{
  "best_metric": 0.08512182533740997,
  "best_model_checkpoint": "./model_dir\\checkpoint-12900",
  "epoch": 4.0,
  "eval_steps": 100,
  "global_step": 19736,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.002026753141467369,
      "grad_norm": 3.2063848972320557,
      "learning_rate": 0.00019989866234292666,
      "loss": 0.5934,
      "step": 10
    },
    {
      "epoch": 0.004053506282934738,
      "grad_norm": 23.631210327148438,
      "learning_rate": 0.00019979732468585327,
      "loss": 0.6803,
      "step": 20
    },
    {
      "epoch": 0.006080259424402107,
      "grad_norm": 0.24470672011375427,
      "learning_rate": 0.0001996959870287799,
      "loss": 0.4146,
      "step": 30
    },
    {
      "epoch": 0.008107012565869477,
      "grad_norm": 0.6175822615623474,
      "learning_rate": 0.00019959464937170654,
      "loss": 0.8121,
      "step": 40
    },
    {
      "epoch": 0.010133765707336847,
      "grad_norm": 0.15328890085220337,
      "learning_rate": 0.00019949331171463316,
      "loss": 0.1752,
      "step": 50
    },
    {
      "epoch": 0.012160518848804215,
      "grad_norm": 13.491938591003418,
      "learning_rate": 0.0001993919740575598,
      "loss": 0.7451,
      "step": 60
    },
    {
      "epoch": 0.014187271990271585,
      "grad_norm": 0.1225113719701767,
      "learning_rate": 0.00019929063640048645,
      "loss": 0.2662,
      "step": 70
    },
    {
      "epoch": 0.016214025131738953,
      "grad_norm": 0.08415812253952026,
      "learning_rate": 0.00019918929874341304,
      "loss": 0.4252,
      "step": 80
    },
    {
      "epoch": 0.018240778273206325,
      "grad_norm": 0.07037778943777084,
      "learning_rate": 0.0001990879610863397,
      "loss": 0.1327,
      "step": 90
    },
    {
      "epoch": 0.020267531414673693,
      "grad_norm": 0.06210090592503548,
      "learning_rate": 0.00019898662342926634,
      "loss": 0.8814,
      "step": 100
    },
    {
      "epoch": 0.020267531414673693,
      "eval_accuracy": 0.76007326007326,
      "eval_loss": 0.9432225227355957,
      "eval_runtime": 46.1647,
      "eval_samples_per_second": 11.827,
      "eval_steps_per_second": 1.495,
      "step": 100
    },
    {
      "epoch": 0.02229428455614106,
      "grad_norm": 0.09677636623382568,
      "learning_rate": 0.00019888528577219296,
      "loss": 0.048,
      "step": 110
    },
    {
      "epoch": 0.02432103769760843,
      "grad_norm": 0.06678306311368942,
      "learning_rate": 0.00019878394811511957,
      "loss": 0.6436,
      "step": 120
    },
    {
      "epoch": 0.0263477908390758,
      "grad_norm": 0.0650462880730629,
      "learning_rate": 0.00019868261045804622,
      "loss": 0.3952,
      "step": 130
    },
    {
      "epoch": 0.02837454398054317,
      "grad_norm": 0.06772352010011673,
      "learning_rate": 0.00019858127280097284,
      "loss": 0.0091,
      "step": 140
    },
    {
      "epoch": 0.030401297122010538,
      "grad_norm": 0.12076238542795181,
      "learning_rate": 0.00019847993514389949,
      "loss": 1.0189,
      "step": 150
    },
    {
      "epoch": 0.032428050263477906,
      "grad_norm": 0.06581564247608185,
      "learning_rate": 0.00019837859748682613,
      "loss": 0.4685,
      "step": 160
    },
    {
      "epoch": 0.03445480340494528,
      "grad_norm": 0.09730742871761322,
      "learning_rate": 0.00019827725982975272,
      "loss": 0.4774,
      "step": 170
    },
    {
      "epoch": 0.03648155654641265,
      "grad_norm": 0.09050973504781723,
      "learning_rate": 0.00019817592217267937,
      "loss": 0.7342,
      "step": 180
    },
    {
      "epoch": 0.038508309687880014,
      "grad_norm": 0.0777946189045906,
      "learning_rate": 0.00019807458451560602,
      "loss": 0.4465,
      "step": 190
    },
    {
      "epoch": 0.040535062829347386,
      "grad_norm": 9.031527519226074,
      "learning_rate": 0.00019797324685853264,
      "loss": 0.4374,
      "step": 200
    },
    {
      "epoch": 0.040535062829347386,
      "eval_accuracy": 0.8864468864468864,
      "eval_loss": 0.4927012622356415,
      "eval_runtime": 50.3717,
      "eval_samples_per_second": 10.839,
      "eval_steps_per_second": 1.37,
      "step": 200
    },
    {
      "epoch": 0.04256181597081476,
      "grad_norm": 0.1537230908870697,
      "learning_rate": 0.00019787190920145928,
      "loss": 0.0122,
      "step": 210
    },
    {
      "epoch": 0.04458856911228212,
      "grad_norm": 0.06793428212404251,
      "learning_rate": 0.0001977705715443859,
      "loss": 0.3605,
      "step": 220
    },
    {
      "epoch": 0.046615322253749494,
      "grad_norm": 0.05865870788693428,
      "learning_rate": 0.00019766923388731252,
      "loss": 0.4666,
      "step": 230
    },
    {
      "epoch": 0.04864207539521686,
      "grad_norm": 0.07028614729642868,
      "learning_rate": 0.00019756789623023917,
      "loss": 0.5873,
      "step": 240
    },
    {
      "epoch": 0.05066882853668423,
      "grad_norm": 13.204813003540039,
      "learning_rate": 0.0001974665585731658,
      "loss": 0.1207,
      "step": 250
    },
    {
      "epoch": 0.0526955816781516,
      "grad_norm": 0.0801391452550888,
      "learning_rate": 0.00019736522091609243,
      "loss": 0.0369,
      "step": 260
    },
    {
      "epoch": 0.05472233481961897,
      "grad_norm": 0.06529464572668076,
      "learning_rate": 0.00019726388325901905,
      "loss": 0.0137,
      "step": 270
    },
    {
      "epoch": 0.05674908796108634,
      "grad_norm": 0.0377487987279892,
      "learning_rate": 0.0001971625456019457,
      "loss": 0.4132,
      "step": 280
    },
    {
      "epoch": 0.05877584110255371,
      "grad_norm": 0.0332893468439579,
      "learning_rate": 0.00019706120794487232,
      "loss": 0.0059,
      "step": 290
    },
    {
      "epoch": 0.060802594244021076,
      "grad_norm": 0.0327690914273262,
      "learning_rate": 0.00019695987028779896,
      "loss": 0.0042,
      "step": 300
    },
    {
      "epoch": 0.060802594244021076,
      "eval_accuracy": 0.9267399267399268,
      "eval_loss": 0.35336464643478394,
      "eval_runtime": 49.466,
      "eval_samples_per_second": 11.038,
      "eval_steps_per_second": 1.395,
      "step": 300
    },
    {
      "epoch": 0.06282934738548845,
      "grad_norm": 0.029724499210715294,
      "learning_rate": 0.00019685853263072558,
      "loss": 0.004,
      "step": 310
    },
    {
      "epoch": 0.06485610052695581,
      "grad_norm": 0.05020628869533539,
      "learning_rate": 0.00019675719497365223,
      "loss": 0.5528,
      "step": 320
    },
    {
      "epoch": 0.06688285366842318,
      "grad_norm": 0.03226783871650696,
      "learning_rate": 0.00019665585731657885,
      "loss": 0.0087,
      "step": 330
    },
    {
      "epoch": 0.06890960680989056,
      "grad_norm": 0.03420436754822731,
      "learning_rate": 0.0001965545196595055,
      "loss": 0.0036,
      "step": 340
    },
    {
      "epoch": 0.07093635995135793,
      "grad_norm": 0.045836128294467926,
      "learning_rate": 0.0001964531820024321,
      "loss": 1.3375,
      "step": 350
    },
    {
      "epoch": 0.0729631130928253,
      "grad_norm": 0.8065479397773743,
      "learning_rate": 0.00019635184434535873,
      "loss": 0.5225,
      "step": 360
    },
    {
      "epoch": 0.07498986623429266,
      "grad_norm": 0.0850081741809845,
      "learning_rate": 0.00019625050668828538,
      "loss": 0.0174,
      "step": 370
    },
    {
      "epoch": 0.07701661937576003,
      "grad_norm": 0.10884834825992584,
      "learning_rate": 0.00019614916903121202,
      "loss": 0.9653,
      "step": 380
    },
    {
      "epoch": 0.0790433725172274,
      "grad_norm": 0.07834824174642563,
      "learning_rate": 0.00019604783137413864,
      "loss": 0.4623,
      "step": 390
    },
    {
      "epoch": 0.08107012565869477,
      "grad_norm": 0.09319596737623215,
      "learning_rate": 0.00019594649371706526,
      "loss": 0.0093,
      "step": 400
    },
    {
      "epoch": 0.08107012565869477,
      "eval_accuracy": 0.9413919413919414,
      "eval_loss": 0.23352867364883423,
      "eval_runtime": 50.3492,
      "eval_samples_per_second": 10.844,
      "eval_steps_per_second": 1.37,
      "step": 400
    },
    {
      "epoch": 0.08309687880016214,
      "grad_norm": 0.06457238644361496,
      "learning_rate": 0.0001958451560599919,
      "loss": 1.0088,
      "step": 410
    },
    {
      "epoch": 0.08512363194162952,
      "grad_norm": 0.10535389930009842,
      "learning_rate": 0.00019574381840291853,
      "loss": 0.0167,
      "step": 420
    },
    {
      "epoch": 0.08715038508309687,
      "grad_norm": 0.12313750386238098,
      "learning_rate": 0.00019564248074584517,
      "loss": 1.3943,
      "step": 430
    },
    {
      "epoch": 0.08917713822456425,
      "grad_norm": 0.22795221209526062,
      "learning_rate": 0.0001955411430887718,
      "loss": 0.4093,
      "step": 440
    },
    {
      "epoch": 0.09120389136603162,
      "grad_norm": 0.09523962438106537,
      "learning_rate": 0.0001954398054316984,
      "loss": 0.0147,
      "step": 450
    },
    {
      "epoch": 0.09323064450749899,
      "grad_norm": 0.13858100771903992,
      "learning_rate": 0.00019533846777462506,
      "loss": 0.8304,
      "step": 460
    },
    {
      "epoch": 0.09525739764896636,
      "grad_norm": 0.2124408781528473,
      "learning_rate": 0.0001952371301175517,
      "loss": 0.011,
      "step": 470
    },
    {
      "epoch": 0.09728415079043372,
      "grad_norm": 0.04959016665816307,
      "learning_rate": 0.00019513579246047832,
      "loss": 0.0067,
      "step": 480
    },
    {
      "epoch": 0.09931090393190109,
      "grad_norm": 0.511265754699707,
      "learning_rate": 0.00019503445480340497,
      "loss": 0.0071,
      "step": 490
    },
    {
      "epoch": 0.10133765707336846,
      "grad_norm": 0.05785638839006424,
      "learning_rate": 0.0001949331171463316,
      "loss": 0.125,
      "step": 500
    },
    {
      "epoch": 0.10133765707336846,
      "eval_accuracy": 0.9285714285714286,
      "eval_loss": 0.36297860741615295,
      "eval_runtime": 50.3719,
      "eval_samples_per_second": 10.839,
      "eval_steps_per_second": 1.37,
      "step": 500
    },
    {
      "epoch": 0.10336441021483583,
      "grad_norm": 0.033664364367723465,
      "learning_rate": 0.0001948317794892582,
      "loss": 0.0036,
      "step": 510
    },
    {
      "epoch": 0.1053911633563032,
      "grad_norm": 0.03444075211882591,
      "learning_rate": 0.00019473044183218485,
      "loss": 1.1187,
      "step": 520
    },
    {
      "epoch": 0.10741791649777058,
      "grad_norm": 0.03646405413746834,
      "learning_rate": 0.0001946291041751115,
      "loss": 0.0067,
      "step": 530
    },
    {
      "epoch": 0.10944466963923793,
      "grad_norm": 0.037678416818380356,
      "learning_rate": 0.0001945277665180381,
      "loss": 0.5384,
      "step": 540
    },
    {
      "epoch": 0.1114714227807053,
      "grad_norm": 9.350079536437988,
      "learning_rate": 0.00019442642886096474,
      "loss": 1.0734,
      "step": 550
    },
    {
      "epoch": 0.11349817592217268,
      "grad_norm": 0.11477504670619965,
      "learning_rate": 0.00019432509120389138,
      "loss": 1.0244,
      "step": 560
    },
    {
      "epoch": 0.11552492906364005,
      "grad_norm": 0.09600544720888138,
      "learning_rate": 0.000194223753546818,
      "loss": 0.014,
      "step": 570
    },
    {
      "epoch": 0.11755168220510742,
      "grad_norm": 0.08096398413181305,
      "learning_rate": 0.00019412241588974465,
      "loss": 0.0123,
      "step": 580
    },
    {
      "epoch": 0.1195784353465748,
      "grad_norm": 0.05573810264468193,
      "learning_rate": 0.00019402107823267127,
      "loss": 0.0086,
      "step": 590
    },
    {
      "epoch": 0.12160518848804215,
      "grad_norm": 0.05102921277284622,
      "learning_rate": 0.0001939197405755979,
      "loss": 0.4924,
      "step": 600
    },
    {
      "epoch": 0.12160518848804215,
      "eval_accuracy": 0.9468864468864469,
      "eval_loss": 0.23736581206321716,
      "eval_runtime": 47.7703,
      "eval_samples_per_second": 11.43,
      "eval_steps_per_second": 1.444,
      "step": 600
    },
    {
      "epoch": 0.12363194162950952,
      "grad_norm": 95.2990951538086,
      "learning_rate": 0.00019381840291852453,
      "loss": 0.2859,
      "step": 610
    },
    {
      "epoch": 0.1256586947709769,
      "grad_norm": 0.043713510036468506,
      "learning_rate": 0.00019371706526145118,
      "loss": 0.4951,
      "step": 620
    },
    {
      "epoch": 0.12768544791244427,
      "grad_norm": 0.057737067341804504,
      "learning_rate": 0.0001936157276043778,
      "loss": 0.9175,
      "step": 630
    },
    {
      "epoch": 0.12971220105391162,
      "grad_norm": 0.07625377923250198,
      "learning_rate": 0.00019351438994730442,
      "loss": 0.287,
      "step": 640
    },
    {
      "epoch": 0.131738954195379,
      "grad_norm": 0.0649242177605629,
      "learning_rate": 0.00019341305229023106,
      "loss": 0.0102,
      "step": 650
    },
    {
      "epoch": 0.13376570733684637,
      "grad_norm": 0.07740580290555954,
      "learning_rate": 0.00019331171463315768,
      "loss": 0.008,
      "step": 660
    },
    {
      "epoch": 0.13579246047831375,
      "grad_norm": 0.043081168085336685,
      "learning_rate": 0.00019321037697608433,
      "loss": 0.0063,
      "step": 670
    },
    {
      "epoch": 0.1378192136197811,
      "grad_norm": 0.03584828972816467,
      "learning_rate": 0.00019310903931901098,
      "loss": 0.0044,
      "step": 680
    },
    {
      "epoch": 0.13984596676124847,
      "grad_norm": 0.05630917102098465,
      "learning_rate": 0.00019300770166193757,
      "loss": 0.5407,
      "step": 690
    },
    {
      "epoch": 0.14187271990271585,
      "grad_norm": 0.037703562527894974,
      "learning_rate": 0.00019290636400486421,
      "loss": 0.0052,
      "step": 700
    },
    {
      "epoch": 0.14187271990271585,
      "eval_accuracy": 0.9487179487179487,
      "eval_loss": 0.20145167410373688,
      "eval_runtime": 46.9732,
      "eval_samples_per_second": 11.624,
      "eval_steps_per_second": 1.469,
      "step": 700
    },
    {
      "epoch": 0.1438994730441832,
      "grad_norm": 0.029574614018201828,
      "learning_rate": 0.00019280502634779086,
      "loss": 0.0041,
      "step": 710
    },
    {
      "epoch": 0.1459262261856506,
      "grad_norm": 0.04035133868455887,
      "learning_rate": 0.00019270368869071748,
      "loss": 0.0039,
      "step": 720
    },
    {
      "epoch": 0.14795297932711796,
      "grad_norm": 0.054472994059324265,
      "learning_rate": 0.0001926023510336441,
      "loss": 1.1169,
      "step": 730
    },
    {
      "epoch": 0.14997973246858531,
      "grad_norm": 8.792170524597168,
      "learning_rate": 0.00019250101337657074,
      "loss": 0.5607,
      "step": 740
    },
    {
      "epoch": 0.1520064856100527,
      "grad_norm": 0.05719525367021561,
      "learning_rate": 0.00019239967571949736,
      "loss": 0.9535,
      "step": 750
    },
    {
      "epoch": 0.15403323875152006,
      "grad_norm": 0.10215990990400314,
      "learning_rate": 0.000192298338062424,
      "loss": 0.0119,
      "step": 760
    },
    {
      "epoch": 0.15605999189298744,
      "grad_norm": 0.049983322620391846,
      "learning_rate": 0.00019219700040535066,
      "loss": 0.0235,
      "step": 770
    },
    {
      "epoch": 0.1580867450344548,
      "grad_norm": 0.044148072600364685,
      "learning_rate": 0.00019209566274827725,
      "loss": 0.4816,
      "step": 780
    },
    {
      "epoch": 0.1601134981759222,
      "grad_norm": 38.11033630371094,
      "learning_rate": 0.0001919943250912039,
      "loss": 0.2388,
      "step": 790
    },
    {
      "epoch": 0.16214025131738954,
      "grad_norm": 0.03907947614789009,
      "learning_rate": 0.00019189298743413054,
      "loss": 0.3738,
      "step": 800
    },
    {
      "epoch": 0.16214025131738954,
      "eval_accuracy": 0.8864468864468864,
      "eval_loss": 0.4199656844139099,
      "eval_runtime": 50.4489,
      "eval_samples_per_second": 10.823,
      "eval_steps_per_second": 1.368,
      "step": 800
    },
    {
      "epoch": 0.1641670044588569,
      "grad_norm": 0.08232033997774124,
      "learning_rate": 0.00019179164977705716,
      "loss": 0.7998,
      "step": 810
    },
    {
      "epoch": 0.1661937576003243,
      "grad_norm": 0.04074680432677269,
      "learning_rate": 0.0001916903121199838,
      "loss": 0.0489,
      "step": 820
    },
    {
      "epoch": 0.16822051074179165,
      "grad_norm": 12.359615325927734,
      "learning_rate": 0.00019158897446291042,
      "loss": 0.2878,
      "step": 830
    },
    {
      "epoch": 0.17024726388325903,
      "grad_norm": 7.305814743041992,
      "learning_rate": 0.00019148763680583704,
      "loss": 0.3854,
      "step": 840
    },
    {
      "epoch": 0.1722740170247264,
      "grad_norm": 0.03477887436747551,
      "learning_rate": 0.0001913862991487637,
      "loss": 0.0468,
      "step": 850
    },
    {
      "epoch": 0.17430077016619375,
      "grad_norm": 0.03299262002110481,
      "learning_rate": 0.00019128496149169034,
      "loss": 0.0444,
      "step": 860
    },
    {
      "epoch": 0.17632752330766113,
      "grad_norm": 0.029569240286946297,
      "learning_rate": 0.00019118362383461696,
      "loss": 0.0036,
      "step": 870
    },
    {
      "epoch": 0.1783542764491285,
      "grad_norm": 0.026648886501789093,
      "learning_rate": 0.00019108228617754357,
      "loss": 0.0497,
      "step": 880
    },
    {
      "epoch": 0.18038102959059588,
      "grad_norm": 0.0252667348831892,
      "learning_rate": 0.00019098094852047022,
      "loss": 0.4905,
      "step": 890
    },
    {
      "epoch": 0.18240778273206323,
      "grad_norm": 0.027241753414273262,
      "learning_rate": 0.00019087961086339684,
      "loss": 0.4533,
      "step": 900
    },
    {
      "epoch": 0.18240778273206323,
      "eval_accuracy": 0.9285714285714286,
      "eval_loss": 0.2573205232620239,
      "eval_runtime": 48.2075,
      "eval_samples_per_second": 11.326,
      "eval_steps_per_second": 1.431,
      "step": 900
    },
    {
      "epoch": 0.1844345358735306,
      "grad_norm": 0.7260812520980835,
      "learning_rate": 0.00019077827320632349,
      "loss": 0.6588,
      "step": 910
    },
    {
      "epoch": 0.18646128901499798,
      "grad_norm": 9.329017639160156,
      "learning_rate": 0.0001906769355492501,
      "loss": 1.0854,
      "step": 920
    },
    {
      "epoch": 0.18848804215646534,
      "grad_norm": 0.18261322379112244,
      "learning_rate": 0.00019057559789217675,
      "loss": 1.0766,
      "step": 930
    },
    {
      "epoch": 0.19051479529793272,
      "grad_norm": 0.22918297350406647,
      "learning_rate": 0.00019047426023510337,
      "loss": 0.2613,
      "step": 940
    },
    {
      "epoch": 0.19254154843940008,
      "grad_norm": 0.05047022923827171,
      "learning_rate": 0.00019037292257803002,
      "loss": 0.0534,
      "step": 950
    },
    {
      "epoch": 0.19456830158086744,
      "grad_norm": 0.03922040015459061,
      "learning_rate": 0.00019027158492095664,
      "loss": 1.0114,
      "step": 960
    },
    {
      "epoch": 0.19659505472233482,
      "grad_norm": 0.10960974544286728,
      "learning_rate": 0.00019017024726388325,
      "loss": 0.8342,
      "step": 970
    },
    {
      "epoch": 0.19862180786380218,
      "grad_norm": 0.1044323593378067,
      "learning_rate": 0.0001900689096068099,
      "loss": 0.8305,
      "step": 980
    },
    {
      "epoch": 0.20064856100526957,
      "grad_norm": 0.12734976410865784,
      "learning_rate": 0.00018996757194973655,
      "loss": 0.4303,
      "step": 990
    },
    {
      "epoch": 0.20267531414673692,
      "grad_norm": 0.07667859643697739,
      "learning_rate": 0.00018986623429266317,
      "loss": 0.027,
      "step": 1000
    },
    {
      "epoch": 0.20267531414673692,
      "eval_accuracy": 0.9120879120879121,
      "eval_loss": 0.34083420038223267,
      "eval_runtime": 50.3862,
      "eval_samples_per_second": 10.836,
      "eval_steps_per_second": 1.369,
      "step": 1000
    },
    {
      "epoch": 0.2047020672882043,
      "grad_norm": 0.04669175669550896,
      "learning_rate": 0.00018976489663558979,
      "loss": 0.3329,
      "step": 1010
    },
    {
      "epoch": 0.20672882042967167,
      "grad_norm": 0.0436105839908123,
      "learning_rate": 0.00018966355897851643,
      "loss": 0.416,
      "step": 1020
    },
    {
      "epoch": 0.20875557357113902,
      "grad_norm": 0.03847363963723183,
      "learning_rate": 0.00018956222132144305,
      "loss": 0.0052,
      "step": 1030
    },
    {
      "epoch": 0.2107823267126064,
      "grad_norm": 0.04070694372057915,
      "learning_rate": 0.0001894608836643697,
      "loss": 0.6871,
      "step": 1040
    },
    {
      "epoch": 0.21280907985407377,
      "grad_norm": 0.044938068836927414,
      "learning_rate": 0.00018935954600729632,
      "loss": 0.3686,
      "step": 1050
    },
    {
      "epoch": 0.21483583299554115,
      "grad_norm": 0.3830691874027252,
      "learning_rate": 0.00018925820835022293,
      "loss": 0.0704,
      "step": 1060
    },
    {
      "epoch": 0.2168625861370085,
      "grad_norm": 0.09455056488513947,
      "learning_rate": 0.00018915687069314958,
      "loss": 0.9455,
      "step": 1070
    },
    {
      "epoch": 0.21888933927847587,
      "grad_norm": 0.03371147811412811,
      "learning_rate": 0.00018905553303607623,
      "loss": 0.3658,
      "step": 1080
    },
    {
      "epoch": 0.22091609241994326,
      "grad_norm": 0.4871658980846405,
      "learning_rate": 0.00018895419537900285,
      "loss": 0.2985,
      "step": 1090
    },
    {
      "epoch": 0.2229428455614106,
      "grad_norm": 0.038024235516786575,
      "learning_rate": 0.0001888528577219295,
      "loss": 0.6685,
      "step": 1100
    },
    {
      "epoch": 0.2229428455614106,
      "eval_accuracy": 0.826007326007326,
      "eval_loss": 0.3140435516834259,
      "eval_runtime": 47.3231,
      "eval_samples_per_second": 11.538,
      "eval_steps_per_second": 1.458,
      "step": 1100
    },
    {
      "epoch": 0.224969598702878,
      "grad_norm": 0.03393388167023659,
      "learning_rate": 0.0001887515200648561,
      "loss": 0.1142,
      "step": 1110
    },
    {
      "epoch": 0.22699635184434536,
      "grad_norm": 0.030963918194174767,
      "learning_rate": 0.00018865018240778273,
      "loss": 0.1054,
      "step": 1120
    },
    {
      "epoch": 0.22902310498581271,
      "grad_norm": 0.036398351192474365,
      "learning_rate": 0.00018854884475070938,
      "loss": 0.6332,
      "step": 1130
    },
    {
      "epoch": 0.2310498581272801,
      "grad_norm": 0.04975204914808273,
      "learning_rate": 0.00018844750709363602,
      "loss": 0.7696,
      "step": 1140
    },
    {
      "epoch": 0.23307661126874746,
      "grad_norm": 0.04488310590386391,
      "learning_rate": 0.00018834616943656262,
      "loss": 0.1347,
      "step": 1150
    },
    {
      "epoch": 0.23510336441021484,
      "grad_norm": 24.52461051940918,
      "learning_rate": 0.00018824483177948926,
      "loss": 0.8782,
      "step": 1160
    },
    {
      "epoch": 0.2371301175516822,
      "grad_norm": 0.0589730367064476,
      "learning_rate": 0.0001881434941224159,
      "loss": 0.4471,
      "step": 1170
    },
    {
      "epoch": 0.2391568706931496,
      "grad_norm": 8.6546630859375,
      "learning_rate": 0.00018804215646534253,
      "loss": 0.5269,
      "step": 1180
    },
    {
      "epoch": 0.24118362383461694,
      "grad_norm": 81.45098876953125,
      "learning_rate": 0.00018794081880826917,
      "loss": 0.4896,
      "step": 1190
    },
    {
      "epoch": 0.2432103769760843,
      "grad_norm": 0.05012829601764679,
      "learning_rate": 0.0001878394811511958,
      "loss": 0.0703,
      "step": 1200
    },
    {
      "epoch": 0.2432103769760843,
      "eval_accuracy": 0.9322344322344323,
      "eval_loss": 0.2424902766942978,
      "eval_runtime": 47.8835,
      "eval_samples_per_second": 11.403,
      "eval_steps_per_second": 1.441,
      "step": 1200
    },
    {
      "epoch": 0.2452371301175517,
      "grad_norm": 0.045358020812273026,
      "learning_rate": 0.0001877381434941224,
      "loss": 0.213,
      "step": 1210
    },
    {
      "epoch": 0.24726388325901905,
      "grad_norm": 8.763317108154297,
      "learning_rate": 0.00018763680583704906,
      "loss": 1.51,
      "step": 1220
    },
    {
      "epoch": 0.24929063640048643,
      "grad_norm": 0.3494032621383667,
      "learning_rate": 0.0001875354681799757,
      "loss": 0.7853,
      "step": 1230
    },
    {
      "epoch": 0.2513173895419538,
      "grad_norm": 0.11207950115203857,
      "learning_rate": 0.00018743413052290232,
      "loss": 0.0299,
      "step": 1240
    },
    {
      "epoch": 0.25334414268342115,
      "grad_norm": 0.05717464163899422,
      "learning_rate": 0.00018733279286582894,
      "loss": 0.0085,
      "step": 1250
    },
    {
      "epoch": 0.25537089582488853,
      "grad_norm": 0.05250593274831772,
      "learning_rate": 0.0001872314552087556,
      "loss": 0.5135,
      "step": 1260
    },
    {
      "epoch": 0.2573976489663559,
      "grad_norm": 0.04397350549697876,
      "learning_rate": 0.0001871301175516822,
      "loss": 0.0056,
      "step": 1270
    },
    {
      "epoch": 0.25942440210782325,
      "grad_norm": 8.682046890258789,
      "learning_rate": 0.00018702877989460885,
      "loss": 1.0666,
      "step": 1280
    },
    {
      "epoch": 0.26145115524929063,
      "grad_norm": 0.07138803601264954,
      "learning_rate": 0.00018692744223753547,
      "loss": 0.5221,
      "step": 1290
    },
    {
      "epoch": 0.263477908390758,
      "grad_norm": 0.1037185862660408,
      "learning_rate": 0.0001868261045804621,
      "loss": 0.9411,
      "step": 1300
    },
    {
      "epoch": 0.263477908390758,
      "eval_accuracy": 0.8223443223443223,
      "eval_loss": 0.7809145450592041,
      "eval_runtime": 49.4934,
      "eval_samples_per_second": 11.032,
      "eval_steps_per_second": 1.394,
      "step": 1300
    },
    {
      "epoch": 0.26550466153222535,
      "grad_norm": 8.10766887664795,
      "learning_rate": 0.00018672476692338874,
      "loss": 0.8295,
      "step": 1310
    },
    {
      "epoch": 0.26753141467369274,
      "grad_norm": 0.18092480301856995,
      "learning_rate": 0.00018662342926631538,
      "loss": 0.0202,
      "step": 1320
    },
    {
      "epoch": 0.2695581678151601,
      "grad_norm": 0.15583205223083496,
      "learning_rate": 0.000186522091609242,
      "loss": 0.3999,
      "step": 1330
    },
    {
      "epoch": 0.2715849209566275,
      "grad_norm": 0.09881691634654999,
      "learning_rate": 0.00018642075395216862,
      "loss": 0.4089,
      "step": 1340
    },
    {
      "epoch": 0.27361167409809484,
      "grad_norm": 0.1042148619890213,
      "learning_rate": 0.00018631941629509527,
      "loss": 1.257,
      "step": 1350
    },
    {
      "epoch": 0.2756384272395622,
      "grad_norm": 7.86333703994751,
      "learning_rate": 0.0001862180786380219,
      "loss": 0.7033,
      "step": 1360
    },
    {
      "epoch": 0.2776651803810296,
      "grad_norm": 7.599421977996826,
      "learning_rate": 0.00018611674098094853,
      "loss": 0.6313,
      "step": 1370
    },
    {
      "epoch": 0.27969193352249694,
      "grad_norm": 0.16084067523479462,
      "learning_rate": 0.00018601540332387518,
      "loss": 0.0262,
      "step": 1380
    },
    {
      "epoch": 0.2817186866639643,
      "grad_norm": 0.09960556775331497,
      "learning_rate": 0.00018591406566680177,
      "loss": 0.3459,
      "step": 1390
    },
    {
      "epoch": 0.2837454398054317,
      "grad_norm": 0.13508349657058716,
      "learning_rate": 0.00018581272800972842,
      "loss": 0.4378,
      "step": 1400
    },
    {
      "epoch": 0.2837454398054317,
      "eval_accuracy": 0.8223443223443223,
      "eval_loss": 0.6968061327934265,
      "eval_runtime": 50.393,
      "eval_samples_per_second": 10.835,
      "eval_steps_per_second": 1.369,
      "step": 1400
    },
    {
      "epoch": 0.2857721929468991,
      "grad_norm": 0.06493504345417023,
      "learning_rate": 0.00018571139035265506,
      "loss": 0.388,
      "step": 1410
    },
    {
      "epoch": 0.2877989460883664,
      "grad_norm": 0.4153689444065094,
      "learning_rate": 0.00018561005269558168,
      "loss": 0.6636,
      "step": 1420
    },
    {
      "epoch": 0.2898256992298338,
      "grad_norm": 0.0951620489358902,
      "learning_rate": 0.00018550871503850833,
      "loss": 0.7434,
      "step": 1430
    },
    {
      "epoch": 0.2918524523713012,
      "grad_norm": 6.5460944175720215,
      "learning_rate": 0.00018540737738143495,
      "loss": 0.2931,
      "step": 1440
    },
    {
      "epoch": 0.2938792055127685,
      "grad_norm": 0.5932949185371399,
      "learning_rate": 0.00018530603972436157,
      "loss": 0.3002,
      "step": 1450
    },
    {
      "epoch": 0.2959059586542359,
      "grad_norm": 8.747719764709473,
      "learning_rate": 0.00018520470206728821,
      "loss": 0.5126,
      "step": 1460
    },
    {
      "epoch": 0.2979327117957033,
      "grad_norm": 0.05845149606466293,
      "learning_rate": 0.00018510336441021486,
      "loss": 0.4434,
      "step": 1470
    },
    {
      "epoch": 0.29995946493717063,
      "grad_norm": 0.23026475310325623,
      "learning_rate": 0.00018500202675314148,
      "loss": 0.6812,
      "step": 1480
    },
    {
      "epoch": 0.301986218078638,
      "grad_norm": 0.06300631165504456,
      "learning_rate": 0.0001849006890960681,
      "loss": 0.5783,
      "step": 1490
    },
    {
      "epoch": 0.3040129712201054,
      "grad_norm": 1.2849607467651367,
      "learning_rate": 0.00018479935143899474,
      "loss": 0.7127,
      "step": 1500
    },
    {
      "epoch": 0.3040129712201054,
      "eval_accuracy": 0.8241758241758241,
      "eval_loss": 0.3293890655040741,
      "eval_runtime": 88.987,
      "eval_samples_per_second": 6.136,
      "eval_steps_per_second": 0.775,
      "step": 1500
    },
    {
      "epoch": 0.3060397243615728,
      "grad_norm": 0.06597457081079483,
      "learning_rate": 0.00018469801378192136,
      "loss": 0.4313,
      "step": 1510
    },
    {
      "epoch": 0.3080664775030401,
      "grad_norm": 0.054177168756723404,
      "learning_rate": 0.000184596676124848,
      "loss": 0.1819,
      "step": 1520
    },
    {
      "epoch": 0.3100932306445075,
      "grad_norm": 0.05142025649547577,
      "learning_rate": 0.00018449533846777463,
      "loss": 0.1027,
      "step": 1530
    },
    {
      "epoch": 0.3121199837859749,
      "grad_norm": 0.595727264881134,
      "learning_rate": 0.00018439400081070127,
      "loss": 0.5553,
      "step": 1540
    },
    {
      "epoch": 0.3141467369274422,
      "grad_norm": 0.044500961899757385,
      "learning_rate": 0.0001842926631536279,
      "loss": 0.5383,
      "step": 1550
    },
    {
      "epoch": 0.3161734900689096,
      "grad_norm": 6.882216453552246,
      "learning_rate": 0.00018419132549655454,
      "loss": 0.4053,
      "step": 1560
    },
    {
      "epoch": 0.318200243210377,
      "grad_norm": 0.06732060760259628,
      "learning_rate": 0.00018408998783948116,
      "loss": 0.765,
      "step": 1570
    },
    {
      "epoch": 0.3202269963518444,
      "grad_norm": 0.07889731228351593,
      "learning_rate": 0.00018398865018240778,
      "loss": 0.1044,
      "step": 1580
    },
    {
      "epoch": 0.3222537494933117,
      "grad_norm": 8.76712703704834,
      "learning_rate": 0.00018388731252533442,
      "loss": 0.5054,
      "step": 1590
    },
    {
      "epoch": 0.3242805026347791,
      "grad_norm": 0.10699472576379776,
      "learning_rate": 0.00018378597486826104,
      "loss": 0.9465,
      "step": 1600
    },
    {
      "epoch": 0.3242805026347791,
      "eval_accuracy": 0.8223443223443223,
      "eval_loss": 0.4912700653076172,
      "eval_runtime": 48.2783,
      "eval_samples_per_second": 11.309,
      "eval_steps_per_second": 1.429,
      "step": 1600
    },
    {
      "epoch": 0.3263072557762465,
      "grad_norm": 0.6591342091560364,
      "learning_rate": 0.0001836846372111877,
      "loss": 0.5213,
      "step": 1610
    },
    {
      "epoch": 0.3283340089177138,
      "grad_norm": 6.809113025665283,
      "learning_rate": 0.0001835832995541143,
      "loss": 0.2711,
      "step": 1620
    },
    {
      "epoch": 0.3303607620591812,
      "grad_norm": 0.08120753616094589,
      "learning_rate": 0.00018348196189704096,
      "loss": 0.1901,
      "step": 1630
    },
    {
      "epoch": 0.3323875152006486,
      "grad_norm": 0.1842392235994339,
      "learning_rate": 0.00018338062423996757,
      "loss": 0.5417,
      "step": 1640
    },
    {
      "epoch": 0.3344142683421159,
      "grad_norm": 0.09670919179916382,
      "learning_rate": 0.00018327928658289422,
      "loss": 0.3437,
      "step": 1650
    },
    {
      "epoch": 0.3364410214835833,
      "grad_norm": 0.04945284500718117,
      "learning_rate": 0.00018317794892582084,
      "loss": 0.1802,
      "step": 1660
    },
    {
      "epoch": 0.3384677746250507,
      "grad_norm": 8.956902503967285,
      "learning_rate": 0.00018307661126874746,
      "loss": 0.9073,
      "step": 1670
    },
    {
      "epoch": 0.34049452776651806,
      "grad_norm": 0.11272208392620087,
      "learning_rate": 0.0001829752736116741,
      "loss": 0.365,
      "step": 1680
    },
    {
      "epoch": 0.3425212809079854,
      "grad_norm": 0.13924777507781982,
      "learning_rate": 0.00018287393595460075,
      "loss": 0.3967,
      "step": 1690
    },
    {
      "epoch": 0.3445480340494528,
      "grad_norm": 0.05262771248817444,
      "learning_rate": 0.00018277259829752737,
      "loss": 0.3834,
      "step": 1700
    },
    {
      "epoch": 0.3445480340494528,
      "eval_accuracy": 0.9047619047619048,
      "eval_loss": 0.25941210985183716,
      "eval_runtime": 47.82,
      "eval_samples_per_second": 11.418,
      "eval_steps_per_second": 1.443,
      "step": 1700
    },
    {
      "epoch": 0.34657478719092016,
      "grad_norm": 7.016679763793945,
      "learning_rate": 0.00018267126064045402,
      "loss": 0.0367,
      "step": 1710
    },
    {
      "epoch": 0.3486015403323875,
      "grad_norm": 9.35397720336914,
      "learning_rate": 0.00018256992298338064,
      "loss": 0.8748,
      "step": 1720
    },
    {
      "epoch": 0.3506282934738549,
      "grad_norm": 0.036801666021347046,
      "learning_rate": 0.00018246858532630725,
      "loss": 0.5593,
      "step": 1730
    },
    {
      "epoch": 0.35265504661532227,
      "grad_norm": 0.040228940546512604,
      "learning_rate": 0.0001823672476692339,
      "loss": 0.0114,
      "step": 1740
    },
    {
      "epoch": 0.35468179975678965,
      "grad_norm": 0.035426534712314606,
      "learning_rate": 0.00018226591001216055,
      "loss": 0.0266,
      "step": 1750
    },
    {
      "epoch": 0.356708552898257,
      "grad_norm": 0.03302001953125,
      "learning_rate": 0.00018216457235508714,
      "loss": 0.0056,
      "step": 1760
    },
    {
      "epoch": 0.35873530603972437,
      "grad_norm": 0.02835068479180336,
      "learning_rate": 0.00018206323469801379,
      "loss": 0.0041,
      "step": 1770
    },
    {
      "epoch": 0.36076205918119175,
      "grad_norm": 0.03482850268483162,
      "learning_rate": 0.00018196189704094043,
      "loss": 0.9573,
      "step": 1780
    },
    {
      "epoch": 0.3627888123226591,
      "grad_norm": 0.03174089640378952,
      "learning_rate": 0.00018186055938386705,
      "loss": 0.5716,
      "step": 1790
    },
    {
      "epoch": 0.36481556546412647,
      "grad_norm": 0.04105382412672043,
      "learning_rate": 0.0001817592217267937,
      "loss": 0.6691,
      "step": 1800
    },
    {
      "epoch": 0.36481556546412647,
      "eval_accuracy": 0.8992673992673993,
      "eval_loss": 0.35374078154563904,
      "eval_runtime": 50.5022,
      "eval_samples_per_second": 10.811,
      "eval_steps_per_second": 1.366,
      "step": 1800
    },
    {
      "epoch": 0.36684231860559385,
      "grad_norm": 0.05565650016069412,
      "learning_rate": 0.00018165788406972032,
      "loss": 0.2785,
      "step": 1810
    },
    {
      "epoch": 0.3688690717470612,
      "grad_norm": 0.038423530757427216,
      "learning_rate": 0.00018155654641264693,
      "loss": 0.0051,
      "step": 1820
    },
    {
      "epoch": 0.37089582488852857,
      "grad_norm": 0.034793466329574585,
      "learning_rate": 0.00018145520875557358,
      "loss": 0.034,
      "step": 1830
    },
    {
      "epoch": 0.37292257802999595,
      "grad_norm": 0.029749181121587753,
      "learning_rate": 0.00018135387109850023,
      "loss": 0.0037,
      "step": 1840
    },
    {
      "epoch": 0.37494933117146334,
      "grad_norm": 0.02845478057861328,
      "learning_rate": 0.00018125253344142685,
      "loss": 0.3519,
      "step": 1850
    },
    {
      "epoch": 0.37697608431293067,
      "grad_norm": 0.026904482394456863,
      "learning_rate": 0.00018115119578435347,
      "loss": 0.0096,
      "step": 1860
    },
    {
      "epoch": 0.37900283745439806,
      "grad_norm": 0.2545231580734253,
      "learning_rate": 0.0001810498581272801,
      "loss": 0.0083,
      "step": 1870
    },
    {
      "epoch": 0.38102959059586544,
      "grad_norm": 0.028374874964356422,
      "learning_rate": 0.00018094852047020673,
      "loss": 0.5834,
      "step": 1880
    },
    {
      "epoch": 0.38305634373733277,
      "grad_norm": 0.03022819012403488,
      "learning_rate": 0.00018084718281313338,
      "loss": 0.5286,
      "step": 1890
    },
    {
      "epoch": 0.38508309687880016,
      "grad_norm": 0.034386951476335526,
      "learning_rate": 0.00018074584515606,
      "loss": 0.3002,
      "step": 1900
    },
    {
      "epoch": 0.38508309687880016,
      "eval_accuracy": 0.9285714285714286,
      "eval_loss": 0.2502322494983673,
      "eval_runtime": 48.7593,
      "eval_samples_per_second": 11.198,
      "eval_steps_per_second": 1.415,
      "step": 1900
    },
    {
      "epoch": 0.38710985002026754,
      "grad_norm": 8.355843544006348,
      "learning_rate": 0.00018064450749898662,
      "loss": 1.2353,
      "step": 1910
    },
    {
      "epoch": 0.3891366031617349,
      "grad_norm": 0.22805503010749817,
      "learning_rate": 0.00018054316984191326,
      "loss": 0.0135,
      "step": 1920
    },
    {
      "epoch": 0.39116335630320226,
      "grad_norm": 9.000629425048828,
      "learning_rate": 0.0001804418321848399,
      "loss": 0.8987,
      "step": 1930
    },
    {
      "epoch": 0.39319010944466964,
      "grad_norm": 0.10103480517864227,
      "learning_rate": 0.00018034049452776653,
      "loss": 0.6317,
      "step": 1940
    },
    {
      "epoch": 0.39521686258613703,
      "grad_norm": 0.053478993475437164,
      "learning_rate": 0.00018023915687069315,
      "loss": 0.0372,
      "step": 1950
    },
    {
      "epoch": 0.39724361572760436,
      "grad_norm": 0.034072987735271454,
      "learning_rate": 0.0001801378192136198,
      "loss": 0.3131,
      "step": 1960
    },
    {
      "epoch": 0.39927036886907175,
      "grad_norm": 0.046674974262714386,
      "learning_rate": 0.0001800364815565464,
      "loss": 0.6352,
      "step": 1970
    },
    {
      "epoch": 0.40129712201053913,
      "grad_norm": 0.03819415718317032,
      "learning_rate": 0.00017993514389947306,
      "loss": 0.0087,
      "step": 1980
    },
    {
      "epoch": 0.40332387515200646,
      "grad_norm": 0.9910769462585449,
      "learning_rate": 0.0001798338062423997,
      "loss": 0.3056,
      "step": 1990
    },
    {
      "epoch": 0.40535062829347385,
      "grad_norm": 0.04858968034386635,
      "learning_rate": 0.0001797324685853263,
      "loss": 0.0473,
      "step": 2000
    },
    {
      "epoch": 0.40535062829347385,
      "eval_accuracy": 0.9322344322344323,
      "eval_loss": 0.23118893802165985,
      "eval_runtime": 49.9277,
      "eval_samples_per_second": 10.936,
      "eval_steps_per_second": 1.382,
      "step": 2000
    },
    {
      "epoch": 0.40737738143494123,
      "grad_norm": 0.025557000190019608,
      "learning_rate": 0.00017963113092825294,
      "loss": 0.2941,
      "step": 2010
    },
    {
      "epoch": 0.4094041345764086,
      "grad_norm": 0.02778341993689537,
      "learning_rate": 0.0001795297932711796,
      "loss": 0.8051,
      "step": 2020
    },
    {
      "epoch": 0.41143088771787595,
      "grad_norm": 0.03390109911561012,
      "learning_rate": 0.0001794284556141062,
      "loss": 0.2695,
      "step": 2030
    },
    {
      "epoch": 0.41345764085934333,
      "grad_norm": 0.026163168251514435,
      "learning_rate": 0.00017932711795703285,
      "loss": 0.2401,
      "step": 2040
    },
    {
      "epoch": 0.4154843940008107,
      "grad_norm": 0.023556355386972427,
      "learning_rate": 0.00017922578029995947,
      "loss": 0.0089,
      "step": 2050
    },
    {
      "epoch": 0.41751114714227805,
      "grad_norm": 0.028165694326162338,
      "learning_rate": 0.0001791244426428861,
      "loss": 0.0059,
      "step": 2060
    },
    {
      "epoch": 0.41953790028374544,
      "grad_norm": 0.036073777824640274,
      "learning_rate": 0.00017902310498581274,
      "loss": 0.5876,
      "step": 2070
    },
    {
      "epoch": 0.4215646534252128,
      "grad_norm": 0.02447659894824028,
      "learning_rate": 0.00017892176732873938,
      "loss": 0.0045,
      "step": 2080
    },
    {
      "epoch": 0.42359140656668015,
      "grad_norm": 0.030884768813848495,
      "learning_rate": 0.000178820429671666,
      "loss": 0.5126,
      "step": 2090
    },
    {
      "epoch": 0.42561815970814754,
      "grad_norm": 0.032383281737565994,
      "learning_rate": 0.00017871909201459262,
      "loss": 0.634,
      "step": 2100
    },
    {
      "epoch": 0.42561815970814754,
      "eval_accuracy": 0.9358974358974359,
      "eval_loss": 0.2406046986579895,
      "eval_runtime": 46.5595,
      "eval_samples_per_second": 11.727,
      "eval_steps_per_second": 1.482,
      "step": 2100
    },
    {
      "epoch": 0.4276449128496149,
      "grad_norm": 0.030924880877137184,
      "learning_rate": 0.00017861775435751927,
      "loss": 0.129,
      "step": 2110
    },
    {
      "epoch": 0.4296716659910823,
      "grad_norm": 0.7038991451263428,
      "learning_rate": 0.0001785164167004459,
      "loss": 0.4845,
      "step": 2120
    },
    {
      "epoch": 0.43169841913254964,
      "grad_norm": 0.02341308444738388,
      "learning_rate": 0.00017841507904337253,
      "loss": 0.0078,
      "step": 2130
    },
    {
      "epoch": 0.433725172274017,
      "grad_norm": 0.024457694962620735,
      "learning_rate": 0.00017831374138629915,
      "loss": 0.7938,
      "step": 2140
    },
    {
      "epoch": 0.4357519254154844,
      "grad_norm": 0.024586468935012817,
      "learning_rate": 0.0001782124037292258,
      "loss": 0.3548,
      "step": 2150
    },
    {
      "epoch": 0.43777867855695174,
      "grad_norm": 0.5639256834983826,
      "learning_rate": 0.00017811106607215242,
      "loss": 0.3112,
      "step": 2160
    },
    {
      "epoch": 0.4398054316984191,
      "grad_norm": 0.026940103620290756,
      "learning_rate": 0.00017800972841507906,
      "loss": 0.5675,
      "step": 2170
    },
    {
      "epoch": 0.4418321848398865,
      "grad_norm": 0.027325104922056198,
      "learning_rate": 0.00017790839075800568,
      "loss": 0.014,
      "step": 2180
    },
    {
      "epoch": 0.4438589379813539,
      "grad_norm": 0.024721914902329445,
      "learning_rate": 0.0001778070531009323,
      "loss": 0.5824,
      "step": 2190
    },
    {
      "epoch": 0.4458856911228212,
