,
      "step": 2620
    },
    {
      "epoch": 4.17,
      "grad_norm": 3.897036075592041,
      "learning_rate": 8.086167800453515e-05,
      "loss": 0.4848,
      "step": 2630
    },
    {
      "epoch": 4.19,
      "grad_norm": 5.990245342254639,
      "learning_rate": 8.040816326530612e-05,
      "loss": 0.5569,
      "step": 2640
    },
    {
      "epoch": 4.21,
      "grad_norm": 4.004576206207275,
      "learning_rate": 7.99546485260771e-05,
      "loss": 0.4929,
      "step": 2650
    },
    {
      "epoch": 4.22,
      "grad_norm": 5.0893754959106445,
      "learning_rate": 7.950113378684808e-05,
      "loss": 0.5564,
      "step": 2660
    },
    {
      "epoch": 4.24,
      "grad_norm": 3.9271037578582764,
      "learning_rate": 7.904761904761905e-05,
      "loss": 0.382,
      "step": 2670
    },
    {
      "epoch": 4.25,
      "grad_norm": 4.929361820220947,
      "learning_rate": 7.859410430839002e-05,
      "loss": 0.5502,
      "step": 2680
    },
    {
      "epoch": 4.27,
      "grad_norm": 4.333121299743652,
      "learning_rate": 7.814058956916101e-05,
      "loss": 0.5877,
      "step": 2690
    },
    {
      "epoch": 4.29,
      "grad_norm": 3.787369728088379,
      "learning_rate": 7.768707482993197e-05,
      "loss": 0.6171,
      "step": 2700
    },
    {
      "epoch": 4.29,
      "eval_accuracy": 0.8103174603174603,
      "eval_loss": 0.6157482266426086,
      "eval_runtime": 91.6071,
      "eval_samples_per_second": 27.509,
      "eval_steps_per_second": 3.439,
      "step": 2700
    },
    {
      "epoch": 4.3,
      "grad_norm": 4.089489936828613,
      "learning_rate": 7.723356009070295e-05,
      "loss": 0.5279,
      "step": 2710
    },
    {
      "epoch": 4.32,
      "grad_norm": 5.687109470367432,
      "learning_rate": 7.678004535147393e-05,
      "loss": 0.4443,
      "step": 2720
    },
    {
      "epoch": 4.33,
      "grad_norm": 3.6936216354370117,
      "learning_rate": 7.632653061224491e-05,
      "loss": 0.533,
      "step": 2730
    },
    {
      "epoch": 4.35,
      "grad_norm": 2.403999090194702,
      "learning_rate": 7.587301587301587e-05,
      "loss": 0.5214,
      "step": 2740
    },
    {
      "epoch": 4.37,
      "grad_norm": 5.553544044494629,
      "learning_rate": 7.541950113378686e-05,
      "loss": 0.5621,
      "step": 2750
    },
    {
      "epoch": 4.38,
      "grad_norm": 2.7590301036834717,
      "learning_rate": 7.496598639455783e-05,
      "loss": 0.4752,
      "step": 2760
    },
    {
      "epoch": 4.4,
      "grad_norm": 6.257105350494385,
      "learning_rate": 7.45124716553288e-05,
      "loss": 0.5196,
      "step": 2770
    },
    {
      "epoch": 4.41,
      "grad_norm": 6.366973400115967,
      "learning_rate": 7.405895691609977e-05,
      "loss": 0.4779,
      "step": 2780
    },
    {
      "epoch": 4.43,
      "grad_norm": 4.746866703033447,
      "learning_rate": 7.360544217687076e-05,
      "loss": 0.4031,
      "step": 2790
    },
    {
      "epoch": 4.44,
      "grad_norm": 3.278367519378662,
      "learning_rate": 7.315192743764173e-05,
      "loss": 0.5401,
      "step": 2800
    },
    {
      "epoch": 4.44,
      "eval_accuracy": 0.8083333333333333,
      "eval_loss": 0.6119987368583679,
      "eval_runtime": 91.505,
      "eval_samples_per_second": 27.539,
      "eval_steps_per_second": 3.442,
      "step": 2800
    },
    {
      "epoch": 4.46,
      "grad_norm": 5.5712504386901855,
      "learning_rate": 7.26984126984127e-05,
      "loss": 0.5565,
      "step": 2810
    },
    {
      "epoch": 4.48,
      "grad_norm": 4.8718767166137695,
      "learning_rate": 7.224489795918368e-05,
      "loss": 0.5219,
      "step": 2820
    },
    {
      "epoch": 4.49,
      "grad_norm": 5.685264587402344,
      "learning_rate": 7.179138321995466e-05,
      "loss": 0.5021,
      "step": 2830
    },
    {
      "epoch": 4.51,
      "grad_norm": 6.191750526428223,
      "learning_rate": 7.133786848072562e-05,
      "loss": 0.5115,
      "step": 2840
    },
    {
      "epoch": 4.52,
      "grad_norm": 5.250429630279541,
      "learning_rate": 7.08843537414966e-05,
      "loss": 0.5805,
      "step": 2850
    },