      "grad_norm": 0.03028850629925728,
      "learning_rate": 0.00017770571544385895,
      "loss": 0.4471,
      "step": 2200
    },
    {
      "epoch": 0.4458856911228212,
      "eval_accuracy": 0.9377289377289377,
      "eval_loss": 0.29827624559402466,
      "eval_runtime": 50.2155,
      "eval_samples_per_second": 10.873,
      "eval_steps_per_second": 1.374,
      "step": 2200
    },
    {
      "epoch": 0.4479124442642886,
      "grad_norm": 44.90025329589844,
      "learning_rate": 0.00017760437778678557,
      "loss": 0.5802,
      "step": 2210
    },
    {
      "epoch": 0.449939197405756,
      "grad_norm": 0.1824452131986618,
      "learning_rate": 0.00017750304012971221,
      "loss": 0.006,
      "step": 2220
    },
    {
      "epoch": 0.4519659505472233,
      "grad_norm": 0.03286247327923775,
      "learning_rate": 0.00017740170247263883,
      "loss": 0.4521,
      "step": 2230
    },
    {
      "epoch": 0.4539927036886907,
      "grad_norm": 0.030796563252806664,
      "learning_rate": 0.00017730036481556548,
      "loss": 0.0075,
      "step": 2240
    },
    {
      "epoch": 0.4560194568301581,
      "grad_norm": 0.025900913402438164,
      "learning_rate": 0.0001771990271584921,
      "loss": 0.0054,
      "step": 2250
    },
    {
      "epoch": 0.45804620997162543,
      "grad_norm": 0.02531251683831215,
      "learning_rate": 0.00017709768950141874,
      "loss": 0.4758,
      "step": 2260
    },
    {
      "epoch": 0.4600729631130928,
      "grad_norm": 2.9233317375183105,
      "learning_rate": 0.00017699635184434536,
      "loss": 0.4245,
      "step": 2270
    },
    {
      "epoch": 0.4620997162545602,
      "grad_norm": 8.508463859558105,
      "learning_rate": 0.00017689501418727198,
      "loss": 0.4203,
      "step": 2280
    },
    {
      "epoch": 0.4641264693960276,
      "grad_norm": 8.093629837036133,
      "learning_rate": 0.00017679367653019863,
      "loss": 1.4054,
      "step": 2290
    },
    {
      "epoch": 0.4661532225374949,
      "grad_norm": 1.5853863954544067,
      "learning_rate": 0.00017669233887312527,
      "loss": 0.3229,
      "step": 2300
    },
    {
      "epoch": 0.4661532225374949,
      "eval_accuracy": 0.9212454212454212,
      "eval_loss": 0.3600928783416748,
      "eval_runtime": 49.7309,
      "eval_samples_per_second": 10.979,
      "eval_steps_per_second": 1.387,
      "step": 2300
    },
    {
      "epoch": 0.4681799756789623,
      "grad_norm": 0.03744541481137276,
      "learning_rate": 0.0001765910012160519,
      "loss": 1.5756,
      "step": 2310
    },
    {
      "epoch": 0.4702067288204297,
      "grad_norm": 8.397408485412598,
      "learning_rate": 0.00017648966355897854,
      "loss": 0.5392,
      "step": 2320
    },
    {
      "epoch": 0.472233481961897,
      "grad_norm": 0.07392023503780365,
      "learning_rate": 0.00017638832590190516,
      "loss": 0.4868,
      "step": 2330
    },
    {
      "epoch": 0.4742602351033644,
      "grad_norm": 0.07620003819465637,
      "learning_rate": 0.00017628698824483178,
      "loss": 0.2934,
      "step": 2340
    },
    {
      "epoch": 0.4762869882448318,
      "grad_norm": 0.05634753778576851,
      "learning_rate": 0.00017618565058775842,
      "loss": 1.0435,
      "step": 2350
    },
    {
      "epoch": 0.4783137413862992,
      "grad_norm": 0.06447988003492355,
      "learning_rate": 0.00017608431293068507,
      "loss": 0.2949,
      "step": 2360
    },
    {
      "epoch": 0.4803404945277665,
      "grad_norm": 0.12075554579496384,
      "learning_rate": 0.00017598297527361166,
      "loss": 0.5441,
      "step": 2370
    },
    {
      "epoch": 0.4823672476692339,
      "grad_norm": 0.04562091454863548,
      "learning_rate": 0.0001758816376165383,
      "loss": 0.2559,
      "step": 2380
    },
    {
      "epoch": 0.4843940008107013,
      "grad_norm": 0.03965690732002258,
      "learning_rate": 0.00017578029995946496,
      "loss": 0.4775,
      "step": 2390
    },
    {
      "epoch": 0.4864207539521686,
      "grad_norm": 0.038196004927158356,
      "learning_rate": 0.00017567896230239157,
      "loss": 0.4769,
      "step": 2400
    },
    {
      "epoch": 0.4864207539521686,
      "eval_accuracy": 0.9010989010989011,
      "eval_loss": 0.2990172803401947,
      "eval_runtime": 48.3946,
      "eval_samples_per_second": 11.282,
      "eval_steps_per_second": 1.426,
      "step": 2400
    },
    {
      "epoch": 0.488447507093636,
      "grad_norm": 0.8857225775718689,
      "learning_rate": 0.00017557762464531822,
      "loss": 0.508,
      "step": 2410
    },
    {
      "epoch": 0.4904742602351034,
      "grad_norm": 0.06662506610155106,
      "learning_rate": 0.00017547628698824484,
      "loss": 0.0342,
      "step": 2420
    },
    {
      "epoch": 0.4925010133765707,
      "grad_norm": 0.06565208733081818,
      "learning_rate": 0.00017537494933117146,
      "loss": 0.2875,
      "step": 2430
    },
    {
      "epoch": 0.4945277665180381,
      "grad_norm": 0.056362636387348175,
      "learning_rate": 0.0001752736116740981,
      "loss": 0.7128,
      "step": 2440
    },
    {
      "epoch": 0.4965545196595055,
      "grad_norm": 0.042020898312330246,
      "learning_rate": 0.00017517227401702475,
      "loss": 0.4941,
      "step": 2450
    },
    {
      "epoch": 0.49858127280097286,
      "grad_norm": 0.11986752599477768,
      "learning_rate": 0.00017507093635995137,
      "loss": 0.3659,
      "step": 2460
    },
    {
      "epoch": 0.5006080259424402,
      "grad_norm": 0.03309104964137077,
      "learning_rate": 0.000174969598702878,
      "loss": 0.0116,
      "step": 2470
    },
    {
      "epoch": 0.5026347790839076,
      "grad_norm": 0.031792882829904556,
      "learning_rate": 0.00017486826104580464,
      "loss": 0.017,
      "step": 2480
    },
    {
      "epoch": 0.5046615322253749,
      "grad_norm": 0.029629742726683617,
      "learning_rate": 0.00017476692338873125,
      "loss": 0.8387,
      "step": 2490
    },
    {
      "epoch": 0.5066882853668423,
      "grad_norm": 0.029320184141397476,
      "learning_rate": 0.0001746655857316579,
      "loss": 0.0135,
      "step": 2500
    },
    {
      "epoch": 0.5066882853668423,
      "eval_accuracy": 0.9029304029304029,
      "eval_loss": 0.31343939900398254,
      "eval_runtime": 46.7566,
      "eval_samples_per_second": 11.678,
      "eval_steps_per_second": 1.476,
      "step": 2500
    },
    {
      "epoch": 0.5087150385083097,
      "grad_norm": 0.03285427764058113,
      "learning_rate": 0.00017456424807458452,
      "loss": 0.3224,
      "step": 2510
    },
    {
      "epoch": 0.5107417916497771,
      "grad_norm": 0.02792639471590519,
      "learning_rate": 0.00017446291041751114,
      "loss": 0.0042,
      "step": 2520
    },
    {
      "epoch": 0.5127685447912445,
      "grad_norm": 0.047437168657779694,
      "learning_rate": 0.00017436157276043779,
      "loss": 0.0038,
      "step": 2530
    },
    {
      "epoch": 0.5147952979327118,
      "grad_norm": 0.025487160310149193,
      "learning_rate": 0.00017426023510336443,
      "loss": 0.0088,
      "step": 2540
    },
    {
      "epoch": 0.5168220510741791,
      "grad_norm": 0.02262968011200428,
      "learning_rate": 0.00017415889744629105,
      "loss": 0.0162,
      "step": 2550
    },
    {
      "epoch": 0.5188488042156465,
      "grad_norm": 0.021806327626109123,
      "learning_rate": 0.00017405755978921767,
      "loss": 0.34,
      "step": 2560
    },
    {
      "epoch": 0.5208755573571139,
      "grad_norm": 0.022118626162409782,
      "learning_rate": 0.00017395622213214432,
      "loss": 0.3285,
      "step": 2570
    },
    {
      "epoch": 0.5229023104985813,
      "grad_norm": 0.021631283685564995,
      "learning_rate": 0.00017385488447507093,
      "loss": 0.0077,
      "step": 2580
    },
    {
      "epoch": 0.5249290636400487,
      "grad_norm": 0.020419137552380562,
      "learning_rate": 0.00017375354681799758,
      "loss": 0.3255,
      "step": 2590
    },
    {
      "epoch": 0.526955816781516,
      "grad_norm": 0.020009351894259453,
      "learning_rate": 0.00017365220916092423,
      "loss": 0.3025,
      "step": 2600
    },
    {
      "epoch": 0.526955816781516,
      "eval_accuracy": 0.9505494505494505,
      "eval_loss": 0.17476053535938263,
      "eval_runtime": 50.4498,
      "eval_samples_per_second": 10.823,
      "eval_steps_per_second": 1.368,
      "step": 2600
    },
    {
      "epoch": 0.5289825699229834,
      "grad_norm": 0.022722359746694565,
      "learning_rate": 0.00017355087150385082,
      "loss": 0.2968,
      "step": 2610
    },
    {
      "epoch": 0.5310093230644507,
      "grad_norm": 0.783506453037262,
      "learning_rate": 0.00017344953384677747,
      "loss": 0.5455,
      "step": 2620
    },
    {
      "epoch": 0.5330360762059181,
      "grad_norm": 0.024192122742533684,
      "learning_rate": 0.0001733481961897041,
      "loss": 0.2437,
      "step": 2630
    },
    {
      "epoch": 0.5350628293473855,
      "grad_norm": 0.01950955204665661,
      "learning_rate": 0.00017324685853263073,
      "loss": 0.2505,
      "step": 2640
    },
    {
      "epoch": 0.5370895824888529,
      "grad_norm": 0.018212512135505676,
      "learning_rate": 0.00017314552087555735,
      "loss": 0.0021,
      "step": 2650
    },
    {
      "epoch": 0.5391163356303202,
      "grad_norm": 0.0288174357265234,
      "learning_rate": 0.000173044183218484,
      "loss": 0.8558,
      "step": 2660
    },
    {
      "epoch": 0.5411430887717876,
      "grad_norm": 0.026978524401783943,
      "learning_rate": 0.00017294284556141062,
      "loss": 0.6232,
      "step": 2670
    },
    {
      "epoch": 0.543169841913255,
      "grad_norm": 7.463168144226074,
      "learning_rate": 0.00017284150790433726,
      "loss": 0.2648,
      "step": 2680
    },
    {
      "epoch": 0.5451965950547223,
      "grad_norm": 0.6724880337715149,
      "learning_rate": 0.0001727401702472639,
      "loss": 0.0268,
      "step": 2690
    },
    {
      "epoch": 0.5472233481961897,
      "grad_norm": 0.2459801882505417,
      "learning_rate": 0.00017263883259019053,
      "loss": 0.0114,
      "step": 2700
    },
    {
      "epoch": 0.5472233481961897,
      "eval_accuracy": 0.9212454212454212,
      "eval_loss": 0.2898261249065399,
      "eval_runtime": 48.0997,
      "eval_samples_per_second": 11.351,
      "eval_steps_per_second": 1.435,
      "step": 2700
    },
    {
      "epoch": 0.5492501013376571,
      "grad_norm": 0.017503296956419945,
      "learning_rate": 0.00017253749493311715,
      "loss": 0.0114,
      "step": 2710
    },
    {
      "epoch": 0.5512768544791244,
      "grad_norm": 0.14966024458408356,
      "learning_rate": 0.0001724361572760438,
      "loss": 0.5774,
      "step": 2720
    },
    {
      "epoch": 0.5533036076205918,
      "grad_norm": 0.05726105347275734,
      "learning_rate": 0.0001723348196189704,
      "loss": 0.0069,
      "step": 2730
    },
    {
      "epoch": 0.5553303607620592,
      "grad_norm": 0.020008716732263565,
      "learning_rate": 0.00017223348196189706,
      "loss": 0.4275,
      "step": 2740
    },
    {
      "epoch": 0.5573571139035266,
      "grad_norm": 0.019402628764510155,
      "learning_rate": 0.00017213214430482368,
      "loss": 0.0041,
      "step": 2750
    },
    {
      "epoch": 0.5593838670449939,
      "grad_norm": 0.029374495148658752,
      "learning_rate": 0.0001720308066477503,
      "loss": 0.0031,
      "step": 2760
    },
    {
      "epoch": 0.5614106201864613,
      "grad_norm": 0.016268255189061165,
      "learning_rate": 0.00017192946899067694,
      "loss": 0.0035,
      "step": 2770
    },
    {
      "epoch": 0.5634373733279286,
      "grad_norm": 0.03707795962691307,
      "learning_rate": 0.0001718281313336036,
      "loss": 1.0901,
      "step": 2780
    },
    {
      "epoch": 0.565464126469396,
      "grad_norm": 0.12352888286113739,
      "learning_rate": 0.0001717267936765302,
      "loss": 0.5252,
      "step": 2790
    },
    {
      "epoch": 0.5674908796108634,
      "grad_norm": 0.028375711292028427,
      "learning_rate": 0.00017162545601945683,
      "loss": 0.1636,
      "step": 2800
    },
    {
      "epoch": 0.5674908796108634,
      "eval_accuracy": 0.9395604395604396,
      "eval_loss": 0.22813718020915985,
      "eval_runtime": 49.0458,
      "eval_samples_per_second": 11.132,
      "eval_steps_per_second": 1.407,
      "step": 2800
    },
    {
      "epoch": 0.5695176327523308,
      "grad_norm": 0.03394703567028046,
      "learning_rate": 0.00017152411836238347,
      "loss": 1.4137,
      "step": 2810
    },
    {
      "epoch": 0.5715443858937982,
      "grad_norm": 0.04633708298206329,
      "learning_rate": 0.0001714227807053101,
      "loss": 0.391,
      "step": 2820
    },
    {
      "epoch": 0.5735711390352655,
      "grad_norm": 0.2281627207994461,
      "learning_rate": 0.00017132144304823674,
      "loss": 0.0142,
      "step": 2830
    },
    {
      "epoch": 0.5755978921767329,
      "grad_norm": 0.034119848161935806,
      "learning_rate": 0.00017122010539116336,
      "loss": 0.6163,
      "step": 2840
    },
    {
      "epoch": 0.5776246453182002,
      "grad_norm": 0.18611551821231842,
      "learning_rate": 0.00017111876773409,
      "loss": 0.0072,
      "step": 2850
    },
    {
      "epoch": 0.5796513984596676,
      "grad_norm": 0.30093681812286377,
      "learning_rate": 0.00017101743007701662,
      "loss": 0.4285,
      "step": 2860
    },
    {
      "epoch": 0.581678151601135,
      "grad_norm": 0.1138378456234932,
      "learning_rate": 0.00017091609241994327,
      "loss": 0.3155,
      "step": 2870
    },
    {
      "epoch": 0.5837049047426024,
      "grad_norm": 0.038984451442956924,
      "learning_rate": 0.0001708147547628699,
      "loss": 0.7609,
      "step": 2880
    },
    {
      "epoch": 0.5857316578840697,
      "grad_norm": 0.031690433621406555,
      "learning_rate": 0.0001707134171057965,
      "loss": 0.8971,
      "step": 2890
    },
    {
      "epoch": 0.587758411025537,
      "grad_norm": 16.161060333251953,
      "learning_rate": 0.00017061207944872315,
      "loss": 0.7427,
      "step": 2900
    },
    {
      "epoch": 0.587758411025537,
      "eval_accuracy": 0.9340659340659341,
      "eval_loss": 0.2333686351776123,
      "eval_runtime": 52.3105,
      "eval_samples_per_second": 10.438,
      "eval_steps_per_second": 1.319,
      "step": 2900
    },
    {
      "epoch": 0.5897851641670044,
      "grad_norm": 0.06748467683792114,
      "learning_rate": 0.0001705107417916498,
      "loss": 0.6921,
      "step": 2910
    },
    {
      "epoch": 0.5918119173084718,
      "grad_norm": 0.04822501912713051,
      "learning_rate": 0.00017040940413457642,
      "loss": 0.0078,
      "step": 2920
    },
    {
      "epoch": 0.5938386704499392,
      "grad_norm": 0.6019286513328552,
      "learning_rate": 0.00017030806647750306,
      "loss": 0.5709,
      "step": 2930
    },
    {
      "epoch": 0.5958654235914066,
      "grad_norm": 0.4931797981262207,
      "learning_rate": 0.00017020672882042968,
      "loss": 0.0105,
      "step": 2940
    },
    {
      "epoch": 0.597892176732874,
      "grad_norm": 0.03483499214053154,
      "learning_rate": 0.0001701053911633563,
      "loss": 0.1893,
      "step": 2950
    },
    {
      "epoch": 0.5999189298743413,
      "grad_norm": 0.030498500913381577,
      "learning_rate": 0.00017000405350628295,
      "loss": 1.0754,
      "step": 2960
    },
    {
      "epoch": 0.6019456830158086,
      "grad_norm": 0.03928288072347641,
      "learning_rate": 0.0001699027158492096,
      "loss": 0.2254,
      "step": 2970
    },
    {
      "epoch": 0.603972436157276,
      "grad_norm": 0.04761834815144539,
      "learning_rate": 0.0001698013781921362,
      "loss": 0.3954,
      "step": 2980
    },
    {
      "epoch": 0.6059991892987434,
      "grad_norm": 0.04778355360031128,
      "learning_rate": 0.00016970004053506283,
      "loss": 0.4868,
      "step": 2990
    },
    {
      "epoch": 0.6080259424402108,
      "grad_norm": 0.04448671266436577,
      "learning_rate": 0.00016959870287798948,
      "loss": 0.0083,
      "step": 3000
    },
    {
      "epoch": 0.6080259424402108,
      "eval_accuracy": 0.9358974358974359,
      "eval_loss": 0.24660906195640564,
      "eval_runtime": 47.9338,
      "eval_samples_per_second": 11.391,
      "eval_steps_per_second": 1.439,
      "step": 3000
    },
    {
      "epoch": 0.6100526955816782,
      "grad_norm": 0.03465289995074272,
      "learning_rate": 0.0001694973652209161,
      "loss": 0.0068,
      "step": 3010
    },
    {
      "epoch": 0.6120794487231456,
      "grad_norm": 0.03001750074326992,
      "learning_rate": 0.00016939602756384274,
      "loss": 0.0083,
      "step": 3020
    },
    {
      "epoch": 0.6141062018646128,
      "grad_norm": 9.022475242614746,
      "learning_rate": 0.00016929468990676936,
      "loss": 0.4394,
      "step": 3030
    },
    {
      "epoch": 0.6161329550060802,
      "grad_norm": 0.029527999460697174,
      "learning_rate": 0.00016919335224969598,
      "loss": 0.0052,
      "step": 3040
    },
    {
      "epoch": 0.6181597081475476,
      "grad_norm": 0.027496349066495895,
      "learning_rate": 0.00016909201459262263,
      "loss": 0.0032,
      "step": 3050
    },
    {
      "epoch": 0.620186461289015,
      "grad_norm": 0.029770225286483765,
      "learning_rate": 0.00016899067693554928,
      "loss": 0.4973,
      "step": 3060
    },
    {
      "epoch": 0.6222132144304824,
      "grad_norm": 0.03310020640492439,
      "learning_rate": 0.0001688893392784759,
      "loss": 0.4599,
      "step": 3070
    },
    {
      "epoch": 0.6242399675719498,
      "grad_norm": 0.24616017937660217,
      "learning_rate": 0.0001687880016214025,
      "loss": 0.009,
      "step": 3080
    },
    {
      "epoch": 0.6262667207134172,
      "grad_norm": 4.76401424407959,
      "learning_rate": 0.00016868666396432916,
      "loss": 0.0128,
      "step": 3090
    },
    {
      "epoch": 0.6282934738548844,
      "grad_norm": 0.01683022826910019,
      "learning_rate": 0.00016858532630725578,
      "loss": 0.0041,
      "step": 3100
    },
    {
      "epoch": 0.6282934738548844,
      "eval_accuracy": 0.9432234432234432,
      "eval_loss": 0.2736949026584625,
      "eval_runtime": 51.3858,
      "eval_samples_per_second": 10.626,
      "eval_steps_per_second": 1.343,
      "step": 3100
    },
    {
      "epoch": 0.6303202269963518,
      "grad_norm": 0.07328706234693527,
      "learning_rate": 0.00016848398865018242,
      "loss": 0.0041,
      "step": 3110
    },
    {
      "epoch": 0.6323469801378192,
      "grad_norm": 0.017396189272403717,
      "learning_rate": 0.00016838265099310904,
      "loss": 0.0032,
      "step": 3120
    },
    {
      "epoch": 0.6343737332792866,
      "grad_norm": 102.0742416381836,
      "learning_rate": 0.00016828131333603566,
      "loss": 0.341,
      "step": 3130
    },
    {
      "epoch": 0.636400486420754,
      "grad_norm": 0.014245778322219849,
      "learning_rate": 0.0001681799756789623,
      "loss": 0.6143,
      "step": 3140
    },
    {
      "epoch": 0.6384272395622214,
      "grad_norm": 0.020878920331597328,
      "learning_rate": 0.00016807863802188896,
      "loss": 0.8261,
      "step": 3150
    },
    {
      "epoch": 0.6404539927036887,
      "grad_norm": 7.475837707519531,
      "learning_rate": 0.00016797730036481557,
      "loss": 0.1837,
      "step": 3160
    },
    {
      "epoch": 0.642480745845156,
      "grad_norm": 0.0842457190155983,
      "learning_rate": 0.0001678759627077422,
      "loss": 0.0093,
      "step": 3170
    },
    {
      "epoch": 0.6445074989866234,
      "grad_norm": 0.018949219956994057,
      "learning_rate": 0.00016777462505066884,
      "loss": 0.3746,
      "step": 3180
    },
    {
      "epoch": 0.6465342521280908,
      "grad_norm": 0.017446573823690414,
      "learning_rate": 0.00016767328739359546,
      "loss": 0.017,
      "step": 3190
    },
    {
      "epoch": 0.6485610052695582,
      "grad_norm": 8.801593780517578,
      "learning_rate": 0.0001675719497365221,
      "loss": 1.7268,
      "step": 3200
    },
    {
      "epoch": 0.6485610052695582,
      "eval_accuracy": 0.9395604395604396,
      "eval_loss": 0.26258817315101624,
      "eval_runtime": 47.8977,
      "eval_samples_per_second": 11.399,
      "eval_steps_per_second": 1.441,
      "step": 3200
    },
    {
      "epoch": 0.6505877584110256,
      "grad_norm": 0.040938105434179306,
      "learning_rate": 0.00016747061207944875,
      "loss": 0.2597,
      "step": 3210
    },
    {
      "epoch": 0.652614511552493,
      "grad_norm": 0.030415307730436325,
      "learning_rate": 0.00016736927442237534,
      "loss": 0.0417,
      "step": 3220
    },
    {
      "epoch": 0.6546412646939602,
      "grad_norm": 0.02822272665798664,
      "learning_rate": 0.000167267936765302,
      "loss": 0.4415,
      "step": 3230
    },
    {
      "epoch": 0.6566680178354276,
      "grad_norm": 0.028088340535759926,
      "learning_rate": 0.00016716659910822864,
      "loss": 0.0044,
      "step": 3240
    },
    {
      "epoch": 0.658694770976895,
      "grad_norm": 0.022151542827486992,
      "learning_rate": 0.00016706526145115525,
      "loss": 0.0065,
      "step": 3250
    },
    {
      "epoch": 0.6607215241183624,
      "grad_norm": 0.10828548669815063,
      "learning_rate": 0.00016696392379408187,
      "loss": 0.0053,
      "step": 3260
    },
    {
      "epoch": 0.6627482772598298,
      "grad_norm": 0.019026095047593117,
      "learning_rate": 0.00016686258613700852,
      "loss": 0.5957,
      "step": 3270
    },
    {
      "epoch": 0.6647750304012972,
      "grad_norm": 9.751636505126953,
      "learning_rate": 0.00016676124847993514,
      "loss": 0.9879,
      "step": 3280
    },
    {
      "epoch": 0.6668017835427645,
      "grad_norm": 0.12755624949932098,
      "learning_rate": 0.00016665991082286179,
      "loss": 0.5395,
      "step": 3290
    },
    {
      "epoch": 0.6688285366842318,
      "grad_norm": 0.21898022294044495,
      "learning_rate": 0.00016655857316578843,
      "loss": 0.0115,
      "step": 3300
    },
    {
      "epoch": 0.6688285366842318,
      "eval_accuracy": 0.9304029304029304,
      "eval_loss": 0.26207664608955383,
      "eval_runtime": 47.5971,
      "eval_samples_per_second": 11.471,
      "eval_steps_per_second": 1.45,
      "step": 3300
    },
    {
      "epoch": 0.6708552898256992,
      "grad_norm": 0.05415915325284004,
      "learning_rate": 0.00016645723550871502,
      "loss": 0.3092,
      "step": 3310
    },
    {
      "epoch": 0.6728820429671666,
      "grad_norm": 0.04886092618107796,
      "learning_rate": 0.00016635589785164167,
      "loss": 0.5327,
      "step": 3320
    },
    {
      "epoch": 0.674908796108634,
      "grad_norm": 0.03909413889050484,
      "learning_rate": 0.00016625456019456832,
      "loss": 0.0087,
      "step": 3330
    },
    {
      "epoch": 0.6769355492501014,
      "grad_norm": 0.040225472301244736,
      "learning_rate": 0.00016615322253749493,
      "loss": 0.7562,
      "step": 3340
    },
    {
      "epoch": 0.6789623023915687,
      "grad_norm": 0.28949564695358276,
      "learning_rate": 0.00016605188488042158,
      "loss": 0.4992,
      "step": 3350
    },
    {
      "epoch": 0.6809890555330361,
      "grad_norm": 0.02482379972934723,
      "learning_rate": 0.0001659505472233482,
      "loss": 0.2469,
      "step": 3360
    },
    {
      "epoch": 0.6830158086745034,
      "grad_norm": 0.9251216053962708,
      "learning_rate": 0.00016584920956627482,
      "loss": 0.8497,
      "step": 3370
    },
    {
      "epoch": 0.6850425618159708,
      "grad_norm": 0.02334374003112316,
      "learning_rate": 0.00016574787190920147,
      "loss": 0.0106,
      "step": 3380
    },
    {
      "epoch": 0.6870693149574382,
      "grad_norm": 0.02161947637796402,
      "learning_rate": 0.0001656465342521281,
      "loss": 0.0092,
      "step": 3390
    },
    {
      "epoch": 0.6890960680989056,
      "grad_norm": 0.3686254322528839,
      "learning_rate": 0.00016554519659505473,
      "loss": 0.6196,
      "step": 3400
    },
    {
      "epoch": 0.6890960680989056,
      "eval_accuracy": 0.9267399267399268,
      "eval_loss": 0.3545539081096649,
      "eval_runtime": 50.3875,
      "eval_samples_per_second": 10.836,
      "eval_steps_per_second": 1.369,
      "step": 3400
    },
    {
      "epoch": 0.6911228212403729,
      "grad_norm": 0.0242935661226511,
      "learning_rate": 0.00016544385893798135,
      "loss": 0.3237,
      "step": 3410
    },
    {
      "epoch": 0.6931495743818403,
      "grad_norm": 0.023281089961528778,
      "learning_rate": 0.000165342521280908,
      "loss": 0.0075,
      "step": 3420
    },
    {
      "epoch": 0.6951763275233077,
      "grad_norm": 15.10773754119873,
      "learning_rate": 0.00016524118362383462,
      "loss": 1.8529,
      "step": 3430
    },
    {
      "epoch": 0.697203080664775,
      "grad_norm": 0.3190120756626129,
      "learning_rate": 0.00016513984596676126,
      "loss": 0.0278,
      "step": 3440
    },
    {
      "epoch": 0.6992298338062424,
      "grad_norm": 0.28799039125442505,
      "learning_rate": 0.00016503850830968788,
      "loss": 1.241,
      "step": 3450
    },
    {
      "epoch": 0.7012565869477098,
      "grad_norm": 0.08853207528591156,
      "learning_rate": 0.00016493717065261453,
      "loss": 0.0218,
      "step": 3460
    },
    {
      "epoch": 0.7032833400891771,
      "grad_norm": 0.05192064866423607,
      "learning_rate": 0.00016483583299554115,
      "loss": 0.7215,
      "step": 3470
    },
    {
      "epoch": 0.7053100932306445,
      "grad_norm": 0.35214605927467346,
      "learning_rate": 0.0001647344953384678,
      "loss": 0.6614,
      "step": 3480
    },
    {
      "epoch": 0.7073368463721119,
      "grad_norm": 0.06966587156057358,
      "learning_rate": 0.0001646331576813944,
      "loss": 0.0155,
      "step": 3490
    },
    {
      "epoch": 0.7093635995135793,
      "grad_norm": 0.04265503212809563,
      "learning_rate": 0.00016453182002432103,
      "loss": 0.0141,
      "step": 3500
    },
    {
      "epoch": 0.7093635995135793,
      "eval_accuracy": 0.9505494505494505,
      "eval_loss": 0.20635519921779633,
      "eval_runtime": 49.0829,
      "eval_samples_per_second": 11.124,
      "eval_steps_per_second": 1.406,
      "step": 3500
    },
    {
      "epoch": 0.7113903526550466,
      "grad_norm": 0.027219530194997787,
      "learning_rate": 0.00016443048236724768,
      "loss": 0.0142,
      "step": 3510
    },
    {
      "epoch": 0.713417105796514,
      "grad_norm": 0.051878347992897034,
      "learning_rate": 0.00016432914471017432,
      "loss": 0.0095,
      "step": 3520
    },
    {
      "epoch": 0.7154438589379813,
      "grad_norm": 0.025991367176175117,
      "learning_rate": 0.00016422780705310094,
      "loss": 0.0053,
      "step": 3530
    },
    {
      "epoch": 0.7174706120794487,
      "grad_norm": 8.773286819458008,
      "learning_rate": 0.0001641264693960276,
      "loss": 1.0328,
      "step": 3540
    },
    {
      "epoch": 0.7194973652209161,
      "grad_norm": 0.09533801674842834,
      "learning_rate": 0.0001640251317389542,
      "loss": 0.0059,
      "step": 3550
    },
    {
      "epoch": 0.7215241183623835,
      "grad_norm": 0.0500728003680706,
      "learning_rate": 0.00016392379408188083,
      "loss": 0.0071,
      "step": 3560
    },
    {
      "epoch": 0.7235508715038508,
      "grad_norm": 0.025775128975510597,
      "learning_rate": 0.00016382245642480747,
      "loss": 0.0076,
      "step": 3570
    },
    {
      "epoch": 0.7255776246453182,
      "grad_norm": 0.03006925620138645,
      "learning_rate": 0.00016372111876773412,
      "loss": 0.0046,
      "step": 3580
    },
    {
      "epoch": 0.7276043777867855,
      "grad_norm": 0.027190232649445534,
      "learning_rate": 0.0001636197811106607,
      "loss": 0.5673,
      "step": 3590
    },
    {
      "epoch": 0.7296311309282529,
      "grad_norm": 0.025452909991145134,
      "learning_rate": 0.00016351844345358736,
      "loss": 0.006,
      "step": 3600
    },
    {
      "epoch": 0.7296311309282529,
      "eval_accuracy": 0.9487179487179487,
      "eval_loss": 0.2203851193189621,
      "eval_runtime": 50.394,
      "eval_samples_per_second": 10.835,
      "eval_steps_per_second": 1.369,
      "step": 3600
    },
    {
      "epoch": 0.7316578840697203,
      "grad_norm": 0.0257217139005661,
      "learning_rate": 0.000163417105796514,
      "loss": 0.0044,
      "step": 3610
    },
    {
      "epoch": 0.7336846372111877,
      "grad_norm": 0.01945357397198677,
      "learning_rate": 0.00016331576813944062,
      "loss": 0.0045,
      "step": 3620
    },
    {
      "epoch": 0.7357113903526551,
      "grad_norm": 0.05564802885055542,
      "learning_rate": 0.00016321443048236727,
      "loss": 0.5722,
      "step": 3630
    },
    {
      "epoch": 0.7377381434941224,
      "grad_norm": 0.06898773461580276,
      "learning_rate": 0.0001631130928252939,
      "loss": 0.0048,
      "step": 3640
    },
    {
      "epoch": 0.7397648966355898,
      "grad_norm": 0.02191758342087269,
      "learning_rate": 0.0001630117551682205,
      "loss": 0.0043,
      "step": 3650
    },
    {
      "epoch": 0.7417916497770571,
      "grad_norm": 0.024975722655653954,
      "learning_rate": 0.00016291041751114715,
      "loss": 0.5345,
      "step": 3660
    },
    {
      "epoch": 0.7438184029185245,
      "grad_norm": 0.057791002094745636,
      "learning_rate": 0.0001628090798540738,
      "loss": 1.3099,
      "step": 3670
    },
    {
      "epoch": 0.7458451560599919,
      "grad_norm": 0.5608110427856445,
      "learning_rate": 0.00016270774219700042,
      "loss": 0.5324,
      "step": 3680
    },
    {
      "epoch": 0.7478719092014593,
      "grad_norm": 0.08051464706659317,
      "learning_rate": 0.00016260640453992704,
      "loss": 0.0108,
      "step": 3690
    },
    {
      "epoch": 0.7498986623429267,
      "grad_norm": 0.022334013134241104,
      "learning_rate": 0.00016250506688285368,
      "loss": 0.0226,
      "step": 3700
    },
    {
      "epoch": 0.7498986623429267,
      "eval_accuracy": 0.945054945054945,
      "eval_loss": 0.2544254660606384,
      "eval_runtime": 50.1882,
      "eval_samples_per_second": 10.879,
      "eval_steps_per_second": 1.375,
      "step": 3700
    },
    {
      "epoch": 0.751925415484394,
      "grad_norm": 0.019915521144866943,
      "learning_rate": 0.0001624037292257803,
      "loss": 0.005,
      "step": 3710
    },
    {
      "epoch": 0.7539521686258613,
      "grad_norm": 0.022011524066329002,
      "learning_rate": 0.00016230239156870695,
      "loss": 0.0046,
      "step": 3720
    },
    {
      "epoch": 0.7559789217673287,
      "grad_norm": 0.018707839772105217,
      "learning_rate": 0.00016220105391163357,
      "loss": 0.3635,
      "step": 3730
    },
    {
      "epoch": 0.7580056749087961,
      "grad_norm": 0.020906759425997734,
      "learning_rate": 0.0001620997162545602,
      "loss": 0.7634,
      "step": 3740
    },
    {
      "epoch": 0.7600324280502635,
      "grad_norm": 0.01719515211880207,
      "learning_rate": 0.00016199837859748683,
      "loss": 0.4532,
      "step": 3750
    },
    {
      "epoch": 0.7620591811917309,
      "grad_norm": 0.02170070819556713,
      "learning_rate": 0.00016189704094041348,
      "loss": 0.0023,
      "step": 3760
    },
    {
      "epoch": 0.7640859343331983,
      "grad_norm": 0.02049507014453411,
      "learning_rate": 0.0001617957032833401,
      "loss": 0.579,
      "step": 3770
    },
    {
      "epoch": 0.7661126874746655,
      "grad_norm": 0.0394478403031826,
      "learning_rate": 0.00016169436562626672,
      "loss": 0.3844,
      "step": 3780
    },
    {
      "epoch": 0.7681394406161329,
      "grad_norm": 0.029683848842978477,
      "learning_rate": 0.00016159302796919336,
      "loss": 0.2356,
      "step": 3790
    },
    {
      "epoch": 0.7701661937576003,
      "grad_norm": 0.027253972366452217,
      "learning_rate": 0.00016149169031211998,
      "loss": 0.0084,
      "step": 3800
    },
    {
      "epoch": 0.7701661937576003,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.16977500915527344,
      "eval_runtime": 49.3971,
      "eval_samples_per_second": 11.053,
      "eval_steps_per_second": 1.397,
      "step": 3800
    },
    {
      "epoch": 0.7721929468990677,
      "grad_norm": 0.05158557370305061,
      "learning_rate": 0.00016139035265504663,
      "loss": 0.3413,
      "step": 3810
    },
    {
      "epoch": 0.7742197000405351,
      "grad_norm": 0.030439468100667,
      "learning_rate": 0.00016128901499797328,
      "loss": 0.5162,
      "step": 3820
    },
    {
      "epoch": 0.7762464531820025,
      "grad_norm": 0.12102861702442169,
      "learning_rate": 0.00016118767734089987,
      "loss": 0.3457,
      "step": 3830
    },
    {
      "epoch": 0.7782732063234697,
      "grad_norm": 0.0236958060413599,
      "learning_rate": 0.0001610863396838265,
      "loss": 0.0053,
      "step": 3840
    },
    {
      "epoch": 0.7802999594649371,
      "grad_norm": 0.08496245741844177,
      "learning_rate": 0.00016098500202675316,
      "loss": 0.2449,
      "step": 3850
    },
    {
      "epoch": 0.7823267126064045,
      "grad_norm": 0.03004252351820469,
      "learning_rate": 0.00016088366436967978,
      "loss": 0.525,
      "step": 3860
    },
    {
      "epoch": 0.7843534657478719,
      "grad_norm": 0.12477823346853256,
      "learning_rate": 0.0001607823267126064,
      "loss": 0.5057,
      "step": 3870
    },
    {
      "epoch": 0.7863802188893393,
      "grad_norm": 0.039420656859874725,
      "learning_rate": 0.00016068098905553304,
      "loss": 0.0204,
      "step": 3880
    },
    {
      "epoch": 0.7884069720308067,
      "grad_norm": 0.026279330253601074,
      "learning_rate": 0.00016057965139845966,
      "loss": 0.409,
      "step": 3890
    },
    {
      "epoch": 0.7904337251722741,
      "grad_norm": 0.020288709551095963,
      "learning_rate": 0.0001604783137413863,
      "loss": 0.0035,
      "step": 3900
    },
    {
      "epoch": 0.7904337251722741,
      "eval_accuracy": 0.9304029304029304,
      "eval_loss": 0.25412970781326294,
      "eval_runtime": 195.8099,
      "eval_samples_per_second": 2.788,
      "eval_steps_per_second": 0.352,
      "step": 3900
    },
    {
      "epoch": 0.7924604783137413,
      "grad_norm": 0.01888447254896164,
      "learning_rate": 0.00016037697608431296,
      "loss": 0.0032,
      "step": 3910
    },
    {
      "epoch": 0.7944872314552087,
      "grad_norm": 0.017191996797919273,
      "learning_rate": 0.00016027563842723955,
      "loss": 0.0196,
      "step": 3920
    },
    {
      "epoch": 0.7965139845966761,
      "grad_norm": 0.01510152593255043,
      "learning_rate": 0.0001601743007701662,
      "loss": 0.0021,
      "step": 3930
    },
    {
      "epoch": 0.7985407377381435,
      "grad_norm": 0.014952914789319038,
      "learning_rate": 0.00016007296311309284,
      "loss": 0.4915,
      "step": 3940
    },
    {
      "epoch": 0.8005674908796109,
      "grad_norm": 0.014625638723373413,
      "learning_rate": 0.00015997162545601946,
      "loss": 0.0787,
      "step": 3950
    },
    {
      "epoch": 0.8025942440210783,
      "grad_norm": 0.022355277091264725,
      "learning_rate": 0.0001598702877989461,
      "loss": 0.9008,
      "step": 3960
    },
    {
      "epoch": 0.8046209971625456,
      "grad_norm": 9.831883430480957,
      "learning_rate": 0.00015976895014187272,
      "loss": 1.1827,
      "step": 3970
    },
    {
      "epoch": 0.8066477503040129,
      "grad_norm": 0.06537922471761703,
      "learning_rate": 0.00015966761248479934,
      "loss": 0.0154,
      "step": 3980
    },
    {
      "epoch": 0.8086745034454803,
      "grad_norm": 0.025931930169463158,
      "learning_rate": 0.000159566274827726,
      "loss": 0.9364,
      "step": 3990
    },
    {
      "epoch": 0.8107012565869477,
      "grad_norm": 0.022608162835240364,
      "learning_rate": 0.00015946493717065264,
      "loss": 0.0137,
      "step": 4000
    },
    {
      "epoch": 0.8107012565869477,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.12349825352430344,
      "eval_runtime": 197.3061,
      "eval_samples_per_second": 2.767,
      "eval_steps_per_second": 0.35,
      "step": 4000
    },
    {
      "epoch": 0.8127280097284151,
      "grad_norm": 0.022211702540516853,
      "learning_rate": 0.00015936359951357925,
      "loss": 0.2224,
      "step": 4010
    },
    {
      "epoch": 0.8147547628698825,
      "grad_norm": 8.373774528503418,
      "learning_rate": 0.00015926226185650587,
      "loss": 0.4282,
      "step": 4020
    },
    {
      "epoch": 0.8167815160113499,
      "grad_norm": 0.02716963365674019,
      "learning_rate": 0.00015916092419943252,
      "loss": 0.0043,
      "step": 4030
    },
    {
      "epoch": 0.8188082691528172,
      "grad_norm": 0.019980791956186295,
      "learning_rate": 0.00015905958654235914,
      "loss": 0.048,
      "step": 4040
    },
    {
      "epoch": 0.8208350222942845,
      "grad_norm": 0.019544512033462524,
      "learning_rate": 0.00015895824888528579,
      "loss": 0.4089,
      "step": 4050
    },
    {
      "epoch": 0.8228617754357519,
      "grad_norm": 0.13491490483283997,
      "learning_rate": 0.0001588569112282124,
      "loss": 0.8966,
      "step": 4060
    },
    {
      "epoch": 0.8248885285772193,
      "grad_norm": 0.03310864046216011,
      "learning_rate": 0.00015875557357113905,
      "loss": 0.7088,
      "step": 4070
    },
    {
      "epoch": 0.8269152817186867,
      "grad_norm": 0.1253669708967209,
      "learning_rate": 0.00015865423591406567,
      "loss": 0.0569,
      "step": 4080
    },
    {
      "epoch": 0.828942034860154,
      "grad_norm": 0.028889838606119156,
      "learning_rate": 0.00015855289825699232,
      "loss": 0.008,
      "step": 4090
    },
    {
      "epoch": 0.8309687880016214,
      "grad_norm": 0.1096082478761673,
      "learning_rate": 0.00015845156059991894,
      "loss": 0.9026,
      "step": 4100
    },
    {
      "epoch": 0.8309687880016214,
      "eval_accuracy": 0.924908424908425,
      "eval_loss": 0.33190029859542847,
      "eval_runtime": 191.8064,
      "eval_samples_per_second": 2.847,
      "eval_steps_per_second": 0.36,
      "step": 4100
    },
    {
      "epoch": 0.8329955411430888,
      "grad_norm": 0.022014208137989044,
      "learning_rate": 0.00015835022294284555,
      "loss": 0.005,
      "step": 4110
    },
    {
      "epoch": 0.8350222942845561,
      "grad_norm": 8.200719833374023,
      "learning_rate": 0.0001582488852857722,
      "loss": 0.3836,
      "step": 4120
    },
    {
      "epoch": 0.8370490474260235,
      "grad_norm": 0.26717376708984375,
      "learning_rate": 0.00015814754762869885,
      "loss": 0.0054,
      "step": 4130
    },
    {
      "epoch": 0.8390758005674909,
      "grad_norm": 0.018976308405399323,
      "learning_rate": 0.00015804620997162547,
      "loss": 0.3529,
      "step": 4140
    },
    {
      "epoch": 0.8411025537089583,
      "grad_norm": 0.020883018150925636,
      "learning_rate": 0.0001579448723145521,
      "loss": 0.4956,
      "step": 4150
    },
    {
      "epoch": 0.8431293068504256,
      "grad_norm": 0.0398993194103241,
      "learning_rate": 0.00015784353465747873,
      "loss": 1.5973,
      "step": 4160
    },
    {
      "epoch": 0.845156059991893,
      "grad_norm": 1.495320439338684,
      "learning_rate": 0.00015774219700040535,
      "loss": 0.0245,
      "step": 4170
    },
    {
      "epoch": 0.8471828131333603,
      "grad_norm": 0.8191264867782593,
      "learning_rate": 0.000157640859343332,
      "loss": 0.0561,
      "step": 4180
    },
    {
      "epoch": 0.8492095662748277,
      "grad_norm": 0.01949804276227951,
      "learning_rate": 0.00015753952168625862,
      "loss": 0.0103,
      "step": 4190
    },
    {
      "epoch": 0.8512363194162951,
      "grad_norm": 0.02161702886223793,
      "learning_rate": 0.00015743818402918523,
      "loss": 0.4531,
      "step": 4200
    },
    {
      "epoch": 0.8512363194162951,
      "eval_accuracy": 0.9413919413919414,
      "eval_loss": 0.2221231460571289,
      "eval_runtime": 191.3883,
      "eval_samples_per_second": 2.853,
      "eval_steps_per_second": 0.361,
      "step": 4200
    },
    {
      "epoch": 0.8532630725577625,
      "grad_norm": 0.16009657084941864,
      "learning_rate": 0.00015733684637211188,
      "loss": 0.0062,
      "step": 4210
    },
    {
      "epoch": 0.8552898256992298,
      "grad_norm": 0.016462478786706924,
      "learning_rate": 0.00015723550871503853,
      "loss": 0.0053,
      "step": 4220
    },
    {
      "epoch": 0.8573165788406972,
      "grad_norm": 0.11075223982334137,
      "learning_rate": 0.00015713417105796515,
      "loss": 0.0055,
      "step": 4230
    },
    {
      "epoch": 0.8593433319821646,
      "grad_norm": 8.982980728149414,
      "learning_rate": 0.0001570328334008918,
      "loss": 0.9514,
      "step": 4240
    },
    {
      "epoch": 0.8613700851236319,
      "grad_norm": 0.023856502026319504,
      "learning_rate": 0.0001569314957438184,
      "loss": 0.2588,
      "step": 4250
    },
    {
      "epoch": 0.8633968382650993,
      "grad_norm": 0.025910574942827225,
      "learning_rate": 0.00015683015808674503,
      "loss": 0.0022,
      "step": 4260
    },
    {
      "epoch": 0.8654235914065667,
      "grad_norm": 0.017551589757204056,
      "learning_rate": 0.00015672882042967168,
      "loss": 0.0036,
      "step": 4270
    },
    {
      "epoch": 0.867450344548034,
      "grad_norm": 0.016716953366994858,
      "learning_rate": 0.00015662748277259832,
      "loss": 0.0411,
      "step": 4280
    },
    {
      "epoch": 0.8694770976895014,
      "grad_norm": 0.017348669469356537,
      "learning_rate": 0.00015652614511552494,
      "loss": 0.4721,
      "step": 4290
    },
    {
      "epoch": 0.8715038508309688,
      "grad_norm": 0.016933787614107132,
      "learning_rate": 0.00015642480745845156,
      "loss": 0.0039,
      "step": 4300
    },
    {
      "epoch": 0.8715038508309688,
      "eval_accuracy": 0.9560439560439561,
      "eval_loss": 0.18227943778038025,
      "eval_runtime": 177.8492,
      "eval_samples_per_second": 3.07,
      "eval_steps_per_second": 0.388,
      "step": 4300
    },
    {
      "epoch": 0.8735306039724362,
      "grad_norm": 0.016385329887270927,
      "learning_rate": 0.0001563234698013782,
      "loss": 0.0029,
      "step": 4310
    },
    {
      "epoch": 0.8755573571139035,
      "grad_norm": 0.015703881159424782,
      "learning_rate": 0.00015622213214430483,
      "loss": 0.0041,
      "step": 4320
    },
    {
      "epoch": 0.8775841102553709,
      "grad_norm": 0.01479947380721569,
      "learning_rate": 0.00015612079448723147,
      "loss": 0.0026,
      "step": 4330
    },
    {
      "epoch": 0.8796108633968382,
      "grad_norm": 0.10049838572740555,
      "learning_rate": 0.0001560194568301581,
      "loss": 0.4659,
      "step": 4340
    },
    {
      "epoch": 0.8816376165383056,
      "grad_norm": 0.014785503037273884,
      "learning_rate": 0.0001559181191730847,
      "loss": 0.0038,
      "step": 4350
    },
    {
      "epoch": 0.883664369679773,
      "grad_norm": 0.017938219010829926,
      "learning_rate": 0.00015581678151601136,
      "loss": 0.0041,
      "step": 4360
    },
    {
      "epoch": 0.8856911228212404,
      "grad_norm": 0.01392845343798399,
      "learning_rate": 0.000155715443858938,
      "loss": 0.0016,
      "step": 4370
    },
    {
      "epoch": 0.8877178759627078,
      "grad_norm": 0.013149906881153584,
      "learning_rate": 0.00015561410620186462,
      "loss": 0.0016,
      "step": 4380
    },
    {
      "epoch": 0.8897446291041751,