    {
      "epoch": 4.54,
      "grad_norm": 5.834912300109863,
      "learning_rate": 7.043083900226758e-05,
      "loss": 0.584,
      "step": 2860
    },
    {
      "epoch": 4.56,
      "grad_norm": 2.52066707611084,
      "learning_rate": 6.997732426303855e-05,
      "loss": 0.3752,
      "step": 2870
    },
    {
      "epoch": 4.57,
      "grad_norm": 2.4411842823028564,
      "learning_rate": 6.952380952380952e-05,
      "loss": 0.5748,
      "step": 2880
    },
    {
      "epoch": 4.59,
      "grad_norm": 4.865506649017334,
      "learning_rate": 6.907029478458051e-05,
      "loss": 0.6119,
      "step": 2890
    },
    {
      "epoch": 4.6,
      "grad_norm": 2.109516143798828,
      "learning_rate": 6.861678004535148e-05,
      "loss": 0.6105,
      "step": 2900
    },
    {
      "epoch": 4.6,
      "eval_accuracy": 0.8325396825396826,
      "eval_loss": 0.5618996024131775,
      "eval_runtime": 90.5734,
      "eval_samples_per_second": 27.823,
      "eval_steps_per_second": 3.478,
      "step": 2900
    },
    {
      "epoch": 4.62,
      "grad_norm": 6.931108474731445,
      "learning_rate": 6.816326530612245e-05,
      "loss": 0.6452,
      "step": 2910
    },
    {
      "epoch": 4.63,
      "grad_norm": 5.011793613433838,
      "learning_rate": 6.770975056689343e-05,
      "loss": 0.3178,
      "step": 2920
    },
    {
      "epoch": 4.65,
      "grad_norm": 2.996354818344116,
      "learning_rate": 6.72562358276644e-05,
      "loss": 0.3436,
      "step": 2930
    },
    {
      "epoch": 4.67,
      "grad_norm": 4.3052754402160645,
      "learning_rate": 6.680272108843538e-05,
      "loss": 0.4743,
      "step": 2940
    },
    {
      "epoch": 4.68,
      "grad_norm": 5.777819633483887,
      "learning_rate": 6.634920634920636e-05,
      "loss": 0.6294,
      "step": 2950
    },
    {
      "epoch": 4.7,
      "grad_norm": 4.6029157638549805,
      "learning_rate": 6.589569160997733e-05,
      "loss": 0.495,
      "step": 2960
    },
    {
      "epoch": 4.71,
      "grad_norm": 3.34942889213562,
      "learning_rate": 6.54421768707483e-05,
      "loss": 0.5216,
      "step": 2970
    },
    {
      "epoch": 4.73,
      "grad_norm": 3.747777223587036,
      "learning_rate": 6.498866213151927e-05,
      "loss": 0.4908,
      "step": 2980
    },
    {
      "epoch": 4.75,
      "grad_norm": 3.217052459716797,
      "learning_rate": 6.453514739229024e-05,
      "loss": 0.5847,
      "step": 2990
    },
    {
      "epoch": 4.76,
      "grad_norm": 2.1214661598205566,
      "learning_rate": 6.408163265306123e-05,
      "loss": 0.7497,
      "step": 3000
    },
    {
      "epoch": 4.76,
      "eval_accuracy": 0.8301587301587302,
      "eval_loss": 0.5859270095825195,
      "eval_runtime": 89.9674,
      "eval_samples_per_second": 28.01,
      "eval_steps_per_second": 3.501,
      "step": 3000
    },
    {
      "epoch": 4.78,
      "grad_norm": 3.495818614959717,
      "learning_rate": 6.36281179138322e-05,
      "loss": 0.4944,
      "step": 3010
    },
    {
      "epoch": 4.79,
      "grad_norm": 6.128993034362793,
      "learning_rate": 6.317460317460318e-05,
      "loss": 0.5474,
      "step": 3020
    },
    {
      "epoch": 4.81,
      "grad_norm": 2.1895864009857178,
      "learning_rate": 6.272108843537415e-05,
      "loss": 0.5447,
      "step": 3030
    },
    {
      "epoch": 4.83,
      "grad_norm": 4.312340259552002,
      "learning_rate": 6.226757369614513e-05,
      "loss": 0.5194,
      "step": 3040
    },
    {
      "epoch": 4.84,
      "grad_norm": 5.473392009735107,
      "learning_rate": 6.181405895691609e-05,
      "loss": 0.5538,
      "step": 3050
    },
    {
      "epoch": 4.86,
      "grad_norm": 2.302345037460327,
      "learning_rate": 6.136054421768708e-05,
      "loss": 0.4733,
      "step": 3060
    },
    {
      "epoch": 4.87,
      "grad_norm": 3.3499696254730225,
      "learning_rate": 6.090702947845806e-05,
      "loss": 0.755,
      "step": 3070
    },
    {
      "epoch": 4.89,
      "grad_norm": 3.382891893386841,
      "learning_rate": 6.045351473922902e-05,
      "loss": 0.4829,
      "step": 3080
    },
    {
      "epoch": 4.9,
      "grad_norm": 3.4998888969421387,
      "learning_rate": 6e-05,
      "loss": 0.4543,
      "step": 3090
    },
    {
      "epoch": 4.92,
      "grad_norm": 3.2224600315093994,
      "learning_rate": 5.954648526077098e-05,
      "loss": 0.4856,
      "step": 3100
    },
    {
      "epoch": 4.92,
      "eval_accuracy": 0.8261904761904761,
      "eval_loss": 0.5833402872085571,
      "eval_runtime": 90.9659,
      "eval_samples_per_second": 27.703,
      "eval_steps_per_second": 3.463,
      "step": 3100
    },
    {
      "epoch": 4.94,
      "grad_norm": 2.1335177421569824,
      "learning_rate": 5.909297052154196e-05,
      "loss": 0.3995,
      "step": 3110
    },
    {
      "epoch": 4.95,
      "grad_norm": 4.156594753265381,
      "learning_rate": 5.8639455782312925e-05,
      "loss": 0.5539,
      "step": 3120
    },
    {
      "epoch": 4.97,
      "grad_norm": 5.811089515686035,
      "learning_rate": 5.8185941043083904e-05,
      "loss": 0.579,
      "step": 3130
    },
    {
      "epoch": 4.98,
      "grad_norm": 1.908595323562622,
      "learning_rate": 5.773242630385488e-05,
      "loss": 0.5315,
      "step": 3140
    },
    {
      "epoch": 5.0,
      "grad_norm": 2.8674533367156982,
      "learning_rate": 5.727891156462585e-05,
      "loss": 0.5577,
      "step": 3150
    },
    {
      "epoch": 5.02,
      "grad_norm": 1.5174939632415771,
      "learning_rate": 5.682539682539683e-05,
      "loss": 0.5052,
      "step": 3160
    },
    {
      "epoch": 5.03,
      "grad_norm": 3.2664871215820312,
      "learning_rate": 5.637188208616781e-05,
      "loss": 0.3135,
      "step": 3170
    },
    {
      "epoch": 5.05,
      "grad_norm": 5.35206937789917,
      "learning_rate": 5.5918367346938786e-05,
      "loss": 0.5859,
      "step": 3180
    },
    {
      "epoch": 5.06,
      "grad_norm": 2.6170833110809326,
      "learning_rate": 5.546485260770975e-05,
      "loss": 0.4511,
      "step": 3190
    },
    {
      "epoch": 5.08,
      "grad_norm": 3.2739861011505127,
      "learning_rate": 5.501133786848073e-05,
      "loss": 0.4959,
      "step": 3200
    },
    {
      "epoch": 5.08,
      "eval_accuracy": 0.832936507936508,
      "eval_loss": 0.5703846216201782,
      "eval_runtime": 92.0072,
      "eval_samples_per_second": 27.389,
      "eval_steps_per_second": 3.424,
      "step": 3200
    },
    {
      "epoch": 5.1,
      "grad_norm": 0.7258143424987793,
      "learning_rate": 5.455782312925171e-05,
      "loss": 0.4922,
      "step": 3210
    },
    {
      "epoch": 5.11,
      "grad_norm": 1.8463611602783203,
      "learning_rate": 5.4104308390022675e-05,
      "loss": 0.3888,
      "step": 3220
    },
    {
      "epoch": 5.13,
      "grad_norm": 4.082030296325684,
      "learning_rate": 5.3650793650793654e-05,
      "loss": 0.4486,
      "step": 3230
    },
    {
      "epoch": 5.14,
      "grad_norm": 1.123199462890625,
      "learning_rate": 5.319727891156463e-05,
      "loss": 0.4229,
      "step": 3240
    },
    {
      "epoch": 5.16,
      "grad_norm": 2.546860694885254,
      "learning_rate": 5.2743764172335605e-05,
      "loss": 0.5514,
      "step": 3250
    },
    {
      "epoch": 5.17,
      "grad_norm": 6.097323417663574,
      "learning_rate": 5.229024943310658e-05,
      "loss": 0.4391,
      "step": 3260
    },
    {
      "epoch": 5.19,
      "grad_norm": 3.1347315311431885,
      "learning_rate": 5.1836734693877557e-05,
      "loss": 0.4455,
      "step": 3270
    },
    {
      "epoch": 5.21,
      "grad_norm": 2.0431602001190186,
      "learning_rate": 5.138321995464853e-05,
      "loss": 0.3076,
      "step": 3280
    },
    {
      "epoch": 5.22,
      "grad_norm": 4.452468395233154,
      "learning_rate": 5.09297052154195e-05,
      "loss": 0.4442,
      "step": 3290
    },
    {
      "epoch": 5.24,
      "grad_norm": 5.282511234283447,
      "learning_rate": 5.047619047619048e-05,
      "loss": 0.4413,
      "step": 3300
    },
    {
      "epoch": 5.24,
      "eval_accuracy": 0.819047619047619,
      "eval_loss": 0.6217456459999084,
      "eval_runtime": 91.9773,
      "eval_samples_per_second": 27.398,
      "eval_steps_per_second": 3.425,