      "grad_norm": 0.012668000534176826,
      "learning_rate": 0.00015551276854479124,
      "loss": 0.0029,
      "step": 4390
    },
    {
      "epoch": 0.8917713822456425,
      "grad_norm": 0.013916357420384884,
      "learning_rate": 0.0001554114308877179,
      "loss": 1.3298,
      "step": 4400
    },
    {
      "epoch": 0.8917713822456425,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.21251553297042847,
      "eval_runtime": 44.7327,
      "eval_samples_per_second": 12.206,
      "eval_steps_per_second": 1.542,
      "step": 4400
    },
    {
      "epoch": 0.8937981353871098,
      "grad_norm": 8.101142883300781,
      "learning_rate": 0.0001553100932306445,
      "loss": 0.3703,
      "step": 4410
    },
    {
      "epoch": 0.8958248885285772,
      "grad_norm": 0.024000072851777077,
      "learning_rate": 0.00015520875557357115,
      "loss": 0.0048,
      "step": 4420
    },
    {
      "epoch": 0.8978516416700446,
      "grad_norm": 0.12596918642520905,
      "learning_rate": 0.0001551074179164978,
      "loss": 1.2688,
      "step": 4430
    },
    {
      "epoch": 0.899878394811512,
      "grad_norm": 0.021065017208456993,
      "learning_rate": 0.0001550060802594244,
      "loss": 0.6219,
      "step": 4440
    },
    {
      "epoch": 0.9019051479529794,
      "grad_norm": 0.017869850620627403,
      "learning_rate": 0.00015490474260235104,
      "loss": 0.0409,
      "step": 4450
    },
    {
      "epoch": 0.9039319010944467,
      "grad_norm": 0.016384383663535118,
      "learning_rate": 0.00015480340494527768,
      "loss": 0.0156,
      "step": 4460
    },
    {
      "epoch": 0.905958654235914,
      "grad_norm": 0.0144584272056818,
      "learning_rate": 0.0001547020672882043,
      "loss": 0.2851,
      "step": 4470
    },
    {
      "epoch": 0.9079854073773814,
      "grad_norm": 0.01379258744418621,
      "learning_rate": 0.00015460072963113092,
      "loss": 0.0049,
      "step": 4480
    },
    {
      "epoch": 0.9100121605188488,
      "grad_norm": 0.01782505214214325,
      "learning_rate": 0.00015449939197405757,
      "loss": 0.5712,
      "step": 4490
    },
    {
      "epoch": 0.9120389136603162,
      "grad_norm": 0.2501247525215149,
      "learning_rate": 0.0001543980543169842,
      "loss": 0.4403,
      "step": 4500
    },
    {
      "epoch": 0.9120389136603162,
      "eval_accuracy": 0.8937728937728938,
      "eval_loss": 0.49004384875297546,
      "eval_runtime": 49.7342,
      "eval_samples_per_second": 10.978,
      "eval_steps_per_second": 1.387,
      "step": 4500
    },
    {
      "epoch": 0.9140656668017836,
      "grad_norm": 0.017710473388433456,
      "learning_rate": 0.00015429671665991083,
      "loss": 0.2776,
      "step": 4510
    },
    {
      "epoch": 0.9160924199432509,
      "grad_norm": 0.018590105697512627,
      "learning_rate": 0.00015419537900283748,
      "loss": 0.3543,
      "step": 4520
    },
    {
      "epoch": 0.9181191730847182,
      "grad_norm": 0.08836981654167175,
      "learning_rate": 0.00015409404134576407,
      "loss": 1.3144,
      "step": 4530
    },
    {
      "epoch": 0.9201459262261856,
      "grad_norm": 0.7396237850189209,
      "learning_rate": 0.00015399270368869072,
      "loss": 0.7759,
      "step": 4540
    },
    {
      "epoch": 0.922172679367653,
      "grad_norm": 0.04432849586009979,
      "learning_rate": 0.00015389136603161736,
      "loss": 0.3832,
      "step": 4550
    },
    {
      "epoch": 0.9241994325091204,
      "grad_norm": 0.030519770458340645,
      "learning_rate": 0.00015379002837454398,
      "loss": 0.3436,
      "step": 4560
    },
    {
      "epoch": 0.9262261856505878,
      "grad_norm": 0.033803071826696396,
      "learning_rate": 0.00015368869071747063,
      "loss": 0.0147,
      "step": 4570
    },
    {
      "epoch": 0.9282529387920552,
      "grad_norm": 0.022388586774468422,
      "learning_rate": 0.00015358735306039725,
      "loss": 0.0106,
      "step": 4580
    },
    {
      "epoch": 0.9302796919335224,
      "grad_norm": 0.017862534150481224,
      "learning_rate": 0.00015348601540332387,
      "loss": 0.0053,
      "step": 4590
    },
    {
      "epoch": 0.9323064450749898,
      "grad_norm": 0.016625819727778435,
      "learning_rate": 0.0001533846777462505,
      "loss": 0.0025,
      "step": 4600
    },
    {
      "epoch": 0.9323064450749898,
      "eval_accuracy": 0.924908424908425,
      "eval_loss": 0.3010227382183075,
      "eval_runtime": 47.4421,
      "eval_samples_per_second": 11.509,
      "eval_steps_per_second": 1.454,
      "step": 4600
    },
    {
      "epoch": 0.9343331982164572,
      "grad_norm": 0.08215595036745071,
      "learning_rate": 0.00015328334008917716,
      "loss": 0.0045,
      "step": 4610
    },
    {
      "epoch": 0.9363599513579246,
      "grad_norm": 0.013064803555607796,
      "learning_rate": 0.00015318200243210378,
      "loss": 0.0085,
      "step": 4620
    },
    {
      "epoch": 0.938386704499392,
      "grad_norm": 0.11635798960924149,
      "learning_rate": 0.0001530806647750304,
      "loss": 0.0041,
      "step": 4630
    },
    {
      "epoch": 0.9404134576408594,
      "grad_norm": 0.012454268522560596,
      "learning_rate": 0.00015297932711795704,
      "loss": 0.0062,
      "step": 4640
    },
    {
      "epoch": 0.9424402107823268,
      "grad_norm": 0.08445250242948532,
      "learning_rate": 0.00015287798946088366,
      "loss": 0.0033,
      "step": 4650
    },
    {
      "epoch": 0.944466963923794,
      "grad_norm": 0.009565513581037521,
      "learning_rate": 0.0001527766518038103,
      "loss": 0.0016,
      "step": 4660
    },
    {
      "epoch": 0.9464937170652614,
      "grad_norm": 0.009623161517083645,
      "learning_rate": 0.00015267531414673693,
      "loss": 0.6146,
      "step": 4670
    },
    {
      "epoch": 0.9485204702067288,
      "grad_norm": 0.046667005866765976,
      "learning_rate": 0.00015257397648966357,
      "loss": 0.4766,
      "step": 4680
    },
    {
      "epoch": 0.9505472233481962,
      "grad_norm": 12.367220878601074,
      "learning_rate": 0.0001524726388325902,
      "loss": 0.8669,
      "step": 4690
    },
    {
      "epoch": 0.9525739764896636,
      "grad_norm": 0.16352809965610504,
      "learning_rate": 0.00015237130117551684,
      "loss": 0.0056,
      "step": 4700
    },
    {
      "epoch": 0.9525739764896636,
      "eval_accuracy": 0.9267399267399268,
      "eval_loss": 0.29778388142585754,
      "eval_runtime": 50.4375,
      "eval_samples_per_second": 10.825,
      "eval_steps_per_second": 1.368,
      "step": 4700
    },
    {
      "epoch": 0.954600729631131,
      "grad_norm": 0.012875114567577839,
      "learning_rate": 0.00015226996351844346,
      "loss": 1.1458,
      "step": 4710
    },
    {
      "epoch": 0.9566274827725983,
      "grad_norm": 0.280487060546875,
      "learning_rate": 0.00015216862586137008,
      "loss": 0.0048,
      "step": 4720
    },
    {
      "epoch": 0.9586542359140656,
      "grad_norm": 0.014041159301996231,
      "learning_rate": 0.00015206728820429672,
      "loss": 0.0014,
      "step": 4730
    },
    {
      "epoch": 0.960680989055533,
      "grad_norm": 0.26647835969924927,
      "learning_rate": 0.00015196595054722337,
      "loss": 0.5942,
      "step": 4740
    },
    {
      "epoch": 0.9627077421970004,
      "grad_norm": 0.4411877691745758,
      "learning_rate": 0.00015186461289015,
      "loss": 1.0128,
      "step": 4750
    },
    {
      "epoch": 0.9647344953384678,
      "grad_norm": 0.20091605186462402,
      "learning_rate": 0.00015176327523307664,
      "loss": 0.4349,
      "step": 4760
    },
    {
      "epoch": 0.9667612484799352,
      "grad_norm": 0.026517393067479134,
      "learning_rate": 0.00015166193757600325,
      "loss": 0.3691,
      "step": 4770
    },
    {
      "epoch": 0.9687880016214026,
      "grad_norm": 0.02195628732442856,
      "learning_rate": 0.00015156059991892987,
      "loss": 0.4035,
      "step": 4780
    },
    {
      "epoch": 0.9708147547628699,
      "grad_norm": 0.020767178386449814,
      "learning_rate": 0.00015145926226185652,
      "loss": 0.3917,
      "step": 4790
    },
    {
      "epoch": 0.9728415079043372,
      "grad_norm": 8.028557777404785,
      "learning_rate": 0.00015135792460478314,
      "loss": 0.3642,
      "step": 4800
    },
    {
      "epoch": 0.9728415079043372,
      "eval_accuracy": 0.945054945054945,
      "eval_loss": 0.21620576083660126,
      "eval_runtime": 48.6161,
      "eval_samples_per_second": 11.231,
      "eval_steps_per_second": 1.419,
      "step": 4800
    },
    {
      "epoch": 0.9748682610458046,
      "grad_norm": 0.03613073378801346,
      "learning_rate": 0.00015125658694770976,
      "loss": 0.3875,
      "step": 4810
    },
    {
      "epoch": 0.976895014187272,
      "grad_norm": 0.391535222530365,
      "learning_rate": 0.0001511552492906364,
      "loss": 0.0072,
      "step": 4820
    },
    {
      "epoch": 0.9789217673287394,
      "grad_norm": 0.020091639831662178,
      "learning_rate": 0.00015105391163356305,
      "loss": 0.0022,
      "step": 4830
    },
    {
      "epoch": 0.9809485204702068,
      "grad_norm": 0.012985313311219215,
      "learning_rate": 0.00015095257397648967,
      "loss": 0.0087,
      "step": 4840
    },
    {
      "epoch": 0.9829752736116741,
      "grad_norm": 0.013414951972663403,
      "learning_rate": 0.00015085123631941632,
      "loss": 0.0016,
      "step": 4850
    },
    {
      "epoch": 0.9850020267531414,
      "grad_norm": 9.259086608886719,
      "learning_rate": 0.00015074989866234294,
      "loss": 0.6416,
      "step": 4860
    },
    {
      "epoch": 0.9870287798946088,
      "grad_norm": 0.02051294408738613,
      "learning_rate": 0.00015064856100526955,
      "loss": 0.002,
      "step": 4870
    },
    {
      "epoch": 0.9890555330360762,
      "grad_norm": 0.015335061587393284,
      "learning_rate": 0.0001505472233481962,
      "loss": 0.0025,
      "step": 4880
    },
    {
      "epoch": 0.9910822861775436,
      "grad_norm": 0.11195827275514603,
      "learning_rate": 0.00015044588569112285,
      "loss": 0.0069,
      "step": 4890
    },
    {
      "epoch": 0.993109039319011,
      "grad_norm": 0.01335314754396677,
      "learning_rate": 0.00015034454803404947,
      "loss": 0.5704,
      "step": 4900
    },
    {
      "epoch": 0.993109039319011,
      "eval_accuracy": 0.9413919413919414,
      "eval_loss": 0.24586544930934906,
      "eval_runtime": 50.4198,
      "eval_samples_per_second": 10.829,
      "eval_steps_per_second": 1.369,
      "step": 4900
    },
    {
      "epoch": 0.9951357924604783,
      "grad_norm": 0.034988854080438614,
      "learning_rate": 0.00015024321037697608,
      "loss": 0.0083,
      "step": 4910
    },
    {
      "epoch": 0.9971625456019457,
      "grad_norm": 0.08549095690250397,
      "learning_rate": 0.00015014187271990273,
      "loss": 0.4143,
      "step": 4920
    },
    {
      "epoch": 0.999189298743413,
      "grad_norm": 7.6238579750061035,
      "learning_rate": 0.00015004053506282935,
      "loss": 0.6845,
      "step": 4930
    },
    {
      "epoch": 1.0012160518848805,
      "grad_norm": 0.051581960171461105,
      "learning_rate": 0.000149939197405756,
      "loss": 0.0079,
      "step": 4940
    },
    {
      "epoch": 1.0032428050263478,
      "grad_norm": 9.605396270751953,
      "learning_rate": 0.00014983785974868262,
      "loss": 0.3086,
      "step": 4950
    },
    {
      "epoch": 1.0052695581678153,
      "grad_norm": 0.012973079457879066,
      "learning_rate": 0.00014973652209160923,
      "loss": 0.3138,
      "step": 4960
    },
    {
      "epoch": 1.0072963113092825,
      "grad_norm": 0.012983009219169617,
      "learning_rate": 0.00014963518443453588,
      "loss": 0.0084,
      "step": 4970
    },
    {
      "epoch": 1.0093230644507498,
      "grad_norm": 0.01481639128178358,
      "learning_rate": 0.00014953384677746253,
      "loss": 0.0036,
      "step": 4980
    },
    {
      "epoch": 1.0113498175922173,
      "grad_norm": 0.019080623984336853,
      "learning_rate": 0.00014943250912038915,
      "loss": 0.0028,
      "step": 4990
    },
    {
      "epoch": 1.0133765707336846,
      "grad_norm": 0.011588463559746742,
      "learning_rate": 0.00014933117146331577,
      "loss": 0.1761,
      "step": 5000
    },
    {
      "epoch": 1.0133765707336846,
      "eval_accuracy": 0.9652014652014652,
      "eval_loss": 0.16739189624786377,
      "eval_runtime": 48.1195,
      "eval_samples_per_second": 11.347,
      "eval_steps_per_second": 1.434,
      "step": 5000
    },
    {
      "epoch": 1.015403323875152,
      "grad_norm": 0.06948883086442947,
      "learning_rate": 0.0001492298338062424,
      "loss": 0.0203,
      "step": 5010
    },
    {
      "epoch": 1.0174300770166194,
      "grad_norm": 0.010312282480299473,
      "learning_rate": 0.00014912849614916903,
      "loss": 0.1966,
      "step": 5020
    },
    {
      "epoch": 1.0194568301580866,
      "grad_norm": 0.03138704597949982,
      "learning_rate": 0.00014902715849209568,
      "loss": 0.6573,
      "step": 5030
    },
    {
      "epoch": 1.0214835832995541,
      "grad_norm": 0.03172450512647629,
      "learning_rate": 0.00014892582083502232,
      "loss": 0.0043,
      "step": 5040
    },
    {
      "epoch": 1.0235103364410214,
      "grad_norm": 0.06547462940216064,
      "learning_rate": 0.00014882448317794891,
      "loss": 0.0034,
      "step": 5050
    },
    {
      "epoch": 1.025537089582489,
      "grad_norm": 0.023958025500178337,
      "learning_rate": 0.00014872314552087556,
      "loss": 0.6314,
      "step": 5060
    },
    {
      "epoch": 1.0275638427239562,
      "grad_norm": 0.022199515253305435,
      "learning_rate": 0.0001486218078638022,
      "loss": 0.0052,
      "step": 5070
    },
    {
      "epoch": 1.0295905958654237,
      "grad_norm": 0.023740138858556747,
      "learning_rate": 0.00014852047020672883,
      "loss": 0.0033,
      "step": 5080
    },
    {
      "epoch": 1.031617349006891,
      "grad_norm": 0.021390412002801895,
      "learning_rate": 0.00014841913254965545,
      "loss": 0.0041,
      "step": 5090
    },
    {
      "epoch": 1.0336441021483582,
      "grad_norm": 0.02551642619073391,
      "learning_rate": 0.0001483177948925821,
      "loss": 0.0023,
      "step": 5100
    },
    {
      "epoch": 1.0336441021483582,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.18546713888645172,
      "eval_runtime": 50.6035,
      "eval_samples_per_second": 10.79,
      "eval_steps_per_second": 1.364,
      "step": 5100
    },
    {
      "epoch": 1.0356708552898257,
      "grad_norm": 0.011156442575156689,
      "learning_rate": 0.0001482164572355087,
      "loss": 0.0023,
      "step": 5110
    },
    {
      "epoch": 1.037697608431293,
      "grad_norm": 0.051858507096767426,
      "learning_rate": 0.00014811511957843536,
      "loss": 0.0026,
      "step": 5120
    },
    {
      "epoch": 1.0397243615727605,
      "grad_norm": 0.011161372996866703,
      "learning_rate": 0.000148013781921362,
      "loss": 0.0013,
      "step": 5130
    },
    {
      "epoch": 1.0417511147142278,
      "grad_norm": 0.009919494390487671,
      "learning_rate": 0.0001479124442642886,
      "loss": 0.0016,
      "step": 5140
    },
    {
      "epoch": 1.0437778678556953,
      "grad_norm": 0.008916059508919716,
      "learning_rate": 0.00014781110660721524,
      "loss": 0.0018,
      "step": 5150
    },
    {
      "epoch": 1.0458046209971625,
      "grad_norm": 0.009410860016942024,
      "learning_rate": 0.0001477097689501419,
      "loss": 0.0017,
      "step": 5160
    },
    {
      "epoch": 1.0478313741386298,
      "grad_norm": 0.008212919346988201,
      "learning_rate": 0.0001476084312930685,
      "loss": 0.001,
      "step": 5170
    },
    {
      "epoch": 1.0498581272800973,
      "grad_norm": 0.0102662593126297,
      "learning_rate": 0.00014750709363599515,
      "loss": 0.0016,
      "step": 5180
    },
    {
      "epoch": 1.0518848804215646,
      "grad_norm": 0.008377410471439362,
      "learning_rate": 0.00014740575597892177,
      "loss": 0.5981,
      "step": 5190
    },
    {
      "epoch": 1.053911633563032,
      "grad_norm": 0.011341779492795467,
      "learning_rate": 0.0001473044183218484,
      "loss": 0.1477,
      "step": 5200
    },
    {
      "epoch": 1.053911633563032,
      "eval_accuracy": 0.9652014652014652,
      "eval_loss": 0.15164771676063538,
      "eval_runtime": 50.7744,
      "eval_samples_per_second": 10.753,
      "eval_steps_per_second": 1.359,
      "step": 5200
    },
    {
      "epoch": 1.0559383867044994,
      "grad_norm": 0.0325639471411705,
      "learning_rate": 0.00014720308066477504,
      "loss": 0.1064,
      "step": 5210
    },
    {
      "epoch": 1.0579651398459669,
      "grad_norm": 0.03364536911249161,
      "learning_rate": 0.00014710174300770168,
      "loss": 0.5205,
      "step": 5220
    },
    {
      "epoch": 1.0599918929874341,
      "grad_norm": 0.034866757690906525,
      "learning_rate": 0.0001470004053506283,
      "loss": 0.0129,
      "step": 5230
    },
    {
      "epoch": 1.0620186461289014,
      "grad_norm": 0.014683687128126621,
      "learning_rate": 0.00014689906769355492,
      "loss": 0.3104,
      "step": 5240
    },
    {
      "epoch": 1.064045399270369,
      "grad_norm": 0.03544352203607559,
      "learning_rate": 0.00014679773003648157,
      "loss": 0.9745,
      "step": 5250
    },
    {
      "epoch": 1.0660721524118362,
      "grad_norm": 0.7537864446640015,
      "learning_rate": 0.0001466963923794082,
      "loss": 0.0038,
      "step": 5260
    },
    {
      "epoch": 1.0680989055533037,
      "grad_norm": 0.013130277395248413,
      "learning_rate": 0.00014659505472233483,
      "loss": 0.2515,
      "step": 5270
    },
    {
      "epoch": 1.070125658694771,
      "grad_norm": 0.013951468281447887,
      "learning_rate": 0.00014649371706526145,
      "loss": 0.5514,
      "step": 5280
    },
    {
      "epoch": 1.0721524118362384,
      "grad_norm": 0.013629689812660217,
      "learning_rate": 0.0001463923794081881,
      "loss": 0.3022,
      "step": 5290
    },
    {
      "epoch": 1.0741791649777057,
      "grad_norm": 0.016600918024778366,
      "learning_rate": 0.00014629104175111472,
      "loss": 0.0034,
      "step": 5300
    },
    {
      "epoch": 1.0741791649777057,
      "eval_accuracy": 0.7326007326007326,
      "eval_loss": 0.81169593334198,
      "eval_runtime": 50.4513,
      "eval_samples_per_second": 10.822,
      "eval_steps_per_second": 1.368,
      "step": 5300
    },
    {
      "epoch": 1.076205918119173,
      "grad_norm": 8.864251136779785,
      "learning_rate": 0.00014618970409404136,
      "loss": 0.8189,
      "step": 5310
    },
    {
      "epoch": 1.0782326712606405,
      "grad_norm": 0.012856297194957733,
      "learning_rate": 0.00014608836643696798,
      "loss": 0.0027,
      "step": 5320
    },
    {
      "epoch": 1.0802594244021078,
      "grad_norm": 0.07995825260877609,
      "learning_rate": 0.0001459870287798946,
      "loss": 0.0022,
      "step": 5330
    },
    {
      "epoch": 1.0822861775435753,
      "grad_norm": 0.0793415978550911,
      "learning_rate": 0.00014588569112282125,
      "loss": 0.168,
      "step": 5340
    },
    {
      "epoch": 1.0843129306850425,
      "grad_norm": 0.010782618075609207,
      "learning_rate": 0.00014578435346574787,
      "loss": 0.0052,
      "step": 5350
    },
    {
      "epoch": 1.08633968382651,
      "grad_norm": 0.01051894761621952,
      "learning_rate": 0.0001456830158086745,
      "loss": 0.4624,
      "step": 5360
    },
    {
      "epoch": 1.0883664369679773,
      "grad_norm": 0.013191080652177334,
      "learning_rate": 0.00014558167815160116,
      "loss": 1.1017,
      "step": 5370
    },
    {
      "epoch": 1.0903931901094446,
      "grad_norm": 0.018109530210494995,
      "learning_rate": 0.00014548034049452778,
      "loss": 0.0047,
      "step": 5380
    },
    {
      "epoch": 1.092419943250912,
      "grad_norm": 0.018270637840032578,
      "learning_rate": 0.0001453790028374544,
      "loss": 0.0019,
      "step": 5390
    },
    {
      "epoch": 1.0944466963923793,
      "grad_norm": 0.020311271771788597,
      "learning_rate": 0.00014527766518038104,
      "loss": 0.4936,
      "step": 5400
    },
    {
      "epoch": 1.0944466963923793,
      "eval_accuracy": 0.9377289377289377,
      "eval_loss": 0.21022464334964752,
      "eval_runtime": 50.3508,
      "eval_samples_per_second": 10.844,
      "eval_steps_per_second": 1.37,
      "step": 5400
    },
    {
      "epoch": 1.0964734495338468,
      "grad_norm": 0.07982522994279861,
      "learning_rate": 0.00014517632752330766,
      "loss": 0.9046,
      "step": 5410
    },
    {
      "epoch": 1.0985002026753141,
      "grad_norm": 0.1394471377134323,
      "learning_rate": 0.00014507498986623428,
      "loss": 0.3592,
      "step": 5420
    },
    {
      "epoch": 1.1005269558167816,
      "grad_norm": 0.03570636734366417,
      "learning_rate": 0.00014497365220916093,
      "loss": 0.3656,
      "step": 5430
    },
    {
      "epoch": 1.102553708958249,
      "grad_norm": 0.01714455708861351,
      "learning_rate": 0.00014487231455208757,
      "loss": 0.0669,
      "step": 5440
    },
    {
      "epoch": 1.1045804620997162,
      "grad_norm": 0.013243530876934528,
      "learning_rate": 0.0001447709768950142,
      "loss": 0.002,
      "step": 5450
    },
    {
      "epoch": 1.1066072152411837,
      "grad_norm": 0.08380251377820969,
      "learning_rate": 0.00014466963923794084,
      "loss": 0.0048,
      "step": 5460
    },
    {
      "epoch": 1.108633968382651,
      "grad_norm": 0.10041346400976181,
      "learning_rate": 0.00014456830158086746,
      "loss": 0.6733,
      "step": 5470
    },
    {
      "epoch": 1.1106607215241184,
      "grad_norm": 0.018171994015574455,
      "learning_rate": 0.00014446696392379408,
      "loss": 0.0029,
      "step": 5480
    },
    {
      "epoch": 1.1126874746655857,
      "grad_norm": 0.02070717327296734,
      "learning_rate": 0.00014436562626672072,
      "loss": 0.3299,
      "step": 5490
    },
    {
      "epoch": 1.1147142278070532,
      "grad_norm": 0.014606601558625698,
      "learning_rate": 0.00014426428860964737,
      "loss": 0.0158,
      "step": 5500
    },
    {
      "epoch": 1.1147142278070532,
      "eval_accuracy": 0.9523809523809523,
      "eval_loss": 0.18859973549842834,
      "eval_runtime": 48.3094,
      "eval_samples_per_second": 11.302,
      "eval_steps_per_second": 1.428,
      "step": 5500
    },
    {
      "epoch": 1.1167409809485205,
      "grad_norm": 0.01225406676530838,
      "learning_rate": 0.000144162950952574,
      "loss": 0.0015,
      "step": 5510
    },
    {
      "epoch": 1.1187677340899878,
      "grad_norm": 0.011000520549714565,
      "learning_rate": 0.0001440616132955006,
      "loss": 0.0194,
      "step": 5520
    },
    {
      "epoch": 1.1207944872314552,
      "grad_norm": 0.011146724224090576,
      "learning_rate": 0.00014396027563842725,
      "loss": 0.0028,
      "step": 5530
    },
    {
      "epoch": 1.1228212403729225,
      "grad_norm": 0.009962116368114948,
      "learning_rate": 0.00014385893798135387,
      "loss": 0.0026,
      "step": 5540
    },
    {
      "epoch": 1.12484799351439,
      "grad_norm": 0.01039500255137682,
      "learning_rate": 0.00014375760032428052,
      "loss": 0.4091,
      "step": 5550
    },
    {
      "epoch": 1.1268747466558573,
      "grad_norm": 0.00947809312492609,
      "learning_rate": 0.00014365626266720714,
      "loss": 0.0041,
      "step": 5560
    },
    {
      "epoch": 1.1289014997973248,
      "grad_norm": 0.01036769337952137,
      "learning_rate": 0.00014355492501013376,
      "loss": 0.331,
      "step": 5570
    },
    {
      "epoch": 1.130928252938792,
      "grad_norm": 18.621246337890625,
      "learning_rate": 0.0001434535873530604,
      "loss": 0.83,
      "step": 5580
    },
    {
      "epoch": 1.1329550060802593,
      "grad_norm": 0.24284754693508148,
      "learning_rate": 0.00014335224969598705,
      "loss": 0.019,
      "step": 5590
    },
    {
      "epoch": 1.1349817592217268,
      "grad_norm": 0.016728529706597328,
      "learning_rate": 0.00014325091203891367,
      "loss": 0.0041,
      "step": 5600
    },
    {
      "epoch": 1.1349817592217268,
      "eval_accuracy": 0.9285714285714286,
      "eval_loss": 0.25444215536117554,
      "eval_runtime": 50.4897,
      "eval_samples_per_second": 10.814,
      "eval_steps_per_second": 1.367,
      "step": 5600
    },
    {
      "epoch": 1.1370085123631941,
      "grad_norm": 0.010051474906504154,
      "learning_rate": 0.0001431495743818403,
      "loss": 0.0043,
      "step": 5610
    },
    {
      "epoch": 1.1390352655046616,
      "grad_norm": 0.008353658951818943,
      "learning_rate": 0.00014304823672476694,
      "loss": 0.0021,
      "step": 5620
    },
    {
      "epoch": 1.1410620186461289,
      "grad_norm": 0.027239592745900154,
      "learning_rate": 0.00014294689906769355,
      "loss": 0.3325,
      "step": 5630
    },
    {
      "epoch": 1.1430887717875962,
      "grad_norm": 0.00991065800189972,
      "learning_rate": 0.0001428455614106202,
      "loss": 0.0096,
      "step": 5640
    },
    {
      "epoch": 1.1451155249290637,
      "grad_norm": 0.01124146580696106,
      "learning_rate": 0.00014274422375354685,
      "loss": 0.7095,
      "step": 5650
    },
    {
      "epoch": 1.147142278070531,
      "grad_norm": 0.009897377341985703,
      "learning_rate": 0.00014264288609647344,
      "loss": 0.0032,
      "step": 5660
    },
    {
      "epoch": 1.1491690312119984,
      "grad_norm": 0.009127453900873661,
      "learning_rate": 0.00014254154843940008,
      "loss": 0.006,
      "step": 5670
    },
    {
      "epoch": 1.1511957843534657,
      "grad_norm": 0.009598911739885807,
      "learning_rate": 0.00014244021078232673,
      "loss": 0.0014,
      "step": 5680
    },
    {
      "epoch": 1.1532225374949332,
      "grad_norm": 0.033580176532268524,
      "learning_rate": 0.00014233887312525335,
      "loss": 0.0021,
      "step": 5690
    },
    {
      "epoch": 1.1552492906364005,
      "grad_norm": 0.012398588471114635,
      "learning_rate": 0.00014223753546817997,
      "loss": 0.7993,
      "step": 5700
    },
    {
      "epoch": 1.1552492906364005,
      "eval_accuracy": 0.9304029304029304,
      "eval_loss": 0.25227564573287964,
      "eval_runtime": 50.1813,
      "eval_samples_per_second": 10.881,
      "eval_steps_per_second": 1.375,
      "step": 5700
    },
    {
      "epoch": 1.157276043777868,
      "grad_norm": 0.021187039092183113,
      "learning_rate": 0.00014213619781110662,
      "loss": 0.0019,
      "step": 5710
    },
    {
      "epoch": 1.1593027969193352,
      "grad_norm": 0.012165443040430546,
      "learning_rate": 0.00014203486015403323,
      "loss": 0.0134,
      "step": 5720
    },
    {
      "epoch": 1.1613295500608025,
      "grad_norm": 0.009807312861084938,
      "learning_rate": 0.00014193352249695988,
      "loss": 0.2478,
      "step": 5730
    },
    {
      "epoch": 1.16335630320227,
      "grad_norm": 0.009756849147379398,
      "learning_rate": 0.00014183218483988653,
      "loss": 0.0014,
      "step": 5740
    },
    {
      "epoch": 1.1653830563437373,
      "grad_norm": 0.008829077705740929,
      "learning_rate": 0.00014173084718281312,
      "loss": 0.0012,
      "step": 5750
    },
    {
      "epoch": 1.1674098094852048,
      "grad_norm": 0.08207603543996811,
      "learning_rate": 0.00014162950952573977,
      "loss": 0.0021,
      "step": 5760
    },
    {
      "epoch": 1.169436562626672,
      "grad_norm": 10.163952827453613,
      "learning_rate": 0.0001415281718686664,
      "loss": 1.0266,
      "step": 5770
    },
    {
      "epoch": 1.1714633157681393,
      "grad_norm": 0.03201476112008095,
      "learning_rate": 0.00014142683421159303,
      "loss": 0.2121,
      "step": 5780
    },
    {
      "epoch": 1.1734900689096068,
      "grad_norm": 0.032649170607328415,
      "learning_rate": 0.00014132549655451968,
      "loss": 0.6479,
      "step": 5790
    },
    {
      "epoch": 1.175516822051074,
      "grad_norm": 4.468822002410889,
      "learning_rate": 0.0001412241588974463,
      "loss": 0.6292,
      "step": 5800
    },
    {
      "epoch": 1.175516822051074,
      "eval_accuracy": 0.945054945054945,
      "eval_loss": 0.1681230217218399,
      "eval_runtime": 47.0611,
      "eval_samples_per_second": 11.602,
      "eval_steps_per_second": 1.466,
      "step": 5800
    },
    {
      "epoch": 1.1775435751925416,
      "grad_norm": 0.01576576940715313,
      "learning_rate": 0.00014112282124037291,
      "loss": 0.0021,
      "step": 5810
    },
    {
      "epoch": 1.1795703283340089,
      "grad_norm": 0.011586645618081093,
      "learning_rate": 0.00014102148358329956,
      "loss": 0.5428,
      "step": 5820
    },
    {
      "epoch": 1.1815970814754764,
      "grad_norm": 0.1455240249633789,
      "learning_rate": 0.0001409201459262262,
      "loss": 0.0091,
      "step": 5830
    },
    {
      "epoch": 1.1836238346169436,
      "grad_norm": 0.014510289765894413,
      "learning_rate": 0.00014081880826915283,
      "loss": 0.0022,
      "step": 5840
    },
    {
      "epoch": 1.1856505877584111,
      "grad_norm": 0.020773572847247124,
      "learning_rate": 0.00014071747061207945,
      "loss": 1.1927,
      "step": 5850
    },
    {
      "epoch": 1.1876773408998784,
      "grad_norm": 10.849150657653809,
      "learning_rate": 0.0001406161329550061,
      "loss": 0.448,
      "step": 5860
    },
    {
      "epoch": 1.1897040940413457,
      "grad_norm": 22.743642807006836,
      "learning_rate": 0.0001405147952979327,
      "loss": 0.4768,
      "step": 5870
    },
    {
      "epoch": 1.1917308471828132,
      "grad_norm": 0.020906543359160423,
      "learning_rate": 0.00014041345764085936,
      "loss": 0.0059,
      "step": 5880
    },
    {
      "epoch": 1.1937576003242805,
      "grad_norm": 0.017099998891353607,
      "learning_rate": 0.00014031211998378598,
      "loss": 0.1922,
      "step": 5890
    },
    {
      "epoch": 1.195784353465748,
      "grad_norm": 0.013171990402042866,
      "learning_rate": 0.0001402107823267126,
      "loss": 0.0048,
      "step": 5900
    },
    {
      "epoch": 1.195784353465748,
      "eval_accuracy": 0.9377289377289377,
      "eval_loss": 0.2745637595653534,
      "eval_runtime": 50.3331,
      "eval_samples_per_second": 10.848,
      "eval_steps_per_second": 1.371,
      "step": 5900
    },
    {
      "epoch": 1.1978111066072152,
      "grad_norm": 1.8413903713226318,
      "learning_rate": 0.00014010944466963924,
      "loss": 0.0059,
      "step": 5910
    },
    {
      "epoch": 1.1998378597486825,
      "grad_norm": 13.45578670501709,
      "learning_rate": 0.0001400081070125659,
      "loss": 1.215,
      "step": 5920
    },
    {
      "epoch": 1.20186461289015,
      "grad_norm": 0.026761075481772423,
      "learning_rate": 0.0001399067693554925,
      "loss": 0.0028,
      "step": 5930
    },
    {
      "epoch": 1.2038913660316173,
      "grad_norm": 0.10492666810750961,
      "learning_rate": 0.00013980543169841913,
      "loss": 0.4688,
      "step": 5940
    },
    {
      "epoch": 1.2059181191730848,
      "grad_norm": 0.02013002149760723,
      "learning_rate": 0.00013970409404134577,
      "loss": 0.0057,
      "step": 5950
    },
    {
      "epoch": 1.207944872314552,
      "grad_norm": 0.01581076718866825,
      "learning_rate": 0.0001396027563842724,
      "loss": 0.0474,
      "step": 5960
    },
    {
      "epoch": 1.2099716254560196,
      "grad_norm": 0.019453393295407295,
      "learning_rate": 0.00013950141872719904,
      "loss": 0.2017,
      "step": 5970
    },
    {
      "epoch": 1.2119983785974868,
      "grad_norm": 0.013611565344035625,
      "learning_rate": 0.00013940008107012566,
      "loss": 0.1209,
      "step": 5980
    },
    {
      "epoch": 1.214025131738954,
      "grad_norm": 0.012966940179467201,
      "learning_rate": 0.0001392987434130523,
      "loss": 0.005,
      "step": 5990
    },
    {
      "epoch": 1.2160518848804216,
      "grad_norm": 8.436431884765625,
      "learning_rate": 0.00013919740575597892,
      "loss": 0.4908,
      "step": 6000
    },
    {
      "epoch": 1.2160518848804216,
      "eval_accuracy": 0.9358974358974359,
      "eval_loss": 0.31936949491500854,
      "eval_runtime": 50.4877,
      "eval_samples_per_second": 10.815,
      "eval_steps_per_second": 1.367,
      "step": 6000
    },
    {
      "epoch": 1.2180786380218889,
      "grad_norm": 0.012801182456314564,
      "learning_rate": 0.00013909606809890557,
      "loss": 0.5063,
      "step": 6010
    },
    {
      "epoch": 1.2201053911633564,
      "grad_norm": 0.012545096687972546,
      "learning_rate": 0.0001389947304418322,
      "loss": 0.0014,
      "step": 6020
    },
    {
      "epoch": 1.2221321443048236,
      "grad_norm": 0.1367848515510559,
      "learning_rate": 0.0001388933927847588,
      "loss": 0.0059,
      "step": 6030
    },
    {
      "epoch": 1.2241588974462911,
      "grad_norm": 0.011788008734583855,
      "learning_rate": 0.00013879205512768545,
      "loss": 0.0029,
      "step": 6040
    },
    {
      "epoch": 1.2261856505877584,
      "grad_norm": 0.13413147628307343,
      "learning_rate": 0.0001386907174706121,
      "loss": 0.0053,
      "step": 6050
    },
    {
      "epoch": 1.2282124037292257,
      "grad_norm": 0.0873718187212944,
      "learning_rate": 0.00013858937981353872,
      "loss": 0.0046,
      "step": 6060
    },
    {
      "epoch": 1.2302391568706932,
      "grad_norm": 0.0356074720621109,
      "learning_rate": 0.00013848804215646536,
      "loss": 0.6459,
      "step": 6070
    },
    {
      "epoch": 1.2322659100121605,
      "grad_norm": 0.07616440206766129,
      "learning_rate": 0.00013838670449939198,
      "loss": 0.0242,
      "step": 6080
    },
    {
      "epoch": 1.234292663153628,
      "grad_norm": 0.013825983740389347,
      "learning_rate": 0.0001382853668423186,
      "loss": 0.2924,
      "step": 6090
    },
    {
      "epoch": 1.2363194162950952,
      "grad_norm": 0.05042441561818123,
      "learning_rate": 0.00013818402918524525,
      "loss": 0.4156,
      "step": 6100
    },
    {
      "epoch": 1.2363194162950952,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.13203047215938568,
      "eval_runtime": 50.4313,
      "eval_samples_per_second": 10.827,
      "eval_steps_per_second": 1.368,
      "step": 6100
    },
    {
      "epoch": 1.2383461694365625,
      "grad_norm": 0.013852720148861408,
      "learning_rate": 0.0001380826915281719,
      "loss": 0.3545,
      "step": 6110
    },
    {
      "epoch": 1.24037292257803,
      "grad_norm": 0.012885493226349354,
      "learning_rate": 0.0001379813538710985,
      "loss": 0.0035,
      "step": 6120
    },
    {
      "epoch": 1.2423996757194973,
      "grad_norm": 0.029491251334547997,
      "learning_rate": 0.00013788001621402513,
      "loss": 0.0025,
      "step": 6130
    },
    {
      "epoch": 1.2444264288609648,
      "grad_norm": 0.011255360208451748,
      "learning_rate": 0.00013777867855695178,
      "loss": 0.0023,
      "step": 6140
    },
    {
      "epoch": 1.246453182002432,
      "grad_norm": 0.02658005803823471,
      "learning_rate": 0.0001376773408998784,
      "loss": 0.0017,
      "step": 6150
    },
    {
      "epoch": 1.2484799351438995,
      "grad_norm": 0.009282359853386879,
      "learning_rate": 0.00013757600324280504,
      "loss": 0.0012,
      "step": 6160
    },
    {
      "epoch": 1.2505066882853668,
      "grad_norm": 0.00913665909320116,
      "learning_rate": 0.00013747466558573166,
      "loss": 0.6614,
      "step": 6170
    },
    {
      "epoch": 1.2525334414268343,
      "grad_norm": 0.011567773297429085,
      "learning_rate": 0.00013737332792865828,
      "loss": 0.6064,
      "step": 6180
    },
    {
      "epoch": 1.2545601945683016,
      "grad_norm": 0.019871024414896965,
      "learning_rate": 0.00013727199027158493,
      "loss": 0.0037,
      "step": 6190
    },
    {
      "epoch": 1.2565869477097689,
      "grad_norm": 0.013100901618599892,
      "learning_rate": 0.00013717065261451157,
      "loss": 0.0056,
      "step": 6200
    },
    {
      "epoch": 1.2565869477097689,
      "eval_accuracy": 0.8992673992673993,
      "eval_loss": 0.3195291757583618,
      "eval_runtime": 49.1699,
      "eval_samples_per_second": 11.104,
      "eval_steps_per_second": 1.403,
      "step": 6200
    },
    {
      "epoch": 1.2586137008512364,
      "grad_norm": 9.371114730834961,
      "learning_rate": 0.0001370693149574382,
      "loss": 0.2707,
      "step": 6210
    },
    {
      "epoch": 1.2606404539927036,
      "grad_norm": 0.009086466394364834,
      "learning_rate": 0.0001369679773003648,
      "loss": 0.116,
      "step": 6220
    },
    {
      "epoch": 1.2626672071341711,
      "grad_norm": 0.3528311252593994,
      "learning_rate": 0.00013686663964329146,
      "loss": 0.0044,
      "step": 6230
    },
    {
      "epoch": 1.2646939602756384,
      "grad_norm": 0.009217889979481697,
      "learning_rate": 0.00013676530198621808,
      "loss": 0.0816,
      "step": 6240
    },
    {
      "epoch": 1.2667207134171057,
      "grad_norm": 0.008680080063641071,
      "learning_rate": 0.00013666396432914472,
      "loss": 0.4132,
      "step": 6250
    },
    {
      "epoch": 1.2687474665585732,
      "grad_norm": 0.008342086337506771,
      "learning_rate": 0.00013656262667207137,
      "loss": 0.0014,
      "step": 6260
    },
    {
      "epoch": 1.2707742197000405,
      "grad_norm": 0.007774224504828453,
      "learning_rate": 0.00013646128901499796,
      "loss": 0.0017,
      "step": 6270
    },
    {
      "epoch": 1.272800972841508,
      "grad_norm": 0.007495061028748751,
      "learning_rate": 0.0001363599513579246,
      "loss": 0.0022,
      "step": 6280
    },
    {
      "epoch": 1.2748277259829752,
      "grad_norm": 0.042415015399456024,
      "learning_rate": 0.00013625861370085125,
      "loss": 0.5376,
      "step": 6290
    },
    {
      "epoch": 1.2768544791244427,
      "grad_norm": 0.008079341612756252,
      "learning_rate": 0.00013615727604377787,
      "loss": 0.0013,
      "step": 6300
    },
    {
      "epoch": 1.2768544791244427,
      "eval_accuracy": 0.9615384615384616,
      "eval_loss": 0.15808889269828796,
      "eval_runtime": 50.3952,
      "eval_samples_per_second": 10.834,
      "eval_steps_per_second": 1.369,
      "step": 6300
    },
    {
      "epoch": 1.27888123226591,
      "grad_norm": 0.007808642461895943,
      "learning_rate": 0.0001360559383867045,
      "loss": 0.0015,
      "step": 6310
    },
    {
      "epoch": 1.2809079854073775,
      "grad_norm": 0.0075371447019279,
      "learning_rate": 0.00013595460072963114,
      "loss": 0.0026,
      "step": 6320
    },
    {
      "epoch": 1.2829347385488448,
      "grad_norm": 0.06489552557468414,
      "learning_rate": 0.00013585326307255776,
      "loss": 0.4667,
      "step": 6330
    },
    {
      "epoch": 1.284961491690312,
      "grad_norm": 0.007573794573545456,
      "learning_rate": 0.0001357519254154844,
      "loss": 0.4561,
      "step": 6340
    },
    {
      "epoch": 1.2869882448317795,
      "grad_norm": 0.008704758249223232,
      "learning_rate": 0.00013565058775841105,
      "loss": 0.7077,
      "step": 6350
    },
    {
      "epoch": 1.2890149979732468,
      "grad_norm": 0.12125850468873978,
      "learning_rate": 0.00013554925010133764,
      "loss": 0.3011,
      "step": 6360
    },
    {
      "epoch": 1.2910417511147143,
      "grad_norm": 0.0689581036567688,
      "learning_rate": 0.0001354479124442643,
      "loss": 0.0025,
      "step": 6370
    },
    {
      "epoch": 1.2930685042561816,
      "grad_norm": 2.5777084827423096,
      "learning_rate": 0.00013534657478719094,
      "loss": 0.0149,
      "step": 6380
    },
    {
      "epoch": 1.2950952573976489,
      "grad_norm": 0.009006716310977936,
      "learning_rate": 0.00013524523713011755,
      "loss": 0.0026,
      "step": 6390
    },
    {
      "epoch": 1.2971220105391164,
      "grad_norm": 0.1382710337638855,
      "learning_rate": 0.0001351438994730442,
      "loss": 0.0027,
      "step": 6400
    },
    {
      "epoch": 1.2971220105391164,
      "eval_accuracy": 0.9413919413919414,
      "eval_loss": 0.2660368084907532,
      "eval_runtime": 47.3575,
      "eval_samples_per_second": 11.529,
      "eval_steps_per_second": 1.457,
      "step": 6400
    },
    {
      "epoch": 1.2991487636805836,
      "grad_norm": 0.008069846779108047,
      "learning_rate": 0.00013504256181597082,
      "loss": 0.0017,
      "step": 6410
    },
    {
      "epoch": 1.3011755168220511,
      "grad_norm": 0.008503837510943413,
      "learning_rate": 0.00013494122415889744,
      "loss": 0.5095,
      "step": 6420
    },
    {
      "epoch": 1.3032022699635184,
      "grad_norm": 0.008521609008312225,
      "learning_rate": 0.00013483988650182408,
      "loss": 0.3073,
      "step": 6430
    },
    {
      "epoch": 1.3052290231049857,
      "grad_norm": 0.008452016860246658,
      "learning_rate": 0.00013473854884475073,
      "loss": 0.0017,
      "step": 6440
    },
    {
      "epoch": 1.3072557762464532,
      "grad_norm": 8.367830276489258,
      "learning_rate": 0.00013463721118767735,
      "loss": 0.4673,
      "step": 6450
    },
    {
      "epoch": 1.3092825293879207,
      "grad_norm": 0.08890511095523834,
      "learning_rate": 0.00013453587353060397,
      "loss": 0.0046,
      "step": 6460
    },
    {
      "epoch": 1.311309282529388,
      "grad_norm": 0.008890490978956223,
      "learning_rate": 0.00013443453587353062,
      "loss": 0.4202,
      "step": 6470
    },
    {
      "epoch": 1.3133360356708552,
      "grad_norm": 0.009985828772187233,
      "learning_rate": 0.00013433319821645723,
      "loss": 0.6943,
      "step": 6480
    },
    {
      "epoch": 1.3153627888123227,
      "grad_norm": 0.13318373262882233,
      "learning_rate": 0.00013423186055938388,
      "loss": 0.0028,
      "step": 6490
    },
    {
      "epoch": 1.31738954195379,
      "grad_norm": 0.011707228608429432,
      "learning_rate": 0.0001341305229023105,
      "loss": 0.1753,
      "step": 6500
    },
    {
      "epoch": 1.31738954195379,
      "eval_accuracy": 0.9560439560439561,
      "eval_loss": 0.18581902980804443,
      "eval_runtime": 62.4093,
      "eval_samples_per_second": 8.749,
      "eval_steps_per_second": 1.106,
      "step": 6500
    },
    {
      "epoch": 1.3194162950952575,
      "grad_norm": 0.022704975679516792,
      "learning_rate": 0.00013402918524523712,
      "loss": 0.0043,
      "step": 6510
    },
    {
      "epoch": 1.3214430482367248,
      "grad_norm": 0.09878182411193848,
      "learning_rate": 0.00013392784758816377,
      "loss": 0.005,
      "step": 6520
    },
    {
      "epoch": 1.323469801378192,
      "grad_norm": 0.010758204385638237,
      "learning_rate": 0.0001338265099310904,
      "loss": 0.005,
      "step": 6530
    },
    {
      "epoch": 1.3254965545196595,
      "grad_norm": 0.08000922948122025,
      "learning_rate": 0.00013372517227401703,
      "loss": 0.0027,
      "step": 6540
    },
    {
      "epoch": 1.3275233076611268,
      "grad_norm": 0.009771626442670822,
      "learning_rate": 0.00013362383461694365,
      "loss": 0.4507,
      "step": 6550
    },
    {
      "epoch": 1.3295500608025943,
      "grad_norm": 0.09330135583877563,
      "learning_rate": 0.0001335224969598703,
      "loss": 0.0652,
      "step": 6560
    },
    {
      "epoch": 1.3315768139440616,
      "grad_norm": 0.06133830174803734,
      "learning_rate": 0.00013342115930279691,
      "loss": 0.0024,
      "step": 6570
    },
    {