      "step": 3300
    },
    {
      "epoch": 5.25,
      "grad_norm": 8.806018829345703,
      "learning_rate": 5.002267573696145e-05,
      "loss": 0.4263,
      "step": 3310
    },
    {
      "epoch": 5.27,
      "grad_norm": 4.782093048095703,
      "learning_rate": 4.9569160997732425e-05,
      "loss": 0.6431,
      "step": 3320
    },
    {
      "epoch": 5.29,
      "grad_norm": 5.115506172180176,
      "learning_rate": 4.9115646258503404e-05,
      "loss": 0.4876,
      "step": 3330
    },
    {
      "epoch": 5.3,
      "grad_norm": 4.416604995727539,
      "learning_rate": 4.8662131519274376e-05,
      "loss": 0.371,
      "step": 3340
    },
    {
      "epoch": 5.32,
      "grad_norm": 3.7609243392944336,
      "learning_rate": 4.820861678004535e-05,
      "loss": 0.3873,
      "step": 3350
    },
    {
      "epoch": 5.33,
      "grad_norm": 4.652336120605469,
      "learning_rate": 4.775510204081633e-05,
      "loss": 0.4134,
      "step": 3360
    },
    {
      "epoch": 5.35,
      "grad_norm": 5.700862407684326,
      "learning_rate": 4.73015873015873e-05,
      "loss": 0.4463,
      "step": 3370
    },
    {
      "epoch": 5.37,
      "grad_norm": 4.792759895324707,
      "learning_rate": 4.684807256235828e-05,
      "loss": 0.4243,
      "step": 3380
    },
    {
      "epoch": 5.38,
      "grad_norm": 4.60031795501709,
      "learning_rate": 4.639455782312925e-05,
      "loss": 0.3602,
      "step": 3390
    },
    {
      "epoch": 5.4,
      "grad_norm": 6.657435417175293,
      "learning_rate": 4.594104308390023e-05,
      "loss": 0.4513,
      "step": 3400
    },
    {
      "epoch": 5.4,
      "eval_accuracy": 0.8293650793650794,
      "eval_loss": 0.5750200748443604,
      "eval_runtime": 96.412,
      "eval_samples_per_second": 26.138,
      "eval_steps_per_second": 3.267,
      "step": 3400
    },
    {
      "epoch": 5.41,
      "grad_norm": 3.526585578918457,
      "learning_rate": 4.54875283446712e-05,
      "loss": 0.374,
      "step": 3410
    },
    {
      "epoch": 5.43,
      "grad_norm": 2.742906332015991,
      "learning_rate": 4.5034013605442174e-05,
      "loss": 0.3822,
      "step": 3420
    },
    {
      "epoch": 5.44,
      "grad_norm": 3.6437864303588867,
      "learning_rate": 4.4580498866213154e-05,
      "loss": 0.4369,
      "step": 3430
    },
    {
      "epoch": 5.46,
      "grad_norm": 2.621497869491577,
      "learning_rate": 4.4126984126984126e-05,
      "loss": 0.3399,
      "step": 3440
    },
    {
      "epoch": 5.48,
      "grad_norm": 1.8672789335250854,
      "learning_rate": 4.3673469387755105e-05,
      "loss": 0.584,
      "step": 3450
    },
    {
      "epoch": 5.49,
      "grad_norm": 1.4387513399124146,
      "learning_rate": 4.321995464852608e-05,
      "loss": 0.4242,
      "step": 3460
    },
    {
      "epoch": 5.51,
      "grad_norm": 0.7556703090667725,
      "learning_rate": 4.2766439909297056e-05,
      "loss": 0.462,
      "step": 3470
    },
    {
      "epoch": 5.52,
      "grad_norm": 2.0518500804901123,
      "learning_rate": 4.231292517006803e-05,
      "loss": 0.4239,
      "step": 3480
    },
    {
      "epoch": 5.54,
      "grad_norm": 4.876095771789551,
      "learning_rate": 4.1859410430839e-05,
      "loss": 0.5488,
      "step": 3490
    },
    {
      "epoch": 5.56,
      "grad_norm": 1.6563076972961426,
      "learning_rate": 4.140589569160998e-05,
      "loss": 0.3987,
      "step": 3500
    },
    {
      "epoch": 5.56,
      "eval_accuracy": 0.8341269841269842,
      "eval_loss": 0.5825861096382141,
      "eval_runtime": 98.7535,
      "eval_samples_per_second": 25.518,
      "eval_steps_per_second": 3.19,
      "step": 3500
    },
    {
      "epoch": 5.57,
      "grad_norm": 6.538485050201416,
      "learning_rate": 4.095238095238095e-05,
      "loss": 0.5583,
      "step": 3510
    },
    {
      "epoch": 5.59,
      "grad_norm": 4.381786823272705,
      "learning_rate": 4.049886621315193e-05,
      "loss": 0.408,
      "step": 3520
    },
    {
      "epoch": 5.6,
      "grad_norm": 3.696018695831299,
      "learning_rate": 4.00453514739229e-05,
      "loss": 0.4421,
      "step": 3530
    },
    {
      "epoch": 5.62,
      "grad_norm": 3.3752057552337646,
      "learning_rate": 3.9591836734693876e-05,
      "loss": 0.4057,
      "step": 3540
    },
    {
      "epoch": 5.63,
      "grad_norm": 6.889618396759033,
      "learning_rate": 3.9138321995464855e-05,
      "loss": 0.4479,
      "step": 3550
    },
    {
      "epoch": 5.65,
      "grad_norm": 5.5090131759643555,
      "learning_rate": 3.868480725623583e-05,
      "loss": 0.6009,
      "step": 3560
    },
    {
      "epoch": 5.67,
      "grad_norm": 2.368633985519409,
      "learning_rate": 3.8231292517006806e-05,
      "loss": 0.4276,
      "step": 3570
    },
    {
      "epoch": 5.68,
      "grad_norm": 4.297762393951416,
      "learning_rate": 3.777777777777778e-05,
      "loss": 0.4368,
      "step": 3580
    },
    {
      "epoch": 5.7,
      "grad_norm": 3.7877378463745117,
      "learning_rate": 3.732426303854876e-05,
      "loss": 0.3481,
      "step": 3590
    },
    {
      "epoch": 5.71,
      "grad_norm": 2.9667625427246094,
      "learning_rate": 3.687074829931973e-05,
      "loss": 0.4395,
      "step": 3600
    },
    {
      "epoch": 5.71,
      "eval_accuracy": 0.8384920634920635,
      "eval_loss": 0.5753714442253113,
      "eval_runtime": 97.0735,
      "eval_samples_per_second": 25.96,
      "eval_steps_per_second": 3.245,
      "step": 3600
    },
    {
      "epoch": 5.73,
      "grad_norm": 5.830836296081543,
      "learning_rate": 3.64172335600907e-05,
      "loss": 0.4859,
      "step": 3610
    },
    {
      "epoch": 5.75,
      "grad_norm": 5.262144088745117,
      "learning_rate": 3.596371882086168e-05,
      "loss": 0.4409,
      "step": 3620
    },
    {
      "epoch": 5.76,
      "grad_norm": 0.9806166291236877,
      "learning_rate": 3.551020408163265e-05,
      "loss": 0.3884,
      "step": 3630
    },
    {
      "epoch": 5.78,
      "grad_norm": 3.1864867210388184,
      "learning_rate": 3.505668934240363e-05,
      "loss": 0.3753,
      "step": 3640
    },
    {
      "epoch": 5.79,
      "grad_norm": 2.679213762283325,
      "learning_rate": 3.4603174603174604e-05,
      "loss": 0.3764,
      "step": 3650
    },
    {
      "epoch": 5.81,
      "grad_norm": 1.3040461540222168,
      "learning_rate": 3.4149659863945583e-05,
      "loss": 0.3632,
      "step": 3660
    },
    {
      "epoch": 5.83,
      "grad_norm": 6.803734302520752,
      "learning_rate": 3.3696145124716556e-05,
      "loss": 0.6695,
      "step": 3670
    },
    {
      "epoch": 5.84,
      "grad_norm": 3.717195510864258,
      "learning_rate": 3.324263038548753e-05,
      "loss": 0.4226,
      "step": 3680
    },
    {
      "epoch": 5.86,
      "grad_norm": 0.3749280273914337,
      "learning_rate": 3.278911564625851e-05,
      "loss": 0.4404,
      "step": 3690
    },
    {
      "epoch": 5.87,
      "grad_norm": 3.4548332691192627,
      "learning_rate": 3.233560090702948e-05,
      "loss": 0.4669,
      "step": 3700
    },
    {
      "epoch": 5.87,
      "eval_accuracy": 0.8357142857142857,
      "eval_loss": 0.5652737021446228,
      "eval_runtime": 98.0382,
      "eval_samples_per_second": 25.704,
      "eval_steps_per_second": 3.213,
      "step": 3700
    },
    {
      "epoch": 5.89,
      "grad_norm": 2.5312252044677734,
      "learning_rate": 3.188208616780046e-05,
      "loss": 0.4786,
      "step": 3710
    },
    {
      "epoch": 5.9,
      "grad_norm": 4.841963291168213,
      "learning_rate": 3.142857142857143e-05,
      "loss": 0.5179,
      "step": 3720
    },
    {
      "epoch": 5.92,
      "grad_norm": 4.394898414611816,
      "learning_rate": 3.097505668934241e-05,
      "loss": 0.5062,
      "step": 3730
    },
    {
      "epoch": 5.94,
      "grad_norm": 4.590970516204834,
      "learning_rate": 3.052154195011338e-05,
      "loss": 0.3639,
      "step": 3740
    },
    {
      "epoch": 5.95,
      "grad_norm": 4.6838459968566895,
      "learning_rate": 3.0068027210884354e-05,
      "loss": 0.3552,
      "step": 3750
    },
    {
      "epoch": 5.97,
      "grad_norm": 4.303285121917725,
      "learning_rate": 2.961451247165533e-05,
      "loss": 0.3526,
      "step": 3760
    },
    {
      "epoch": 5.98,