      "epoch": 1.3336035670855289,
      "grad_norm": 0.00886420626193285,
      "learning_rate": 0.00013331982164572356,
      "loss": 0.0819,
      "step": 6580
    },
    {
      "epoch": 1.3356303202269963,
      "grad_norm": 0.008621442131698132,
      "learning_rate": 0.00013321848398865018,
      "loss": 0.0025,
      "step": 6590
    },
    {
      "epoch": 1.3376570733684638,
      "grad_norm": 0.06374801695346832,
      "learning_rate": 0.00013311714633157683,
      "loss": 0.0013,
      "step": 6600
    },
    {
      "epoch": 1.3376570733684638,
      "eval_accuracy": 0.9615384615384616,
      "eval_loss": 0.2017831653356552,
      "eval_runtime": 50.4548,
      "eval_samples_per_second": 10.822,
      "eval_steps_per_second": 1.368,
      "step": 6600
    },
    {
      "epoch": 1.3396838265099311,
      "grad_norm": 0.0080784372985363,
      "learning_rate": 0.00013301580867450345,
      "loss": 0.0009,
      "step": 6610
    },
    {
      "epoch": 1.3417105796513984,
      "grad_norm": 0.007808597292751074,
      "learning_rate": 0.0001329144710174301,
      "loss": 0.0021,
      "step": 6620
    },
    {
      "epoch": 1.343737332792866,
      "grad_norm": 0.053149402141571045,
      "learning_rate": 0.0001328131333603567,
      "loss": 0.0034,
      "step": 6630
    },
    {
      "epoch": 1.3457640859343332,
      "grad_norm": 0.04929700493812561,
      "learning_rate": 0.00013271179570328333,
      "loss": 0.0013,
      "step": 6640
    },
    {
      "epoch": 1.3477908390758007,
      "grad_norm": 0.00766234565526247,
      "learning_rate": 0.00013261045804620998,
      "loss": 0.0013,
      "step": 6650
    },
    {
      "epoch": 1.349817592217268,
      "grad_norm": 0.007020989898592234,
      "learning_rate": 0.00013250912038913662,
      "loss": 0.5308,
      "step": 6660
    },
    {
      "epoch": 1.3518443453587352,
      "grad_norm": 0.007394090294837952,
      "learning_rate": 0.00013240778273206324,
      "loss": 0.0021,
      "step": 6670
    },
    {
      "epoch": 1.3538710985002027,
      "grad_norm": 0.065384142100811,
      "learning_rate": 0.0001323064450749899,
      "loss": 0.6402,
      "step": 6680
    },
    {
      "epoch": 1.35589785164167,
      "grad_norm": 0.04742232337594032,
      "learning_rate": 0.0001322051074179165,
      "loss": 0.0041,
      "step": 6690
    },
    {
      "epoch": 1.3579246047831375,
      "grad_norm": 0.06092911213636398,
      "learning_rate": 0.00013210376976084313,
      "loss": 0.0033,
      "step": 6700
    },
    {
      "epoch": 1.3579246047831375,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.14746183156967163,
      "eval_runtime": 46.8652,
      "eval_samples_per_second": 11.65,
      "eval_steps_per_second": 1.472,
      "step": 6700
    },
    {
      "epoch": 1.3599513579246048,
      "grad_norm": 0.011696516536176205,
      "learning_rate": 0.00013200243210376977,
      "loss": 0.2504,
      "step": 6710
    },
    {
      "epoch": 1.361978111066072,
      "grad_norm": 0.00928343553096056,
      "learning_rate": 0.00013190109444669642,
      "loss": 0.0039,
      "step": 6720
    },
    {
      "epoch": 1.3640048642075395,
      "grad_norm": 0.007892207242548466,
      "learning_rate": 0.00013179975678962304,
      "loss": 0.1108,
      "step": 6730
    },
    {
      "epoch": 1.366031617349007,
      "grad_norm": 0.0072936443611979485,
      "learning_rate": 0.00013169841913254966,
      "loss": 0.0009,
      "step": 6740
    },
    {
      "epoch": 1.3680583704904743,
      "grad_norm": 0.00823760312050581,
      "learning_rate": 0.0001315970814754763,
      "loss": 0.4251,
      "step": 6750
    },
    {
      "epoch": 1.3700851236319416,
      "grad_norm": 14.980917930603027,
      "learning_rate": 0.00013149574381840292,
      "loss": 0.6206,
      "step": 6760
    },
    {
      "epoch": 1.372111876773409,
      "grad_norm": 0.01617688313126564,
      "learning_rate": 0.00013139440616132957,
      "loss": 0.0022,
      "step": 6770
    },
    {
      "epoch": 1.3741386299148763,
      "grad_norm": 0.04971484839916229,
      "learning_rate": 0.0001312930685042562,
      "loss": 0.0049,
      "step": 6780
    },
    {
      "epoch": 1.3761653830563438,
      "grad_norm": 0.01129236165434122,
      "learning_rate": 0.0001311917308471828,
      "loss": 0.0038,
      "step": 6790
    },
    {
      "epoch": 1.3781921361978111,
      "grad_norm": 0.009702612645924091,
      "learning_rate": 0.00013109039319010945,
      "loss": 0.0037,
      "step": 6800
    },
    {
      "epoch": 1.3781921361978111,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.141746386885643,
      "eval_runtime": 50.2101,
      "eval_samples_per_second": 10.874,
      "eval_steps_per_second": 1.374,
      "step": 6800
    },
    {
      "epoch": 1.3802188893392784,
      "grad_norm": 0.008086517453193665,
      "learning_rate": 0.0001309890555330361,
      "loss": 0.5346,
      "step": 6810
    },
    {
      "epoch": 1.3822456424807459,
      "grad_norm": 0.06050160527229309,
      "learning_rate": 0.00013088771787596272,
      "loss": 0.0015,
      "step": 6820
    },
    {
      "epoch": 1.3842723956222132,
      "grad_norm": 0.0070461998693645,
      "learning_rate": 0.00013078638021888934,
      "loss": 0.0022,
      "step": 6830
    },
    {
      "epoch": 1.3862991487636807,
      "grad_norm": 0.05749473720788956,
      "learning_rate": 0.00013068504256181598,
      "loss": 0.002,
      "step": 6840
    },
    {
      "epoch": 1.388325901905148,
      "grad_norm": 10.301835060119629,
      "learning_rate": 0.0001305837049047426,
      "loss": 1.1992,
      "step": 6850
    },
    {
      "epoch": 1.3903526550466152,
      "grad_norm": 0.009321840479969978,
      "learning_rate": 0.00013048236724766925,
      "loss": 0.4717,
      "step": 6860
    },
    {
      "epoch": 1.3923794081880827,
      "grad_norm": 0.0103806983679533,
      "learning_rate": 0.0001303810295905959,
      "loss": 0.0021,
      "step": 6870
    },
    {
      "epoch": 1.3944061613295502,
      "grad_norm": 0.012516554445028305,
      "learning_rate": 0.00013027969193352249,
      "loss": 0.3355,
      "step": 6880
    },
    {
      "epoch": 1.3964329144710175,
      "grad_norm": 0.011678949929773808,
      "learning_rate": 0.00013017835427644913,
      "loss": 0.0031,
      "step": 6890
    },
    {
      "epoch": 1.3984596676124847,
      "grad_norm": 0.04873785376548767,
      "learning_rate": 0.00013007701661937578,
      "loss": 1.2775,
      "step": 6900
    },
    {
      "epoch": 1.3984596676124847,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.11005932837724686,
      "eval_runtime": 52.1773,
      "eval_samples_per_second": 10.464,
      "eval_steps_per_second": 1.322,
      "step": 6900
    },
    {
      "epoch": 1.4004864207539522,
      "grad_norm": 0.34723198413848877,
      "learning_rate": 0.0001299756789623024,
      "loss": 0.149,
      "step": 6910
    },
    {
      "epoch": 1.4025131738954195,
      "grad_norm": 0.059217892587184906,
      "learning_rate": 0.00012987434130522902,
      "loss": 0.0068,
      "step": 6920
    },
    {
      "epoch": 1.404539927036887,
      "grad_norm": 0.025736402720212936,
      "learning_rate": 0.00012977300364815566,
      "loss": 0.8352,
      "step": 6930
    },
    {
      "epoch": 1.4065666801783543,
      "grad_norm": 0.01956155151128769,
      "learning_rate": 0.00012967166599108228,
      "loss": 0.4539,
      "step": 6940
    },
    {
      "epoch": 1.4085934333198216,
      "grad_norm": 0.020265795290470123,
      "learning_rate": 0.00012957032833400893,
      "loss": 0.0025,
      "step": 6950
    },
    {
      "epoch": 1.410620186461289,
      "grad_norm": 0.015172087587416172,
      "learning_rate": 0.00012946899067693557,
      "loss": 0.0048,
      "step": 6960
    },
    {
      "epoch": 1.4126469396027563,
      "grad_norm": 0.044815607368946075,
      "learning_rate": 0.00012936765301986217,
      "loss": 0.9364,
      "step": 6970
    },
    {
      "epoch": 1.4146736927442238,
      "grad_norm": 0.043388959020376205,
      "learning_rate": 0.0001292663153627888,
      "loss": 0.0073,
      "step": 6980
    },
    {
      "epoch": 1.416700445885691,
      "grad_norm": 0.024978777393698692,
      "learning_rate": 0.00012916497770571546,
      "loss": 0.0102,
      "step": 6990
    },
    {
      "epoch": 1.4187271990271584,
      "grad_norm": 0.08704657852649689,
      "learning_rate": 0.00012906364004864208,
      "loss": 0.0051,
      "step": 7000
    },
    {
      "epoch": 1.4187271990271584,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.12920933961868286,
      "eval_runtime": 48.3802,
      "eval_samples_per_second": 11.286,
      "eval_steps_per_second": 1.426,
      "step": 7000
    },
    {
      "epoch": 1.4207539521686259,
      "grad_norm": 0.15106026828289032,
      "learning_rate": 0.00012896230239156872,
      "loss": 0.2045,
      "step": 7010
    },
    {
      "epoch": 1.4227807053100932,
      "grad_norm": 0.07075616717338562,
      "learning_rate": 0.00012886096473449534,
      "loss": 0.0048,
      "step": 7020
    },
    {
      "epoch": 1.4248074584515606,
      "grad_norm": 0.008650687523186207,
      "learning_rate": 0.00012875962707742196,
      "loss": 0.4909,
      "step": 7030
    },
    {
      "epoch": 1.426834211593028,
      "grad_norm": 0.011124524287879467,
      "learning_rate": 0.0001286582894203486,
      "loss": 0.0019,
      "step": 7040
    },
    {
      "epoch": 1.4288609647344954,
      "grad_norm": 0.008949482813477516,
      "learning_rate": 0.00012855695176327525,
      "loss": 0.0034,
      "step": 7050
    },
    {
      "epoch": 1.4308877178759627,
      "grad_norm": 0.0093899667263031,
      "learning_rate": 0.00012845561410620185,
      "loss": 0.0011,
      "step": 7060
    },
    {
      "epoch": 1.4329144710174302,
      "grad_norm": 0.0081386873498559,
      "learning_rate": 0.0001283542764491285,
      "loss": 0.0068,
      "step": 7070
    },
    {
      "epoch": 1.4349412241588975,
      "grad_norm": 0.06856881827116013,
      "learning_rate": 0.00012825293879205514,
      "loss": 0.0031,
      "step": 7080
    },
    {
      "epoch": 1.4369679773003647,
      "grad_norm": 0.007543250452727079,
      "learning_rate": 0.00012815160113498176,
      "loss": 0.4818,
      "step": 7090
    },
    {
      "epoch": 1.4389947304418322,
      "grad_norm": 0.008494116365909576,
      "learning_rate": 0.0001280502634779084,
      "loss": 0.4954,
      "step": 7100
    },
    {
      "epoch": 1.4389947304418322,
      "eval_accuracy": 0.9468864468864469,
      "eval_loss": 0.24729128181934357,
      "eval_runtime": 50.3955,
      "eval_samples_per_second": 10.834,
      "eval_steps_per_second": 1.369,
      "step": 7100
    },
    {
      "epoch": 1.4410214835832995,
      "grad_norm": 0.008369691669940948,
      "learning_rate": 0.00012794892582083502,
      "loss": 0.4182,
      "step": 7110
    },
    {
      "epoch": 1.443048236724767,
      "grad_norm": 0.00896457303315401,
      "learning_rate": 0.00012784758816376164,
      "loss": 0.0027,
      "step": 7120
    },
    {
      "epoch": 1.4450749898662343,
      "grad_norm": 0.008388702757656574,
      "learning_rate": 0.0001277462505066883,
      "loss": 0.153,
      "step": 7130
    },
    {
      "epoch": 1.4471017430077016,
      "grad_norm": 0.011490442790091038,
      "learning_rate": 0.00012764491284961494,
      "loss": 0.6053,
      "step": 7140
    },
    {
      "epoch": 1.449128496149169,
      "grad_norm": 0.010425391606986523,
      "learning_rate": 0.00012754357519254155,
      "loss": 0.0012,
      "step": 7150
    },
    {
      "epoch": 1.4511552492906363,
      "grad_norm": 0.01710696704685688,
      "learning_rate": 0.00012744223753546817,
      "loss": 0.0036,
      "step": 7160
    },
    {
      "epoch": 1.4531820024321038,
      "grad_norm": 10.217206001281738,
      "learning_rate": 0.00012734089987839482,
      "loss": 0.2024,
      "step": 7170
    },
    {
      "epoch": 1.455208755573571,
      "grad_norm": 0.10198117792606354,
      "learning_rate": 0.00012723956222132144,
      "loss": 0.0135,
      "step": 7180
    },
    {
      "epoch": 1.4572355087150384,
      "grad_norm": 0.007917650043964386,
      "learning_rate": 0.00012713822456424808,
      "loss": 0.0073,
      "step": 7190
    },
    {
      "epoch": 1.4592622618565059,
      "grad_norm": 0.00752667523920536,
      "learning_rate": 0.0001270368869071747,
      "loss": 0.1533,
      "step": 7200
    },
    {
      "epoch": 1.4592622618565059,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.11806527525186539,
      "eval_runtime": 50.3295,
      "eval_samples_per_second": 10.849,
      "eval_steps_per_second": 1.371,
      "step": 7200
    },
    {
      "epoch": 1.4612890149979734,
      "grad_norm": 0.0070377918891608715,
      "learning_rate": 0.00012693554925010135,
      "loss": 0.001,
      "step": 7210
    },
    {
      "epoch": 1.4633157681394406,
      "grad_norm": 0.007255935110151768,
      "learning_rate": 0.00012683421159302797,
      "loss": 0.4828,
      "step": 7220
    },
    {
      "epoch": 1.465342521280908,
      "grad_norm": 0.00714620528742671,
      "learning_rate": 0.00012673287393595462,
      "loss": 0.003,
      "step": 7230
    },
    {
      "epoch": 1.4673692744223754,
      "grad_norm": 0.006977812387049198,
      "learning_rate": 0.00012663153627888123,
      "loss": 0.0008,
      "step": 7240
    },
    {
      "epoch": 1.4693960275638427,
      "grad_norm": 0.00690856808796525,
      "learning_rate": 0.00012653019862180785,
      "loss": 0.4458,
      "step": 7250
    },
    {
      "epoch": 1.4714227807053102,
      "grad_norm": 0.007223729509860277,
      "learning_rate": 0.0001264288609647345,
      "loss": 0.14,
      "step": 7260
    },
    {
      "epoch": 1.4734495338467775,
      "grad_norm": 0.00870873499661684,
      "learning_rate": 0.00012632752330766115,
      "loss": 0.7125,
      "step": 7270
    },
    {
      "epoch": 1.4754762869882447,
      "grad_norm": 0.00891156680881977,
      "learning_rate": 0.00012622618565058777,
      "loss": 0.0041,
      "step": 7280
    },
    {
      "epoch": 1.4775030401297122,
      "grad_norm": 0.008706452324986458,
      "learning_rate": 0.0001261248479935144,
      "loss": 0.0053,
      "step": 7290
    },
    {
      "epoch": 1.4795297932711795,
      "grad_norm": 0.00838247500360012,
      "learning_rate": 0.00012602351033644103,
      "loss": 0.0022,
      "step": 7300
    },
    {
      "epoch": 1.4795297932711795,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.15122626721858978,
      "eval_runtime": 50.3826,
      "eval_samples_per_second": 10.837,
      "eval_steps_per_second": 1.37,
      "step": 7300
    },
    {
      "epoch": 1.481556546412647,
      "grad_norm": 0.009303655475378036,
      "learning_rate": 0.00012592217267936765,
      "loss": 0.004,
      "step": 7310
    },
    {
      "epoch": 1.4835832995541143,
      "grad_norm": 0.07870160788297653,
      "learning_rate": 0.0001258208350222943,
      "loss": 0.7105,
      "step": 7320
    },
    {
      "epoch": 1.4856100526955816,
      "grad_norm": 0.01071919035166502,
      "learning_rate": 0.00012571949736522094,
      "loss": 0.0052,
      "step": 7330
    },
    {
      "epoch": 1.487636805837049,
      "grad_norm": 0.00973090622574091,
      "learning_rate": 0.00012561815970814756,
      "loss": 0.4796,
      "step": 7340
    },
    {
      "epoch": 1.4896635589785165,
      "grad_norm": 0.01041113119572401,
      "learning_rate": 0.00012551682205107418,
      "loss": 0.4586,
      "step": 7350
    },
    {
      "epoch": 1.4916903121199838,
      "grad_norm": 0.010993756353855133,
      "learning_rate": 0.00012541548439400083,
      "loss": 0.4192,
      "step": 7360
    },
    {
      "epoch": 1.493717065261451,
      "grad_norm": 0.18665625154972076,
      "learning_rate": 0.00012531414673692745,
      "loss": 0.0035,
      "step": 7370
    },
    {
      "epoch": 1.4957438184029186,
      "grad_norm": 0.013825108297169209,
      "learning_rate": 0.0001252128090798541,
      "loss": 0.5293,
      "step": 7380
    },
    {
      "epoch": 1.4977705715443859,
      "grad_norm": 0.1872238665819168,
      "learning_rate": 0.0001251114714227807,
      "loss": 0.0109,
      "step": 7390
    },
    {
      "epoch": 1.4997973246858534,
      "grad_norm": 0.14896440505981445,
      "learning_rate": 0.00012501013376570733,
      "loss": 0.005,
      "step": 7400
    },
    {
      "epoch": 1.4997973246858534,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.13286960124969482,
      "eval_runtime": 47.115,
      "eval_samples_per_second": 11.589,
      "eval_steps_per_second": 1.465,
      "step": 7400
    },
    {
      "epoch": 1.5018240778273206,
      "grad_norm": 14.306511878967285,
      "learning_rate": 0.00012490879610863398,
      "loss": 0.0343,
      "step": 7410
    },
    {
      "epoch": 1.503850830968788,
      "grad_norm": 0.09887898713350296,
      "learning_rate": 0.00012480745845156062,
      "loss": 0.0047,
      "step": 7420
    },
    {
      "epoch": 1.5058775841102554,
      "grad_norm": 0.2640922963619232,
      "learning_rate": 0.00012470612079448724,
      "loss": 0.0021,
      "step": 7430
    },
    {
      "epoch": 1.507904337251723,
      "grad_norm": 0.008840842172503471,
      "learning_rate": 0.00012460478313741386,
      "loss": 0.0047,
      "step": 7440
    },
    {
      "epoch": 1.5099310903931902,
      "grad_norm": 0.01149743888527155,
      "learning_rate": 0.0001245034454803405,
      "loss": 0.9103,
      "step": 7450
    },
    {
      "epoch": 1.5119578435346575,
      "grad_norm": 0.12512584030628204,
      "learning_rate": 0.00012440210782326713,
      "loss": 0.0066,
      "step": 7460
    },
    {
      "epoch": 1.5139845966761247,
      "grad_norm": 0.009386054240167141,
      "learning_rate": 0.00012430077016619377,
      "loss": 0.0011,
      "step": 7470
    },
    {
      "epoch": 1.5160113498175922,
      "grad_norm": 0.0302512738853693,
      "learning_rate": 0.00012419943250912042,
      "loss": 0.0039,
      "step": 7480
    },
    {
      "epoch": 1.5180381029590597,
      "grad_norm": 0.008296155370771885,
      "learning_rate": 0.000124098094852047,
      "loss": 0.0021,
      "step": 7490
    },
    {
      "epoch": 1.520064856100527,
      "grad_norm": 0.00832971278578043,
      "learning_rate": 0.00012399675719497366,
      "loss": 0.4396,
      "step": 7500
    },
    {
      "epoch": 1.520064856100527,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.12192816287279129,
      "eval_runtime": 49.4588,
      "eval_samples_per_second": 11.039,
      "eval_steps_per_second": 1.395,
      "step": 7500
    },
    {
      "epoch": 1.5220916092419943,
      "grad_norm": 0.008520863950252533,
      "learning_rate": 0.0001238954195379003,
      "loss": 0.0027,
      "step": 7510
    },
    {
      "epoch": 1.5241183623834615,
      "grad_norm": 0.008221836760640144,
      "learning_rate": 0.00012379408188082692,
      "loss": 0.0025,
      "step": 7520
    },
    {
      "epoch": 1.526145115524929,
      "grad_norm": 0.00811388622969389,
      "learning_rate": 0.00012369274422375354,
      "loss": 0.7087,
      "step": 7530
    },
    {
      "epoch": 1.5281718686663965,
      "grad_norm": 0.00921147596091032,
      "learning_rate": 0.0001235914065666802,
      "loss": 0.001,
      "step": 7540
    },
    {
      "epoch": 1.5301986218078638,
      "grad_norm": 0.010357416234910488,
      "learning_rate": 0.0001234900689096068,
      "loss": 0.3959,
      "step": 7550
    },
    {
      "epoch": 1.532225374949331,
      "grad_norm": 0.011035448871552944,
      "learning_rate": 0.00012338873125253345,
      "loss": 0.0078,
      "step": 7560
    },
    {
      "epoch": 1.5342521280907986,
      "grad_norm": 0.04455064609646797,
      "learning_rate": 0.0001232873935954601,
      "loss": 0.0059,
      "step": 7570
    },
    {
      "epoch": 1.5362788812322659,
      "grad_norm": 0.009438133798539639,
      "learning_rate": 0.0001231860559383867,
      "loss": 0.0043,
      "step": 7580
    },
    {
      "epoch": 1.5383056343737334,
      "grad_norm": 0.008560802787542343,
      "learning_rate": 0.00012308471828131334,
      "loss": 0.0024,
      "step": 7590
    },
    {
      "epoch": 1.5403323875152006,
      "grad_norm": 0.008284498006105423,
      "learning_rate": 0.00012298338062423998,
      "loss": 0.0044,
      "step": 7600
    },
    {
      "epoch": 1.5403323875152006,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.16649296879768372,
      "eval_runtime": 50.3329,
      "eval_samples_per_second": 10.848,
      "eval_steps_per_second": 1.371,
      "step": 7600
    },
    {
      "epoch": 1.542359140656668,
      "grad_norm": 0.00793454609811306,
      "learning_rate": 0.0001228820429671666,
      "loss": 0.0009,
      "step": 7610
    },
    {
      "epoch": 1.5443858937981354,
      "grad_norm": 0.09009437263011932,
      "learning_rate": 0.00012278070531009325,
      "loss": 0.0029,
      "step": 7620
    },
    {
      "epoch": 1.546412646939603,
      "grad_norm": 0.008107963018119335,
      "learning_rate": 0.00012267936765301987,
      "loss": 0.0008,
      "step": 7630
    },
    {
      "epoch": 1.5484394000810702,
      "grad_norm": 8.214654922485352,
      "learning_rate": 0.00012257802999594649,
      "loss": 0.8614,
      "step": 7640
    },
    {
      "epoch": 1.5504661532225374,
      "grad_norm": 0.008868742734193802,
      "learning_rate": 0.00012247669233887313,
      "loss": 0.4323,
      "step": 7650
    },
    {
      "epoch": 1.5524929063640047,
      "grad_norm": 0.009688987396657467,
      "learning_rate": 0.00012237535468179978,
      "loss": 0.362,
      "step": 7660
    },
    {
      "epoch": 1.5545196595054722,
      "grad_norm": 0.010521828196942806,
      "learning_rate": 0.00012227401702472637,
      "loss": 0.0141,
      "step": 7670
    },
    {
      "epoch": 1.5565464126469397,
      "grad_norm": 0.22994591295719147,
      "learning_rate": 0.00012217267936765302,
      "loss": 0.0142,
      "step": 7680
    },
    {
      "epoch": 1.558573165788407,
      "grad_norm": 0.008071416057646275,
      "learning_rate": 0.00012207134171057966,
      "loss": 0.0036,
      "step": 7690
    },
    {
      "epoch": 1.5605999189298743,
      "grad_norm": 9.246078491210938,
      "learning_rate": 0.0001219700040535063,
      "loss": 0.7054,
      "step": 7700
    },
    {
      "epoch": 1.5605999189298743,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.16522414982318878,
      "eval_runtime": 46.6457,
      "eval_samples_per_second": 11.705,
      "eval_steps_per_second": 1.479,
      "step": 7700
    },
    {
      "epoch": 1.5626266720713418,
      "grad_norm": 39.2440185546875,
      "learning_rate": 0.00012186866639643293,
      "loss": 0.8085,
      "step": 7710
    },
    {
      "epoch": 1.564653425212809,
      "grad_norm": 0.1413354128599167,
      "learning_rate": 0.00012176732873935955,
      "loss": 0.0097,
      "step": 7720
    },
    {
      "epoch": 1.5666801783542765,
      "grad_norm": 0.018545445054769516,
      "learning_rate": 0.00012166599108228618,
      "loss": 0.0032,
      "step": 7730
    },
    {
      "epoch": 1.5687069314957438,
      "grad_norm": 0.02521946094930172,
      "learning_rate": 0.00012156465342521281,
      "loss": 0.0039,
      "step": 7740
    },
    {
      "epoch": 1.570733684637211,
      "grad_norm": 0.019712034612894058,
      "learning_rate": 0.00012146331576813945,
      "loss": 0.4125,
      "step": 7750
    },
    {
      "epoch": 1.5727604377786786,
      "grad_norm": 0.012678617611527443,
      "learning_rate": 0.00012136197811106609,
      "loss": 0.437,
      "step": 7760
    },
    {
      "epoch": 1.574787190920146,
      "grad_norm": 0.014548375271260738,
      "learning_rate": 0.0001212606404539927,
      "loss": 0.6042,
      "step": 7770
    },
    {
      "epoch": 1.5768139440616133,
      "grad_norm": 0.014556125737726688,
      "learning_rate": 0.00012115930279691934,
      "loss": 0.0912,
      "step": 7780
    },
    {
      "epoch": 1.5788406972030806,
      "grad_norm": 0.014982403255999088,
      "learning_rate": 0.00012105796513984598,
      "loss": 0.4238,
      "step": 7790
    },
    {
      "epoch": 1.580867450344548,
      "grad_norm": 0.014122478663921356,
      "learning_rate": 0.00012095662748277261,
      "loss": 0.4057,
      "step": 7800
    },
    {
      "epoch": 1.580867450344548,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.16829834878444672,
      "eval_runtime": 50.1526,
      "eval_samples_per_second": 10.887,
      "eval_steps_per_second": 1.376,
      "step": 7800
    },
    {
      "epoch": 1.5828942034860154,
      "grad_norm": 0.042958226054906845,
      "learning_rate": 0.00012085528982569923,
      "loss": 0.3522,
      "step": 7810
    },
    {
      "epoch": 1.584920956627483,
      "grad_norm": 0.014224754646420479,
      "learning_rate": 0.00012075395216862586,
      "loss": 0.6036,
      "step": 7820
    },
    {
      "epoch": 1.5869477097689502,
      "grad_norm": 0.012702335603535175,
      "learning_rate": 0.00012065261451155249,
      "loss": 0.0226,
      "step": 7830
    },
    {
      "epoch": 1.5889744629104174,
      "grad_norm": 0.012445378117263317,
      "learning_rate": 0.00012055127685447914,
      "loss": 0.0063,
      "step": 7840
    },
    {
      "epoch": 1.5910012160518847,
      "grad_norm": 0.011407758109271526,
      "learning_rate": 0.00012044993919740577,
      "loss": 0.0057,
      "step": 7850
    },
    {
      "epoch": 1.5930279691933522,
      "grad_norm": 0.3679606318473816,
      "learning_rate": 0.00012034860154033238,
      "loss": 1.0166,
      "step": 7860
    },
    {
      "epoch": 1.5950547223348197,
      "grad_norm": 0.013010966591536999,
      "learning_rate": 0.00012024726388325902,
      "loss": 0.0016,
      "step": 7870
    },
    {
      "epoch": 1.597081475476287,
      "grad_norm": 0.013456660322844982,
      "learning_rate": 0.00012014592622618566,
      "loss": 0.0016,
      "step": 7880
    },
    {
      "epoch": 1.5991082286177543,
      "grad_norm": 0.012532263062894344,
      "learning_rate": 0.00012004458856911229,
      "loss": 0.009,
      "step": 7890
    },
    {
      "epoch": 1.6011349817592218,
      "grad_norm": 0.01159537211060524,
      "learning_rate": 0.00011994325091203894,
      "loss": 0.011,
      "step": 7900
    },
    {
      "epoch": 1.6011349817592218,
      "eval_accuracy": 0.9285714285714286,
      "eval_loss": 0.3927004933357239,
      "eval_runtime": 50.1818,
      "eval_samples_per_second": 10.88,
      "eval_steps_per_second": 1.375,
      "step": 7900
    },
    {
      "epoch": 1.6031617349006893,
      "grad_norm": 0.010766562074422836,
      "learning_rate": 0.00011984191325496554,
      "loss": 0.0061,
      "step": 7910
    },
    {
      "epoch": 1.6051884880421565,
      "grad_norm": 0.015047757886350155,
      "learning_rate": 0.00011974057559789217,
      "loss": 1.1934,
      "step": 7920
    },
    {
      "epoch": 1.6072152411836238,
      "grad_norm": 0.025979992002248764,
      "learning_rate": 0.00011963923794081882,
      "loss": 0.0091,
      "step": 7930
    },
    {
      "epoch": 1.609241994325091,
      "grad_norm": 0.025664396584033966,
      "learning_rate": 0.00011953790028374545,
      "loss": 0.0938,
      "step": 7940
    },
    {
      "epoch": 1.6112687474665586,
      "grad_norm": 0.026306096464395523,
      "learning_rate": 0.00011943656262667207,
      "loss": 0.8206,
      "step": 7950
    },
    {
      "epoch": 1.613295500608026,
      "grad_norm": 0.01710491254925728,
      "learning_rate": 0.0001193352249695987,
      "loss": 0.4074,
      "step": 7960
    },
    {
      "epoch": 1.6153222537494933,
      "grad_norm": 0.20820653438568115,
      "learning_rate": 0.00011923388731252534,
      "loss": 0.007,
      "step": 7970
    },
    {
      "epoch": 1.6173490068909606,
      "grad_norm": 0.22726105153560638,
      "learning_rate": 0.00011913254965545197,
      "loss": 0.8665,
      "step": 7980
    },
    {
      "epoch": 1.619375760032428,
      "grad_norm": 0.017417067661881447,
      "learning_rate": 0.00011903121199837862,
      "loss": 0.0103,
      "step": 7990
    },
    {
      "epoch": 1.6214025131738954,
      "grad_norm": 0.022518666461110115,
      "learning_rate": 0.00011892987434130522,
      "loss": 0.7,
      "step": 8000
    },
    {
      "epoch": 1.6214025131738954,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.09989867359399796,
      "eval_runtime": 47.4891,
      "eval_samples_per_second": 11.497,
      "eval_steps_per_second": 1.453,
      "step": 8000
    },
    {
      "epoch": 1.6234292663153629,
      "grad_norm": 0.037107568234205246,
      "learning_rate": 0.00011882853668423187,
      "loss": 0.0115,
      "step": 8010
    },
    {
      "epoch": 1.6254560194568302,
      "grad_norm": 0.2722521722316742,
      "learning_rate": 0.0001187271990271585,
      "loss": 0.0076,
      "step": 8020
    },
    {
      "epoch": 1.6274827725982974,
      "grad_norm": 0.020260510966181755,
      "learning_rate": 0.00011862586137008513,
      "loss": 0.003,
      "step": 8030
    },
    {
      "epoch": 1.629509525739765,
      "grad_norm": 11.66630744934082,
      "learning_rate": 0.00011852452371301177,
      "loss": 1.0787,
      "step": 8040
    },
    {
      "epoch": 1.6315362788812324,
      "grad_norm": 0.03827909007668495,
      "learning_rate": 0.00011842318605593838,
      "loss": 0.0067,
      "step": 8050
    },
    {
      "epoch": 1.6335630320226997,
      "grad_norm": 0.05085739120841026,
      "learning_rate": 0.00011832184839886502,
      "loss": 0.0078,
      "step": 8060
    },
    {
      "epoch": 1.635589785164167,
      "grad_norm": 0.03725253790616989,
      "learning_rate": 0.00011822051074179166,
      "loss": 0.2087,
      "step": 8070
    },
    {
      "epoch": 1.6376165383056343,
      "grad_norm": 0.016044335439801216,
      "learning_rate": 0.0001181191730847183,
      "loss": 0.0063,
      "step": 8080
    },
    {
      "epoch": 1.6396432914471017,
      "grad_norm": 0.013224687427282333,
      "learning_rate": 0.00011801783542764493,
      "loss": 0.0591,
      "step": 8090
    },
    {
      "epoch": 1.6416700445885692,
      "grad_norm": 0.012949744239449501,
      "learning_rate": 0.00011791649777057155,
      "loss": 0.0026,
      "step": 8100
    },
    {
      "epoch": 1.6416700445885692,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.12492057681083679,
      "eval_runtime": 47.5509,
      "eval_samples_per_second": 11.482,
      "eval_steps_per_second": 1.451,
      "step": 8100
    },
    {
      "epoch": 1.6436967977300365,
      "grad_norm": 0.011045355349779129,
      "learning_rate": 0.00011781516011349818,
      "loss": 0.0017,
      "step": 8110
    },
    {
      "epoch": 1.6457235508715038,
      "grad_norm": 0.010615518316626549,
      "learning_rate": 0.00011771382245642481,
      "loss": 0.0013,
      "step": 8120
    },
    {
      "epoch": 1.647750304012971,
      "grad_norm": 10.780515670776367,
      "learning_rate": 0.00011761248479935145,
      "loss": 1.5084,
      "step": 8130
    },
    {
      "epoch": 1.6497770571544386,
      "grad_norm": 0.023448778316378593,
      "learning_rate": 0.00011751114714227806,
      "loss": 0.0028,
      "step": 8140
    },
    {
      "epoch": 1.651803810295906,
      "grad_norm": 0.014660656452178955,
      "learning_rate": 0.0001174098094852047,
      "loss": 0.0034,
      "step": 8150
    },
    {
      "epoch": 1.6538305634373733,
      "grad_norm": 0.16463139653205872,
      "learning_rate": 0.00011730847182813134,
      "loss": 0.7985,
      "step": 8160
    },
    {
      "epoch": 1.6558573165788406,
      "grad_norm": 0.020764421671628952,
      "learning_rate": 0.00011720713417105798,
      "loss": 0.5618,
      "step": 8170
    },
    {
      "epoch": 1.657884069720308,
      "grad_norm": 7.786919116973877,
      "learning_rate": 0.00011710579651398461,
      "loss": 0.4688,
      "step": 8180
    },
    {
      "epoch": 1.6599108228617754,
      "grad_norm": 0.15201693773269653,
      "learning_rate": 0.00011700445885691123,
      "loss": 0.3611,
      "step": 8190
    },
    {
      "epoch": 1.6619375760032429,
      "grad_norm": 0.01921539194881916,
      "learning_rate": 0.00011690312119983786,
      "loss": 0.002,
      "step": 8200
    },
    {
      "epoch": 1.6619375760032429,
      "eval_accuracy": 0.9615384615384616,
      "eval_loss": 0.13859277963638306,
      "eval_runtime": 50.3004,
      "eval_samples_per_second": 10.855,
      "eval_steps_per_second": 1.372,
      "step": 8200
    },
    {
      "epoch": 1.6639643291447102,
      "grad_norm": 0.015072059817612171,
      "learning_rate": 0.00011680178354276449,
      "loss": 0.5388,
      "step": 8210
    },
    {
      "epoch": 1.6659910822861774,
      "grad_norm": 0.018651235848665237,
      "learning_rate": 0.00011670044588569114,
      "loss": 0.3912,
      "step": 8220
    },
    {
      "epoch": 1.668017835427645,
      "grad_norm": 0.029343340545892715,
      "learning_rate": 0.00011659910822861777,
      "loss": 0.0036,
      "step": 8230
    },
    {
      "epoch": 1.6700445885691124,
      "grad_norm": 0.017087263986468315,
      "learning_rate": 0.00011649777057154439,
      "loss": 0.0055,
      "step": 8240
    },
    {
      "epoch": 1.6720713417105797,
      "grad_norm": 0.22589299082756042,
      "learning_rate": 0.00011639643291447102,
      "loss": 0.0078,
      "step": 8250
    },
    {
      "epoch": 1.674098094852047,
      "grad_norm": 0.012997474521398544,
      "learning_rate": 0.00011629509525739766,
      "loss": 0.0047,
      "step": 8260
    },
    {
      "epoch": 1.6761248479935142,
      "grad_norm": 0.011784421280026436,
      "learning_rate": 0.00011619375760032429,
      "loss": 0.0058,
      "step": 8270
    },
    {
      "epoch": 1.6781516011349817,
      "grad_norm": 0.014172839932143688,
      "learning_rate": 0.00011609241994325091,
      "loss": 0.0069,
      "step": 8280
    },
    {
      "epoch": 1.6801783542764492,
      "grad_norm": 0.010487750172615051,
      "learning_rate": 0.00011599108228617754,
      "loss": 0.0028,
      "step": 8290
    },
    {
      "epoch": 1.6822051074179165,
      "grad_norm": 0.010051158256828785,
      "learning_rate": 0.00011588974462910417,
      "loss": 0.0041,
      "step": 8300
    },
    {
      "epoch": 1.6822051074179165,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.11753609031438828,
      "eval_runtime": 50.388,
      "eval_samples_per_second": 10.836,
      "eval_steps_per_second": 1.369,
      "step": 8300
    },
    {
      "epoch": 1.6842318605593838,
      "grad_norm": 0.10530339926481247,
      "learning_rate": 0.00011578840697203082,
      "loss": 0.0048,
      "step": 8310
    },
    {
      "epoch": 1.6862586137008513,
      "grad_norm": 0.009275587275624275,
      "learning_rate": 0.00011568706931495745,
      "loss": 0.6973,
      "step": 8320
    },
    {
      "epoch": 1.6882853668423186,
      "grad_norm": 0.012001892551779747,
      "learning_rate": 0.00011558573165788407,
      "loss": 0.0023,
      "step": 8330
    },
    {
      "epoch": 1.690312119983786,
      "grad_norm": 12.442451477050781,
      "learning_rate": 0.0001154843940008107,
      "loss": 0.5785,
      "step": 8340
    },
    {
      "epoch": 1.6923388731252533,
      "grad_norm": 0.012126659043133259,
      "learning_rate": 0.00011538305634373734,
      "loss": 0.4018,
      "step": 8350
    },
    {
      "epoch": 1.6943656262667206,
      "grad_norm": 0.011285288259387016,
      "learning_rate": 0.00011528171868666397,
      "loss": 0.0054,
      "step": 8360
    },
    {
      "epoch": 1.696392379408188,
      "grad_norm": 0.011082268320024014,
      "learning_rate": 0.00011518038102959062,
      "loss": 0.005,
      "step": 8370
    },
    {
      "epoch": 1.6984191325496556,
      "grad_norm": 0.04278941825032234,
      "learning_rate": 0.00011507904337251722,
      "loss": 0.0043,
      "step": 8380
    },
    {
      "epoch": 1.7004458856911229,
      "grad_norm": 0.10877048969268799,
      "learning_rate": 0.00011497770571544387,
      "loss": 0.0023,
      "step": 8390
    },
    {
      "epoch": 1.7024726388325901,
      "grad_norm": 0.08552572131156921,
      "learning_rate": 0.0001148763680583705,
      "loss": 0.0034,
      "step": 8400
    },
    {
      "epoch": 1.7024726388325901,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.11600035429000854,
      "eval_runtime": 49.2969,
      "eval_samples_per_second": 11.076,
      "eval_steps_per_second": 1.4,
      "step": 8400
    },
    {
      "epoch": 1.7044993919740574,
      "grad_norm": 0.0091655058786273,
      "learning_rate": 0.00011477503040129713,
      "loss": 0.001,
      "step": 8410
    },
    {
      "epoch": 1.706526145115525,
      "grad_norm": 0.09666013717651367,
      "learning_rate": 0.00011467369274422375,
      "loss": 0.1692,
      "step": 8420
    },
    {
      "epoch": 1.7085528982569924,
      "grad_norm": 0.00899573601782322,
      "learning_rate": 0.00011457235508715038,
      "loss": 0.003,
      "step": 8430
    },
    {
      "epoch": 1.7105796513984597,
      "grad_norm": 9.241072654724121,
      "learning_rate": 0.00011447101743007702,
      "loss": 0.7043,
      "step": 8440
    },
    {
      "epoch": 1.712606404539927,
      "grad_norm": 0.010043786838650703,
      "learning_rate": 0.00011436967977300366,
      "loss": 0.0019,
      "step": 8450
    },
    {
      "epoch": 1.7146331576813945,
      "grad_norm": 0.01046903245151043,
      "learning_rate": 0.0001142683421159303,
      "loss": 0.4672,
      "step": 8460
    },
    {
      "epoch": 1.7166599108228617,
      "grad_norm": 0.011509249918162823,
      "learning_rate": 0.0001141670044588569,
      "loss": 0.0028,
      "step": 8470
    },
    {
      "epoch": 1.7186866639643292,
      "grad_norm": 0.012271828949451447,
      "learning_rate": 0.00011406566680178355,
      "loss": 0.4348,
      "step": 8480
    },
    {
      "epoch": 1.7207134171057965,
      "grad_norm": 0.010808630846440792,
      "learning_rate": 0.00011396432914471018,
      "loss": 0.0422,
      "step": 8490
    },
    {
      "epoch": 1.7227401702472638,
      "grad_norm": 0.1264539510011673,
      "learning_rate": 0.00011386299148763681,
      "loss": 0.0041,
      "step": 8500
    },
    {
      "epoch": 1.7227401702472638,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.20973940193653107,
      "eval_runtime": 45.8027,
      "eval_samples_per_second": 11.921,
      "eval_steps_per_second": 1.506,
      "step": 8500
    },
    {
      "epoch": 1.7247669233887313,
      "grad_norm": 0.11836584657430649,
      "learning_rate": 0.00011376165383056346,
      "loss": 0.5827,
      "step": 8510
    },
    {
      "epoch": 1.7267936765301988,
      "grad_norm": 0.1095990464091301,
      "learning_rate": 0.00011366031617349006,
      "loss": 0.0078,
      "step": 8520
    },
    {
      "epoch": 1.728820429671666,
      "grad_norm": 0.05297279730439186,
      "learning_rate": 0.0001135589785164167,
      "loss": 0.5844,
      "step": 8530
    },
    {
      "epoch": 1.7308471828131333,
      "grad_norm": 0.10899315029382706,
      "learning_rate": 0.00011345764085934334,
      "loss": 0.6763,
      "step": 8540
    },
    {
      "epoch": 1.7328739359546006,
      "grad_norm": 0.023719044402241707,
      "learning_rate": 0.00011335630320226998,
      "loss": 0.4364,
      "step": 8550
    },
    {
      "epoch": 1.734900689096068,
      "grad_norm": 0.014032549224793911,
      "learning_rate": 0.0001132549655451966,
      "loss": 0.0061,
      "step": 8560
    },
    {
      "epoch": 1.7369274422375356,
      "grad_norm": 0.010516401380300522,
      "learning_rate": 0.00011315362788812323,
      "loss": 0.0762,
      "step": 8570
    },
    {
      "epoch": 1.7389541953790029,
      "grad_norm": 0.13084004819393158,
      "learning_rate": 0.00011305229023104986,
      "loss": 0.0242,
      "step": 8580
    },
    {
      "epoch": 1.7409809485204701,
      "grad_norm": 0.010349682532250881,
      "learning_rate": 0.00011295095257397649,
      "loss": 0.2937,
      "step": 8590
    },
    {
      "epoch": 1.7430077016619374,
      "grad_norm": 0.009873668663203716,
      "learning_rate": 0.00011284961491690314,
      "loss": 0.3303,
      "step": 8600
    },
    {
      "epoch": 1.7430077016619374,
      "eval_accuracy": 0.9597069597069597,
      "eval_loss": 0.15265528857707977,
      "eval_runtime": 49.6316,
      "eval_samples_per_second": 11.001,
      "eval_steps_per_second": 1.39,
      "step": 8600
    },
    {
      "epoch": 1.745034454803405,
      "grad_norm": 0.5044683218002319,
      "learning_rate": 0.00011274827725982974,
      "loss": 1.0479,
      "step": 8610
    },
    {
      "epoch": 1.7470612079448724,
      "grad_norm": 0.17945395410060883,
      "learning_rate": 0.00011264693960275639,
      "loss": 0.4109,
      "step": 8620
    },
    {
      "epoch": 1.7490879610863397,
      "grad_norm": 8.084101676940918,
      "learning_rate": 0.00011254560194568302,
      "loss": 0.3745,
      "step": 8630
    },
    {
      "epoch": 1.751114714227807,
      "grad_norm": 0.015885835513472557,
      "learning_rate": 0.00011244426428860966,
      "loss": 0.0023,
      "step": 8640
    },
    {
      "epoch": 1.7531414673692745,
      "grad_norm": 0.035967472940683365,
      "learning_rate": 0.00011234292663153629,
      "loss": 0.3532,
      "step": 8650
    },
    {
      "epoch": 1.755168220510742,
      "grad_norm": 0.01129199843853712,
      "learning_rate": 0.00011224158897446291,
      "loss": 0.006,
      "step": 8660
    },
    {
      "epoch": 1.7571949736522092,
      "grad_norm": 0.294033408164978,
      "learning_rate": 0.00011214025131738954,
      "loss": 0.0176,
      "step": 8670
    },
    {
      "epoch": 1.7592217267936765,
      "grad_norm": 0.00972708035260439,
      "learning_rate": 0.00011203891366031619,
      "loss": 0.0105,
      "step": 8680
    },
    {
      "epoch": 1.7612484799351438,
      "grad_norm": 9.236714363098145,
      "learning_rate": 0.00011193757600324282,
      "loss": 0.7035,
      "step": 8690
    },
    {
      "epoch": 1.7632752330766113,
      "grad_norm": 0.010941145941615105,
      "learning_rate": 0.00011183623834616945,
      "loss": 0.006,
      "step": 8700
    },
    {
      "epoch": 1.7632752330766113,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.13893744349479675,
      "eval_runtime": 49.6277,
      "eval_samples_per_second": 11.002,
      "eval_steps_per_second": 1.39,
      "step": 8700
    },
    {
      "epoch": 1.7653019862180788,
      "grad_norm": 0.016714446246623993,
      "learning_rate": 0.00011173490068909607,
      "loss": 0.9278,
      "step": 8710
    },
    {
      "epoch": 1.767328739359546,
      "grad_norm": 0.01921900361776352,
      "learning_rate": 0.0001116335630320227,
      "loss": 0.0035,
      "step": 8720
    },
    {
      "epoch": 1.7693554925010133,
      "grad_norm": 0.02214176394045353,
      "learning_rate": 0.00011153222537494934,
      "loss": 0.0057,
      "step": 8730
    },
    {
      "epoch": 1.7713822456424806,
      "grad_norm": 0.024326831102371216,
      "learning_rate": 0.00011143088771787597,
      "loss": 1.0052,
      "step": 8740
    },
    {
      "epoch": 1.773408998783948,
      "grad_norm": 0.20547142624855042,
      "learning_rate": 0.00011132955006080259,
      "loss": 0.3104,
      "step": 8750
    },
    {
      "epoch": 1.7754357519254156,
      "grad_norm": 0.15221790969371796,
      "learning_rate": 0.00011122821240372922,
      "loss": 0.0205,
      "step": 8760
    },
    {
      "epoch": 1.7774625050668829,
      "grad_norm": 0.014965363778173923,
      "learning_rate": 0.00011112687474665587,