      "grad_norm": 6.06368350982666,
      "learning_rate": 2.9160997732426306e-05,
      "loss": 0.3266,
      "step": 3770
    },
    {
      "epoch": 6.0,
      "grad_norm": 4.715054988861084,
      "learning_rate": 2.870748299319728e-05,
      "loss": 0.4612,
      "step": 3780
    },
    {
      "epoch": 6.02,
      "grad_norm": 3.029750108718872,
      "learning_rate": 2.8253968253968253e-05,
      "loss": 0.3938,
      "step": 3790
    },
    {
      "epoch": 6.03,
      "grad_norm": 6.53373384475708,
      "learning_rate": 2.7800453514739233e-05,
      "loss": 0.4005,
      "step": 3800
    },
    {
      "epoch": 6.03,
      "eval_accuracy": 0.8376984126984127,
      "eval_loss": 0.542424201965332,
      "eval_runtime": 92.2864,
      "eval_samples_per_second": 27.306,
      "eval_steps_per_second": 3.413,
      "step": 3800
    },
    {
      "epoch": 6.05,
      "grad_norm": 2.6752915382385254,
      "learning_rate": 2.7346938775510205e-05,
      "loss": 0.311,
      "step": 3810
    },
    {
      "epoch": 6.06,
      "grad_norm": 5.5302863121032715,
      "learning_rate": 2.6893424036281177e-05,
      "loss": 0.4392,
      "step": 3820
    },
    {
      "epoch": 6.08,
      "grad_norm": 3.140334367752075,
      "learning_rate": 2.6439909297052156e-05,
      "loss": 0.3301,
      "step": 3830
    },
    {
      "epoch": 6.1,
      "grad_norm": 7.531820774078369,
      "learning_rate": 2.598639455782313e-05,
      "loss": 0.4208,
      "step": 3840
    },
    {
      "epoch": 6.11,
      "grad_norm": 4.852546215057373,
      "learning_rate": 2.5532879818594107e-05,
      "loss": 0.3858,
      "step": 3850
    },
    {
      "epoch": 6.13,
      "grad_norm": 6.866227149963379,
      "learning_rate": 2.507936507936508e-05,
      "loss": 0.4882,
      "step": 3860
    },
    {
      "epoch": 6.14,
      "grad_norm": 2.053617477416992,
      "learning_rate": 2.4625850340136055e-05,
      "loss": 0.3131,
      "step": 3870
    },
    {
      "epoch": 6.16,
      "grad_norm": 3.9806272983551025,
      "learning_rate": 2.417233560090703e-05,
      "loss": 0.4078,
      "step": 3880
    },
    {
      "epoch": 6.17,
      "grad_norm": 3.6454029083251953,
      "learning_rate": 2.3718820861678007e-05,
      "loss": 0.4362,
      "step": 3890
    },
    {
      "epoch": 6.19,
      "grad_norm": 7.963522434234619,
      "learning_rate": 2.326530612244898e-05,
      "loss": 0.4457,
      "step": 3900
    },
    {
      "epoch": 6.19,
      "eval_accuracy": 0.8392857142857143,
      "eval_loss": 0.5619771480560303,
      "eval_runtime": 91.8917,
      "eval_samples_per_second": 27.424,
      "eval_steps_per_second": 3.428,
      "step": 3900
    },
    {
      "epoch": 6.21,
      "grad_norm": 0.9147405624389648,
      "learning_rate": 2.2811791383219955e-05,
      "loss": 0.278,
      "step": 3910
    },
    {
      "epoch": 6.22,
      "grad_norm": 7.86151647567749,
      "learning_rate": 2.235827664399093e-05,
      "loss": 0.3866,
      "step": 3920
    },
    {
      "epoch": 6.24,
      "grad_norm": 5.7085371017456055,
      "learning_rate": 2.1904761904761906e-05,
      "loss": 0.3633,
      "step": 3930
    },
    {
      "epoch": 6.25,
      "grad_norm": 4.3009772300720215,
      "learning_rate": 2.145124716553288e-05,
      "loss": 0.3932,
      "step": 3940
    },
    {
      "epoch": 6.27,
      "grad_norm": 0.3253590166568756,
      "learning_rate": 2.0997732426303857e-05,
      "loss": 0.3084,
      "step": 3950
    },
    {
      "epoch": 6.29,
      "grad_norm": 4.880115509033203,
      "learning_rate": 2.0544217687074833e-05,
      "loss": 0.3723,
      "step": 3960
    },
    {
      "epoch": 6.3,
      "grad_norm": 5.515774250030518,
      "learning_rate": 2.0090702947845805e-05,
      "loss": 0.4312,
      "step": 3970
    },
    {
      "epoch": 6.32,
      "grad_norm": 4.72555685043335,
      "learning_rate": 1.963718820861678e-05,
      "loss": 0.2305,
      "step": 3980
    },
    {
      "epoch": 6.33,
      "grad_norm": 0.13027626276016235,
      "learning_rate": 1.9183673469387756e-05,
      "loss": 0.3037,
      "step": 3990
    },
    {
      "epoch": 6.35,