      "loss": 0.005,
      "step": 8770
    },
    {
      "epoch": 1.7794892582083501,
      "grad_norm": 0.01209799014031887,
      "learning_rate": 0.0001110255370895825,
      "loss": 0.005,
      "step": 8780
    },
    {
      "epoch": 1.7815160113498176,
      "grad_norm": 0.011696490459144115,
      "learning_rate": 0.00011092419943250913,
      "loss": 0.0054,
      "step": 8790
    },
    {
      "epoch": 1.7835427644912851,
      "grad_norm": 0.010011350736021996,
      "learning_rate": 0.00011082286177543575,
      "loss": 0.0012,
      "step": 8800
    },
    {
      "epoch": 1.7835427644912851,
      "eval_accuracy": 0.9597069597069597,
      "eval_loss": 0.17990480363368988,
      "eval_runtime": 48.1871,
      "eval_samples_per_second": 11.331,
      "eval_steps_per_second": 1.432,
      "step": 8800
    },
    {
      "epoch": 1.7855695176327524,
      "grad_norm": 0.0209858026355505,
      "learning_rate": 0.00011072152411836238,
      "loss": 0.5705,
      "step": 8810
    },
    {
      "epoch": 1.7875962707742197,
      "grad_norm": 0.010832561179995537,
      "learning_rate": 0.00011062018646128902,
      "loss": 0.0075,
      "step": 8820
    },
    {
      "epoch": 1.789623023915687,
      "grad_norm": 0.009939854964613914,
      "learning_rate": 0.00011051884880421566,
      "loss": 0.0049,
      "step": 8830
    },
    {
      "epoch": 1.7916497770571544,
      "grad_norm": 0.009512858465313911,
      "learning_rate": 0.0001104175111471423,
      "loss": 0.4469,
      "step": 8840
    },
    {
      "epoch": 1.793676530198622,
      "grad_norm": 0.009505028836429119,
      "learning_rate": 0.0001103161734900689,
      "loss": 0.0025,
      "step": 8850
    },
    {
      "epoch": 1.7957032833400892,
      "grad_norm": 0.011000646278262138,
      "learning_rate": 0.00011021483583299555,
      "loss": 0.6492,
      "step": 8860
    },
    {
      "epoch": 1.7977300364815565,
      "grad_norm": 0.011823964305222034,
      "learning_rate": 0.00011011349817592218,
      "loss": 0.0043,
      "step": 8870
    },
    {
      "epoch": 1.7997567896230238,
      "grad_norm": 0.028389113023877144,
      "learning_rate": 0.00011001216051884881,
      "loss": 0.002,
      "step": 8880
    },
    {
      "epoch": 1.8017835427644913,
      "grad_norm": 0.0190489050000906,
      "learning_rate": 0.00010991082286177543,
      "loss": 0.0061,
      "step": 8890
    },
    {
      "epoch": 1.8038102959059588,
      "grad_norm": 0.027871625497937202,
      "learning_rate": 0.00010980948520470206,
      "loss": 0.0027,
      "step": 8900
    },
    {
      "epoch": 1.8038102959059588,
      "eval_accuracy": 0.9615384615384616,
      "eval_loss": 0.17173990607261658,
      "eval_runtime": 210.3245,
      "eval_samples_per_second": 2.596,
      "eval_steps_per_second": 0.328,
      "step": 8900
    },
    {
      "epoch": 1.805837049047426,
      "grad_norm": 0.00957685150206089,
      "learning_rate": 0.0001097081475476287,
      "loss": 0.0012,
      "step": 8910
    },
    {
      "epoch": 1.8078638021888933,
      "grad_norm": 0.013015580363571644,
      "learning_rate": 0.00010960680989055534,
      "loss": 0.0042,
      "step": 8920
    },
    {
      "epoch": 1.8098905553303608,
      "grad_norm": 0.008935632184147835,
      "learning_rate": 0.00010950547223348198,
      "loss": 0.0018,
      "step": 8930
    },
    {
      "epoch": 1.811917308471828,
      "grad_norm": 0.011125830933451653,
      "learning_rate": 0.0001094041345764086,
      "loss": 0.0022,
      "step": 8940
    },
    {
      "epoch": 1.8139440616132956,
      "grad_norm": 0.008295311592519283,
      "learning_rate": 0.00010930279691933523,
      "loss": 0.409,
      "step": 8950
    },
    {
      "epoch": 1.8159708147547629,
      "grad_norm": 0.007981337606906891,
      "learning_rate": 0.00010920145926226186,
      "loss": 0.0026,
      "step": 8960
    },
    {
      "epoch": 1.8179975678962301,
      "grad_norm": 0.007811954244971275,
      "learning_rate": 0.00010910012160518849,
      "loss": 0.0025,
      "step": 8970
    },
    {
      "epoch": 1.8200243210376976,
      "grad_norm": 0.011638646014034748,
      "learning_rate": 0.00010899878394811514,
      "loss": 0.3995,
      "step": 8980
    },
    {
      "epoch": 1.8220510741791651,
      "grad_norm": 0.06227179989218712,
      "learning_rate": 0.00010889744629104174,
      "loss": 0.0019,
      "step": 8990
    },
    {
      "epoch": 1.8240778273206324,
      "grad_norm": 8.30929183959961,
      "learning_rate": 0.00010879610863396839,
      "loss": 0.4926,
      "step": 9000
    },
    {
      "epoch": 1.8240778273206324,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.15174922347068787,
      "eval_runtime": 114.8648,
      "eval_samples_per_second": 4.753,
      "eval_steps_per_second": 0.601,
      "step": 9000
    },
    {
      "epoch": 1.8261045804620997,
      "grad_norm": 0.17263716459274292,
      "learning_rate": 0.00010869477097689502,
      "loss": 0.485,
      "step": 9010
    },
    {
      "epoch": 1.828131333603567,
      "grad_norm": 0.008078324608504772,
      "learning_rate": 0.00010859343331982166,
      "loss": 0.0021,
      "step": 9020
    },
    {
      "epoch": 1.8301580867450344,
      "grad_norm": 0.008292094804346561,
      "learning_rate": 0.00010849209566274828,
      "loss": 0.0048,
      "step": 9030
    },
    {
      "epoch": 1.832184839886502,
      "grad_norm": 0.007516876794397831,
      "learning_rate": 0.00010839075800567491,
      "loss": 0.0019,
      "step": 9040
    },
    {
      "epoch": 1.8342115930279692,
      "grad_norm": 0.007887504994869232,
      "learning_rate": 0.00010828942034860154,
      "loss": 0.0009,
      "step": 9050
    },
    {
      "epoch": 1.8362383461694365,
      "grad_norm": 0.015964463353157043,
      "learning_rate": 0.00010818808269152819,
      "loss": 0.0028,
      "step": 9060
    },
    {
      "epoch": 1.838265099310904,
      "grad_norm": 0.1481909453868866,
      "learning_rate": 0.00010808674503445482,
      "loss": 0.0033,
      "step": 9070
    },
    {
      "epoch": 1.8402918524523713,
      "grad_norm": 0.006927825044840574,
      "learning_rate": 0.00010798540737738143,
      "loss": 0.0031,
      "step": 9080
    },
    {
      "epoch": 1.8423186055938388,
      "grad_norm": 0.006547117605805397,
      "learning_rate": 0.00010788406972030807,
      "loss": 0.0015,
      "step": 9090
    },
    {
      "epoch": 1.844345358735306,
      "grad_norm": 0.006231117062270641,
      "learning_rate": 0.0001077827320632347,
      "loss": 0.0023,
      "step": 9100
    },
    {
      "epoch": 1.844345358735306,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.12722833454608917,
      "eval_runtime": 71.9284,
      "eval_samples_per_second": 7.591,
      "eval_steps_per_second": 0.959,
      "step": 9100
    },
    {
      "epoch": 1.8463721118767733,
      "grad_norm": 0.006408121902495623,
      "learning_rate": 0.00010768139440616134,
      "loss": 0.0019,
      "step": 9110
    },
    {
      "epoch": 1.8483988650182408,
      "grad_norm": 0.006610712967813015,
      "learning_rate": 0.00010758005674908798,
      "loss": 0.7366,
      "step": 9120
    },
    {
      "epoch": 1.8504256181597083,
      "grad_norm": 0.060888879001140594,
      "learning_rate": 0.00010747871909201459,
      "loss": 0.0028,
      "step": 9130
    },
    {
      "epoch": 1.8524523713011756,
      "grad_norm": 0.0076481616124510765,
      "learning_rate": 0.00010737738143494122,
      "loss": 0.0022,
      "step": 9140
    },
    {
      "epoch": 1.8544791244426428,
      "grad_norm": 0.006980573292821646,
      "learning_rate": 0.00010727604377786787,
      "loss": 0.0014,
      "step": 9150
    },
    {
      "epoch": 1.8565058775841101,
      "grad_norm": 0.006991061381995678,
      "learning_rate": 0.0001071747061207945,
      "loss": 0.0021,
      "step": 9160
    },
    {
      "epoch": 1.8585326307255776,
      "grad_norm": 0.007438084110617638,
      "learning_rate": 0.00010707336846372112,
      "loss": 0.0014,
      "step": 9170
    },
    {
      "epoch": 1.8605593838670451,
      "grad_norm": 0.006888444069772959,
      "learning_rate": 0.00010697203080664775,
      "loss": 0.0014,
      "step": 9180
    },
    {
      "epoch": 1.8625861370085124,
      "grad_norm": 0.006462920922785997,
      "learning_rate": 0.00010687069314957438,
      "loss": 0.0013,
      "step": 9190
    },
    {
      "epoch": 1.8646128901499797,
      "grad_norm": 0.04772758483886719,
      "learning_rate": 0.00010676935549250102,
      "loss": 0.5028,
      "step": 9200
    },
    {
      "epoch": 1.8646128901499797,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.1444089710712433,
      "eval_runtime": 76.7389,
      "eval_samples_per_second": 7.115,
      "eval_steps_per_second": 0.899,
      "step": 9200
    },
    {
      "epoch": 1.866639643291447,
      "grad_norm": 0.007817757315933704,
      "learning_rate": 0.00010666801783542766,
      "loss": 0.7305,
      "step": 9210
    },
    {
      "epoch": 1.8686663964329144,
      "grad_norm": 0.00858191680163145,
      "learning_rate": 0.00010656668017835427,
      "loss": 1.0045,
      "step": 9220
    },
    {
      "epoch": 1.870693149574382,
      "grad_norm": 0.009446145966649055,
      "learning_rate": 0.00010646534252128091,
      "loss": 0.002,
      "step": 9230
    },
    {
      "epoch": 1.8727199027158492,
      "grad_norm": 0.00917446706444025,
      "learning_rate": 0.00010636400486420755,
      "loss": 0.0083,
      "step": 9240
    },
    {
      "epoch": 1.8747466558573165,
      "grad_norm": 0.09278376400470734,
      "learning_rate": 0.00010626266720713418,
      "loss": 0.5391,
      "step": 9250
    },
    {
      "epoch": 1.876773408998784,
      "grad_norm": 0.009064766578376293,
      "learning_rate": 0.00010616132955006081,
      "loss": 0.0019,
      "step": 9260
    },
    {
      "epoch": 1.8788001621402515,
      "grad_norm": 0.009625071659684181,
      "learning_rate": 0.00010605999189298743,
      "loss": 0.909,
      "step": 9270
    },
    {
      "epoch": 1.8808269152817187,
      "grad_norm": 0.011897668242454529,
      "learning_rate": 0.00010595865423591406,
      "loss": 0.4662,
      "step": 9280
    },
    {
      "epoch": 1.882853668423186,
      "grad_norm": 0.10990961641073227,
      "learning_rate": 0.0001058573165788407,
      "loss": 0.0138,
      "step": 9290
    },
    {
      "epoch": 1.8848804215646533,
      "grad_norm": 0.01036153919994831,
      "learning_rate": 0.00010575597892176734,
      "loss": 0.0051,
      "step": 9300
    },
    {
      "epoch": 1.8848804215646533,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.1275521218776703,
      "eval_runtime": 72.9524,
      "eval_samples_per_second": 7.484,
      "eval_steps_per_second": 0.946,
      "step": 9300
    },
    {
      "epoch": 1.8869071747061208,
      "grad_norm": 0.0937432125210762,
      "learning_rate": 0.00010565464126469395,
      "loss": 0.0087,
      "step": 9310
    },
    {
      "epoch": 1.8889339278475883,
      "grad_norm": 0.009895688854157925,
      "learning_rate": 0.0001055533036076206,
      "loss": 0.2518,
      "step": 9320
    },
    {
      "epoch": 1.8909606809890556,
      "grad_norm": 0.008279825560748577,
      "learning_rate": 0.00010545196595054723,
      "loss": 0.0021,
      "step": 9330
    },
    {
      "epoch": 1.8929874341305228,
      "grad_norm": 0.0076844836585223675,
      "learning_rate": 0.00010535062829347386,
      "loss": 0.0019,
      "step": 9340
    },
    {
      "epoch": 1.8950141872719901,
      "grad_norm": 0.07048050314188004,
      "learning_rate": 0.00010524929063640049,
      "loss": 0.0024,
      "step": 9350
    },
    {
      "epoch": 1.8970409404134576,
      "grad_norm": 8.189104080200195,
      "learning_rate": 0.00010514795297932711,
      "loss": 0.9113,
      "step": 9360
    },
    {
      "epoch": 1.899067693554925,
      "grad_norm": 0.008056996390223503,
      "learning_rate": 0.00010504661532225374,
      "loss": 0.0046,
      "step": 9370
    },
    {
      "epoch": 1.9010944466963924,
      "grad_norm": 0.007996568456292152,
      "learning_rate": 0.00010494527766518039,
      "loss": 0.005,
      "step": 9380
    },
    {
      "epoch": 1.9031211998378597,
      "grad_norm": 0.00770143186673522,
      "learning_rate": 0.00010484394000810702,
      "loss": 0.0045,
      "step": 9390
    },
    {
      "epoch": 1.9051479529793272,
      "grad_norm": 0.007663055323064327,
      "learning_rate": 0.00010474260235103366,
      "loss": 0.0019,
      "step": 9400
    },
    {
      "epoch": 1.9051479529793272,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.15504136681556702,
      "eval_runtime": 78.4367,
      "eval_samples_per_second": 6.961,
      "eval_steps_per_second": 0.88,
      "step": 9400
    },
    {
      "epoch": 1.9071747061207946,
      "grad_norm": 0.00760328583419323,
      "learning_rate": 0.00010464126469396028,
      "loss": 0.0018,
      "step": 9410
    },
    {
      "epoch": 1.909201459262262,
      "grad_norm": 21.01438331604004,
      "learning_rate": 0.00010453992703688691,
      "loss": 0.3983,
      "step": 9420
    },
    {
      "epoch": 1.9112282124037292,
      "grad_norm": 0.007572650909423828,
      "learning_rate": 0.00010443858937981354,
      "loss": 0.4565,
      "step": 9430
    },
    {
      "epoch": 1.9132549655451965,
      "grad_norm": 0.007722463458776474,
      "learning_rate": 0.00010433725172274019,
      "loss": 0.0009,
      "step": 9440
    },
    {
      "epoch": 1.915281718686664,
      "grad_norm": 0.007536469493061304,
      "learning_rate": 0.00010423591406566682,
      "loss": 0.0027,
      "step": 9450
    },
    {
      "epoch": 1.9173084718281315,
      "grad_norm": 0.007839085534214973,
      "learning_rate": 0.00010413457640859343,
      "loss": 0.4254,
      "step": 9460
    },
    {
      "epoch": 1.9193352249695987,
      "grad_norm": 0.12683309614658356,
      "learning_rate": 0.00010403323875152007,
      "loss": 0.0037,
      "step": 9470
    },
    {
      "epoch": 1.921361978111066,
      "grad_norm": 0.007982700131833553,
      "learning_rate": 0.0001039319010944467,
      "loss": 0.0038,
      "step": 9480
    },
    {
      "epoch": 1.9233887312525333,
      "grad_norm": 0.007812978699803352,
      "learning_rate": 0.00010383056343737334,
      "loss": 0.3642,
      "step": 9490
    },
    {
      "epoch": 1.9254154843940008,
      "grad_norm": 0.0074735949747264385,
      "learning_rate": 0.00010372922578029996,
      "loss": 0.0052,
      "step": 9500
    },
    {
      "epoch": 1.9254154843940008,
      "eval_accuracy": 0.9633699633699634,
      "eval_loss": 0.19582439959049225,
      "eval_runtime": 76.1295,
      "eval_samples_per_second": 7.172,
      "eval_steps_per_second": 0.906,
      "step": 9500
    },
    {
      "epoch": 1.9274422375354683,
      "grad_norm": 0.007177373860031366,
      "learning_rate": 0.00010362788812322659,
      "loss": 0.005,
      "step": 9510
    },
    {
      "epoch": 1.9294689906769356,
      "grad_norm": 0.007034731563180685,
      "learning_rate": 0.00010352655046615322,
      "loss": 0.428,
      "step": 9520
    },
    {
      "epoch": 1.9314957438184028,
      "grad_norm": 0.00731532322242856,
      "learning_rate": 0.00010342521280907987,
      "loss": 0.0059,
      "step": 9530
    },
    {
      "epoch": 1.9335224969598703,
      "grad_norm": 0.007122396025806665,
      "learning_rate": 0.0001033238751520065,
      "loss": 0.6688,
      "step": 9540
    },
    {
      "epoch": 1.9355492501013376,
      "grad_norm": 0.009541909210383892,
      "learning_rate": 0.00010322253749493312,
      "loss": 0.0027,
      "step": 9550
    },
    {
      "epoch": 1.937576003242805,
      "grad_norm": 0.1792389452457428,
      "learning_rate": 0.00010312119983785975,
      "loss": 0.0067,
      "step": 9560
    },
    {
      "epoch": 1.9396027563842724,
      "grad_norm": 0.010249803774058819,
      "learning_rate": 0.00010301986218078638,
      "loss": 0.3941,
      "step": 9570
    },
    {
      "epoch": 1.9416295095257396,
      "grad_norm": 0.11628035455942154,
      "learning_rate": 0.00010291852452371302,
      "loss": 0.4732,
      "step": 9580
    },
    {
      "epoch": 1.9436562626672071,
      "grad_norm": 0.10727599263191223,
      "learning_rate": 0.00010281718686663966,
      "loss": 0.7003,
      "step": 9590
    },
    {
      "epoch": 1.9456830158086746,
      "grad_norm": 0.168293759226799,
      "learning_rate": 0.00010271584920956627,
      "loss": 0.0099,
      "step": 9600
    },
    {
      "epoch": 1.9456830158086746,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.13594172894954681,
      "eval_runtime": 78.0607,
      "eval_samples_per_second": 6.995,
      "eval_steps_per_second": 0.884,
      "step": 9600
    },
    {
      "epoch": 1.947709768950142,
      "grad_norm": 0.014889145269989967,
      "learning_rate": 0.00010261451155249291,
      "loss": 0.0028,
      "step": 9610
    },
    {
      "epoch": 1.9497365220916092,
      "grad_norm": 0.1032806932926178,
      "learning_rate": 0.00010251317389541955,
      "loss": 0.4153,
      "step": 9620
    },
    {
      "epoch": 1.9517632752330765,
      "grad_norm": 0.0114668607711792,
      "learning_rate": 0.00010241183623834618,
      "loss": 0.0057,
      "step": 9630
    },
    {
      "epoch": 1.953790028374544,
      "grad_norm": 0.18307086825370789,
      "learning_rate": 0.0001023104985812728,
      "loss": 0.0052,
      "step": 9640
    },
    {
      "epoch": 1.9558167815160115,
      "grad_norm": 0.009420246817171574,
      "learning_rate": 0.00010220916092419943,
      "loss": 0.0061,
      "step": 9650
    },
    {
      "epoch": 1.9578435346574787,
      "grad_norm": 0.01083473488688469,
      "learning_rate": 0.00010210782326712606,
      "loss": 0.833,
      "step": 9660
    },
    {
      "epoch": 1.959870287798946,
      "grad_norm": 0.17974765598773956,
      "learning_rate": 0.00010200648561005271,
      "loss": 0.0044,
      "step": 9670
    },
    {
      "epoch": 1.9618970409404135,
      "grad_norm": 0.10389303416013718,
      "learning_rate": 0.00010190514795297934,
      "loss": 0.0036,
      "step": 9680
    },
    {
      "epoch": 1.9639237940818808,
      "grad_norm": 0.010213226079940796,
      "learning_rate": 0.00010180381029590595,
      "loss": 0.0028,
      "step": 9690
    },
    {
      "epoch": 1.9659505472233483,
      "grad_norm": 0.009096471592783928,
      "learning_rate": 0.0001017024726388326,
      "loss": 0.3494,
      "step": 9700
    },
    {
      "epoch": 1.9659505472233483,
      "eval_accuracy": 0.9542124542124543,
      "eval_loss": 0.19689708948135376,
      "eval_runtime": 77.5829,
      "eval_samples_per_second": 7.038,
      "eval_steps_per_second": 0.889,
      "step": 9700
    },
    {
      "epoch": 1.9679773003648156,
      "grad_norm": 0.007500569336116314,
      "learning_rate": 0.00010160113498175923,
      "loss": 0.5272,
      "step": 9710
    },
    {
      "epoch": 1.9700040535062828,
      "grad_norm": 0.007903238758444786,
      "learning_rate": 0.00010149979732468586,
      "loss": 0.0036,
      "step": 9720
    },
    {
      "epoch": 1.9720308066477503,
      "grad_norm": 0.022083204239606857,
      "learning_rate": 0.00010139845966761249,
      "loss": 0.0023,
      "step": 9730
    },
    {
      "epoch": 1.9740575597892178,
      "grad_norm": 0.00797079224139452,
      "learning_rate": 0.00010129712201053911,
      "loss": 0.884,
      "step": 9740
    },
    {
      "epoch": 1.976084312930685,
      "grad_norm": 0.010292446240782738,
      "learning_rate": 0.00010119578435346574,
      "loss": 0.0073,
      "step": 9750
    },
    {
      "epoch": 1.9781110660721524,
      "grad_norm": 0.00823873933404684,
      "learning_rate": 0.00010109444669639239,
      "loss": 0.3514,
      "step": 9760
    },
    {
      "epoch": 1.9801378192136196,
      "grad_norm": 0.10817509144544601,
      "learning_rate": 0.00010099310903931902,
      "loss": 0.0078,
      "step": 9770
    },
    {
      "epoch": 1.9821645723550871,
      "grad_norm": 0.007184948306530714,
      "learning_rate": 0.00010089177138224564,
      "loss": 0.0009,
      "step": 9780
    },
    {
      "epoch": 1.9841913254965546,
      "grad_norm": 0.13583801686763763,
      "learning_rate": 0.00010079043372517228,
      "loss": 0.4471,
      "step": 9790
    },
    {
      "epoch": 1.986218078638022,
      "grad_norm": 0.11831188201904297,
      "learning_rate": 0.00010068909606809891,
      "loss": 0.0035,
      "step": 9800
    },
    {
      "epoch": 1.986218078638022,
      "eval_accuracy": 0.9578754578754579,
      "eval_loss": 0.16712024807929993,
      "eval_runtime": 77.7527,
      "eval_samples_per_second": 7.022,
      "eval_steps_per_second": 0.887,
      "step": 9800
    },
    {
      "epoch": 1.9882448317794892,
      "grad_norm": 0.17543888092041016,
      "learning_rate": 0.00010058775841102554,
      "loss": 1.7896,
      "step": 9810
    },
    {
      "epoch": 1.9902715849209567,
      "grad_norm": 0.01244261208921671,
      "learning_rate": 0.00010048642075395219,
      "loss": 0.0038,
      "step": 9820
    },
    {
      "epoch": 1.992298338062424,
      "grad_norm": 0.010985905304551125,
      "learning_rate": 0.00010038508309687879,
      "loss": 0.0011,
      "step": 9830
    },
    {
      "epoch": 1.9943250912038915,
      "grad_norm": 0.0096734669059515,
      "learning_rate": 0.00010028374543980544,
      "loss": 0.0092,
      "step": 9840
    },
    {
      "epoch": 1.9963518443453587,
      "grad_norm": 0.009174306876957417,
      "learning_rate": 0.00010018240778273207,
      "loss": 0.0039,
      "step": 9850
    },
    {
      "epoch": 1.998378597486826,
      "grad_norm": 0.010896787978708744,
      "learning_rate": 0.0001000810701256587,
      "loss": 0.0142,
      "step": 9860
    },
    {
      "epoch": 2.0004053506282933,
      "grad_norm": 0.008190103806555271,
      "learning_rate": 9.997973246858532e-05,
      "loss": 0.0048,
      "step": 9870
    },
    {
      "epoch": 2.002432103769761,
      "grad_norm": 0.007770847994834185,
      "learning_rate": 9.987839481151197e-05,
      "loss": 0.0025,
      "step": 9880
    },
    {
      "epoch": 2.0044588569112283,
      "grad_norm": 0.009759979322552681,
      "learning_rate": 9.977705715443859e-05,
      "loss": 0.3902,
      "step": 9890
    },
    {
      "epoch": 2.0064856100526955,
      "grad_norm": 0.007514787372201681,
      "learning_rate": 9.967571949736522e-05,
      "loss": 0.0025,
      "step": 9900
    },
    {
      "epoch": 2.0064856100526955,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.14347095787525177,
      "eval_runtime": 72.6553,
      "eval_samples_per_second": 7.515,
      "eval_steps_per_second": 0.95,
      "step": 9900
    },
    {
      "epoch": 2.008512363194163,
      "grad_norm": 0.007548211142420769,
      "learning_rate": 9.957438184029185e-05,
      "loss": 0.0066,
      "step": 9910
    },
    {
      "epoch": 2.0105391163356305,
      "grad_norm": 0.11043214052915573,
      "learning_rate": 9.947304418321849e-05,
      "loss": 0.0049,
      "step": 9920
    },
    {
      "epoch": 2.012565869477098,
      "grad_norm": 0.006975914351642132,
      "learning_rate": 9.937170652614512e-05,
      "loss": 0.5866,
      "step": 9930
    },
    {
      "epoch": 2.014592622618565,
      "grad_norm": 0.015638014301657677,
      "learning_rate": 9.927036886907175e-05,
      "loss": 0.4344,
      "step": 9940
    },
    {
      "epoch": 2.0166193757600324,
      "grad_norm": 0.006982761435210705,
      "learning_rate": 9.916903121199838e-05,
      "loss": 0.0098,
      "step": 9950
    },
    {
      "epoch": 2.0186461289014996,
      "grad_norm": 0.006579644978046417,
      "learning_rate": 9.906769355492502e-05,
      "loss": 0.0094,
      "step": 9960
    },
    {
      "epoch": 2.0206728820429674,
      "grad_norm": 0.006210808642208576,
      "learning_rate": 9.896635589785165e-05,
      "loss": 0.0073,
      "step": 9970
    },
    {
      "epoch": 2.0226996351844346,
      "grad_norm": 0.030406000092625618,
      "learning_rate": 9.886501824077827e-05,
      "loss": 0.0041,
      "step": 9980
    },
    {
      "epoch": 2.024726388325902,
      "grad_norm": 0.011690155602991581,
      "learning_rate": 9.876368058370491e-05,
      "loss": 0.0042,
      "step": 9990
    },
    {
      "epoch": 2.026753141467369,
      "grad_norm": 0.005466008093208075,
      "learning_rate": 9.866234292663155e-05,
      "loss": 0.0006,
      "step": 10000
    },
    {
      "epoch": 2.026753141467369,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.11869247257709503,
      "eval_runtime": 72.4776,
      "eval_samples_per_second": 7.533,
      "eval_steps_per_second": 0.952,
      "step": 10000
    },
    {
      "epoch": 2.0287798946088365,
      "grad_norm": 0.00618654815480113,
      "learning_rate": 9.856100526955817e-05,
      "loss": 0.0014,
      "step": 10010
    },
    {
      "epoch": 2.030806647750304,
      "grad_norm": 0.005324477329850197,
      "learning_rate": 9.845966761248481e-05,
      "loss": 0.0024,
      "step": 10020
    },
    {
      "epoch": 2.0328334008917714,
      "grad_norm": 0.005492259282618761,
      "learning_rate": 9.835832995541143e-05,
      "loss": 0.0006,
      "step": 10030
    },
    {
      "epoch": 2.0348601540332387,
      "grad_norm": 0.005066926125437021,
      "learning_rate": 9.825699229833806e-05,
      "loss": 0.0006,
      "step": 10040
    },
    {
      "epoch": 2.036886907174706,
      "grad_norm": 0.004997451324015856,
      "learning_rate": 9.81556546412647e-05,
      "loss": 0.0005,
      "step": 10050
    },
    {
      "epoch": 2.0389136603161733,
      "grad_norm": 0.004906692076474428,
      "learning_rate": 9.805431698419133e-05,
      "loss": 0.0031,
      "step": 10060
    },
    {
      "epoch": 2.040940413457641,
      "grad_norm": 0.005268965382128954,
      "learning_rate": 9.795297932711796e-05,
      "loss": 0.417,
      "step": 10070
    },
    {
      "epoch": 2.0429671665991083,
      "grad_norm": 0.13274605572223663,
      "learning_rate": 9.78516416700446e-05,
      "loss": 0.002,
      "step": 10080
    },
    {
      "epoch": 2.0449939197405755,
      "grad_norm": 0.13358217477798462,
      "learning_rate": 9.775030401297123e-05,
      "loss": 0.4915,
      "step": 10090
    },
    {
      "epoch": 2.047020672882043,
      "grad_norm": 0.0064069111831486225,
      "learning_rate": 9.764896635589785e-05,
      "loss": 0.0035,
      "step": 10100
    },
    {
      "epoch": 2.047020672882043,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.13034148514270782,
      "eval_runtime": 75.3202,
      "eval_samples_per_second": 7.249,
      "eval_steps_per_second": 0.916,
      "step": 10100
    },
    {
      "epoch": 2.0490474260235105,
      "grad_norm": 0.09453519433736801,
      "learning_rate": 9.754762869882449e-05,
      "loss": 0.4684,
      "step": 10110
    },
    {
      "epoch": 2.051074179164978,
      "grad_norm": 0.005555831361562014,
      "learning_rate": 9.744629104175111e-05,
      "loss": 0.0018,
      "step": 10120
    },
    {
      "epoch": 2.053100932306445,
      "grad_norm": 0.005330225918442011,
      "learning_rate": 9.734495338467774e-05,
      "loss": 0.0028,
      "step": 10130
    },
    {
      "epoch": 2.0551276854479124,
      "grad_norm": 0.14626197516918182,
      "learning_rate": 9.724361572760439e-05,
      "loss": 0.0023,
      "step": 10140
    },
    {
      "epoch": 2.0571544385893796,
      "grad_norm": 0.005088937468826771,
      "learning_rate": 9.714227807053101e-05,
      "loss": 0.0016,
      "step": 10150
    },
    {
      "epoch": 2.0591811917308473,
      "grad_norm": 0.005233987234532833,
      "learning_rate": 9.704094041345764e-05,
      "loss": 0.0054,
      "step": 10160
    },
    {
      "epoch": 2.0612079448723146,
      "grad_norm": 0.00554394256323576,
      "learning_rate": 9.693960275638428e-05,
      "loss": 0.4676,
      "step": 10170
    },
    {
      "epoch": 2.063234698013782,
      "grad_norm": 0.00520316231995821,
      "learning_rate": 9.683826509931091e-05,
      "loss": 0.0026,
      "step": 10180
    },
    {
      "epoch": 2.065261451155249,
      "grad_norm": 0.005012698005884886,
      "learning_rate": 9.673692744223754e-05,
      "loss": 0.0019,
      "step": 10190
    },
    {
      "epoch": 2.0672882042967164,
      "grad_norm": 0.005969386547803879,
      "learning_rate": 9.663558978516417e-05,
      "loss": 0.7492,
      "step": 10200
    },
    {
      "epoch": 2.0672882042967164,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.12935897707939148,
      "eval_runtime": 78.1995,
      "eval_samples_per_second": 6.982,
      "eval_steps_per_second": 0.882,
      "step": 10200
    },
    {
      "epoch": 2.069314957438184,
      "grad_norm": 0.006063271313905716,
      "learning_rate": 9.65342521280908e-05,
      "loss": 0.002,
      "step": 10210
    },
    {
      "epoch": 2.0713417105796514,
      "grad_norm": 0.15411558747291565,
      "learning_rate": 9.643291447101744e-05,
      "loss": 0.0025,
      "step": 10220
    },
    {
      "epoch": 2.0733684637211187,
      "grad_norm": 0.006028542760759592,
      "learning_rate": 9.633157681394407e-05,
      "loss": 0.0028,
      "step": 10230
    },
    {
      "epoch": 2.075395216862586,
      "grad_norm": 0.005989754106849432,
      "learning_rate": 9.623023915687069e-05,
      "loss": 0.0017,
      "step": 10240
    },
    {
      "epoch": 2.0774219700040537,
      "grad_norm": 0.006351362448185682,
      "learning_rate": 9.612890149979734e-05,
      "loss": 0.0033,
      "step": 10250
    },
    {
      "epoch": 2.079448723145521,
      "grad_norm": 0.7296304702758789,
      "learning_rate": 9.602756384272396e-05,
      "loss": 0.7193,
      "step": 10260
    },
    {
      "epoch": 2.0814754762869883,
      "grad_norm": 0.010890481062233448,
      "learning_rate": 9.592622618565059e-05,
      "loss": 1.4063,
      "step": 10270
    },
    {
      "epoch": 2.0835022294284555,
      "grad_norm": 0.014627876691520214,
      "learning_rate": 9.582488852857723e-05,
      "loss": 1.5828,
      "step": 10280
    },
    {
      "epoch": 2.085528982569923,
      "grad_norm": 0.012853405438363552,
      "learning_rate": 9.572355087150385e-05,
      "loss": 0.0077,
      "step": 10290
    },
    {
      "epoch": 2.0875557357113905,
      "grad_norm": 0.06252937763929367,
      "learning_rate": 9.562221321443049e-05,
      "loss": 0.0154,
      "step": 10300
    },
    {
      "epoch": 2.0875557357113905,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11084937304258347,
      "eval_runtime": 74.4902,
      "eval_samples_per_second": 7.33,
      "eval_steps_per_second": 0.926,
      "step": 10300
    },
    {
      "epoch": 2.089582488852858,
      "grad_norm": 0.012860951945185661,
      "learning_rate": 9.552087555735712e-05,
      "loss": 0.005,
      "step": 10310
    },
    {
      "epoch": 2.091609241994325,
      "grad_norm": 0.01152266189455986,
      "learning_rate": 9.541953790028375e-05,
      "loss": 0.3453,
      "step": 10320
    },
    {
      "epoch": 2.0936359951357923,
      "grad_norm": 0.00889595691114664,
      "learning_rate": 9.531820024321037e-05,
      "loss": 0.0035,
      "step": 10330
    },
    {
      "epoch": 2.0956627482772596,
      "grad_norm": 0.04616453871130943,
      "learning_rate": 9.521686258613702e-05,
      "loss": 0.0013,
      "step": 10340
    },
    {
      "epoch": 2.0976895014187273,
      "grad_norm": 0.010016735643148422,
      "learning_rate": 9.511552492906365e-05,
      "loss": 0.0084,
      "step": 10350
    },
    {
      "epoch": 2.0997162545601946,
      "grad_norm": 0.17551860213279724,
      "learning_rate": 9.501418727199027e-05,
      "loss": 0.0062,
      "step": 10360
    },
    {
      "epoch": 2.101743007701662,
      "grad_norm": 0.02717851661145687,
      "learning_rate": 9.491284961491692e-05,
      "loss": 0.2781,
      "step": 10370
    },
    {
      "epoch": 2.103769760843129,
      "grad_norm": 0.006723467726260424,
      "learning_rate": 9.481151195784353e-05,
      "loss": 0.0025,
      "step": 10380
    },
    {
      "epoch": 2.105796513984597,
      "grad_norm": 0.006483188830316067,
      "learning_rate": 9.471017430077017e-05,
      "loss": 0.0064,
      "step": 10390
    },
    {
      "epoch": 2.107823267126064,
      "grad_norm": 0.006499973591417074,
      "learning_rate": 9.460883664369681e-05,
      "loss": 0.0007,
      "step": 10400
    },
    {
      "epoch": 2.107823267126064,
      "eval_accuracy": 0.9487179487179487,
      "eval_loss": 0.26751798391342163,
      "eval_runtime": 72.3009,
      "eval_samples_per_second": 7.552,
      "eval_steps_per_second": 0.954,
      "step": 10400
    },
    {
      "epoch": 2.1098500202675314,
      "grad_norm": 0.006514677777886391,
      "learning_rate": 9.450749898662343e-05,
      "loss": 0.0027,
      "step": 10410
    },
    {
      "epoch": 2.1118767734089987,
      "grad_norm": 0.0059182727709412575,
      "learning_rate": 9.440616132955006e-05,
      "loss": 0.0032,
      "step": 10420
    },
    {
      "epoch": 2.113903526550466,
      "grad_norm": 0.09711989760398865,
      "learning_rate": 9.43048236724767e-05,
      "loss": 0.0029,
      "step": 10430
    },
    {
      "epoch": 2.1159302796919337,
      "grad_norm": 0.006037118844687939,
      "learning_rate": 9.420348601540333e-05,
      "loss": 0.0006,
      "step": 10440
    },
    {
      "epoch": 2.117957032833401,
      "grad_norm": 0.005589130334556103,
      "learning_rate": 9.410214835832995e-05,
      "loss": 0.2581,
      "step": 10450
    },
    {
      "epoch": 2.1199837859748683,
      "grad_norm": 0.08527911454439163,
      "learning_rate": 9.40008107012566e-05,
      "loss": 0.0016,
      "step": 10460
    },
    {
      "epoch": 2.1220105391163355,
      "grad_norm": 0.07996934652328491,
      "learning_rate": 9.389947304418323e-05,
      "loss": 0.4869,
      "step": 10470
    },
    {
      "epoch": 2.124037292257803,
      "grad_norm": 0.007227790541946888,
      "learning_rate": 9.379813538710985e-05,
      "loss": 0.4646,
      "step": 10480
    },
    {
      "epoch": 2.1260640453992705,
      "grad_norm": 0.0073805213905870914,
      "learning_rate": 9.36967977300365e-05,
      "loss": 0.0031,
      "step": 10490
    },
    {
      "epoch": 2.128090798540738,
      "grad_norm": 0.00691114692017436,
      "learning_rate": 9.359546007296311e-05,
      "loss": 0.0008,
      "step": 10500
    },
    {
      "epoch": 2.128090798540738,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.1334153413772583,
      "eval_runtime": 77.7167,
      "eval_samples_per_second": 7.026,
      "eval_steps_per_second": 0.888,
      "step": 10500
    },
    {
      "epoch": 2.130117551682205,
      "grad_norm": 0.00719085056334734,
      "learning_rate": 9.349412241588974e-05,
      "loss": 0.0061,
      "step": 10510
    },
    {
      "epoch": 2.1321443048236723,
      "grad_norm": 0.005878064781427383,
      "learning_rate": 9.339278475881638e-05,
      "loss": 0.0079,
      "step": 10520
    },
    {
      "epoch": 2.1341710579651396,
      "grad_norm": 0.005655771121382713,
      "learning_rate": 9.329144710174301e-05,
      "loss": 0.003,
      "step": 10530
    },
    {
      "epoch": 2.1361978111066073,
      "grad_norm": 0.07498996704816818,
      "learning_rate": 9.319010944466964e-05,
      "loss": 0.0046,
      "step": 10540
    },
    {
      "epoch": 2.1382245642480746,
      "grad_norm": 0.005195173434913158,
      "learning_rate": 9.308877178759628e-05,
      "loss": 0.0014,
      "step": 10550
    },
    {
      "epoch": 2.140251317389542,
      "grad_norm": 0.1367482841014862,
      "learning_rate": 9.298743413052291e-05,
      "loss": 0.0019,
      "step": 10560
    },
    {
      "epoch": 2.142278070531009,
      "grad_norm": 0.005297408904880285,
      "learning_rate": 9.288609647344954e-05,
      "loss": 0.0027,
      "step": 10570
    },
    {
      "epoch": 2.144304823672477,
      "grad_norm": 0.006349307019263506,
      "learning_rate": 9.278475881637617e-05,
      "loss": 0.0016,
      "step": 10580
    },
    {
      "epoch": 2.146331576813944,
      "grad_norm": 0.005633942317217588,
      "learning_rate": 9.268342115930279e-05,
      "loss": 0.2652,
      "step": 10590
    },
    {
      "epoch": 2.1483583299554114,
      "grad_norm": 0.05479590594768524,
      "learning_rate": 9.258208350222944e-05,
      "loss": 0.003,
      "step": 10600
    },
    {
      "epoch": 2.1483583299554114,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.15828333795070648,
      "eval_runtime": 76.828,
      "eval_samples_per_second": 7.107,
      "eval_steps_per_second": 0.898,
      "step": 10600
    },
    {
      "epoch": 2.1503850830968787,
      "grad_norm": 0.00457912078127265,
      "learning_rate": 9.248074584515607e-05,
      "loss": 0.003,
      "step": 10610
    },
    {
      "epoch": 2.152411836238346,
      "grad_norm": 0.005090583115816116,
      "learning_rate": 9.237940818808269e-05,
      "loss": 0.7639,
      "step": 10620
    },
    {
      "epoch": 2.1544385893798137,
      "grad_norm": 0.0057753403671085835,
      "learning_rate": 9.227807053100934e-05,
      "loss": 0.0011,
      "step": 10630
    },
    {
      "epoch": 2.156465342521281,
      "grad_norm": 0.006544235162436962,
      "learning_rate": 9.217673287393596e-05,
      "loss": 0.0009,
      "step": 10640
    },
    {
      "epoch": 2.1584920956627482,
      "grad_norm": 0.005472972523421049,
      "learning_rate": 9.207539521686259e-05,
      "loss": 0.0006,
      "step": 10650
    },
    {
      "epoch": 2.1605188488042155,
      "grad_norm": 0.005556443706154823,
      "learning_rate": 9.197405755978922e-05,
      "loss": 0.7453,
      "step": 10660
    },
    {
      "epoch": 2.1625456019456832,
      "grad_norm": 0.007693535648286343,
      "learning_rate": 9.187271990271585e-05,
      "loss": 0.0014,
      "step": 10670
    },
    {
      "epoch": 2.1645723550871505,
      "grad_norm": 0.007847932167351246,
      "learning_rate": 9.177138224564249e-05,
      "loss": 0.0015,
      "step": 10680
    },
    {
      "epoch": 2.166599108228618,
      "grad_norm": 0.016400208696722984,
      "learning_rate": 9.167004458856912e-05,
      "loss": 0.0014,
      "step": 10690
    },
    {
      "epoch": 2.168625861370085,
      "grad_norm": 0.010732585564255714,
      "learning_rate": 9.156870693149575e-05,
      "loss": 0.4043,
      "step": 10700
    },
    {
      "epoch": 2.168625861370085,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.11980610340833664,
      "eval_runtime": 76.3749,
      "eval_samples_per_second": 7.149,
      "eval_steps_per_second": 0.903,
      "step": 10700
    },
    {
      "epoch": 2.1706526145115523,
      "grad_norm": 0.04904749244451523,
      "learning_rate": 9.146736927442237e-05,
      "loss": 0.0034,
      "step": 10710
    },
    {
      "epoch": 2.17267936765302,
      "grad_norm": 873.9728393554688,
      "learning_rate": 9.136603161734902e-05,
      "loss": 0.157,
      "step": 10720
    },
    {
      "epoch": 2.1747061207944873,
      "grad_norm": 0.006011638790369034,
      "learning_rate": 9.126469396027564e-05,
      "loss": 0.0012,
      "step": 10730
    },
    {
      "epoch": 2.1767328739359546,
      "grad_norm": 0.040896765887737274,
      "learning_rate": 9.116335630320227e-05,
      "loss": 0.0016,
      "step": 10740
    },
    {
      "epoch": 2.178759627077422,
      "grad_norm": 0.005535133183002472,
      "learning_rate": 9.106201864612892e-05,
      "loss": 0.0016,
      "step": 10750
    },
    {
      "epoch": 2.180786380218889,
      "grad_norm": 0.0057787708938121796,
      "learning_rate": 9.096068098905553e-05,
      "loss": 0.0018,
      "step": 10760
    },
    {
      "epoch": 2.182813133360357,
      "grad_norm": 0.007440904155373573,
      "learning_rate": 9.085934333198217e-05,
      "loss": 1.3857,
      "step": 10770
    },
    {
      "epoch": 2.184839886501824,
      "grad_norm": 0.008765467442572117,
      "learning_rate": 9.07580056749088e-05,
      "loss": 0.5295,
      "step": 10780
    },
    {
      "epoch": 2.1868666396432914,
      "grad_norm": 0.009373875334858894,
      "learning_rate": 9.065666801783543e-05,
      "loss": 0.005,
      "step": 10790
    },
    {
      "epoch": 2.1888933927847587,
      "grad_norm": 0.007794825825840235,
      "learning_rate": 9.055533036076206e-05,
      "loss": 0.0016,
      "step": 10800
    },
    {
      "epoch": 2.1888933927847587,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.11303602159023285,
      "eval_runtime": 71.211,
      "eval_samples_per_second": 7.667,
      "eval_steps_per_second": 0.969,
      "step": 10800
    },
    {
      "epoch": 2.190920145926226,
      "grad_norm": 0.0074144043028354645,
      "learning_rate": 9.04539927036887e-05,
      "loss": 0.0015,
      "step": 10810
    },
    {
      "epoch": 2.1929468990676937,
      "grad_norm": 0.00861959159374237,
      "learning_rate": 9.035265504661533e-05,
      "loss": 0.0017,
      "step": 10820
    },
    {
      "epoch": 2.194973652209161,
      "grad_norm": 0.008004290983080864,
      "learning_rate": 9.025131738954196e-05,
      "loss": 0.6036,
      "step": 10830
    },
    {
      "epoch": 2.1970004053506282,
      "grad_norm": 0.05160319060087204,
      "learning_rate": 9.01499797324686e-05,
      "loss": 0.0024,
      "step": 10840
    },
    {
      "epoch": 2.1990271584920955,
      "grad_norm": 0.03136659041047096,
      "learning_rate": 9.004864207539521e-05,
      "loss": 0.511,
      "step": 10850
    },
    {
      "epoch": 2.2010539116335632,
      "grad_norm": 0.14971128106117249,
      "learning_rate": 8.994730441832186e-05,
      "loss": 0.9823,
      "step": 10860
    },
    {
      "epoch": 2.2030806647750305,
      "grad_norm": 0.009087011218070984,
      "learning_rate": 8.984596676124848e-05,
      "loss": 0.0022,
      "step": 10870
    },
    {
      "epoch": 2.205107417916498,
      "grad_norm": 0.010233234614133835,
      "learning_rate": 8.974462910417511e-05,
      "loss": 0.005,
      "step": 10880
    },
    {
      "epoch": 2.207134171057965,
      "grad_norm": 0.008126436732709408,
      "learning_rate": 8.964329144710175e-05,
      "loss": 0.0085,
      "step": 10890
    },
    {
      "epoch": 2.2091609241994323,
      "grad_norm": 0.010633064433932304,
      "learning_rate": 8.954195379002838e-05,
      "loss": 0.0033,
      "step": 10900
    },
    {
      "epoch": 2.2091609241994323,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11015180498361588,
      "eval_runtime": 75.0803,
      "eval_samples_per_second": 7.272,
      "eval_steps_per_second": 0.919,
      "step": 10900
    },
    {
      "epoch": 2.2111876773409,
      "grad_norm": 0.008036792278289795,
      "learning_rate": 8.944061613295501e-05,
      "loss": 0.3998,
      "step": 10910
    },
    {
      "epoch": 2.2132144304823673,
      "grad_norm": 0.17050723731517792,
      "learning_rate": 8.933927847588164e-05,
      "loss": 1.1443,
      "step": 10920
    },
    {
      "epoch": 2.2152411836238346,
      "grad_norm": 0.030761348083615303,
      "learning_rate": 8.923794081880828e-05,
      "loss": 0.5031,
      "step": 10930
    },
    {
      "epoch": 2.217267936765302,
      "grad_norm": 0.01277527492493391,
      "learning_rate": 8.91366031617349e-05,
      "loss": 0.0014,
      "step": 10940
    },
    {
      "epoch": 2.2192946899067696,
      "grad_norm": 0.01332076545804739,
      "learning_rate": 8.903526550466154e-05,
      "loss": 0.0056,
      "step": 10950
    },
    {
      "epoch": 2.221321443048237,
      "grad_norm": 0.0114201745018363,
      "learning_rate": 8.893392784758817e-05,
      "loss": 0.0034,
      "step": 10960
    },
    {
      "epoch": 2.223348196189704,
      "grad_norm": 0.01028998102992773,
      "learning_rate": 8.883259019051479e-05,
      "loss": 0.0012,
      "step": 10970
    },
    {
      "epoch": 2.2253749493311714,
      "grad_norm": 0.011788029223680496,
      "learning_rate": 8.873125253344144e-05,
      "loss": 0.8934,
      "step": 10980
    },
    {
      "epoch": 2.2274017024726387,
      "grad_norm": 0.012349162250757217,
      "learning_rate": 8.862991487636806e-05,
      "loss": 0.0175,
      "step": 10990
    },
    {
      "epoch": 2.2294284556141064,
      "grad_norm": 0.026122385635972023,
      "learning_rate": 8.852857721929469e-05,
      "loss": 1.0287,
      "step": 11000
    },
    {
      "epoch": 2.2294284556141064,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.10534922033548355,
      "eval_runtime": 70.5584,
      "eval_samples_per_second": 7.738,
      "eval_steps_per_second": 0.978,
      "step": 11000
    },
    {
      "epoch": 2.2314552087555737,
      "grad_norm": 0.027139287441968918,
      "learning_rate": 8.842723956222134e-05,
      "loss": 0.0108,
      "step": 11010
    },
    {
      "epoch": 2.233481961897041,
      "grad_norm": 0.0179609265178442,
      "learning_rate": 8.832590190514796e-05,
      "loss": 0.3606,
      "step": 11020
    },
    {
      "epoch": 2.2355087150385082,
      "grad_norm": 0.015641842037439346,
      "learning_rate": 8.822456424807459e-05,
      "loss": 0.0066,
      "step": 11030
    },
    {
      "epoch": 2.2375354681799755,
      "grad_norm": 0.013537859544157982,
      "learning_rate": 8.812322659100122e-05,
      "loss": 0.0059,
      "step": 11040
    },
    {
      "epoch": 2.2395622213214432,
      "grad_norm": 0.012414119206368923,
      "learning_rate": 8.802188893392785e-05,
      "loss": 0.0056,
      "step": 11050
    },
    {
      "epoch": 2.2415889744629105,
      "grad_norm": 0.015200314112007618,
      "learning_rate": 8.792055127685447e-05,
      "loss": 0.9405,
      "step": 11060
    },
    {
      "epoch": 2.2436157276043778,
      "grad_norm": 0.020387277007102966,
      "learning_rate": 8.781921361978112e-05,
      "loss": 0.0144,
      "step": 11070
    },
    {
      "epoch": 2.245642480745845,
      "grad_norm": 0.016295185312628746,
      "learning_rate": 8.771787596270775e-05,
      "loss": 0.4575,
      "step": 11080
    },
    {
      "epoch": 2.2476692338873123,
      "grad_norm": 0.405525267124176,
      "learning_rate": 8.761653830563437e-05,
      "loss": 0.0092,
      "step": 11090
    },
    {
      "epoch": 2.24969598702878,
      "grad_norm": 0.016881616786122322,
      "learning_rate": 8.751520064856102e-05,
      "loss": 0.3159,
      "step": 11100
    },
    {
      "epoch": 2.24969598702878,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10043869912624359,
      "eval_runtime": 72.9585,
      "eval_samples_per_second": 7.484,
      "eval_steps_per_second": 0.946,
      "step": 11100
    },
    {
      "epoch": 2.2517227401702473,
      "grad_norm": 0.01376478374004364,
      "learning_rate": 8.741386299148764e-05,
      "loss": 0.0095,
      "step": 11110
    },
    {
      "epoch": 2.2537494933117146,
      "grad_norm": 0.030010223388671875,
      "learning_rate": 8.731252533441427e-05,
      "loss": 0.0126,
      "step": 11120
    },
    {
      "epoch": 2.255776246453182,
      "grad_norm": 0.011137278750538826,
      "learning_rate": 8.72111876773409e-05,
      "loss": 0.0038,
      "step": 11130
    },
    {
      "epoch": 2.2578029995946496,
      "grad_norm": 0.015156904235482216,
      "learning_rate": 8.710985002026753e-05,
      "loss": 0.7007,
      "step": 11140
    },
    {
      "epoch": 2.259829752736117,
      "grad_norm": 0.010496138595044613,
      "learning_rate": 8.700851236319417e-05,
      "loss": 0.003,
      "step": 11150
    },
    {
      "epoch": 2.261856505877584,
      "grad_norm": 0.0970601812005043,
      "learning_rate": 8.69071747061208e-05,
      "loss": 0.0035,
      "step": 11160
    },
    {
      "epoch": 2.2638832590190514,
      "grad_norm": 0.009098607115447521,
      "learning_rate": 8.680583704904743e-05,
      "loss": 0.0043,
      "step": 11170
    },
    {
      "epoch": 2.2659100121605187,
      "grad_norm": 0.008969028480350971,
      "learning_rate": 8.670449939197406e-05,
      "loss": 0.0059,
      "step": 11180
    },
    {
      "epoch": 2.2679367653019864,
      "grad_norm": 0.009530116803944111,
      "learning_rate": 8.66031617349007e-05,
      "loss": 0.0047,
      "step": 11190
    },
    {
      "epoch": 2.2699635184434537,
      "grad_norm": 0.008103832602500916,
      "learning_rate": 8.650182407782732e-05,
      "loss": 0.0464,
      "step": 11200
    },
    {
      "epoch": 2.2699635184434537,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11814778298139572,
      "eval_runtime": 74.7548,
      "eval_samples_per_second": 7.304,
      "eval_steps_per_second": 0.923,
      "step": 11200
    },
    {
      "epoch": 2.271990271584921,
      "grad_norm": 0.008212698623538017,
      "learning_rate": 8.640048642075396e-05,
      "loss": 0.0033,
      "step": 11210
    },
    {
      "epoch": 2.2740170247263882,
      "grad_norm": 0.007732476573437452,
      "learning_rate": 8.62991487636806e-05,
      "loss": 0.7119,
      "step": 11220
    },
    {
      "epoch": 2.276043777867856,
      "grad_norm": 0.009567365050315857,
      "learning_rate": 8.619781110660721e-05,
      "loss": 0.1782,
      "step": 11230
    },
    {
      "epoch": 2.278070531009323,
      "grad_norm": 0.05973416939377785,
      "learning_rate": 8.609647344953386e-05,
      "loss": 0.0023,
      "step": 11240
    },
    {
      "epoch": 2.2800972841507905,
      "grad_norm": 0.009354268200695515,
      "learning_rate": 8.599513579246048e-05,
      "loss": 0.0025,
      "step": 11250
    },
    {
      "epoch": 2.2821240372922578,
      "grad_norm": 0.00866016000509262,
      "learning_rate": 8.589379813538711e-05,
      "loss": 0.001,
      "step": 11260
    },
    {
      "epoch": 2.284150790433725,
      "grad_norm": 0.008696874603629112,
      "learning_rate": 8.579246047831375e-05,
      "loss": 0.0033,
      "step": 11270
    },
    {
      "epoch": 2.2861775435751923,
      "grad_norm": 0.00813909713178873,
      "learning_rate": 8.569112282124038e-05,
      "loss": 0.002,
      "step": 11280
    },
    {
      "epoch": 2.28820429671666,
      "grad_norm": 0.00973751861602068,
      "learning_rate": 8.558978516416701e-05,
      "loss": 0.0014,
      "step": 11290
    },
    {
      "epoch": 2.2902310498581273,
      "grad_norm": 0.007876581512391567,
      "learning_rate": 8.548844750709364e-05,
      "loss": 0.002,
      "step": 11300
    },
    {
      "epoch": 2.2902310498581273,
      "eval_accuracy": 0.9560439560439561,
      "eval_loss": 0.2652477025985718,
      "eval_runtime": 76.3473,
      "eval_samples_per_second": 7.152,
      "eval_steps_per_second": 0.904,
      "step": 11300
    },
    {
      "epoch": 2.2922578029995946,
      "grad_norm": 0.00753940362483263,
      "learning_rate": 8.538710985002028e-05,
      "loss": 0.0013,
      "step": 11310
    },
    {
      "epoch": 2.294284556141062,
      "grad_norm": 0.007245345041155815,
      "learning_rate": 8.52857721929469e-05,
      "loss": 0.0008,
      "step": 11320
    },
    {
      "epoch": 2.2963113092825296,
      "grad_norm": 0.00759044848382473,
      "learning_rate": 8.518443453587354e-05,
      "loss": 0.5261,
      "step": 11330
    },
    {
      "epoch": 2.298338062423997,
      "grad_norm": 0.0077063278295099735,
      "learning_rate": 8.508309687880016e-05,
      "loss": 0.0019,
      "step": 11340
    },
    {
      "epoch": 2.300364815565464,
      "grad_norm": 0.007679214235395193,
      "learning_rate": 8.498175922172679e-05,
      "loss": 0.001,
      "step": 11350
    },
    {
      "epoch": 2.3023915687069314,
      "grad_norm": 0.0077296835370361805,
      "learning_rate": 8.488042156465344e-05,
      "loss": 0.0015,
      "step": 11360
    },
    {
      "epoch": 2.3044183218483987,
      "grad_norm": 0.007177960593253374,
      "learning_rate": 8.477908390758006e-05,
      "loss": 0.0019,
      "step": 11370
    },
    {
      "epoch": 2.3064450749898664,
      "grad_norm": 0.006858570966869593,
      "learning_rate": 8.467774625050669e-05,
      "loss": 0.002,
      "step": 11380
    },
    {
      "epoch": 2.3084718281313337,
      "grad_norm": 0.03824995085597038,
      "learning_rate": 8.457640859343332e-05,
      "loss": 0.0011,
      "step": 11390
    },
    {
      "epoch": 2.310498581272801,
      "grad_norm": 0.007602210622280836,
      "learning_rate": 8.447507093635996e-05,
      "loss": 0.0758,
      "step": 11400
    },
    {
      "epoch": 2.310498581272801,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.141328364610672,
      "eval_runtime": 73.9667,
      "eval_samples_per_second": 7.382,
      "eval_steps_per_second": 0.933,
      "step": 11400
    },
    {
      "epoch": 2.312525334414268,
      "grad_norm": 0.008061299100518227,
      "learning_rate": 8.437373327928659e-05,
      "loss": 0.0011,
      "step": 11410
    },
    {
      "epoch": 2.314552087555736,
      "grad_norm": 0.24199163913726807,
      "learning_rate": 8.427239562221322e-05,
      "loss": 0.0025,
      "step": 11420
    },
    {
      "epoch": 2.316578840697203,
      "grad_norm": 13.275392532348633,
      "learning_rate": 8.417105796513985e-05,
      "loss": 1.0208,
      "step": 11430
    },
    {
      "epoch": 2.3186055938386705,
      "grad_norm": 0.006914615165442228,
      "learning_rate": 8.406972030806647e-05,
      "loss": 0.0017,
      "step": 11440
    },
    {
      "epoch": 2.3206323469801378,
      "grad_norm": 0.009104193188250065,
      "learning_rate": 8.396838265099312e-05,
      "loss": 0.2027,
      "step": 11450
    },
    {
      "epoch": 2.322659100121605,
      "grad_norm": 0.007759319618344307,
      "learning_rate": 8.386704499391974e-05,
      "loss": 0.2647,
      "step": 11460
    },
    {
      "epoch": 2.3246858532630723,
      "grad_norm": 0.04744374752044678,
      "learning_rate": 8.376570733684637e-05,
      "loss": 0.0013,
      "step": 11470
    },
    {
      "epoch": 2.32671260640454,
      "grad_norm": 0.08690156787633896,
      "learning_rate": 8.3664369679773e-05,
      "loss": 0.476,
      "step": 11480
    },
    {
      "epoch": 2.3287393595460073,
      "grad_norm": 0.007856318727135658,
      "learning_rate": 8.356303202269964e-05,
      "loss": 0.5006,
      "step": 11490
    },
    {
      "epoch": 2.3307661126874746,
      "grad_norm": 0.008044347167015076,
      "learning_rate": 8.346169436562627e-05,
      "loss": 0.0027,
      "step": 11500
    },
    {
      "epoch": 2.3307661126874746,
      "eval_accuracy": 0.945054945054945,
      "eval_loss": 0.2024707794189453,
      "eval_runtime": 213.0607,
      "eval_samples_per_second": 2.563,
      "eval_steps_per_second": 0.324,
      "step": 11500
    },
    {
      "epoch": 2.332792865828942,
      "grad_norm": 0.011505112051963806,
      "learning_rate": 8.33603567085529e-05,
      "loss": 0.1192,
      "step": 11510
    },
    {
      "epoch": 2.3348196189704096,
      "grad_norm": 0.007121453061699867,
      "learning_rate": 8.325901905147953e-05,
      "loss": 0.0023,
      "step": 11520
    },
    {
      "epoch": 2.336846372111877,
      "grad_norm": 0.007184705231338739,
      "learning_rate": 8.315768139440617e-05,
      "loss": 0.0047,
      "step": 11530
    },
    {
      "epoch": 2.338873125253344,
      "grad_norm": 0.007320650387555361,
      "learning_rate": 8.30563437373328e-05,
      "loss": 0.4688,
      "step": 11540
    },
    {
      "epoch": 2.3408998783948114,
      "grad_norm": 0.0069977049715816975,
      "learning_rate": 8.295500608025942e-05,
      "loss": 0.0026,
      "step": 11550
    },
    {
      "epoch": 2.3429266315362787,
      "grad_norm": 0.006968265399336815,
      "learning_rate": 8.285366842318606e-05,
      "loss": 0.0027,
      "step": 11560
    },
    {
      "epoch": 2.3449533846777464,
      "grad_norm": 0.007060995325446129,
      "learning_rate": 8.27523307661127e-05,
      "loss": 0.0021,
      "step": 11570
    },
    {
      "epoch": 2.3469801378192137,
      "grad_norm": 0.006562945432960987,
      "learning_rate": 8.265099310903932e-05,
      "loss": 0.0038,
      "step": 11580
    },
    {
      "epoch": 2.349006890960681,
      "grad_norm": 0.00626459950581193,
      "learning_rate": 8.254965545196596e-05,
      "loss": 0.0037,
      "step": 11590
    },
    {
      "epoch": 2.351033644102148,
      "grad_norm": 0.006127441301941872,
      "learning_rate": 8.244831779489258e-05,
      "loss": 0.0011,
      "step": 11600
    },
    {
      "epoch": 2.351033644102148,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.13722333312034607,
      "eval_runtime": 211.759,
      "eval_samples_per_second": 2.578,
      "eval_steps_per_second": 0.326,
      "step": 11600
    },
    {
      "epoch": 2.353060397243616,
      "grad_norm": 6.2088303565979,
      "learning_rate": 8.234698013781921e-05,
      "loss": 0.0371,
      "step": 11610
    },
    {
      "epoch": 2.355087150385083,
      "grad_norm": 0.005822502542287111,
      "learning_rate": 8.224564248074585e-05,
      "loss": 0.0017,
      "step": 11620
    },
    {
      "epoch": 2.3571139035265505,
      "grad_norm": 0.005627756007015705,
      "learning_rate": 8.214430482367248e-05,
      "loss": 0.3425,
      "step": 11630
    },
    {
      "epoch": 2.3591406566680178,
      "grad_norm": 0.005736709106713533,
      "learning_rate": 8.204296716659911e-05,
      "loss": 0.0006,
      "step": 11640
    },
    {
      "epoch": 2.361167409809485,
      "grad_norm": 0.04625217244029045,
      "learning_rate": 8.194162950952575e-05,
      "loss": 0.0028,
      "step": 11650
    },
    {
      "epoch": 2.3631941629509527,
      "grad_norm": 0.005438953638076782,
      "learning_rate": 8.184029185245238e-05,
      "loss": 0.0018,
      "step": 11660
    },
    {
      "epoch": 2.36522091609242,
      "grad_norm": 0.005321810021996498,
      "learning_rate": 8.1738954195379e-05,
      "loss": 0.001,
      "step": 11670
    },
    {
      "epoch": 2.3672476692338873,
      "grad_norm": 0.0051677110604941845,
      "learning_rate": 8.163761653830564e-05,
      "loss": 0.001,
      "step": 11680
    },
    {
      "epoch": 2.3692744223753546,
      "grad_norm": 0.03902814909815788,
      "learning_rate": 8.153627888123228e-05,
      "loss": 0.3135,
      "step": 11690
    },
    {
      "epoch": 2.3713011755168223,
      "grad_norm": 0.005199107341468334,
      "learning_rate": 8.14349412241589e-05,
      "loss": 0.0009,
      "step": 11700
    },
    {
      "epoch": 2.3713011755168223,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.1458211988210678,
      "eval_runtime": 213.7889,
      "eval_samples_per_second": 2.554,
      "eval_steps_per_second": 0.323,
      "step": 11700
    },
    {
      "epoch": 2.3733279286582896,
      "grad_norm": 0.005133099388331175,
      "learning_rate": 8.133360356708554e-05,
      "loss": 0.0018,
      "step": 11710
    },
    {
      "epoch": 2.375354681799757,
      "grad_norm": 0.03676588460803032,
      "learning_rate": 8.123226591001216e-05,
      "loss": 0.0013,
      "step": 11720
    },
    {
      "epoch": 2.377381434941224,
      "grad_norm": 0.03504160791635513,
      "learning_rate": 8.113092825293879e-05,
      "loss": 0.002,
      "step": 11730
    },
    {
      "epoch": 2.3794081880826914,
      "grad_norm": 0.005366514436900616,
      "learning_rate": 8.102959059586543e-05,
      "loss": 0.7592,
      "step": 11740
    },
    {
      "epoch": 2.3814349412241587,
      "grad_norm": 0.03350529819726944,
      "learning_rate": 8.092825293879206e-05,
      "loss": 0.0017,
      "step": 11750
    },
    {
      "epoch": 2.3834616943656264,
      "grad_norm": 0.006684424821287394,
      "learning_rate": 8.082691528171869e-05,
      "loss": 0.5015,
      "step": 11760
    },
    {
      "epoch": 2.3854884475070937,
      "grad_norm": 0.006355280987918377,
      "learning_rate": 8.072557762464532e-05,
      "loss": 0.0642,
      "step": 11770
    },
    {
      "epoch": 2.387515200648561,
      "grad_norm": 0.006372395902872086,
      "learning_rate": 8.062423996757196e-05,
      "loss": 0.9822,
      "step": 11780
    },
    {
      "epoch": 2.389541953790028,
      "grad_norm": 0.006969653069972992,
      "learning_rate": 8.052290231049859e-05,
      "loss": 0.0021,
      "step": 11790
    },
    {
      "epoch": 2.391568706931496,
      "grad_norm": 0.006968590896576643,
      "learning_rate": 8.042156465342522e-05,
      "loss": 0.4178,
      "step": 11800
    },
    {
      "epoch": 2.391568706931496,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.1403357833623886,
      "eval_runtime": 44.9689,
      "eval_samples_per_second": 12.142,
      "eval_steps_per_second": 1.534,
      "step": 11800
    },
    {
      "epoch": 2.393595460072963,
      "grad_norm": 0.00706483842805028,
      "learning_rate": 8.032022699635184e-05,
      "loss": 0.0016,
      "step": 11810
    },
    {
      "epoch": 2.3956222132144305,
      "grad_norm": 0.09024324268102646,
      "learning_rate": 8.021888933927849e-05,
      "loss": 0.4716,
      "step": 11820
    },
    {
      "epoch": 2.3976489663558977,
      "grad_norm": 0.007200425956398249,
      "learning_rate": 8.011755168220512e-05,
      "loss": 0.0019,
      "step": 11830
    },
    {
      "epoch": 2.399675719497365,
      "grad_norm": 0.007590134162455797,
      "learning_rate": 8.001621402513174e-05,
      "loss": 0.2847,
      "step": 11840
    },
    {
      "epoch": 2.4017024726388327,
      "grad_norm": 0.007439048029482365,
      "learning_rate": 7.991487636805838e-05,
      "loss": 0.0008,
      "step": 11850
    },
    {
      "epoch": 2.4037292257803,
      "grad_norm": 0.1115642786026001,
      "learning_rate": 7.9813538710985e-05,
      "loss": 0.0023,
      "step": 11860
    },
    {
      "epoch": 2.4057559789217673,
      "grad_norm": 0.007191861514002085,
      "learning_rate": 7.971220105391164e-05,
      "loss": 0.0034,
      "step": 11870
    },
    {
      "epoch": 2.4077827320632346,
      "grad_norm": 0.006929817609488964,
      "learning_rate": 7.961086339683827e-05,
      "loss": 0.0021,
      "step": 11880
    },
    {
      "epoch": 2.4098094852047023,
      "grad_norm": 0.006752875633537769,
      "learning_rate": 7.95095257397649e-05,
      "loss": 0.0239,
      "step": 11890
    },
    {
      "epoch": 2.4118362383461696,
      "grad_norm": 0.006513989996165037,
      "learning_rate": 7.940818808269153e-05,
      "loss": 0.0028,
      "step": 11900
    },
    {
      "epoch": 2.4118362383461696,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.14057356119155884,
      "eval_runtime": 48.5905,
      "eval_samples_per_second": 11.237,
      "eval_steps_per_second": 1.42,
      "step": 11900
    },
    {
      "epoch": 2.413862991487637,
      "grad_norm": 0.006521328818053007,
      "learning_rate": 7.930685042561817e-05,
      "loss": 0.0024,
      "step": 11910
    },
    {
      "epoch": 2.415889744629104,
      "grad_norm": 0.006255956832319498,
      "learning_rate": 7.92055127685448e-05,
      "loss": 0.0007,
      "step": 11920
    },
    {
      "epoch": 2.4179164977705714,
      "grad_norm": 0.08518156409263611,
      "learning_rate": 7.910417511147142e-05,
      "loss": 0.4114,
      "step": 11930
    },
    {
      "epoch": 2.419943250912039,
      "grad_norm": 0.006490825209766626,
      "learning_rate": 7.900283745439806e-05,
      "loss": 0.0018,
      "step": 11940
    },
    {
      "epoch": 2.4219700040535064,
      "grad_norm": 0.09310487657785416,
      "learning_rate": 7.890149979732468e-05,
      "loss": 0.0018,
      "step": 11950
    },
    {
      "epoch": 2.4239967571949737,
      "grad_norm": 0.006273780949413776,
      "learning_rate": 7.880016214025132e-05,
      "loss": 0.2179,
      "step": 11960
    },
    {
      "epoch": 2.426023510336441,
      "grad_norm": 0.00689974520355463,
      "learning_rate": 7.869882448317796e-05,
      "loss": 0.7144,
      "step": 11970
    },
    {
      "epoch": 2.428050263477908,
      "grad_norm": 0.22196179628372192,
      "learning_rate": 7.859748682610458e-05,
      "loss": 0.0045,
      "step": 11980
    },
    {
      "epoch": 2.430077016619376,
      "grad_norm": 0.007610929664224386,
      "learning_rate": 7.849614916903121e-05,
      "loss": 0.0233,
      "step": 11990
    },
    {
      "epoch": 2.432103769760843,
      "grad_norm": 0.006953901145607233,
      "learning_rate": 7.839481151195785e-05,
      "loss": 0.0009,
      "step": 12000
    },
    {
      "epoch": 2.432103769760843,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.12951794266700745,
      "eval_runtime": 47.4956,
      "eval_samples_per_second": 11.496,
      "eval_steps_per_second": 1.453,
      "step": 12000
    },
    {
      "epoch": 2.4341305229023105,
      "grad_norm": 0.006798309274017811,
      "learning_rate": 7.829347385488448e-05,
      "loss": 0.0008,
      "step": 12010
    },
    {
      "epoch": 2.4361572760437777,
      "grad_norm": 0.006593131460249424,
      "learning_rate": 7.81921361978111e-05,
      "loss": 0.0009,
      "step": 12020
    },
    {
      "epoch": 2.438184029185245,
      "grad_norm": 0.008275761269032955,
      "learning_rate": 7.809079854073775e-05,
      "loss": 0.0022,
      "step": 12030
    },
    {
      "epoch": 2.4402107823267127,
      "grad_norm": 0.0106047997251153,
      "learning_rate": 7.798946088366438e-05,
      "loss": 0.0701,
      "step": 12040
    },
    {
      "epoch": 2.44223753546818,
      "grad_norm": 0.006123432423919439,
      "learning_rate": 7.7888123226591e-05,
      "loss": 0.7337,
      "step": 12050
    },
    {
      "epoch": 2.4442642886096473,
      "grad_norm": 0.01519776787608862,
      "learning_rate": 7.778678556951764e-05,
      "loss": 0.0023,
      "step": 12060
    },
    {
      "epoch": 2.4462910417511146,
      "grad_norm": 0.007234047167003155,
      "learning_rate": 7.768544791244426e-05,
      "loss": 0.0016,
      "step": 12070
    },
    {
      "epoch": 2.4483177948925823,
      "grad_norm": 0.00756498659029603,
      "learning_rate": 7.75841102553709e-05,
      "loss": 0.0022,
      "step": 12080
    },
    {
      "epoch": 2.4503445480340496,
      "grad_norm": 0.006812751293182373,
      "learning_rate": 7.748277259829753e-05,
      "loss": 0.0015,
      "step": 12090
    },
    {
      "epoch": 2.452371301175517,
      "grad_norm": 0.06002921611070633,
      "learning_rate": 7.738143494122416e-05,
      "loss": 0.002,
      "step": 12100
    },
    {
      "epoch": 2.452371301175517,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.16850095987319946,
      "eval_runtime": 48.7693,
      "eval_samples_per_second": 11.196,
      "eval_steps_per_second": 1.415,
      "step": 12100
    },
    {
      "epoch": 2.454398054316984,
      "grad_norm": 0.0065709855407476425,
      "learning_rate": 7.728009728415079e-05,
      "loss": 0.506,
      "step": 12110
    },
    {
      "epoch": 2.4564248074584514,
      "grad_norm": 0.06784232705831528,
      "learning_rate": 7.717875962707743e-05,
      "loss": 0.6716,
      "step": 12120
    },
    {
      "epoch": 2.458451560599919,
      "grad_norm": 0.05798913538455963,
      "learning_rate": 7.707742197000406e-05,
      "loss": 0.0025,
      "step": 12130
    },
    {
      "epoch": 2.4604783137413864,
      "grad_norm": 0.01616346277296543,
      "learning_rate": 7.697608431293069e-05,
      "loss": 0.0051,
      "step": 12140
    },
    {
      "epoch": 2.4625050668828536,
      "grad_norm": 0.01163223572075367,
      "learning_rate": 7.687474665585732e-05,
      "loss": 0.6955,
      "step": 12150
    },
    {
      "epoch": 2.464531820024321,
      "grad_norm": 0.1036323606967926,
      "learning_rate": 7.677340899878394e-05,
      "loss": 0.5214,
      "step": 12160
    },
    {
      "epoch": 2.4665585731657886,
      "grad_norm": 8.167608261108398,
      "learning_rate": 7.667207134171059e-05,
      "loss": 0.4512,
      "step": 12170
    },
    {
      "epoch": 2.468585326307256,
      "grad_norm": 0.02810136415064335,
      "learning_rate": 7.657073368463722e-05,
      "loss": 0.0084,
      "step": 12180
    },
    {
      "epoch": 2.470612079448723,
      "grad_norm": 0.015877937898039818,
      "learning_rate": 7.646939602756384e-05,
      "loss": 0.1126,
      "step": 12190
    },
    {
      "epoch": 2.4726388325901905,
      "grad_norm": 0.01663248986005783,
      "learning_rate": 7.636805837049049e-05,
      "loss": 0.0022,
      "step": 12200
    },
    {
      "epoch": 2.4726388325901905,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.11507423967123032,
      "eval_runtime": 50.3537,
      "eval_samples_per_second": 10.843,
      "eval_steps_per_second": 1.37,
      "step": 12200
    },
    {
      "epoch": 2.4746655857316577,
      "grad_norm": 0.011700361035764217,
      "learning_rate": 7.62667207134171e-05,
      "loss": 0.0029,
      "step": 12210
    },
    {
      "epoch": 2.476692338873125,
      "grad_norm": 0.011930271051824093,
      "learning_rate": 7.616538305634374e-05,
      "loss": 0.0015,
      "step": 12220
    },
    {
      "epoch": 2.4787190920145927,
      "grad_norm": 0.009713798761367798,
      "learning_rate": 7.606404539927037e-05,
      "loss": 0.002,
      "step": 12230
    },
    {
      "epoch": 2.48074584515606,
      "grad_norm": 0.05965464562177658,
      "learning_rate": 7.5962707742197e-05,
      "loss": 0.0048,
      "step": 12240
    },
    {
      "epoch": 2.4827725982975273,
      "grad_norm": 0.007966786623001099,
      "learning_rate": 7.586137008512364e-05,
      "loss": 0.0019,
      "step": 12250
    },
    {
      "epoch": 2.4847993514389946,
      "grad_norm": 0.055097125470638275,
      "learning_rate": 7.576003242805027e-05,
      "loss": 0.0015,
      "step": 12260
    },
    {
      "epoch": 2.4868261045804623,
      "grad_norm": 0.012019754387438297,
      "learning_rate": 7.56586947709769e-05,
      "loss": 0.0015,
      "step": 12270
    },
    {
      "epoch": 2.4888528577219295,
      "grad_norm": 0.051623955368995667,
      "learning_rate": 7.555735711390352e-05,
      "loss": 0.0032,
      "step": 12280
    },
    {
      "epoch": 2.490879610863397,
      "grad_norm": 0.0482669472694397,
      "learning_rate": 7.545601945683017e-05,
      "loss": 0.0018,
      "step": 12290
    },
    {
      "epoch": 2.492906364004864,
      "grad_norm": 0.0074604470282793045,
      "learning_rate": 7.535468179975679e-05,
      "loss": 0.0008,
      "step": 12300
    },
    {
      "epoch": 2.492906364004864,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.16353316605091095,
      "eval_runtime": 47.0606,
      "eval_samples_per_second": 11.602,
      "eval_steps_per_second": 1.466,
      "step": 12300
    },
    {
      "epoch": 2.4949331171463314,
      "grad_norm": 0.007159892935305834,
      "learning_rate": 7.525334414268342e-05,
      "loss": 0.724,
      "step": 12310
    },
    {
      "epoch": 2.496959870287799,
      "grad_norm": 0.007703261915594339,
      "learning_rate": 7.515200648561006e-05,
      "loss": 0.0439,
      "step": 12320
    },
    {
      "epoch": 2.4989866234292664,
      "grad_norm": 0.007789842318743467,
      "learning_rate": 7.505066882853668e-05,
      "loss": 0.0026,
      "step": 12330
    },
    {
      "epoch": 2.5010133765707336,
      "grad_norm": 0.008039126172661781,
      "learning_rate": 7.494933117146332e-05,
      "loss": 0.0014,
      "step": 12340
    },
    {
      "epoch": 2.503040129712201,
      "grad_norm": 0.014271967113018036,
      "learning_rate": 7.484799351438995e-05,
      "loss": 0.7042,
      "step": 12350
    },
    {
      "epoch": 2.5050668828536686,
      "grad_norm": 0.01031290553510189,
      "learning_rate": 7.474665585731658e-05,
      "loss": 0.0018,
      "step": 12360
    },
    {
      "epoch": 2.507093635995136,
      "grad_norm": 0.013794993981719017,
      "learning_rate": 7.464531820024321e-05,
      "loss": 0.6638,
      "step": 12370
    },
    {
      "epoch": 2.509120389136603,
      "grad_norm": 0.013239707797765732,
      "learning_rate": 7.454398054316985e-05,
      "loss": 0.0018,
      "step": 12380
    },
    {
      "epoch": 2.5111471422780705,
      "grad_norm": 0.04807324707508087,
      "learning_rate": 7.444264288609648e-05,
      "loss": 0.4963,
      "step": 12390
    },
    {
      "epoch": 2.5131738954195377,
      "grad_norm": 0.016311677172780037,
      "learning_rate": 7.434130522902311e-05,
      "loss": 0.0035,
      "step": 12400
    },
    {
      "epoch": 2.5131738954195377,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.12831450998783112,
      "eval_runtime": 50.0658,
      "eval_samples_per_second": 10.906,
      "eval_steps_per_second": 1.378,
      "step": 12400
    },
    {
      "epoch": 2.515200648561005,
      "grad_norm": 0.025094080716371536,
      "learning_rate": 7.423996757194975e-05,
      "loss": 0.4348,
      "step": 12410
    },
    {
      "epoch": 2.5172274017024727,
      "grad_norm": 0.013695972040295601,
      "learning_rate": 7.413862991487636e-05,
      "loss": 0.0023,
      "step": 12420
    },
    {
      "epoch": 2.51925415484394,
      "grad_norm": 0.011631296947598457,
      "learning_rate": 7.403729225780301e-05,
      "loss": 0.0025,
      "step": 12430
    },
    {
      "epoch": 2.5212809079854073,
      "grad_norm": 0.04988066852092743,
      "learning_rate": 7.393595460072964e-05,
      "loss": 0.0041,
      "step": 12440
    },
    {
      "epoch": 2.523307661126875,
      "grad_norm": 0.010438222438097,
      "learning_rate": 7.383461694365626e-05,
      "loss": 0.0022,
      "step": 12450
    },
    {
      "epoch": 2.5253344142683423,
      "grad_norm": 0.06062883883714676,
      "learning_rate": 7.37332792865829e-05,
      "loss": 0.0017,
      "step": 12460
    },
    {
      "epoch": 2.5273611674098095,
      "grad_norm": 0.011333596892654896,
      "learning_rate": 7.363194162950953e-05,
      "loss": 0.6328,
      "step": 12470
    },
    {
      "epoch": 2.529387920551277,
      "grad_norm": 0.012408534996211529,
      "learning_rate": 7.353060397243616e-05,
      "loss": 0.0014,
      "step": 12480
    },
    {
      "epoch": 2.531414673692744,
      "grad_norm": 0.013357922434806824,
      "learning_rate": 7.342926631536279e-05,
      "loss": 0.5256,
      "step": 12490
    },
    {
      "epoch": 2.5334414268342114,
      "grad_norm": 0.0201270692050457,
      "learning_rate": 7.332792865828943e-05,
      "loss": 0.7689,
      "step": 12500
    },
    {
      "epoch": 2.5334414268342114,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.15511265397071838,
      "eval_runtime": 52.8918,
      "eval_samples_per_second": 10.323,
      "eval_steps_per_second": 1.305,
      "step": 12500
    },
    {
      "epoch": 2.535468179975679,
      "grad_norm": 8.252181053161621,
      "learning_rate": 7.322659100121606e-05,
      "loss": 0.5807,
      "step": 12510
    },
    {
      "epoch": 2.5374949331171464,
      "grad_norm": 0.08722558617591858,
      "learning_rate": 7.312525334414269e-05,
      "loss": 0.0678,
      "step": 12520
    },
    {
      "epoch": 2.5395216862586136,
      "grad_norm": 0.024868663400411606,
      "learning_rate": 7.302391568706932e-05,
      "loss": 0.0043,
      "step": 12530
    },
    {
      "epoch": 2.541548439400081,
      "grad_norm": 0.07840926200151443,
      "learning_rate": 7.292257802999594e-05,
      "loss": 0.0038,
      "step": 12540
    },
    {
      "epoch": 2.5435751925415486,
      "grad_norm": 0.017033569514751434,
      "learning_rate": 7.282124037292259e-05,
      "loss": 0.0487,
      "step": 12550
    },
    {
      "epoch": 2.545601945683016,
      "grad_norm": 0.01854672282934189,
      "learning_rate": 7.271990271584921e-05,
      "loss": 0.0032,
      "step": 12560
    },
    {
      "epoch": 2.547628698824483,
      "grad_norm": 0.011858698911964893,
      "learning_rate": 7.261856505877584e-05,
      "loss": 0.2406,
      "step": 12570
    },
    {
      "epoch": 2.5496554519659504,
      "grad_norm": 0.016108671203255653,
      "learning_rate": 7.251722740170249e-05,
      "loss": 0.0027,
      "step": 12580
    },
    {
      "epoch": 2.5516822051074177,
      "grad_norm": 8.133279800415039,
      "learning_rate": 7.24158897446291e-05,
      "loss": 0.4846,
      "step": 12590
    },
    {
      "epoch": 2.5537089582488854,
      "grad_norm": 0.010792690329253674,
      "learning_rate": 7.231455208755574e-05,
      "loss": 0.0126,
      "step": 12600
    },
    {
      "epoch": 2.5537089582488854,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11439267545938492,
      "eval_runtime": 137.7708,
      "eval_samples_per_second": 3.963,
      "eval_steps_per_second": 0.501,
      "step": 12600
    },
    {
      "epoch": 2.5557357113903527,
      "grad_norm": 0.009957440197467804,
      "learning_rate": 7.221321443048237e-05,
      "loss": 0.0145,
      "step": 12610
    },
    {
      "epoch": 2.55776246453182,
      "grad_norm": 0.009772456251084805,
      "learning_rate": 7.2111876773409e-05,
      "loss": 0.4205,
      "step": 12620
    },
    {
      "epoch": 2.5597892176732873,
      "grad_norm": 0.011758706532418728,
      "learning_rate": 7.201053911633562e-05,
      "loss": 0.0012,
      "step": 12630
    },
    {
      "epoch": 2.561815970814755,
      "grad_norm": 0.00956796109676361,
      "learning_rate": 7.190920145926227e-05,
      "loss": 0.0021,
      "step": 12640
    },
    {
      "epoch": 2.5638427239562223,
      "grad_norm": 0.009638584218919277,
      "learning_rate": 7.18078638021889e-05,
      "loss": 0.0011,
      "step": 12650
    },
    {
      "epoch": 2.5658694770976895,
      "grad_norm": 0.009001140482723713,
      "learning_rate": 7.170652614511552e-05,
      "loss": 0.0036,
      "step": 12660
    },
    {
      "epoch": 2.567896230239157,
      "grad_norm": 0.009449226781725883,
      "learning_rate": 7.160518848804217e-05,
      "loss": 0.0062,
      "step": 12670
    },
    {
      "epoch": 2.569922983380624,
      "grad_norm": 0.008211339823901653,
      "learning_rate": 7.150385083096879e-05,
      "loss": 0.0026,
      "step": 12680
    },
    {
      "epoch": 2.5719497365220914,
      "grad_norm": 0.008610163815319538,
      "learning_rate": 7.140251317389542e-05,
      "loss": 0.0018,
      "step": 12690
    },
    {
      "epoch": 2.573976489663559,
      "grad_norm": 0.00778192188590765,
      "learning_rate": 7.130117551682205e-05,
      "loss": 0.0028,
      "step": 12700
    },
    {
      "epoch": 2.573976489663559,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.09194227308034897,
      "eval_runtime": 198.1506,
      "eval_samples_per_second": 2.755,
      "eval_steps_per_second": 0.348,
      "step": 12700
    },
    {
      "epoch": 2.5760032428050263,
      "grad_norm": 0.055679794400930405,
      "learning_rate": 7.119983785974868e-05,
      "loss": 0.0034,
      "step": 12710
    },
    {
      "epoch": 2.5780299959464936,
      "grad_norm": 0.015648189932107925,
      "learning_rate": 7.109850020267532e-05,
      "loss": 1.0479,
      "step": 12720
    },
    {
      "epoch": 2.5800567490879613,
      "grad_norm": 0.012088973075151443,
      "learning_rate": 7.099716254560195e-05,
      "loss": 0.4879,
      "step": 12730
    },
    {
      "epoch": 2.5820835022294286,
      "grad_norm": 0.00957775954157114,
      "learning_rate": 7.089582488852858e-05,
      "loss": 0.0027,
      "step": 12740
    },
    {
      "epoch": 2.584110255370896,
      "grad_norm": 0.011500762775540352,
      "learning_rate": 7.079448723145521e-05,
      "loss": 0.2668,
      "step": 12750
    },
    {
      "epoch": 2.586137008512363,
      "grad_norm": 0.014193236827850342,
      "learning_rate": 7.069314957438185e-05,
      "loss": 0.0033,
      "step": 12760
    },
    {
      "epoch": 2.5881637616538304,
      "grad_norm": 0.010394740849733353,
      "learning_rate": 7.059181191730847e-05,
      "loss": 0.004,
      "step": 12770
    },
    {
      "epoch": 2.5901905147952977,
      "grad_norm": 0.013128154911100864,
      "learning_rate": 7.049047426023511e-05,
      "loss": 0.0064,
      "step": 12780
    },
    {
      "epoch": 2.5922172679367654,
      "grad_norm": 0.009156018495559692,
      "learning_rate": 7.038913660316175e-05,
      "loss": 0.4746,
      "step": 12790
    },
    {
      "epoch": 2.5942440210782327,
      "grad_norm": 1.7480326890945435,
      "learning_rate": 7.028779894608836e-05,
      "loss": 0.0053,
      "step": 12800
    },
    {
      "epoch": 2.5942440210782327,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11324365437030792,
      "eval_runtime": 165.9605,
      "eval_samples_per_second": 3.29,
      "eval_steps_per_second": 0.416,
      "step": 12800
    },
    {
      "epoch": 2.5962707742197,
      "grad_norm": 0.010218513198196888,
      "learning_rate": 7.018646128901501e-05,
      "loss": 0.005,
      "step": 12810
    },
    {
      "epoch": 2.5982975273611673,
      "grad_norm": 0.008802374824881554,
      "learning_rate": 7.008512363194163e-05,
      "loss": 0.0019,
      "step": 12820
    },
    {
      "epoch": 2.600324280502635,
      "grad_norm": 0.07323005795478821,
      "learning_rate": 6.998378597486826e-05,
      "loss": 0.002,
      "step": 12830
    },
    {
      "epoch": 2.6023510336441023,
      "grad_norm": 0.010269319638609886,
      "learning_rate": 6.98824483177949e-05,
      "loss": 0.0011,
      "step": 12840
    },
    {
      "epoch": 2.6043777867855695,
      "grad_norm": 0.08123511075973511,
      "learning_rate": 6.978111066072153e-05,
      "loss": 0.4715,
      "step": 12850
    },
    {
      "epoch": 2.606404539927037,
      "grad_norm": 0.010554922744631767,
      "learning_rate": 6.967977300364816e-05,
      "loss": 0.003,
      "step": 12860
    },
    {
      "epoch": 2.608431293068504,
      "grad_norm": 0.008621037937700748,
      "learning_rate": 6.957843534657479e-05,
      "loss": 0.0038,
      "step": 12870
    },
    {
      "epoch": 2.6104580462099714,
      "grad_norm": 0.07454310357570648,
      "learning_rate": 6.947709768950143e-05,
      "loss": 0.0046,
      "step": 12880
    },
    {
      "epoch": 2.612484799351439,
      "grad_norm": 0.0076058050617575645,
      "learning_rate": 6.937576003242804e-05,
      "loss": 0.0055,
      "step": 12890
    },
    {
      "epoch": 2.6145115524929063,
      "grad_norm": 0.008823697455227375,
      "learning_rate": 6.927442237535469e-05,
      "loss": 0.0018,
      "step": 12900
    },
    {
      "epoch": 2.6145115524929063,
      "eval_accuracy": 0.9853479853479854,
      "eval_loss": 0.08512182533740997,
      "eval_runtime": 249.2668,
      "eval_samples_per_second": 2.19,
      "eval_steps_per_second": 0.277,
      "step": 12900
    },
    {
      "epoch": 2.6165383056343736,
      "grad_norm": 0.00716985110193491,
      "learning_rate": 6.917308471828131e-05,
      "loss": 0.0009,
      "step": 12910
    },
    {
      "epoch": 2.6185650587758413,
      "grad_norm": 0.007007387932389975,
      "learning_rate": 6.907174706120794e-05,
      "loss": 0.0029,
      "step": 12920
    },
    {
      "epoch": 2.6205918119173086,
      "grad_norm": 0.05765289068222046,
      "learning_rate": 6.897040940413459e-05,
      "loss": 0.0021,
      "step": 12930
    },
    {
      "epoch": 2.622618565058776,
      "grad_norm": 0.007231696508824825,
      "learning_rate": 6.886907174706121e-05,
      "loss": 0.0016,
      "step": 12940
    },
    {
      "epoch": 2.624645318200243,
      "grad_norm": 0.007360757794231176,
      "learning_rate": 6.876773408998784e-05,
      "loss": 0.4605,
      "step": 12950
    },
    {
      "epoch": 2.6266720713417104,
      "grad_norm": 0.017173467203974724,
      "learning_rate": 6.866639643291447e-05,
      "loss": 0.0015,
      "step": 12960
    },
    {
      "epoch": 2.6286988244831777,
      "grad_norm": 0.11444433033466339,
      "learning_rate": 6.85650587758411e-05,
      "loss": 0.67,
      "step": 12970
    },
    {
      "epoch": 2.6307255776246454,
      "grad_norm": 0.011185628361999989,
      "learning_rate": 6.846372111876774e-05,
      "loss": 0.4956,
      "step": 12980
    },
    {
      "epoch": 2.6327523307661127,
      "grad_norm": 8.056800842285156,
      "learning_rate": 6.836238346169437e-05,
      "loss": 0.4779,
      "step": 12990
    },
    {
      "epoch": 2.63477908390758,
      "grad_norm": 0.008451711386442184,
      "learning_rate": 6.8261045804621e-05,
      "loss": 0.0014,
      "step": 13000
    },
    {
      "epoch": 2.63477908390758,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10948704928159714,
      "eval_runtime": 371.0896,
      "eval_samples_per_second": 1.471,
      "eval_steps_per_second": 0.186,
      "step": 13000
    },
    {
      "epoch": 2.6368058370490477,
      "grad_norm": 0.013107103295624256,
      "learning_rate": 6.815970814754764e-05,
      "loss": 0.0042,
      "step": 13010
    },
    {
      "epoch": 2.638832590190515,
      "grad_norm": 0.008205901831388474,
      "learning_rate": 6.805837049047427e-05,
      "loss": 0.5378,
      "step": 13020
    },
    {
      "epoch": 2.6408593433319822,
      "grad_norm": 0.13924755156040192,
      "learning_rate": 6.795703283340089e-05,
      "loss": 0.0067,
      "step": 13030
    },
    {
      "epoch": 2.6428860964734495,
      "grad_norm": 0.07655428349971771,
      "learning_rate": 6.785569517632752e-05,
      "loss": 0.0051,
      "step": 13040
    },
    {
      "epoch": 2.644912849614917,
      "grad_norm": 0.011031039990484715,
      "learning_rate": 6.775435751925417e-05,
      "loss": 0.002,
      "step": 13050
    },
    {
      "epoch": 2.646939602756384,
      "grad_norm": 0.007966967299580574,
      "learning_rate": 6.765301986218079e-05,
      "loss": 0.0025,
      "step": 13060
    },
    {
      "epoch": 2.648966355897852,
      "grad_norm": 0.062473565340042114,
      "learning_rate": 6.755168220510742e-05,
      "loss": 0.0038,
      "step": 13070
    },
    {
      "epoch": 2.650993109039319,
      "grad_norm": 0.006963303312659264,
      "learning_rate": 6.745034454803405e-05,
      "loss": 0.0017,
      "step": 13080
    },
    {
      "epoch": 2.6530198621807863,
      "grad_norm": 0.009901291690766811,
      "learning_rate": 6.734900689096068e-05,
      "loss": 0.003,
      "step": 13090
    },
    {
      "epoch": 2.6550466153222536,
      "grad_norm": 0.05340264365077019,
      "learning_rate": 6.724766923388732e-05,
      "loss": 0.0017,
      "step": 13100
    },
    {
      "epoch": 2.6550466153222536,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.08778306096792221,
      "eval_runtime": 355.0894,
      "eval_samples_per_second": 1.538,
      "eval_steps_per_second": 0.194,
      "step": 13100
    },
    {
      "epoch": 2.6570733684637213,
      "grad_norm": 0.006709767039865255,
      "learning_rate": 6.714633157681395e-05,
      "loss": 0.0019,
      "step": 13110
    },
    {
      "epoch": 2.6591001216051886,
      "grad_norm": 0.008395565673708916,
      "learning_rate": 6.704499391974058e-05,
      "loss": 0.0057,
      "step": 13120
    },
    {
      "epoch": 2.661126874746656,
      "grad_norm": 0.00617511011660099,
      "learning_rate": 6.694365626266721e-05,
      "loss": 0.0012,
      "step": 13130
    },
    {
      "epoch": 2.663153627888123,
      "grad_norm": 0.00602354621514678,
      "learning_rate": 6.684231860559385e-05,
      "loss": 0.0043,
      "step": 13140
    },
    {
      "epoch": 2.6651803810295904,
      "grad_norm": 0.007556139025837183,
      "learning_rate": 6.674098094852047e-05,
      "loss": 0.0013,
      "step": 13150
    },
    {
      "epoch": 2.6672071341710577,
      "grad_norm": 0.007209232077002525,
      "learning_rate": 6.663964329144711e-05,
      "loss": 0.0022,