      "grad_norm": 4.166107654571533,
      "learning_rate": 1.8730158730158732e-05,
      "loss": 0.3693,
      "step": 4000
    },
    {
      "epoch": 6.35,
      "eval_accuracy": 0.8412698412698413,
      "eval_loss": 0.5411426424980164,
      "eval_runtime": 91.5694,
      "eval_samples_per_second": 27.52,
      "eval_steps_per_second": 3.44,
      "step": 4000
    },
    {
      "epoch": 6.37,
      "grad_norm": 3.1201322078704834,
      "learning_rate": 1.8276643990929708e-05,
      "loss": 0.3472,
      "step": 4010
    },
    {
      "epoch": 6.38,
      "grad_norm": 5.773028373718262,
      "learning_rate": 1.7823129251700683e-05,
      "loss": 0.376,
      "step": 4020
    },
    {
      "epoch": 6.4,
      "grad_norm": 4.617589473724365,
      "learning_rate": 1.736961451247166e-05,
      "loss": 0.4044,
      "step": 4030
    },
    {
      "epoch": 6.41,
      "grad_norm": 2.2319741249084473,
      "learning_rate": 1.691609977324263e-05,
      "loss": 0.2603,
      "step": 4040
    },
    {
      "epoch": 6.43,
      "grad_norm": 4.363328456878662,
      "learning_rate": 1.6462585034013607e-05,
      "loss": 0.3736,
      "step": 4050
    },
    {
      "epoch": 6.44,
      "grad_norm": 1.4275892972946167,
      "learning_rate": 1.6009070294784583e-05,
      "loss": 0.2221,
      "step": 4060
    },
    {
      "epoch": 6.46,
      "grad_norm": 1.1315315961837769,
      "learning_rate": 1.5555555555555555e-05,
      "loss": 0.2594,
      "step": 4070
    },
    {
      "epoch": 6.48,
      "grad_norm": 2.161555051803589,
      "learning_rate": 1.5102040816326532e-05,
      "loss": 0.3224,
      "step": 4080
    },
    {
      "epoch": 6.49,
      "grad_norm": 5.551158428192139,
      "learning_rate": 1.4648526077097508e-05,
      "loss": 0.2861,
      "step": 4090
    },
    {
      "epoch": 6.51,
      "grad_norm": 4.5997419357299805,
      "learning_rate": 1.419501133786848e-05,
      "loss": 0.2933,
      "step": 4100
    },
    {
      "epoch": 6.51,
      "eval_accuracy": 0.8484126984126984,
      "eval_loss": 0.5324992537498474,
      "eval_runtime": 92.004,
      "eval_samples_per_second": 27.39,
      "eval_steps_per_second": 3.424,
      "step": 4100
    },
    {
      "epoch": 6.52,
      "grad_norm": 4.983388900756836,
      "learning_rate": 1.3741496598639456e-05,
      "loss": 0.3076,
      "step": 4110
    },
    {
      "epoch": 6.54,
      "grad_norm": 7.082801818847656,
      "learning_rate": 1.3287981859410432e-05,
      "loss": 0.4137,
      "step": 4120
    },
    {
      "epoch": 6.56,
      "grad_norm": 5.25059700012207,
      "learning_rate": 1.2834467120181407e-05,
      "loss": 0.3036,
      "step": 4130
    },
    {
      "epoch": 6.57,
      "grad_norm": 4.526573181152344,
      "learning_rate": 1.2380952380952381e-05,
      "loss": 0.3122,
      "step": 4140
    },
    {
      "epoch": 6.59,
      "grad_norm": 3.637713670730591,
      "learning_rate": 1.1927437641723357e-05,
      "loss": 0.4162,
      "step": 4150
    },
    {
      "epoch": 6.6,
      "grad_norm": 7.545217514038086,
      "learning_rate": 1.147392290249433e-05,
      "loss": 0.5283,
      "step": 4160
    },
    {
      "epoch": 6.62,
      "grad_norm": 3.526754856109619,
      "learning_rate": 1.1020408163265306e-05,
      "loss": 0.4345,
      "step": 4170
    },
    {
      "epoch": 6.63,
      "grad_norm": 5.48629093170166,
      "learning_rate": 1.056689342403628e-05,
      "loss": 0.3755,
      "step": 4180
    },
    {
      "epoch": 6.65,
      "grad_norm": 3.289555788040161,
      "learning_rate": 1.0113378684807256e-05,
      "loss": 0.3603,
      "step": 4190
    },
    {
      "epoch": 6.67,
      "grad_norm": 3.701138496398926,
      "learning_rate": 9.659863945578232e-06,
      "loss": 0.2603,
      "step": 4200
    },
    {
      "epoch": 6.67,
      "eval_accuracy": 0.8476190476190476,
      "eval_loss": 0.5360472202301025,
      "eval_runtime": 94.6291,
      "eval_samples_per_second": 26.63,
      "eval_steps_per_second": 3.329,
      "step": 4200
    },
    {
      "epoch": 6.68,
      "grad_norm": 4.533103942871094,
      "learning_rate": 9.206349206349207e-06,
      "loss": 0.3483,