      "step": 13160
    },
    {
      "epoch": 2.6692338873125254,
      "grad_norm": 0.005784170236438513,
      "learning_rate": 6.653830563437373e-05,
      "loss": 0.0016,
      "step": 13170
    },
    {
      "epoch": 2.6712606404539927,
      "grad_norm": 0.005957774352282286,
      "learning_rate": 6.643696797730036e-05,
      "loss": 0.5298,
      "step": 13180
    },
    {
      "epoch": 2.67328739359546,
      "grad_norm": 0.005831047426909208,
      "learning_rate": 6.633563032022701e-05,
      "loss": 0.0017,
      "step": 13190
    },
    {
      "epoch": 2.6753141467369277,
      "grad_norm": 0.005900830961763859,
      "learning_rate": 6.623429266315363e-05,
      "loss": 0.0014,
      "step": 13200
    },
    {
      "epoch": 2.6753141467369277,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.13224680721759796,
      "eval_runtime": 340.8015,
      "eval_samples_per_second": 1.602,
      "eval_steps_per_second": 0.202,
      "step": 13200
    },
    {
      "epoch": 2.677340899878395,
      "grad_norm": 0.006084654945880175,
      "learning_rate": 6.613295500608026e-05,
      "loss": 1.0179,
      "step": 13210
    },
    {
      "epoch": 2.6793676530198622,
      "grad_norm": 0.009283014573156834,
      "learning_rate": 6.60316173490069e-05,
      "loss": 0.0455,
      "step": 13220
    },
    {
      "epoch": 2.6813944061613295,
      "grad_norm": 0.008484777063131332,
      "learning_rate": 6.593027969193353e-05,
      "loss": 0.4282,
      "step": 13230
    },
    {
      "epoch": 2.683421159302797,
      "grad_norm": 0.007272159215062857,
      "learning_rate": 6.582894203486015e-05,
      "loss": 0.0012,
      "step": 13240
    },
    {
      "epoch": 2.685447912444264,
      "grad_norm": 0.009423608891665936,
      "learning_rate": 6.572760437778679e-05,
      "loss": 0.0036,
      "step": 13250
    },
    {
      "epoch": 2.687474665585732,
      "grad_norm": 0.022705376148223877,
      "learning_rate": 6.562626672071343e-05,
      "loss": 0.0117,
      "step": 13260
    },
    {
      "epoch": 2.689501418727199,
      "grad_norm": 0.006629382260143757,
      "learning_rate": 6.552492906364004e-05,
      "loss": 0.0037,
      "step": 13270
    },
    {
      "epoch": 2.6915281718686663,
      "grad_norm": 0.007442939560860395,
      "learning_rate": 6.542359140656669e-05,
      "loss": 0.0023,
      "step": 13280
    },
    {
      "epoch": 2.6935549250101336,
      "grad_norm": 0.007592672016471624,
      "learning_rate": 6.532225374949331e-05,
      "loss": 0.1412,
      "step": 13290
    },
    {
      "epoch": 2.6955816781516013,
      "grad_norm": 0.006130645051598549,
      "learning_rate": 6.522091609241994e-05,
      "loss": 0.0015,
      "step": 13300
    },
    {
      "epoch": 2.6955816781516013,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10588613152503967,
      "eval_runtime": 458.6185,
      "eval_samples_per_second": 1.191,
      "eval_steps_per_second": 0.15,
      "step": 13300
    },
    {
      "epoch": 2.6976084312930686,
      "grad_norm": 0.007999964989721775,
      "learning_rate": 6.511957843534658e-05,
      "loss": 0.0015,
      "step": 13310
    },
    {
      "epoch": 2.699635184434536,
      "grad_norm": 0.007326509803533554,
      "learning_rate": 6.501824077827321e-05,
      "loss": 0.4529,
      "step": 13320
    },
    {
      "epoch": 2.701661937576003,
      "grad_norm": 0.005822496023029089,
      "learning_rate": 6.491690312119984e-05,
      "loss": 0.0014,
      "step": 13330
    },
    {
      "epoch": 2.7036886907174704,
      "grad_norm": 8.135007858276367,
      "learning_rate": 6.481556546412647e-05,
      "loss": 0.5238,
      "step": 13340
    },
    {
      "epoch": 2.705715443858938,
      "grad_norm": 0.06068016588687897,
      "learning_rate": 6.47142278070531e-05,
      "loss": 0.0021,
      "step": 13350
    },
    {
      "epoch": 2.7077421970004054,
      "grad_norm": 0.06504343450069427,
      "learning_rate": 6.461289014997974e-05,
      "loss": 0.0025,
      "step": 13360
    },
    {
      "epoch": 2.7097689501418727,
      "grad_norm": 0.005796231795102358,
      "learning_rate": 6.451155249290637e-05,
      "loss": 0.0015,
      "step": 13370
    },
    {
      "epoch": 2.71179570328334,
      "grad_norm": 17.916643142700195,
      "learning_rate": 6.441021483583299e-05,
      "loss": 0.5551,
      "step": 13380
    },
    {
      "epoch": 2.7138224564248077,
      "grad_norm": 0.006149032153189182,
      "learning_rate": 6.430887717875964e-05,
      "loss": 0.4887,
      "step": 13390
    },
    {
      "epoch": 2.715849209566275,
      "grad_norm": 0.019423088058829308,
      "learning_rate": 6.420753952168627e-05,
      "loss": 0.0036,
      "step": 13400
    },
    {
      "epoch": 2.715849209566275,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.09267744421958923,
      "eval_runtime": 347.4343,
      "eval_samples_per_second": 1.572,
      "eval_steps_per_second": 0.199,
      "step": 13400
    },
    {
      "epoch": 2.7178759627077422,
      "grad_norm": 0.006651645991951227,
      "learning_rate": 6.410620186461289e-05,
      "loss": 0.0018,
      "step": 13410
    },
    {
      "epoch": 2.7199027158492095,
      "grad_norm": 0.006167796906083822,
      "learning_rate": 6.400486420753953e-05,
      "loss": 0.0033,
      "step": 13420
    },
    {
      "epoch": 2.721929468990677,
      "grad_norm": 0.009791787713766098,
      "learning_rate": 6.390352655046615e-05,
      "loss": 0.0017,
      "step": 13430
    },
    {
      "epoch": 2.723956222132144,
      "grad_norm": 0.009656204842031002,
      "learning_rate": 6.380218889339279e-05,
      "loss": 0.7369,
      "step": 13440
    },
    {
      "epoch": 2.7259829752736118,
      "grad_norm": 0.010334176942706108,
      "learning_rate": 6.370085123631942e-05,
      "loss": 0.5731,
      "step": 13450
    },
    {
      "epoch": 2.728009728415079,
      "grad_norm": 0.007303066086024046,
      "learning_rate": 6.359951357924605e-05,
      "loss": 0.0024,
      "step": 13460
    },
    {
      "epoch": 2.7300364815565463,
      "grad_norm": 0.007056308910250664,
      "learning_rate": 6.349817592217268e-05,
      "loss": 0.0022,
      "step": 13470
    },
    {
      "epoch": 2.732063234698014,
      "grad_norm": 0.00947339553385973,
      "learning_rate": 6.339683826509932e-05,
      "loss": 0.0035,
      "step": 13480
    },
    {
      "epoch": 2.7340899878394813,
      "grad_norm": 0.011178678832948208,
      "learning_rate": 6.329550060802595e-05,
      "loss": 0.0029,
      "step": 13490
    },
    {
      "epoch": 2.7361167409809486,
      "grad_norm": 0.006489630322903395,
      "learning_rate": 6.319416295095257e-05,
      "loss": 0.0051,
      "step": 13500
    },
    {
      "epoch": 2.7361167409809486,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10085071623325348,
      "eval_runtime": 58.9004,
      "eval_samples_per_second": 9.27,
      "eval_steps_per_second": 1.171,
      "step": 13500
    },
    {
      "epoch": 2.738143494122416,
      "grad_norm": 0.055221959948539734,
      "learning_rate": 6.309282529387921e-05,
      "loss": 0.0041,
      "step": 13510
    },
    {
      "epoch": 2.740170247263883,
      "grad_norm": 0.006161363795399666,
      "learning_rate": 6.299148763680583e-05,
      "loss": 0.002,
      "step": 13520
    },
    {
      "epoch": 2.7421970004053504,
      "grad_norm": 0.006010277196764946,
      "learning_rate": 6.289014997973247e-05,
      "loss": 0.0019,
      "step": 13530
    },
    {
      "epoch": 2.744223753546818,
      "grad_norm": 0.007737209089100361,
      "learning_rate": 6.278881232265911e-05,
      "loss": 0.0026,
      "step": 13540
    },
    {
      "epoch": 2.7462505066882854,
      "grad_norm": 0.04785401001572609,
      "learning_rate": 6.268747466558573e-05,
      "loss": 0.0029,
      "step": 13550
    },
    {
      "epoch": 2.7482772598297527,
      "grad_norm": 0.0055003599263727665,
      "learning_rate": 6.258613700851236e-05,
      "loss": 0.002,
      "step": 13560
    },
    {
      "epoch": 2.75030401297122,
      "grad_norm": 0.044067516922950745,
      "learning_rate": 6.2484799351439e-05,
      "loss": 0.0024,
      "step": 13570
    },
    {
      "epoch": 2.7523307661126877,
      "grad_norm": 0.005313580390065908,
      "learning_rate": 6.238346169436563e-05,
      "loss": 0.4124,
      "step": 13580
    },
    {
      "epoch": 2.754357519254155,
      "grad_norm": 0.005581700708717108,
      "learning_rate": 6.228212403729226e-05,
      "loss": 0.4932,
      "step": 13590
    },
    {
      "epoch": 2.7563842723956222,
      "grad_norm": 0.005542232189327478,
      "learning_rate": 6.21807863802189e-05,
      "loss": 0.0028,
      "step": 13600
    },
    {
      "epoch": 2.7563842723956222,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.16798269748687744,
      "eval_runtime": 53.1296,
      "eval_samples_per_second": 10.277,
      "eval_steps_per_second": 1.299,
      "step": 13600
    },
    {
      "epoch": 2.7584110255370895,
      "grad_norm": 0.046516843140125275,
      "learning_rate": 6.207944872314553e-05,
      "loss": 0.0017,
      "step": 13610
    },
    {
      "epoch": 2.7604377786785568,
      "grad_norm": 0.005435958504676819,
      "learning_rate": 6.197811106607215e-05,
      "loss": 0.3577,
      "step": 13620
    },
    {
      "epoch": 2.762464531820024,
      "grad_norm": 0.006663784850388765,
      "learning_rate": 6.187677340899879e-05,
      "loss": 0.0013,
      "step": 13630
    },
    {
      "epoch": 2.7644912849614918,
      "grad_norm": 0.005261593498289585,
      "learning_rate": 6.177543575192541e-05,
      "loss": 0.0023,
      "step": 13640
    },
    {
      "epoch": 2.766518038102959,
      "grad_norm": 0.04430309310555458,
      "learning_rate": 6.167409809485204e-05,
      "loss": 0.0013,
      "step": 13650
    },
    {
      "epoch": 2.7685447912444263,
      "grad_norm": 0.015064161270856857,
      "learning_rate": 6.157276043777868e-05,
      "loss": 0.0052,
      "step": 13660
    },
    {
      "epoch": 2.770571544385894,
      "grad_norm": 0.006480317562818527,
      "learning_rate": 6.147142278070531e-05,
      "loss": 0.7568,
      "step": 13670
    },
    {
      "epoch": 2.7725982975273613,
      "grad_norm": 0.007233656011521816,
      "learning_rate": 6.137008512363194e-05,
      "loss": 0.001,
      "step": 13680
    },
    {
      "epoch": 2.7746250506688286,
      "grad_norm": 0.005915817804634571,
      "learning_rate": 6.126874746655858e-05,
      "loss": 0.4029,
      "step": 13690
    },
    {
      "epoch": 2.776651803810296,
      "grad_norm": 0.006141826510429382,
      "learning_rate": 6.116740980948521e-05,
      "loss": 0.6951,
      "step": 13700
    },
    {
      "epoch": 2.776651803810296,
      "eval_accuracy": 0.9487179487179487,
      "eval_loss": 0.24969354271888733,
      "eval_runtime": 53.1917,
      "eval_samples_per_second": 10.265,
      "eval_steps_per_second": 1.297,
      "step": 13700
    },
    {
      "epoch": 2.778678556951763,
      "grad_norm": 0.006940559484064579,
      "learning_rate": 6.106607215241184e-05,
      "loss": 0.0013,
      "step": 13710
    },
    {
      "epoch": 2.7807053100932304,
      "grad_norm": 0.007143535651266575,
      "learning_rate": 6.096473449533847e-05,
      "loss": 0.0016,
      "step": 13720
    },
    {
      "epoch": 2.782732063234698,
      "grad_norm": 0.007185479626059532,
      "learning_rate": 6.0863396838265106e-05,
      "loss": 0.487,
      "step": 13730
    },
    {
      "epoch": 2.7847588163761654,
      "grad_norm": 0.5434855818748474,
      "learning_rate": 6.076205918119173e-05,
      "loss": 0.0092,
      "step": 13740
    },
    {
      "epoch": 2.7867855695176327,
      "grad_norm": 0.007091503124684095,
      "learning_rate": 6.066072152411837e-05,
      "loss": 0.002,
      "step": 13750
    },
    {
      "epoch": 2.7888123226591004,
      "grad_norm": 0.04187440127134323,
      "learning_rate": 6.0559383867045e-05,
      "loss": 0.4058,
      "step": 13760
    },
    {
      "epoch": 2.7908390758005677,
      "grad_norm": 0.009630071930587292,
      "learning_rate": 6.045804620997163e-05,
      "loss": 0.0018,
      "step": 13770
    },
    {
      "epoch": 2.792865828942035,
      "grad_norm": 0.2932543456554413,
      "learning_rate": 6.0356708552898255e-05,
      "loss": 0.0048,
      "step": 13780
    },
    {
      "epoch": 2.794892582083502,
      "grad_norm": 0.00625465577468276,
      "learning_rate": 6.0255370895824895e-05,
      "loss": 0.0029,
      "step": 13790
    },
    {
      "epoch": 2.7969193352249695,
      "grad_norm": 0.22124755382537842,
      "learning_rate": 6.015403323875153e-05,
      "loss": 0.0096,
      "step": 13800
    },
    {
      "epoch": 2.7969193352249695,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.1137545183300972,
      "eval_runtime": 50.9606,
      "eval_samples_per_second": 10.714,
      "eval_steps_per_second": 1.354,
      "step": 13800
    },
    {
      "epoch": 2.7989460883664368,
      "grad_norm": 0.009009423665702343,
      "learning_rate": 6.005269558167815e-05,
      "loss": 0.531,
      "step": 13810
    },
    {
      "epoch": 2.8009728415079045,
      "grad_norm": 0.009486384689807892,
      "learning_rate": 5.995135792460479e-05,
      "loss": 0.4189,
      "step": 13820
    },
    {
      "epoch": 2.8029995946493718,
      "grad_norm": 0.015886301174759865,
      "learning_rate": 5.985002026753141e-05,
      "loss": 0.3436,
      "step": 13830
    },
    {
      "epoch": 2.805026347790839,
      "grad_norm": 0.006443299353122711,
      "learning_rate": 5.974868261045805e-05,
      "loss": 0.0014,
      "step": 13840
    },
    {
      "epoch": 2.8070531009323063,
      "grad_norm": 0.04074406623840332,
      "learning_rate": 5.964734495338468e-05,
      "loss": 0.0028,
      "step": 13850
    },
    {
      "epoch": 2.809079854073774,
      "grad_norm": 0.011449450626969337,
      "learning_rate": 5.954600729631131e-05,
      "loss": 0.0035,
      "step": 13860
    },
    {
      "epoch": 2.8111066072152413,
      "grad_norm": 0.006783784367144108,
      "learning_rate": 5.944466963923795e-05,
      "loss": 0.5468,
      "step": 13870
    },
    {
      "epoch": 2.8131333603567086,
      "grad_norm": 0.02318020910024643,
      "learning_rate": 5.9343331982164575e-05,
      "loss": 0.2205,
      "step": 13880
    },
    {
      "epoch": 2.815160113498176,
      "grad_norm": 0.006445059087127447,
      "learning_rate": 5.924199432509121e-05,
      "loss": 0.0045,
      "step": 13890
    },
    {
      "epoch": 2.817186866639643,
      "grad_norm": 0.006976704113185406,
      "learning_rate": 5.9140656668017834e-05,
      "loss": 0.5063,
      "step": 13900
    },
    {
      "epoch": 2.817186866639643,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.1150827407836914,
      "eval_runtime": 54.029,
      "eval_samples_per_second": 10.106,
      "eval_steps_per_second": 1.277,
      "step": 13900
    },
    {
      "epoch": 2.8192136197811104,
      "grad_norm": 0.09443381428718567,
      "learning_rate": 5.903931901094447e-05,
      "loss": 0.0142,
      "step": 13910
    },
    {
      "epoch": 2.821240372922578,
      "grad_norm": 0.013332713395357132,
      "learning_rate": 5.89379813538711e-05,
      "loss": 0.1031,
      "step": 13920
    },
    {
      "epoch": 2.8232671260640454,
      "grad_norm": 0.006388837471604347,
      "learning_rate": 5.883664369679773e-05,
      "loss": 0.0024,
      "step": 13930
    },
    {
      "epoch": 2.8252938792055127,
      "grad_norm": 0.006204861216247082,
      "learning_rate": 5.873530603972437e-05,
      "loss": 0.0024,
      "step": 13940
    },
    {
      "epoch": 2.8273206323469804,
      "grad_norm": 0.007994485087692738,
      "learning_rate": 5.8633968382651e-05,
      "loss": 0.0014,
      "step": 13950
    },
    {
      "epoch": 2.8293473854884477,
      "grad_norm": 0.0065392907708883286,
      "learning_rate": 5.853263072557763e-05,
      "loss": 0.7336,
      "step": 13960
    },
    {
      "epoch": 2.831374138629915,
      "grad_norm": 0.009553118608891964,
      "learning_rate": 5.8431293068504255e-05,
      "loss": 0.0073,
      "step": 13970
    },
    {
      "epoch": 2.833400891771382,
      "grad_norm": 0.06365495920181274,
      "learning_rate": 5.8329955411430895e-05,
      "loss": 0.0045,
      "step": 13980
    },
    {
      "epoch": 2.8354276449128495,
      "grad_norm": 0.006651031784713268,
      "learning_rate": 5.8228617754357514e-05,
      "loss": 0.0029,
      "step": 13990
    },
    {
      "epoch": 2.8374543980543168,
      "grad_norm": 0.0455886535346508,
      "learning_rate": 5.812728009728415e-05,
      "loss": 0.0026,
      "step": 14000
    },
    {
      "epoch": 2.8374543980543168,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11792446672916412,
      "eval_runtime": 53.4448,
      "eval_samples_per_second": 10.216,
      "eval_steps_per_second": 1.291,
      "step": 14000
    },
    {
      "epoch": 2.8394811511957845,
      "grad_norm": 0.006337382830679417,
      "learning_rate": 5.802594244021079e-05,
      "loss": 0.0024,
      "step": 14010
    },
    {
      "epoch": 2.8415079043372518,
      "grad_norm": 0.006261122412979603,
      "learning_rate": 5.792460478313741e-05,
      "loss": 0.0025,
      "step": 14020
    },
    {
      "epoch": 2.843534657478719,
      "grad_norm": 0.00638343533501029,
      "learning_rate": 5.782326712606405e-05,
      "loss": 0.5445,
      "step": 14030
    },
    {
      "epoch": 2.8455614106201863,
      "grad_norm": 9.283822059631348,
      "learning_rate": 5.772192946899068e-05,
      "loss": 1.2487,
      "step": 14040
    },
    {
      "epoch": 2.847588163761654,
      "grad_norm": 0.09864260256290436,
      "learning_rate": 5.762059181191731e-05,
      "loss": 0.0025,
      "step": 14050
    },
    {
      "epoch": 2.8496149169031213,
      "grad_norm": 0.06132243201136589,
      "learning_rate": 5.7519254154843936e-05,
      "loss": 0.8716,
      "step": 14060
    },
    {
      "epoch": 2.8516416700445886,
      "grad_norm": 0.009238564409315586,
      "learning_rate": 5.7417916497770575e-05,
      "loss": 0.0028,
      "step": 14070
    },
    {
      "epoch": 2.853668423186056,
      "grad_norm": 0.008571449667215347,
      "learning_rate": 5.731657884069721e-05,
      "loss": 0.4832,
      "step": 14080
    },
    {
      "epoch": 2.855695176327523,
      "grad_norm": 0.00848014559596777,
      "learning_rate": 5.7215241183623834e-05,
      "loss": 0.0027,
      "step": 14090
    },
    {
      "epoch": 2.857721929468991,
      "grad_norm": 0.008391969837248325,
      "learning_rate": 5.711390352655047e-05,
      "loss": 0.0041,
      "step": 14100
    },
    {
      "epoch": 2.857721929468991,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.12661290168762207,
      "eval_runtime": 50.5324,
      "eval_samples_per_second": 10.805,
      "eval_steps_per_second": 1.365,
      "step": 14100
    },
    {
      "epoch": 2.859748682610458,
      "grad_norm": 0.011006497777998447,
      "learning_rate": 5.70125658694771e-05,
      "loss": 0.0089,
      "step": 14110
    },
    {
      "epoch": 2.8617754357519254,
      "grad_norm": 0.065226711332798,
      "learning_rate": 5.691122821240373e-05,
      "loss": 1.1768,
      "step": 14120
    },
    {
      "epoch": 2.8638021888933927,
      "grad_norm": 0.009555099532008171,
      "learning_rate": 5.680989055533036e-05,
      "loss": 0.0033,
      "step": 14130
    },
    {
      "epoch": 2.8658289420348604,
      "grad_norm": 0.01057603769004345,
      "learning_rate": 5.6708552898257e-05,
      "loss": 0.4643,
      "step": 14140
    },
    {
      "epoch": 2.8678556951763277,
      "grad_norm": 0.2201487123966217,
      "learning_rate": 5.660721524118363e-05,
      "loss": 0.342,
      "step": 14150
    },
    {
      "epoch": 2.869882448317795,
      "grad_norm": 0.012544561177492142,
      "learning_rate": 5.6505877584110255e-05,
      "loss": 0.0799,
      "step": 14160
    },
    {
      "epoch": 2.871909201459262,
      "grad_norm": 0.009598630480468273,
      "learning_rate": 5.6404539927036895e-05,
      "loss": 0.0052,
      "step": 14170
    },
    {
      "epoch": 2.8739359546007295,
      "grad_norm": 0.019055837765336037,
      "learning_rate": 5.630320226996352e-05,
      "loss": 0.0024,
      "step": 14180
    },
    {
      "epoch": 2.8759627077421968,
      "grad_norm": 0.009083470329642296,
      "learning_rate": 5.620186461289015e-05,
      "loss": 0.0024,
      "step": 14190
    },
    {
      "epoch": 2.8779894608836645,
      "grad_norm": 0.00886489450931549,
      "learning_rate": 5.610052695581678e-05,
      "loss": 0.0019,
      "step": 14200
    },
    {
      "epoch": 2.8779894608836645,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.09981222450733185,
      "eval_runtime": 50.1764,
      "eval_samples_per_second": 10.882,
      "eval_steps_per_second": 1.375,
      "step": 14200
    },
    {
      "epoch": 2.8800162140251317,
      "grad_norm": 0.008524172939360142,
      "learning_rate": 5.599918929874341e-05,
      "loss": 0.0062,
      "step": 14210
    },
    {
      "epoch": 2.882042967166599,
      "grad_norm": 0.01295129768550396,
      "learning_rate": 5.589785164167005e-05,
      "loss": 0.0023,
      "step": 14220
    },
    {
      "epoch": 2.8840697203080667,
      "grad_norm": 0.008135986514389515,
      "learning_rate": 5.579651398459668e-05,
      "loss": 0.0021,
      "step": 14230
    },
    {
      "epoch": 2.886096473449534,
      "grad_norm": 0.007830134592950344,
      "learning_rate": 5.569517632752331e-05,
      "loss": 0.0028,
      "step": 14240
    },
    {
      "epoch": 2.8881232265910013,
      "grad_norm": 0.018854841589927673,
      "learning_rate": 5.5593838670449936e-05,
      "loss": 0.0023,
      "step": 14250
    },
    {
      "epoch": 2.8901499797324686,
      "grad_norm": 0.009230035357177258,
      "learning_rate": 5.5492501013376575e-05,
      "loss": 0.445,
      "step": 14260
    },
    {
      "epoch": 2.892176732873936,
      "grad_norm": 0.007842128165066242,
      "learning_rate": 5.53911633563032e-05,
      "loss": 0.4464,
      "step": 14270
    },
    {
      "epoch": 2.894203486015403,
      "grad_norm": 0.7170781493186951,
      "learning_rate": 5.5289825699229834e-05,
      "loss": 0.0045,
      "step": 14280
    },
    {
      "epoch": 2.896230239156871,
      "grad_norm": 38.8918571472168,
      "learning_rate": 5.518848804215647e-05,
      "loss": 0.3268,
      "step": 14290
    },
    {
      "epoch": 2.898256992298338,
      "grad_norm": 0.007682281080633402,
      "learning_rate": 5.50871503850831e-05,
      "loss": 0.0038,
      "step": 14300
    },
    {
      "epoch": 2.898256992298338,
      "eval_accuracy": 0.9652014652014652,
      "eval_loss": 0.12904779613018036,
      "eval_runtime": 50.6457,
      "eval_samples_per_second": 10.781,
      "eval_steps_per_second": 1.362,
      "step": 14300
    },
    {
      "epoch": 2.9002837454398054,
      "grad_norm": 0.007591232657432556,
      "learning_rate": 5.498581272800973e-05,
      "loss": 0.4361,
      "step": 14310
    },
    {
      "epoch": 2.9023104985812727,
      "grad_norm": 0.007921284064650536,
      "learning_rate": 5.488447507093636e-05,
      "loss": 0.0033,
      "step": 14320
    },
    {
      "epoch": 2.9043372517227404,
      "grad_norm": 0.008791577070951462,
      "learning_rate": 5.4783137413863e-05,
      "loss": 0.4804,
      "step": 14330
    },
    {
      "epoch": 2.9063640048642077,
      "grad_norm": 0.008087759837508202,
      "learning_rate": 5.468179975678963e-05,
      "loss": 0.2424,
      "step": 14340
    },
    {
      "epoch": 2.908390758005675,
      "grad_norm": 0.007720464374870062,
      "learning_rate": 5.4580462099716255e-05,
      "loss": 0.0122,
      "step": 14350
    },
    {
      "epoch": 2.910417511147142,
      "grad_norm": 0.05355573073029518,
      "learning_rate": 5.4479124442642895e-05,
      "loss": 0.0032,
      "step": 14360
    },
    {
      "epoch": 2.9124442642886095,
      "grad_norm": 0.007251413073390722,
      "learning_rate": 5.437778678556952e-05,
      "loss": 0.0034,
      "step": 14370
    },
    {
      "epoch": 2.9144710174300767,
      "grad_norm": 0.0070653860457241535,
      "learning_rate": 5.427644912849615e-05,
      "loss": 0.0156,
      "step": 14380
    },
    {
      "epoch": 2.9164977705715445,
      "grad_norm": 5.426998138427734,
      "learning_rate": 5.417511147142278e-05,
      "loss": 0.5408,
      "step": 14390
    },
    {
      "epoch": 2.9185245237130117,
      "grad_norm": 0.11581844836473465,
      "learning_rate": 5.407377381434942e-05,
      "loss": 0.0131,
      "step": 14400
    },
    {
      "epoch": 2.9185245237130117,
      "eval_accuracy": 0.9413919413919414,
      "eval_loss": 0.19978734850883484,
      "eval_runtime": 49.0837,
      "eval_samples_per_second": 11.124,
      "eval_steps_per_second": 1.406,
      "step": 14400
    },
    {
      "epoch": 2.920551276854479,
      "grad_norm": 0.08270927518606186,
      "learning_rate": 5.397243615727605e-05,
      "loss": 0.4666,
      "step": 14410
    },
    {
      "epoch": 2.9225780299959467,
      "grad_norm": 0.008222842589020729,
      "learning_rate": 5.387109850020268e-05,
      "loss": 0.4219,
      "step": 14420
    },
    {
      "epoch": 2.924604783137414,
      "grad_norm": 0.007847205735743046,
      "learning_rate": 5.376976084312931e-05,
      "loss": 0.0029,
      "step": 14430
    },
    {
      "epoch": 2.9266315362788813,
      "grad_norm": 0.007569338195025921,
      "learning_rate": 5.3668423186055936e-05,
      "loss": 0.0156,
      "step": 14440
    },
    {
      "epoch": 2.9286582894203486,
      "grad_norm": 0.045873142778873444,
      "learning_rate": 5.3567085528982575e-05,
      "loss": 0.0032,
      "step": 14450
    },
    {
      "epoch": 2.930685042561816,
      "grad_norm": 1.2365529537200928,
      "learning_rate": 5.34657478719092e-05,
      "loss": 0.0261,
      "step": 14460
    },
    {
      "epoch": 2.932711795703283,
      "grad_norm": 0.007316601928323507,
      "learning_rate": 5.3364410214835834e-05,
      "loss": 0.4418,
      "step": 14470
    },
    {
      "epoch": 2.934738548844751,
      "grad_norm": 0.007185041438788176,
      "learning_rate": 5.326307255776247e-05,
      "loss": 0.0022,
      "step": 14480
    },
    {
      "epoch": 2.936765301986218,
      "grad_norm": 0.007051927037537098,
      "learning_rate": 5.31617349006891e-05,
      "loss": 0.2142,
      "step": 14490
    },
    {
      "epoch": 2.9387920551276854,
      "grad_norm": 0.007010738831013441,
      "learning_rate": 5.306039724361573e-05,
      "loss": 0.0037,
      "step": 14500
    },
    {
      "epoch": 2.9387920551276854,
      "eval_accuracy": 0.9633699633699634,
      "eval_loss": 0.1214241087436676,
      "eval_runtime": 50.1191,
      "eval_samples_per_second": 10.894,
      "eval_steps_per_second": 1.377,
      "step": 14500
    },
    {
      "epoch": 2.9408188082691527,
      "grad_norm": 0.006798572838306427,
      "learning_rate": 5.295905958654236e-05,
      "loss": 0.1482,
      "step": 14510
    },
    {
      "epoch": 2.9428455614106204,
      "grad_norm": 0.08644963055849075,
      "learning_rate": 5.2857721929469e-05,
      "loss": 0.002,
      "step": 14520
    },
    {
      "epoch": 2.9448723145520876,
      "grad_norm": 0.006649952381849289,
      "learning_rate": 5.275638427239562e-05,
      "loss": 0.0037,
      "step": 14530
    },
    {
      "epoch": 2.946899067693555,
      "grad_norm": 1.2725645303726196,
      "learning_rate": 5.2655046615322255e-05,
      "loss": 0.0225,
      "step": 14540
    },
    {
      "epoch": 2.948925820835022,
      "grad_norm": 0.07782497256994247,
      "learning_rate": 5.2553708958248895e-05,
      "loss": 0.2055,
      "step": 14550
    },
    {
      "epoch": 2.9509525739764895,
      "grad_norm": 0.006569635588675737,
      "learning_rate": 5.245237130117552e-05,
      "loss": 0.4006,
      "step": 14560
    },
    {
      "epoch": 2.952979327117957,
      "grad_norm": 0.00675079133361578,
      "learning_rate": 5.2351033644102153e-05,
      "loss": 0.0029,
      "step": 14570
    },
    {
      "epoch": 2.9550060802594245,
      "grad_norm": 0.006548064760863781,
      "learning_rate": 5.224969598702878e-05,
      "loss": 0.0007,
      "step": 14580
    },
    {
      "epoch": 2.9570328334008917,
      "grad_norm": 0.006818473804742098,
      "learning_rate": 5.214835832995542e-05,
      "loss": 0.0018,
      "step": 14590
    },
    {
      "epoch": 2.959059586542359,
      "grad_norm": 0.0069591510109603405,
      "learning_rate": 5.204702067288204e-05,
      "loss": 0.2382,
      "step": 14600
    },
    {
      "epoch": 2.959059586542359,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10965513437986374,
      "eval_runtime": 50.308,
      "eval_samples_per_second": 10.853,
      "eval_steps_per_second": 1.372,
      "step": 14600
    },
    {
      "epoch": 2.9610863396838267,
      "grad_norm": 0.00643087737262249,
      "learning_rate": 5.194568301580868e-05,
      "loss": 0.0027,
      "step": 14610
    },
    {
      "epoch": 2.963113092825294,
      "grad_norm": 0.006223716307431459,
      "learning_rate": 5.184434535873532e-05,
      "loss": 0.0055,
      "step": 14620
    },
    {
      "epoch": 2.9651398459667613,
      "grad_norm": 0.005989754106849432,
      "learning_rate": 5.1743007701661936e-05,
      "loss": 0.0033,
      "step": 14630
    },
    {
      "epoch": 2.9671665991082286,
      "grad_norm": 0.00613937946036458,
      "learning_rate": 5.1641670044588575e-05,
      "loss": 0.0039,
      "step": 14640
    },
    {
      "epoch": 2.969193352249696,
      "grad_norm": 0.006323641631752253,
      "learning_rate": 5.15403323875152e-05,
      "loss": 0.0007,
      "step": 14650
    },
    {
      "epoch": 2.971220105391163,
      "grad_norm": 0.005707310978323221,
      "learning_rate": 5.1438994730441834e-05,
      "loss": 0.0006,
      "step": 14660
    },
    {
      "epoch": 2.973246858532631,
      "grad_norm": 0.06553935259580612,
      "learning_rate": 5.133765707336846e-05,
      "loss": 0.0021,
      "step": 14670
    },
    {
      "epoch": 2.975273611674098,
      "grad_norm": 0.06091271713376045,
      "learning_rate": 5.12363194162951e-05,
      "loss": 0.0032,
      "step": 14680
    },
    {
      "epoch": 2.9773003648155654,
      "grad_norm": 0.0053886366076767445,
      "learning_rate": 5.113498175922173e-05,
      "loss": 0.0006,
      "step": 14690
    },
    {
      "epoch": 2.979327117957033,
      "grad_norm": 0.005328685510903597,
      "learning_rate": 5.103364410214836e-05,
      "loss": 0.0021,
      "step": 14700
    },
    {
      "epoch": 2.979327117957033,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.11521938443183899,
      "eval_runtime": 50.3755,
      "eval_samples_per_second": 10.839,
      "eval_steps_per_second": 1.37,
      "step": 14700
    },
    {
      "epoch": 2.9813538710985004,
      "grad_norm": 2.63281512260437,
      "learning_rate": 5.0932306445075e-05,
      "loss": 0.5353,
      "step": 14710
    },
    {
      "epoch": 2.9833806242399676,
      "grad_norm": 0.006335753481835127,
      "learning_rate": 5.083096878800162e-05,
      "loss": 0.0014,
      "step": 14720
    },
    {
      "epoch": 2.985407377381435,
      "grad_norm": 0.005788266658782959,
      "learning_rate": 5.0729631130928255e-05,
      "loss": 0.4826,
      "step": 14730
    },
    {
      "epoch": 2.987434130522902,
      "grad_norm": 0.006027384661138058,
      "learning_rate": 5.062829347385488e-05,
      "loss": 0.0022,
      "step": 14740
    },
    {
      "epoch": 2.9894608836643695,
      "grad_norm": 0.0832933858036995,
      "learning_rate": 5.052695581678152e-05,
      "loss": 0.4548,
      "step": 14750
    },
    {
      "epoch": 2.991487636805837,
      "grad_norm": 7.829352855682373,
      "learning_rate": 5.0425618159708153e-05,
      "loss": 0.445,
      "step": 14760
    },
    {
      "epoch": 2.9935143899473045,
      "grad_norm": 0.006316567305475473,
      "learning_rate": 5.032428050263478e-05,
      "loss": 0.0139,
      "step": 14770
    },
    {
      "epoch": 2.9955411430887717,
      "grad_norm": 0.006239940412342548,
      "learning_rate": 5.022294284556142e-05,
      "loss": 0.0034,
      "step": 14780
    },
    {
      "epoch": 2.997567896230239,
      "grad_norm": 0.006074379198253155,
      "learning_rate": 5.012160518848804e-05,
      "loss": 0.0007,
      "step": 14790
    },
    {
      "epoch": 2.9995946493717067,
      "grad_norm": 0.005999103654175997,
      "learning_rate": 5.002026753141468e-05,
      "loss": 0.002,
      "step": 14800
    },
    {
      "epoch": 2.9995946493717067,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10014810413122177,
      "eval_runtime": 47.3674,
      "eval_samples_per_second": 11.527,
      "eval_steps_per_second": 1.457,
      "step": 14800
    },
    {
      "epoch": 3.001621402513174,
      "grad_norm": 0.007480887696146965,
      "learning_rate": 4.991892987434131e-05,
      "loss": 0.7399,
      "step": 14810
    },
    {
      "epoch": 3.0036481556546413,
      "grad_norm": 0.006860152818262577,
      "learning_rate": 4.9817592217267936e-05,
      "loss": 0.0018,
      "step": 14820
    },
    {
      "epoch": 3.0056749087961085,
      "grad_norm": 0.007421591319143772,
      "learning_rate": 4.971625456019457e-05,
      "loss": 0.0041,
      "step": 14830
    },
    {
      "epoch": 3.007701661937576,
      "grad_norm": 0.08642849326133728,
      "learning_rate": 4.96149169031212e-05,
      "loss": 0.0028,
      "step": 14840
    },
    {
      "epoch": 3.0097284150790435,
      "grad_norm": 0.08337121456861496,
      "learning_rate": 4.9513579246047834e-05,
      "loss": 0.0038,
      "step": 14850
    },
    {
      "epoch": 3.011755168220511,
      "grad_norm": 0.0063841864466667175,
      "learning_rate": 4.9412241588974466e-05,
      "loss": 0.0129,
      "step": 14860
    },
    {
      "epoch": 3.013781921361978,
      "grad_norm": 0.005981700029224157,
      "learning_rate": 4.93109039319011e-05,
      "loss": 0.0016,
      "step": 14870
    },
    {
      "epoch": 3.0158086745034454,
      "grad_norm": 0.005900870077311993,
      "learning_rate": 4.920956627482773e-05,
      "loss": 0.0065,
      "step": 14880
    },
    {
      "epoch": 3.0178354276449126,
      "grad_norm": 0.005930820945650339,
      "learning_rate": 4.910822861775436e-05,
      "loss": 0.2251,
      "step": 14890
    },
    {
      "epoch": 3.0198621807863804,
      "grad_norm": 0.07632249593734741,
      "learning_rate": 4.900689096068099e-05,
      "loss": 0.0027,
      "step": 14900
    },
    {
      "epoch": 3.0198621807863804,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.12906181812286377,
      "eval_runtime": 50.1474,
      "eval_samples_per_second": 10.888,
      "eval_steps_per_second": 1.376,
      "step": 14900
    },
    {
      "epoch": 3.0218889339278476,
      "grad_norm": 0.00579514866694808,
      "learning_rate": 4.890555330360762e-05,
      "loss": 0.0148,
      "step": 14910
    },
    {
      "epoch": 3.023915687069315,
      "grad_norm": 9.295763969421387,
      "learning_rate": 4.8804215646534255e-05,
      "loss": 0.4507,
      "step": 14920
    },
    {
      "epoch": 3.025942440210782,
      "grad_norm": 0.07488545775413513,
      "learning_rate": 4.870287798946089e-05,
      "loss": 0.0024,
      "step": 14930
    },
    {
      "epoch": 3.02796919335225,
      "grad_norm": 0.005740451160818338,
      "learning_rate": 4.860154033238752e-05,
      "loss": 0.0026,
      "step": 14940
    },
    {
      "epoch": 3.029995946493717,
      "grad_norm": 0.07854526489973068,
      "learning_rate": 4.850020267531415e-05,
      "loss": 0.0015,
      "step": 14950
    },
    {
      "epoch": 3.0320226996351844,
      "grad_norm": 0.07284305989742279,
      "learning_rate": 4.839886501824078e-05,
      "loss": 0.0024,
      "step": 14960
    },
    {
      "epoch": 3.0340494527766517,
      "grad_norm": 0.0056227208115160465,
      "learning_rate": 4.829752736116741e-05,
      "loss": 0.4737,
      "step": 14970
    },
    {
      "epoch": 3.036076205918119,
      "grad_norm": 0.006293569691479206,
      "learning_rate": 4.8196189704094045e-05,
      "loss": 0.4649,
      "step": 14980
    },
    {
      "epoch": 3.0381029590595867,
      "grad_norm": 0.006045538932085037,
      "learning_rate": 4.809485204702068e-05,
      "loss": 0.0006,
      "step": 14990
    },
    {
      "epoch": 3.040129712201054,
      "grad_norm": 0.006128450855612755,
      "learning_rate": 4.799351438994731e-05,
      "loss": 0.971,
      "step": 15000
    },
    {
      "epoch": 3.040129712201054,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.16169553995132446,
      "eval_runtime": 47.2487,
      "eval_samples_per_second": 11.556,
      "eval_steps_per_second": 1.46,
      "step": 15000
    },
    {
      "epoch": 3.0421564653425213,
      "grad_norm": 0.007412407547235489,
      "learning_rate": 4.7892176732873936e-05,
      "loss": 0.0007,
      "step": 15010
    },
    {
      "epoch": 3.0441832184839885,
      "grad_norm": 0.006472380366176367,
      "learning_rate": 4.779083907580057e-05,
      "loss": 0.0035,
      "step": 15020
    },
    {
      "epoch": 3.046209971625456,
      "grad_norm": 0.2207396775484085,
      "learning_rate": 4.76895014187272e-05,
      "loss": 0.0244,
      "step": 15030
    },
    {
      "epoch": 3.0482367247669235,
      "grad_norm": 0.006490557920187712,
      "learning_rate": 4.7588163761653834e-05,
      "loss": 0.002,
      "step": 15040
    },
    {
      "epoch": 3.050263477908391,
      "grad_norm": 0.00590917281806469,
      "learning_rate": 4.748682610458046e-05,
      "loss": 0.0016,
      "step": 15050
    },
    {
      "epoch": 3.052290231049858,
      "grad_norm": 0.09242843836545944,
      "learning_rate": 4.73854884475071e-05,
      "loss": 0.0071,
      "step": 15060
    },
    {
      "epoch": 3.0543169841913254,
      "grad_norm": 0.005696085747331381,
      "learning_rate": 4.728415079043373e-05,
      "loss": 0.003,
      "step": 15070
    },
    {
      "epoch": 3.056343737332793,
      "grad_norm": 0.005512133240699768,
      "learning_rate": 4.718281313336036e-05,
      "loss": 0.0008,
      "step": 15080
    },
    {
      "epoch": 3.0583704904742604,
      "grad_norm": 0.0054388162679970264,
      "learning_rate": 4.708147547628699e-05,
      "loss": 0.0037,
      "step": 15090
    },
    {
      "epoch": 3.0603972436157276,
      "grad_norm": 0.005437575746327639,
      "learning_rate": 4.698013781921362e-05,
      "loss": 0.0024,
      "step": 15100
    },
    {
      "epoch": 3.0603972436157276,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.12453050911426544,
      "eval_runtime": 50.2348,
      "eval_samples_per_second": 10.869,
      "eval_steps_per_second": 1.374,
      "step": 15100
    },
    {
      "epoch": 3.062423996757195,
      "grad_norm": 0.07207438349723816,
      "learning_rate": 4.687880016214025e-05,
      "loss": 0.0023,
      "step": 15110
    },
    {
      "epoch": 3.064450749898662,
      "grad_norm": 0.07433387637138367,
      "learning_rate": 4.677746250506688e-05,
      "loss": 0.0024,
      "step": 15120
    },
    {
      "epoch": 3.06647750304013,
      "grad_norm": 0.06784066557884216,
      "learning_rate": 4.667612484799352e-05,
      "loss": 0.0025,
      "step": 15130
    },
    {
      "epoch": 3.068504256181597,
      "grad_norm": 0.005310414358973503,
      "learning_rate": 4.657478719092015e-05,
      "loss": 0.0006,
      "step": 15140
    },
    {
      "epoch": 3.0705310093230644,
      "grad_norm": 0.06503447145223618,
      "learning_rate": 4.647344953384678e-05,
      "loss": 0.0036,
      "step": 15150
    },
    {
      "epoch": 3.0725577624645317,
      "grad_norm": 0.0050943815149366856,
      "learning_rate": 4.637211187677341e-05,
      "loss": 0.002,
      "step": 15160
    },
    {
      "epoch": 3.074584515605999,
      "grad_norm": 0.0048345597460865974,
      "learning_rate": 4.6270774219700045e-05,
      "loss": 0.0027,
      "step": 15170
    },
    {
      "epoch": 3.0766112687474667,
      "grad_norm": 0.004860189743340015,
      "learning_rate": 4.616943656262667e-05,
      "loss": 0.0005,
      "step": 15180
    },
    {
      "epoch": 3.078638021888934,
      "grad_norm": 0.0678575411438942,
      "learning_rate": 4.606809890555331e-05,
      "loss": 0.002,
      "step": 15190
    },
    {
      "epoch": 3.0806647750304013,
      "grad_norm": 0.004794835112988949,
      "learning_rate": 4.596676124847994e-05,
      "loss": 0.0172,
      "step": 15200
    },
    {
      "epoch": 3.0806647750304013,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.12464291602373123,
      "eval_runtime": 51.9311,
      "eval_samples_per_second": 10.514,
      "eval_steps_per_second": 1.329,
      "step": 15200
    },
    {
      "epoch": 3.0826915281718685,
      "grad_norm": 0.0046084243804216385,
      "learning_rate": 4.586542359140657e-05,
      "loss": 0.0013,
      "step": 15210
    },
    {
      "epoch": 3.084718281313336,
      "grad_norm": 0.004596021492034197,
      "learning_rate": 4.57640859343332e-05,
      "loss": 0.4799,
      "step": 15220
    },
    {
      "epoch": 3.0867450344548035,
      "grad_norm": 0.004740960896015167,
      "learning_rate": 4.5662748277259834e-05,
      "loss": 0.002,
      "step": 15230
    },
    {
      "epoch": 3.088771787596271,
      "grad_norm": 0.004619854502379894,
      "learning_rate": 4.556141062018646e-05,
      "loss": 0.0076,
      "step": 15240
    },
    {
      "epoch": 3.090798540737738,
      "grad_norm": 0.00916003342717886,
      "learning_rate": 4.546007296311309e-05,
      "loss": 0.0012,
      "step": 15250
    },
    {
      "epoch": 3.0928252938792054,
      "grad_norm": 0.004485053010284901,
      "learning_rate": 4.535873530603973e-05,
      "loss": 0.0012,
      "step": 15260
    },
    {
      "epoch": 3.094852047020673,
      "grad_norm": 0.004415922798216343,
      "learning_rate": 4.525739764896636e-05,
      "loss": 0.0073,
      "step": 15270
    },
    {
      "epoch": 3.0968788001621403,
      "grad_norm": 0.054408419877290726,
      "learning_rate": 4.515605999189299e-05,
      "loss": 0.002,
      "step": 15280
    },
    {
      "epoch": 3.0989055533036076,
      "grad_norm": 0.02117004431784153,
      "learning_rate": 4.505472233481962e-05,
      "loss": 0.0007,
      "step": 15290
    },
    {
      "epoch": 3.100932306445075,
      "grad_norm": 0.004247212782502174,
      "learning_rate": 4.495338467774625e-05,
      "loss": 0.0016,
      "step": 15300
    },
    {
      "epoch": 3.100932306445075,
      "eval_accuracy": 0.9633699633699634,
      "eval_loss": 0.16279546916484833,
      "eval_runtime": 48.6438,
      "eval_samples_per_second": 11.224,
      "eval_steps_per_second": 1.418,
      "step": 15300
    },
    {
      "epoch": 3.102959059586542,
      "grad_norm": 0.04946291446685791,
      "learning_rate": 4.485204702067288e-05,
      "loss": 0.0022,
      "step": 15310
    },
    {
      "epoch": 3.10498581272801,
      "grad_norm": 0.0042871409095823765,
      "learning_rate": 4.4750709363599514e-05,
      "loss": 0.0016,
      "step": 15320
    },
    {
      "epoch": 3.107012565869477,
      "grad_norm": 0.00467030331492424,
      "learning_rate": 4.464937170652615e-05,
      "loss": 0.0015,
      "step": 15330
    },
    {
      "epoch": 3.1090393190109444,
      "grad_norm": 0.007411759812384844,
      "learning_rate": 4.454803404945278e-05,
      "loss": 0.0025,
      "step": 15340
    },
    {
      "epoch": 3.1110660721524117,
      "grad_norm": 0.004603350069373846,
      "learning_rate": 4.444669639237941e-05,
      "loss": 0.0011,
      "step": 15350