      "step": 4210
    },
    {
      "epoch": 6.7,
      "grad_norm": 6.598170280456543,
      "learning_rate": 8.752834467120181e-06,
      "loss": 0.4429,
      "step": 4220
    },
    {
      "epoch": 6.71,
      "grad_norm": 3.7935194969177246,
      "learning_rate": 8.299319727891157e-06,
      "loss": 0.2892,
      "step": 4230
    },
    {
      "epoch": 6.73,
      "grad_norm": 4.123006343841553,
      "learning_rate": 7.845804988662133e-06,
      "loss": 0.3482,
      "step": 4240
    },
    {
      "epoch": 6.75,
      "grad_norm": 0.6726520657539368,
      "learning_rate": 7.392290249433107e-06,
      "loss": 0.3813,
      "step": 4250
    },
    {
      "epoch": 6.76,
      "grad_norm": 7.85457181930542,
      "learning_rate": 6.938775510204082e-06,
      "loss": 0.3737,
      "step": 4260
    },
    {
      "epoch": 6.78,
      "grad_norm": 1.485037922859192,
      "learning_rate": 6.485260770975057e-06,
      "loss": 0.2864,
      "step": 4270
    },
    {
      "epoch": 6.79,
      "grad_norm": 2.535890817642212,
      "learning_rate": 6.031746031746032e-06,
      "loss": 0.2424,
      "step": 4280
    },
    {
      "epoch": 6.81,
      "grad_norm": 1.3501815795898438,
      "learning_rate": 5.578231292517007e-06,
      "loss": 0.3013,
      "step": 4290
    },
    {
      "epoch": 6.83,
      "grad_norm": 2.961240291595459,
      "learning_rate": 5.124716553287982e-06,
      "loss": 0.3364,
      "step": 4300
    },
    {
      "epoch": 6.83,
      "eval_accuracy": 0.8496031746031746,
      "eval_loss": 0.5302808284759521,
      "eval_runtime": 99.1164,
      "eval_samples_per_second": 25.425,
      "eval_steps_per_second": 3.178,
      "step": 4300
    },
    {
      "epoch": 6.84,
      "grad_norm": 3.593552589416504,
      "learning_rate": 4.671201814058957e-06,
      "loss": 0.3273,
      "step": 4310
    },
    {
      "epoch": 6.86,
      "grad_norm": 3.5346760749816895,
      "learning_rate": 4.217687074829932e-06,
      "loss": 0.3345,
      "step": 4320
    },
    {
      "epoch": 6.87,
      "grad_norm": 3.0319371223449707,
      "learning_rate": 3.764172335600907e-06,
      "loss": 0.4083,
      "step": 4330
    },
    {
      "epoch": 6.89,
      "grad_norm": 2.639801263809204,
      "learning_rate": 3.310657596371882e-06,
      "loss": 0.2986,
      "step": 4340
    },
    {
      "epoch": 6.9,
      "grad_norm": 1.1220003366470337,
      "learning_rate": 2.8571428571428573e-06,
      "loss": 0.3088,
      "step": 4350
    },
    {
      "epoch": 6.92,
      "grad_norm": 5.006831645965576,
      "learning_rate": 2.4036281179138325e-06,
      "loss": 0.4156,
      "step": 4360
    },
    {
      "epoch": 6.94,
      "grad_norm": 4.62202262878418,
      "learning_rate": 1.9501133786848073e-06,
      "loss": 0.3888,
      "step": 4370
    },
    {
      "epoch": 6.95,
      "grad_norm": 6.1172590255737305,
      "learning_rate": 1.4965986394557823e-06,
      "loss": 0.3449,
      "step": 4380
    },
    {
      "epoch": 6.97,
      "grad_norm": 6.193305969238281,
      "learning_rate": 1.0430839002267576e-06,
      "loss": 0.3253,
      "step": 4390
    },
    {
      "epoch": 6.98,
      "grad_norm": 1.3442612886428833,
      "learning_rate": 5.895691609977325e-07,
      "loss": 0.3639,
      "step": 4400
    },
    {
      "epoch": 6.98,
      "eval_accuracy": 0.8492063492063492,
      "eval_loss": 0.5315643548965454,
      "eval_runtime": 94.3134,
      "eval_samples_per_second": 26.719,
      "eval_steps_per_second": 3.34,
      "step": 4400
    },
    {
      "epoch": 7.0,
      "grad_norm": 4.398177146911621,
      "learning_rate": 1.3605442176870747e-07,
      "loss": 0.2582,
      "step": 4410
    },
    {
      "epoch": 7.0,
      "step": 4410,
      "total_flos": 5.468471871363809e+18,
      "train_loss": 0.7062997149772384,
      "train_runtime": 5016.6762,
      "train_samples_per_second": 14.065,
      "train_steps_per_second": 0.879
    }
  ],
  "logging_steps": 10,
  "max_steps": 4410,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 7,
  "save_steps": 100,
  "total_flos": 5.468471871363809e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}