    },
    {
      "epoch": 3.1130928252938794,
      "grad_norm": 0.0041840276680886745,
      "learning_rate": 4.4345358735306045e-05,
      "loss": 0.0014,
      "step": 15360
    },
    {
      "epoch": 3.1151195784353467,
      "grad_norm": 0.004000730812549591,
      "learning_rate": 4.424402107823267e-05,
      "loss": 0.529,
      "step": 15370
    },
    {
      "epoch": 3.117146331576814,
      "grad_norm": 0.004089189227670431,
      "learning_rate": 4.41426834211593e-05,
      "loss": 0.001,
      "step": 15380
    },
    {
      "epoch": 3.1191730847182813,
      "grad_norm": 0.0041065397672355175,
      "learning_rate": 4.4041345764085936e-05,
      "loss": 0.001,
      "step": 15390
    },
    {
      "epoch": 3.1211998378597485,
      "grad_norm": 0.004117593169212341,
      "learning_rate": 4.394000810701257e-05,
      "loss": 0.0016,
      "step": 15400
    },
    {
      "epoch": 3.1211998378597485,
      "eval_accuracy": 0.9633699633699634,
      "eval_loss": 0.16209031641483307,
      "eval_runtime": 50.0913,
      "eval_samples_per_second": 10.9,
      "eval_steps_per_second": 1.377,
      "step": 15400
    },
    {
      "epoch": 3.1232265910012162,
      "grad_norm": 0.004019442014396191,
      "learning_rate": 4.38386704499392e-05,
      "loss": 0.0012,
      "step": 15410
    },
    {
      "epoch": 3.1252533441426835,
      "grad_norm": 0.0039987810887396336,
      "learning_rate": 4.3737332792865834e-05,
      "loss": 0.0015,
      "step": 15420
    },
    {
      "epoch": 3.127280097284151,
      "grad_norm": 0.004105426371097565,
      "learning_rate": 4.363599513579246e-05,
      "loss": 0.5071,
      "step": 15430
    },
    {
      "epoch": 3.129306850425618,
      "grad_norm": 0.004296373575925827,
      "learning_rate": 4.353465747871909e-05,
      "loss": 0.4613,
      "step": 15440
    },
    {
      "epoch": 3.1313336035670853,
      "grad_norm": 0.004371760878711939,
      "learning_rate": 4.3433319821645725e-05,
      "loss": 0.002,
      "step": 15450
    },
    {
      "epoch": 3.133360356708553,
      "grad_norm": 0.004319510888308287,
      "learning_rate": 4.333198216457236e-05,
      "loss": 0.0012,
      "step": 15460
    },
    {
      "epoch": 3.1353871098500203,
      "grad_norm": 0.004228163044899702,
      "learning_rate": 4.323064450749899e-05,
      "loss": 0.0023,
      "step": 15470
    },
    {
      "epoch": 3.1374138629914876,
      "grad_norm": 0.004289459902793169,
      "learning_rate": 4.312930685042562e-05,
      "loss": 0.2547,
      "step": 15480
    },
    {
      "epoch": 3.139440616132955,
      "grad_norm": 0.004262127447873354,
      "learning_rate": 4.3027969193352255e-05,
      "loss": 0.4993,
      "step": 15490
    },
    {
      "epoch": 3.141467369274422,
      "grad_norm": 0.004445136059075594,
      "learning_rate": 4.292663153627888e-05,
      "loss": 0.0005,
      "step": 15500
    },
    {
      "epoch": 3.141467369274422,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11043673753738403,
      "eval_runtime": 50.316,
      "eval_samples_per_second": 10.851,
      "eval_steps_per_second": 1.371,
      "step": 15500
    },
    {
      "epoch": 3.14349412241589,
      "grad_norm": 0.004357473459094763,
      "learning_rate": 4.2825293879205514e-05,
      "loss": 0.0013,
      "step": 15510
    },
    {
      "epoch": 3.145520875557357,
      "grad_norm": 0.004336953163146973,
      "learning_rate": 4.272395622213215e-05,
      "loss": 0.0045,
      "step": 15520
    },
    {
      "epoch": 3.1475476286988244,
      "grad_norm": 0.0772530660033226,
      "learning_rate": 4.262261856505878e-05,
      "loss": 0.003,
      "step": 15530
    },
    {
      "epoch": 3.1495743818402917,
      "grad_norm": 0.06546340137720108,
      "learning_rate": 4.252128090798541e-05,
      "loss": 0.0028,
      "step": 15540
    },
    {
      "epoch": 3.1516011349817594,
      "grad_norm": 0.004146865103393793,
      "learning_rate": 4.2419943250912045e-05,
      "loss": 0.0021,
      "step": 15550
    },
    {
      "epoch": 3.1536278881232267,
      "grad_norm": 0.0040993038564920425,
      "learning_rate": 4.231860559383867e-05,
      "loss": 0.0011,
      "step": 15560
    },
    {
      "epoch": 3.155654641264694,
      "grad_norm": 0.004080483224242926,
      "learning_rate": 4.22172679367653e-05,
      "loss": 0.0004,
      "step": 15570
    },
    {
      "epoch": 3.1576813944061612,
      "grad_norm": 0.004077650140970945,
      "learning_rate": 4.2115930279691936e-05,
      "loss": 0.0005,
      "step": 15580
    },
    {
      "epoch": 3.1597081475476285,
      "grad_norm": 0.003945386037230492,
      "learning_rate": 4.201459262261856e-05,
      "loss": 0.3534,
      "step": 15590
    },
    {
      "epoch": 3.1617349006890962,
      "grad_norm": 0.004126140847802162,
      "learning_rate": 4.19132549655452e-05,
      "loss": 0.3195,
      "step": 15600
    },
    {
      "epoch": 3.1617349006890962,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.1447170376777649,
      "eval_runtime": 48.1002,
      "eval_samples_per_second": 11.351,
      "eval_steps_per_second": 1.435,
      "step": 15600
    },
    {
      "epoch": 3.1637616538305635,
      "grad_norm": 0.06581421196460724,
      "learning_rate": 4.1811917308471834e-05,
      "loss": 0.0053,
      "step": 15610
    },
    {
      "epoch": 3.165788406972031,
      "grad_norm": 0.004257733467966318,
      "learning_rate": 4.171057965139846e-05,
      "loss": 0.0005,
      "step": 15620
    },
    {
      "epoch": 3.167815160113498,
      "grad_norm": 0.0041100881062448025,
      "learning_rate": 4.160924199432509e-05,
      "loss": 0.0029,
      "step": 15630
    },
    {
      "epoch": 3.1698419132549653,
      "grad_norm": 0.004436058457940817,
      "learning_rate": 4.1507904337251725e-05,
      "loss": 0.7637,
      "step": 15640
    },
    {
      "epoch": 3.171868666396433,
      "grad_norm": 0.061650391668081284,
      "learning_rate": 4.140656668017836e-05,
      "loss": 0.0012,
      "step": 15650
    },
    {
      "epoch": 3.1738954195379003,
      "grad_norm": 0.004761091433465481,
      "learning_rate": 4.1305229023104983e-05,
      "loss": 0.0091,
      "step": 15660
    },
    {
      "epoch": 3.1759221726793676,
      "grad_norm": 0.07118819653987885,
      "learning_rate": 4.120389136603162e-05,
      "loss": 0.0028,
      "step": 15670
    },
    {
      "epoch": 3.177948925820835,
      "grad_norm": 0.004552809055894613,
      "learning_rate": 4.1102553708958255e-05,
      "loss": 0.4552,
      "step": 15680
    },
    {
      "epoch": 3.1799756789623026,
      "grad_norm": 0.0048391795717179775,
      "learning_rate": 4.100121605188488e-05,
      "loss": 0.0026,
      "step": 15690
    },
    {
      "epoch": 3.18200243210377,
      "grad_norm": 0.005941119510680437,
      "learning_rate": 4.0899878394811514e-05,
      "loss": 2.3502,
      "step": 15700
    },
    {
      "epoch": 3.18200243210377,
      "eval_accuracy": 0.9652014652014652,
      "eval_loss": 0.1826866865158081,
      "eval_runtime": 50.1588,
      "eval_samples_per_second": 10.885,
      "eval_steps_per_second": 1.376,
      "step": 15700
    },
    {
      "epoch": 3.184029185245237,
      "grad_norm": 0.006943140644580126,
      "learning_rate": 4.079854073773815e-05,
      "loss": 0.0019,
      "step": 15710
    },
    {
      "epoch": 3.1860559383867044,
      "grad_norm": 0.00895757507532835,
      "learning_rate": 4.069720308066477e-05,
      "loss": 0.002,
      "step": 15720
    },
    {
      "epoch": 3.1880826915281717,
      "grad_norm": 0.0076103429310023785,
      "learning_rate": 4.0595865423591405e-05,
      "loss": 0.0024,
      "step": 15730
    },
    {
      "epoch": 3.1901094446696394,
      "grad_norm": 0.009083851240575314,
      "learning_rate": 4.0494527766518045e-05,
      "loss": 0.0073,
      "step": 15740
    },
    {
      "epoch": 3.1921361978111067,
      "grad_norm": 0.009438318200409412,
      "learning_rate": 4.039319010944467e-05,
      "loss": 0.4983,
      "step": 15750
    },
    {
      "epoch": 3.194162950952574,
      "grad_norm": 0.1173553317785263,
      "learning_rate": 4.02918524523713e-05,
      "loss": 0.003,
      "step": 15760
    },
    {
      "epoch": 3.1961897040940412,
      "grad_norm": 0.006215721368789673,
      "learning_rate": 4.0190514795297936e-05,
      "loss": 0.0022,
      "step": 15770
    },
    {
      "epoch": 3.1982164572355085,
      "grad_norm": 0.0618855357170105,
      "learning_rate": 4.008917713822457e-05,
      "loss": 0.0015,
      "step": 15780
    },
    {
      "epoch": 3.2002432103769762,
      "grad_norm": 0.06262574344873428,
      "learning_rate": 3.9987839481151194e-05,
      "loss": 0.0015,
      "step": 15790
    },
    {
      "epoch": 3.2022699635184435,
      "grad_norm": 0.006081962957978249,
      "learning_rate": 3.988650182407783e-05,
      "loss": 0.4252,
      "step": 15800
    },
    {
      "epoch": 3.2022699635184435,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.10769849270582199,
      "eval_runtime": 47.2943,
      "eval_samples_per_second": 11.545,
      "eval_steps_per_second": 1.459,
      "step": 15800
    },
    {
      "epoch": 3.204296716659911,
      "grad_norm": 0.010130221024155617,
      "learning_rate": 3.978516416700446e-05,
      "loss": 0.002,
      "step": 15810
    },
    {
      "epoch": 3.206323469801378,
      "grad_norm": 0.005564684513956308,
      "learning_rate": 3.968382650993109e-05,
      "loss": 0.0064,
      "step": 15820
    },
    {
      "epoch": 3.2083502229428458,
      "grad_norm": 0.005339638330042362,
      "learning_rate": 3.9582488852857725e-05,
      "loss": 0.4571,
      "step": 15830
    },
    {
      "epoch": 3.210376976084313,
      "grad_norm": 0.005142607726156712,
      "learning_rate": 3.948115119578436e-05,
      "loss": 0.0022,
      "step": 15840
    },
    {
      "epoch": 3.2124037292257803,
      "grad_norm": 0.005214919801801443,
      "learning_rate": 3.9379813538710983e-05,
      "loss": 0.002,
      "step": 15850
    },
    {
      "epoch": 3.2144304823672476,
      "grad_norm": 0.005264309234917164,
      "learning_rate": 3.9278475881637616e-05,
      "loss": 0.0036,
      "step": 15860
    },
    {
      "epoch": 3.216457235508715,
      "grad_norm": 0.0050828661769628525,
      "learning_rate": 3.9177138224564256e-05,
      "loss": 0.0007,
      "step": 15870
    },
    {
      "epoch": 3.2184839886501826,
      "grad_norm": 10.684004783630371,
      "learning_rate": 3.907580056749088e-05,
      "loss": 0.318,
      "step": 15880
    },
    {
      "epoch": 3.22051074179165,
      "grad_norm": 0.004819140769541264,
      "learning_rate": 3.8974462910417514e-05,
      "loss": 0.0013,
      "step": 15890
    },
    {
      "epoch": 3.222537494933117,
      "grad_norm": 0.07204253226518631,
      "learning_rate": 3.887312525334415e-05,
      "loss": 0.0042,
      "step": 15900
    },
    {
      "epoch": 3.222537494933117,
      "eval_accuracy": 0.9706959706959707,
      "eval_loss": 0.1431271880865097,
      "eval_runtime": 50.1685,
      "eval_samples_per_second": 10.883,
      "eval_steps_per_second": 1.375,
      "step": 15900
    },
    {
      "epoch": 3.2245642480745844,
      "grad_norm": 0.004766241647303104,
      "learning_rate": 3.877178759627077e-05,
      "loss": 0.0006,
      "step": 15910
    },
    {
      "epoch": 3.2265910012160517,
      "grad_norm": 0.004663664381951094,
      "learning_rate": 3.8670449939197405e-05,
      "loss": 0.0055,
      "step": 15920
    },
    {
      "epoch": 3.2286177543575194,
      "grad_norm": 0.004675132222473621,
      "learning_rate": 3.856911228212404e-05,
      "loss": 0.0017,
      "step": 15930
    },
    {
      "epoch": 3.2306445074989867,
      "grad_norm": 0.007203600835055113,
      "learning_rate": 3.846777462505067e-05,
      "loss": 0.0018,
      "step": 15940
    },
    {
      "epoch": 3.232671260640454,
      "grad_norm": 0.005374888889491558,
      "learning_rate": 3.83664369679773e-05,
      "loss": 0.0022,
      "step": 15950
    },
    {
      "epoch": 3.2346980137819212,
      "grad_norm": 0.049246896058321,
      "learning_rate": 3.8265099310903936e-05,
      "loss": 0.0015,
      "step": 15960
    },
    {
      "epoch": 3.2367247669233885,
      "grad_norm": 0.004231675993651152,
      "learning_rate": 3.816376165383057e-05,
      "loss": 0.0086,
      "step": 15970
    },
    {
      "epoch": 3.2387515200648562,
      "grad_norm": 0.004210128914564848,
      "learning_rate": 3.8062423996757194e-05,
      "loss": 0.0005,
      "step": 15980
    },
    {
      "epoch": 3.2407782732063235,
      "grad_norm": 0.04477294534444809,
      "learning_rate": 3.796108633968383e-05,
      "loss": 0.0027,
      "step": 15990
    },
    {
      "epoch": 3.2428050263477908,
      "grad_norm": 0.00412526773288846,
      "learning_rate": 3.785974868261046e-05,
      "loss": 1.0207,
      "step": 16000
    },
    {
      "epoch": 3.2428050263477908,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.1287493109703064,
      "eval_runtime": 46.4503,
      "eval_samples_per_second": 11.754,
      "eval_steps_per_second": 1.485,
      "step": 16000
    },
    {
      "epoch": 3.244831779489258,
      "grad_norm": 0.009321167133748531,
      "learning_rate": 3.775841102553709e-05,
      "loss": 0.0041,
      "step": 16010
    },
    {
      "epoch": 3.2468585326307258,
      "grad_norm": 0.004468958359211683,
      "learning_rate": 3.7657073368463725e-05,
      "loss": 0.0038,
      "step": 16020
    },
    {
      "epoch": 3.248885285772193,
      "grad_norm": 0.004842129535973072,
      "learning_rate": 3.755573571139036e-05,
      "loss": 0.0043,
      "step": 16030
    },
    {
      "epoch": 3.2509120389136603,
      "grad_norm": 0.004659067839384079,
      "learning_rate": 3.7454398054316983e-05,
      "loss": 0.7708,
      "step": 16040
    },
    {
      "epoch": 3.2529387920551276,
      "grad_norm": 0.005155695136636496,
      "learning_rate": 3.7353060397243616e-05,
      "loss": 0.0012,
      "step": 16050
    },
    {
      "epoch": 3.254965545196595,
      "grad_norm": 0.0048374454490840435,
      "learning_rate": 3.725172274017025e-05,
      "loss": 0.0019,
      "step": 16060
    },
    {
      "epoch": 3.2569922983380626,
      "grad_norm": 0.004809567239135504,
      "learning_rate": 3.715038508309688e-05,
      "loss": 0.0614,
      "step": 16070
    },
    {
      "epoch": 3.25901905147953,
      "grad_norm": 0.05183807387948036,
      "learning_rate": 3.7049047426023514e-05,
      "loss": 0.003,
      "step": 16080
    },
    {
      "epoch": 3.261045804620997,
      "grad_norm": 0.004665842745453119,
      "learning_rate": 3.694770976895015e-05,
      "loss": 0.0012,
      "step": 16090
    },
    {
      "epoch": 3.2630725577624644,
      "grad_norm": 0.0047615994699299335,
      "learning_rate": 3.684637211187677e-05,
      "loss": 0.5064,
      "step": 16100
    },
    {
      "epoch": 3.2630725577624644,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.16628408432006836,
      "eval_runtime": 49.9971,
      "eval_samples_per_second": 10.921,
      "eval_steps_per_second": 1.38,
      "step": 16100
    },
    {
      "epoch": 3.265099310903932,
      "grad_norm": 0.004774071741849184,
      "learning_rate": 3.6745034454803405e-05,
      "loss": 0.0025,
      "step": 16110
    },
    {
      "epoch": 3.2671260640453994,
      "grad_norm": 0.004758866038173437,
      "learning_rate": 3.664369679773004e-05,
      "loss": 0.0019,
      "step": 16120
    },
    {
      "epoch": 3.2691528171868667,
      "grad_norm": 0.0046609085984528065,
      "learning_rate": 3.654235914065667e-05,
      "loss": 0.0012,
      "step": 16130
    },
    {
      "epoch": 3.271179570328334,
      "grad_norm": 0.004649462644010782,
      "learning_rate": 3.6441021483583296e-05,
      "loss": 0.0019,
      "step": 16140
    },
    {
      "epoch": 3.2732063234698012,
      "grad_norm": 0.004584324546158314,
      "learning_rate": 3.6339683826509936e-05,
      "loss": 0.0012,
      "step": 16150
    },
    {
      "epoch": 3.2752330766112685,
      "grad_norm": 0.004533680621534586,
      "learning_rate": 3.623834616943657e-05,
      "loss": 0.0011,
      "step": 16160
    },
    {
      "epoch": 3.277259829752736,
      "grad_norm": 0.004560311324894428,
      "learning_rate": 3.6137008512363194e-05,
      "loss": 0.0011,
      "step": 16170
    },
    {
      "epoch": 3.2792865828942035,
      "grad_norm": 0.0044805011712014675,
      "learning_rate": 3.603567085528983e-05,
      "loss": 0.5046,
      "step": 16180
    },
    {
      "epoch": 3.2813133360356708,
      "grad_norm": 0.05615738034248352,
      "learning_rate": 3.593433319821646e-05,
      "loss": 0.0024,
      "step": 16190
    },
    {
      "epoch": 3.283340089177138,
      "grad_norm": 0.004519937094300985,
      "learning_rate": 3.5832995541143086e-05,
      "loss": 0.0018,
      "step": 16200
    },
    {
      "epoch": 3.283340089177138,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.13267415761947632,
      "eval_runtime": 49.914,
      "eval_samples_per_second": 10.939,
      "eval_steps_per_second": 1.382,
      "step": 16200
    },
    {
      "epoch": 3.2853668423186058,
      "grad_norm": 0.05049968510866165,
      "learning_rate": 3.5731657884069725e-05,
      "loss": 0.0021,
      "step": 16210
    },
    {
      "epoch": 3.287393595460073,
      "grad_norm": 0.048404913395643234,
      "learning_rate": 3.563032022699636e-05,
      "loss": 0.0043,
      "step": 16220
    },
    {
      "epoch": 3.2894203486015403,
      "grad_norm": 0.004592543467879295,
      "learning_rate": 3.5528982569922983e-05,
      "loss": 0.6482,
      "step": 16230
    },
    {
      "epoch": 3.2914471017430076,
      "grad_norm": 0.004584785085171461,
      "learning_rate": 3.5427644912849616e-05,
      "loss": 0.0017,
      "step": 16240
    },
    {
      "epoch": 3.293473854884475,
      "grad_norm": 0.004494310822337866,
      "learning_rate": 3.532630725577625e-05,
      "loss": 0.0012,
      "step": 16250
    },
    {
      "epoch": 3.2955006080259426,
      "grad_norm": 0.10379710048437119,
      "learning_rate": 3.522496959870288e-05,
      "loss": 0.0021,
      "step": 16260
    },
    {
      "epoch": 3.29752736116741,
      "grad_norm": 0.047005441039800644,
      "learning_rate": 3.512363194162951e-05,
      "loss": 0.002,
      "step": 16270
    },
    {
      "epoch": 3.299554114308877,
      "grad_norm": 0.04951295629143715,
      "learning_rate": 3.502229428455615e-05,
      "loss": 0.3844,
      "step": 16280
    },
    {
      "epoch": 3.3015808674503444,
      "grad_norm": 0.004487763158977032,
      "learning_rate": 3.492095662748278e-05,
      "loss": 0.0017,
      "step": 16290
    },
    {
      "epoch": 3.303607620591812,
      "grad_norm": 0.0049188267439603806,
      "learning_rate": 3.4819618970409405e-05,
      "loss": 0.0006,
      "step": 16300
    },
    {
      "epoch": 3.303607620591812,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11632156372070312,
      "eval_runtime": 47.9658,
      "eval_samples_per_second": 11.383,
      "eval_steps_per_second": 1.439,
      "step": 16300
    },
    {
      "epoch": 3.3056343737332794,
      "grad_norm": 0.004525529686361551,
      "learning_rate": 3.471828131333604e-05,
      "loss": 0.002,
      "step": 16310
    },
    {
      "epoch": 3.3076611268747467,
      "grad_norm": 0.02013551816344261,
      "learning_rate": 3.461694365626267e-05,
      "loss": 0.0012,
      "step": 16320
    },
    {
      "epoch": 3.309687880016214,
      "grad_norm": 0.052169717848300934,
      "learning_rate": 3.4515605999189296e-05,
      "loss": 0.514,
      "step": 16330
    },
    {
      "epoch": 3.311714633157681,
      "grad_norm": 0.007503776345402002,
      "learning_rate": 3.441426834211593e-05,
      "loss": 0.0013,
      "step": 16340
    },
    {
      "epoch": 3.313741386299149,
      "grad_norm": 0.022679857909679413,
      "learning_rate": 3.431293068504257e-05,
      "loss": 0.4985,
      "step": 16350
    },
    {
      "epoch": 3.315768139440616,
      "grad_norm": 0.0106239328160882,
      "learning_rate": 3.4211593027969194e-05,
      "loss": 0.7458,
      "step": 16360
    },
    {
      "epoch": 3.3177948925820835,
      "grad_norm": 0.005137187894433737,
      "learning_rate": 3.411025537089583e-05,
      "loss": 0.0036,
      "step": 16370
    },
    {
      "epoch": 3.3198216457235508,
      "grad_norm": 0.06359566748142242,
      "learning_rate": 3.400891771382246e-05,
      "loss": 0.3599,
      "step": 16380
    },
    {
      "epoch": 3.3218483988650185,
      "grad_norm": 0.0068815951235592365,
      "learning_rate": 3.3907580056749086e-05,
      "loss": 0.0072,
      "step": 16390
    },
    {
      "epoch": 3.3238751520064858,
      "grad_norm": 0.0051819924265146255,
      "learning_rate": 3.380624239967572e-05,
      "loss": 0.0039,
      "step": 16400
    },
    {
      "epoch": 3.3238751520064858,
      "eval_accuracy": 0.9725274725274725,
      "eval_loss": 0.141290545463562,
      "eval_runtime": 50.1791,
      "eval_samples_per_second": 10.881,
      "eval_steps_per_second": 1.375,
      "step": 16400
    },
    {
      "epoch": 3.325901905147953,
      "grad_norm": 0.005205860361456871,
      "learning_rate": 3.370490474260235e-05,
      "loss": 0.0076,
      "step": 16410
    },
    {
      "epoch": 3.3279286582894203,
      "grad_norm": 0.08553090691566467,
      "learning_rate": 3.3603567085528983e-05,
      "loss": 0.4323,
      "step": 16420
    },
    {
      "epoch": 3.3299554114308876,
      "grad_norm": 0.06025693565607071,
      "learning_rate": 3.3502229428455616e-05,
      "loss": 0.0022,
      "step": 16430
    },
    {
      "epoch": 3.331982164572355,
      "grad_norm": 0.056943122297525406,
      "learning_rate": 3.340089177138225e-05,
      "loss": 0.0013,
      "step": 16440
    },
    {
      "epoch": 3.3340089177138226,
      "grad_norm": 0.007948963902890682,
      "learning_rate": 3.329955411430888e-05,
      "loss": 0.0039,
      "step": 16450
    },
    {
      "epoch": 3.33603567085529,
      "grad_norm": 0.004892650060355663,
      "learning_rate": 3.319821645723551e-05,
      "loss": 0.0013,
      "step": 16460
    },
    {
      "epoch": 3.338062423996757,
      "grad_norm": 0.004903930705040693,
      "learning_rate": 3.309687880016214e-05,
      "loss": 0.3424,
      "step": 16470
    },
    {
      "epoch": 3.3400891771382244,
      "grad_norm": 0.007834755815565586,
      "learning_rate": 3.299554114308877e-05,
      "loss": 0.3003,
      "step": 16480
    },
    {
      "epoch": 3.342115930279692,
      "grad_norm": 0.0051131355576217175,
      "learning_rate": 3.2894203486015405e-05,
      "loss": 0.0013,
      "step": 16490
    },
    {
      "epoch": 3.3441426834211594,
      "grad_norm": 0.005035320296883583,
      "learning_rate": 3.279286582894204e-05,
      "loss": 0.5045,
      "step": 16500
    },
    {
      "epoch": 3.3441426834211594,
      "eval_accuracy": 0.9688644688644689,
      "eval_loss": 0.1572037637233734,
      "eval_runtime": 47.102,
      "eval_samples_per_second": 11.592,
      "eval_steps_per_second": 1.465,
      "step": 16500
    },
    {
      "epoch": 3.3461694365626267,
      "grad_norm": 0.07073163986206055,
      "learning_rate": 3.269152817186867e-05,
      "loss": 0.752,
      "step": 16510
    },
    {
      "epoch": 3.348196189704094,
      "grad_norm": 0.007176402490586042,
      "learning_rate": 3.2590190514795296e-05,
      "loss": 0.0015,
      "step": 16520
    },
    {
      "epoch": 3.350222942845561,
      "grad_norm": 0.007812405005097389,
      "learning_rate": 3.248885285772193e-05,
      "loss": 0.0084,
      "step": 16530
    },
    {
      "epoch": 3.352249695987029,
      "grad_norm": 0.008438252843916416,
      "learning_rate": 3.238751520064856e-05,
      "loss": 0.0031,
      "step": 16540
    },
    {
      "epoch": 3.354276449128496,
      "grad_norm": 0.007445335853844881,
      "learning_rate": 3.2286177543575194e-05,
      "loss": 0.0021,
      "step": 16550
    },
    {
      "epoch": 3.3563032022699635,
      "grad_norm": 0.005647340323776007,
      "learning_rate": 3.218483988650183e-05,
      "loss": 0.2166,
      "step": 16560
    },
    {
      "epoch": 3.3583299554114308,
      "grad_norm": 0.007124828174710274,
      "learning_rate": 3.208350222942846e-05,
      "loss": 0.7467,
      "step": 16570
    },
    {
      "epoch": 3.3603567085528985,
      "grad_norm": 0.007737248204648495,
      "learning_rate": 3.1982164572355086e-05,
      "loss": 0.0014,
      "step": 16580
    },
    {
      "epoch": 3.3623834616943657,
      "grad_norm": 0.005917510949075222,
      "learning_rate": 3.188082691528172e-05,
      "loss": 0.0061,
      "step": 16590
    },
    {
      "epoch": 3.364410214835833,
      "grad_norm": 0.006020793225616217,
      "learning_rate": 3.177948925820835e-05,
      "loss": 0.0069,
      "step": 16600
    },
    {
      "epoch": 3.364410214835833,
      "eval_accuracy": 0.967032967032967,
      "eval_loss": 0.1553422510623932,
      "eval_runtime": 50.2992,
      "eval_samples_per_second": 10.855,
      "eval_steps_per_second": 1.372,
      "step": 16600
    },
    {
      "epoch": 3.3664369679773003,
      "grad_norm": 0.005671895109117031,
      "learning_rate": 3.1678151601134983e-05,
      "loss": 0.0403,
      "step": 16610
    },
    {
      "epoch": 3.3684637211187676,
      "grad_norm": 0.005752407480031252,
      "learning_rate": 3.1576813944061616e-05,
      "loss": 0.4096,
      "step": 16620
    },
    {
      "epoch": 3.3704904742602353,
      "grad_norm": 0.005818086210638285,
      "learning_rate": 3.147547628698825e-05,
      "loss": 1.4527,
      "step": 16630
    },
    {
      "epoch": 3.3725172274017026,
      "grad_norm": 0.007470723241567612,
      "learning_rate": 3.137413862991488e-05,
      "loss": 0.0081,
      "step": 16640
    },
    {
      "epoch": 3.37454398054317,
      "grad_norm": 0.007930467836558819,
      "learning_rate": 3.127280097284151e-05,
      "loss": 0.0048,
      "step": 16650
    },
    {
      "epoch": 3.376570733684637,
      "grad_norm": 0.008607808500528336,
      "learning_rate": 3.117146331576814e-05,
      "loss": 0.0042,
      "step": 16660
    },
    {
      "epoch": 3.3785974868261044,
      "grad_norm": 0.006221098359674215,
      "learning_rate": 3.107012565869477e-05,
      "loss": 0.0011,
      "step": 16670
    },
    {
      "epoch": 3.380624239967572,
      "grad_norm": 0.010589617304503918,
      "learning_rate": 3.09687880016214e-05,
      "loss": 0.0032,
      "step": 16680
    },
    {
      "epoch": 3.3826509931090394,
      "grad_norm": 0.005919576622545719,
      "learning_rate": 3.086745034454804e-05,
      "loss": 0.3659,
      "step": 16690
    },
    {
      "epoch": 3.3846777462505067,
      "grad_norm": 0.010437965393066406,
      "learning_rate": 3.076611268747467e-05,
      "loss": 0.0058,
      "step": 16700
    },
    {
      "epoch": 3.3846777462505067,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10221080482006073,
      "eval_runtime": 47.2291,
      "eval_samples_per_second": 11.561,
      "eval_steps_per_second": 1.461,
      "step": 16700
    },
    {
      "epoch": 3.386704499391974,
      "grad_norm": 0.005819872487336397,
      "learning_rate": 3.0664775030401296e-05,
      "loss": 0.0029,
      "step": 16710
    },
    {
      "epoch": 3.388731252533441,
      "grad_norm": 0.059754569083452225,
      "learning_rate": 3.056343737332793e-05,
      "loss": 0.0032,
      "step": 16720
    },
    {
      "epoch": 3.390758005674909,
      "grad_norm": 0.007293600123375654,
      "learning_rate": 3.0462099716254562e-05,
      "loss": 0.0026,
      "step": 16730
    },
    {
      "epoch": 3.392784758816376,
      "grad_norm": 0.0075224339962005615,
      "learning_rate": 3.036076205918119e-05,
      "loss": 0.0026,
      "step": 16740
    },
    {
      "epoch": 3.3948115119578435,
      "grad_norm": 0.006849432829767466,
      "learning_rate": 3.0259424402107824e-05,
      "loss": 0.7453,
      "step": 16750
    },
    {
      "epoch": 3.3968382650993107,
      "grad_norm": 0.00680953124538064,
      "learning_rate": 3.015808674503446e-05,
      "loss": 0.0026,
      "step": 16760
    },
    {
      "epoch": 3.3988650182407785,
      "grad_norm": 0.006028030067682266,
      "learning_rate": 3.005674908796109e-05,
      "loss": 0.0014,
      "step": 16770
    },
    {
      "epoch": 3.4008917713822457,
      "grad_norm": 0.00616206182166934,
      "learning_rate": 2.9955411430887718e-05,
      "loss": 0.0038,
      "step": 16780
    },
    {
      "epoch": 3.402918524523713,
      "grad_norm": 0.006383887026458979,
      "learning_rate": 2.985407377381435e-05,
      "loss": 0.0016,
      "step": 16790
    },
    {
      "epoch": 3.4049452776651803,
      "grad_norm": 0.006572426296770573,
      "learning_rate": 2.975273611674098e-05,
      "loss": 0.006,
      "step": 16800
    },
    {
      "epoch": 3.4049452776651803,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.09933307021856308,
      "eval_runtime": 48.804,
      "eval_samples_per_second": 11.188,
      "eval_steps_per_second": 1.414,
      "step": 16800
    },
    {
      "epoch": 3.4069720308066476,
      "grad_norm": 0.005607839208096266,
      "learning_rate": 2.9651398459667613e-05,
      "loss": 0.0017,
      "step": 16810
    },
    {
      "epoch": 3.4089987839481153,
      "grad_norm": 0.005661617498844862,
      "learning_rate": 2.9550060802594242e-05,
      "loss": 0.0034,
      "step": 16820
    },
    {
      "epoch": 3.4110255370895826,
      "grad_norm": 0.03819294273853302,
      "learning_rate": 2.9448723145520878e-05,
      "loss": 0.0017,
      "step": 16830
    },
    {
      "epoch": 3.41305229023105,
      "grad_norm": 0.005522882100194693,
      "learning_rate": 2.934738548844751e-05,
      "loss": 0.0016,
      "step": 16840
    },
    {
      "epoch": 3.415079043372517,
      "grad_norm": 0.005343144293874502,
      "learning_rate": 2.924604783137414e-05,
      "loss": 0.5167,
      "step": 16850
    },
    {
      "epoch": 3.417105796513985,
      "grad_norm": 0.055113568902015686,
      "learning_rate": 2.9144710174300773e-05,
      "loss": 0.5119,
      "step": 16860
    },
    {
      "epoch": 3.419132549655452,
      "grad_norm": 0.06563153862953186,
      "learning_rate": 2.9043372517227402e-05,
      "loss": 0.9829,
      "step": 16870
    },
    {
      "epoch": 3.4211593027969194,
      "grad_norm": 0.02681180089712143,
      "learning_rate": 2.894203486015403e-05,
      "loss": 0.0017,
      "step": 16880
    },
    {
      "epoch": 3.4231860559383867,
      "grad_norm": 0.00725955655798316,
      "learning_rate": 2.8840697203080667e-05,
      "loss": 0.587,
      "step": 16890
    },
    {
      "epoch": 3.425212809079854,
      "grad_norm": 0.007704930379986763,
      "learning_rate": 2.87393595460073e-05,
      "loss": 0.002,
      "step": 16900
    },
    {
      "epoch": 3.425212809079854,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.09541966766119003,
      "eval_runtime": 50.1127,
      "eval_samples_per_second": 10.895,
      "eval_steps_per_second": 1.377,
      "step": 16900
    },
    {
      "epoch": 3.427239562221321,
      "grad_norm": 0.06465483456850052,
      "learning_rate": 2.863802188893393e-05,
      "loss": 0.0066,
      "step": 16910
    },
    {
      "epoch": 3.429266315362789,
      "grad_norm": 8.011322021484375,
      "learning_rate": 2.8536684231860562e-05,
      "loss": 0.4889,
      "step": 16920
    },
    {
      "epoch": 3.431293068504256,
      "grad_norm": 0.061154309660196304,
      "learning_rate": 2.843534657478719e-05,
      "loss": 0.0023,
      "step": 16930
    },
    {
      "epoch": 3.4333198216457235,
      "grad_norm": 0.006984674371778965,
      "learning_rate": 2.8334008917713824e-05,
      "loss": 0.0016,
      "step": 16940
    },
    {
      "epoch": 3.4353465747871907,
      "grad_norm": 0.013019927777349949,
      "learning_rate": 2.8232671260640453e-05,
      "loss": 0.4919,
      "step": 16950
    },
    {
      "epoch": 3.4373733279286585,
      "grad_norm": 0.006715769879519939,
      "learning_rate": 2.813133360356709e-05,
      "loss": 0.003,
      "step": 16960
    },
    {
      "epoch": 3.4394000810701257,
      "grad_norm": 0.0073410640470683575,
      "learning_rate": 2.802999594649372e-05,
      "loss": 0.9557,
      "step": 16970
    },
    {
      "epoch": 3.441426834211593,
      "grad_norm": 0.007587122730910778,
      "learning_rate": 2.792865828942035e-05,
      "loss": 0.0015,
      "step": 16980
    },
    {
      "epoch": 3.4434535873530603,
      "grad_norm": 0.007860773243010044,
      "learning_rate": 2.782732063234698e-05,
      "loss": 0.0043,
      "step": 16990
    },
    {
      "epoch": 3.4454803404945276,
      "grad_norm": 0.030218765139579773,
      "learning_rate": 2.7725982975273613e-05,
      "loss": 0.0082,
      "step": 17000
    },
    {
      "epoch": 3.4454803404945276,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.09758459776639938,
      "eval_runtime": 47.6392,
      "eval_samples_per_second": 11.461,
      "eval_steps_per_second": 1.448,
      "step": 17000
    },
    {
      "epoch": 3.4475070936359953,
      "grad_norm": 0.007678163703531027,
      "learning_rate": 2.7624645318200242e-05,
      "loss": 0.0026,
      "step": 17010
    },
    {
      "epoch": 3.4495338467774626,
      "grad_norm": 0.006676503922790289,
      "learning_rate": 2.7523307661126875e-05,
      "loss": 0.0029,
      "step": 17020
    },
    {
      "epoch": 3.45156059991893,
      "grad_norm": 0.011629360727965832,
      "learning_rate": 2.742197000405351e-05,
      "loss": 0.0026,
      "step": 17030
    },
    {
      "epoch": 3.453587353060397,
      "grad_norm": 0.0064224074594676495,
      "learning_rate": 2.732063234698014e-05,
      "loss": 0.0011,
      "step": 17040
    },
    {
      "epoch": 3.455614106201865,
      "grad_norm": 0.009052238427102566,
      "learning_rate": 2.7219294689906773e-05,
      "loss": 0.0019,
      "step": 17050
    },
    {
      "epoch": 3.457640859343332,
      "grad_norm": 0.006074480712413788,
      "learning_rate": 2.7117957032833402e-05,
      "loss": 0.0119,
      "step": 17060
    },
    {
      "epoch": 3.4596676124847994,
      "grad_norm": 0.00601073307916522,
      "learning_rate": 2.701661937576003e-05,
      "loss": 0.0026,
      "step": 17070
    },
    {
      "epoch": 3.4616943656262666,
      "grad_norm": 0.006008122581988573,
      "learning_rate": 2.6915281718686664e-05,
      "loss": 0.003,
      "step": 17080
    },
    {
      "epoch": 3.463721118767734,
      "grad_norm": 0.006180489901453257,
      "learning_rate": 2.6813944061613293e-05,
      "loss": 0.0022,
      "step": 17090
    },
    {
      "epoch": 3.4657478719092016,
      "grad_norm": 0.005895602982491255,
      "learning_rate": 2.671260640453993e-05,
      "loss": 0.0029,
      "step": 17100
    },
    {
      "epoch": 3.4657478719092016,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.0977577343583107,
      "eval_runtime": 48.4339,
      "eval_samples_per_second": 11.273,
      "eval_steps_per_second": 1.425,
      "step": 17100
    },
    {
      "epoch": 3.467774625050669,
      "grad_norm": 0.06110497564077377,
      "learning_rate": 2.6611268747466562e-05,
      "loss": 0.0112,
      "step": 17110
    },
    {
      "epoch": 3.469801378192136,
      "grad_norm": 0.0056753759272396564,
      "learning_rate": 2.650993109039319e-05,
      "loss": 0.0034,
      "step": 17120
    },
    {
      "epoch": 3.4718281313336035,
      "grad_norm": 0.006453051697462797,
      "learning_rate": 2.6408593433319824e-05,
      "loss": 0.2768,
      "step": 17130
    },
    {
      "epoch": 3.473854884475071,
      "grad_norm": 0.005545321386307478,
      "learning_rate": 2.6307255776246453e-05,
      "loss": 0.0034,
      "step": 17140
    },
    {
      "epoch": 3.4758816376165385,
      "grad_norm": 0.005507095716893673,
      "learning_rate": 2.6205918119173086e-05,
      "loss": 0.0008,
      "step": 17150
    },
    {
      "epoch": 3.4779083907580057,
      "grad_norm": 0.005461165215820074,
      "learning_rate": 2.6104580462099715e-05,
      "loss": 0.0039,
      "step": 17160
    },
    {
      "epoch": 3.479935143899473,
      "grad_norm": 0.07790371030569077,
      "learning_rate": 2.600324280502635e-05,
      "loss": 0.0028,
      "step": 17170
    },
    {
      "epoch": 3.4819618970409403,
      "grad_norm": 0.005248835310339928,
      "learning_rate": 2.590190514795298e-05,
      "loss": 0.0019,
      "step": 17180
    },
    {
      "epoch": 3.4839886501824076,
      "grad_norm": 0.006307022646069527,
      "learning_rate": 2.5800567490879613e-05,
      "loss": 0.5085,
      "step": 17190
    },
    {
      "epoch": 3.4860154033238753,
      "grad_norm": 0.005371470469981432,
      "learning_rate": 2.5699229833806242e-05,
      "loss": 0.0008,
      "step": 17200
    },
    {
      "epoch": 3.4860154033238753,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.09734708815813065,
      "eval_runtime": 50.2153,
      "eval_samples_per_second": 10.873,
      "eval_steps_per_second": 1.374,
      "step": 17200
    },
    {
      "epoch": 3.4880421564653425,
      "grad_norm": 0.05594426393508911,
      "learning_rate": 2.5597892176732875e-05,
      "loss": 0.002,
      "step": 17210
    },
    {
      "epoch": 3.49006890960681,
      "grad_norm": 0.05673331022262573,
      "learning_rate": 2.5496554519659504e-05,
      "loss": 0.0019,
      "step": 17220
    },
    {
      "epoch": 3.492095662748277,
      "grad_norm": 0.005430069286376238,
      "learning_rate": 2.539521686258614e-05,
      "loss": 0.0031,
      "step": 17230
    },
    {
      "epoch": 3.494122415889745,
      "grad_norm": 8.014272689819336,
      "learning_rate": 2.5293879205512773e-05,
      "loss": 0.5023,
      "step": 17240
    },
    {
      "epoch": 3.496149169031212,
      "grad_norm": 0.006552092730998993,
      "learning_rate": 2.5192541548439402e-05,
      "loss": 0.0026,
      "step": 17250
    },
    {
      "epoch": 3.4981759221726794,
      "grad_norm": 0.0060019418597221375,
      "learning_rate": 2.5091203891366035e-05,
      "loss": 0.0018,
      "step": 17260
    },
    {
      "epoch": 3.5002026753141466,
      "grad_norm": 0.0064520156010985374,
      "learning_rate": 2.4989866234292664e-05,
      "loss": 0.7485,
      "step": 17270
    },
    {
      "epoch": 3.502229428455614,
      "grad_norm": 0.059147171676158905,
      "learning_rate": 2.4888528577219296e-05,
      "loss": 0.0029,
      "step": 17280
    },
    {
      "epoch": 3.5042561815970816,
      "grad_norm": 0.005694244988262653,
      "learning_rate": 2.478719092014593e-05,
      "loss": 0.0013,
      "step": 17290
    },
    {
      "epoch": 3.506282934738549,
      "grad_norm": 0.005751201882958412,
      "learning_rate": 2.468585326307256e-05,
      "loss": 0.0014,
      "step": 17300
    },
    {
      "epoch": 3.506282934738549,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.09791267663240433,
      "eval_runtime": 46.9778,
      "eval_samples_per_second": 11.623,
      "eval_steps_per_second": 1.469,
      "step": 17300
    },
    {
      "epoch": 3.508309687880016,
      "grad_norm": 0.05871247500181198,
      "learning_rate": 2.458451560599919e-05,
      "loss": 0.457,
      "step": 17310
    },
    {
      "epoch": 3.5103364410214835,
      "grad_norm": 0.006226697936654091,
      "learning_rate": 2.4483177948925824e-05,
      "loss": 0.0022,
      "step": 17320
    },
    {
      "epoch": 3.512363194162951,
      "grad_norm": 0.00761007983237505,
      "learning_rate": 2.4381840291852453e-05,
      "loss": 0.0013,
      "step": 17330
    },
    {
      "epoch": 3.5143899473044184,
      "grad_norm": 0.00598880834877491,
      "learning_rate": 2.4280502634779086e-05,
      "loss": 0.4966,
      "step": 17340
    },
    {
      "epoch": 3.5164167004458857,
      "grad_norm": 0.006399375386536121,
      "learning_rate": 2.4179164977705718e-05,
      "loss": 0.0006,
      "step": 17350
    },
    {
      "epoch": 3.518443453587353,
      "grad_norm": 0.0067553287371993065,
      "learning_rate": 2.4077827320632347e-05,
      "loss": 0.0017,
      "step": 17360
    },
    {
      "epoch": 3.5204702067288203,
      "grad_norm": 0.006269978359341621,
      "learning_rate": 2.397648966355898e-05,
      "loss": 0.0846,
      "step": 17370
    },
    {
      "epoch": 3.5224969598702875,
      "grad_norm": 0.005716945044696331,
      "learning_rate": 2.387515200648561e-05,
      "loss": 0.1505,
      "step": 17380
    },
    {
      "epoch": 3.5245237130117553,
      "grad_norm": 0.005705594085156918,
      "learning_rate": 2.3773814349412242e-05,
      "loss": 0.0044,
      "step": 17390
    },
    {
      "epoch": 3.5265504661532225,
      "grad_norm": 0.01609107106924057,
      "learning_rate": 2.3672476692338875e-05,
      "loss": 0.0008,
      "step": 17400
    },
    {
      "epoch": 3.5265504661532225,
      "eval_accuracy": 0.9743589743589743,
      "eval_loss": 0.11509870737791061,
      "eval_runtime": 50.238,
      "eval_samples_per_second": 10.868,
      "eval_steps_per_second": 1.373,
      "step": 17400
    },
    {
      "epoch": 3.52857721929469,
      "grad_norm": 0.005944013595581055,
      "learning_rate": 2.3571139035265504e-05,
      "loss": 0.0013,
      "step": 17410
    },
    {
      "epoch": 3.5306039724361575,
      "grad_norm": 0.3377014398574829,
      "learning_rate": 2.346980137819214e-05,
      "loss": 0.0054,
      "step": 17420
    },
    {
      "epoch": 3.532630725577625,
      "grad_norm": 0.005403846502304077,
      "learning_rate": 2.336846372111877e-05,
      "loss": 0.0033,
      "step": 17430
    },
    {
      "epoch": 3.534657478719092,
      "grad_norm": 0.0827544778585434,
      "learning_rate": 2.32671260640454e-05,
      "loss": 0.0028,
      "step": 17440
    },
    {
      "epoch": 3.5366842318605594,
      "grad_norm": 0.006947341840714216,
      "learning_rate": 2.3165788406972035e-05,
      "loss": 0.493,
      "step": 17450
    },
    {
      "epoch": 3.5387109850020266,
      "grad_norm": 0.005381606053560972,
      "learning_rate": 2.3064450749898664e-05,
      "loss": 0.0018,
      "step": 17460
    },
    {
      "epoch": 3.540737738143494,
      "grad_norm": 0.00517958914861083,
      "learning_rate": 2.2963113092825293e-05,
      "loss": 0.0037,
      "step": 17470
    },
    {
      "epoch": 3.5427644912849616,
      "grad_norm": 0.0052943965420126915,
      "learning_rate": 2.2861775435751926e-05,
      "loss": 0.0015,
      "step": 17480
    },
    {
      "epoch": 3.544791244426429,
      "grad_norm": 0.0053497194312512875,
      "learning_rate": 2.276043777867856e-05,
      "loss": 0.0013,
      "step": 17490
    },
    {
      "epoch": 3.546817997567896,
      "grad_norm": 0.005343489348888397,
      "learning_rate": 2.265910012160519e-05,
      "loss": 0.0023,
      "step": 17500
    },
    {
      "epoch": 3.546817997567896,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10926567018032074,
      "eval_runtime": 50.1658,
      "eval_samples_per_second": 10.884,
      "eval_steps_per_second": 1.375,
      "step": 17500
    },
    {
      "epoch": 3.5488447507093634,
      "grad_norm": 0.05171739682555199,
      "learning_rate": 2.255776246453182e-05,
      "loss": 0.0032,
      "step": 17510
    },
    {
      "epoch": 3.550871503850831,
      "grad_norm": 0.004982843995094299,
      "learning_rate": 2.2456424807458453e-05,
      "loss": 0.2611,
      "step": 17520
    },
    {
      "epoch": 3.5528982569922984,
      "grad_norm": 0.4834737181663513,
      "learning_rate": 2.2355087150385086e-05,
      "loss": 0.0103,
      "step": 17530
    },
    {
      "epoch": 3.5549250101337657,
      "grad_norm": 0.05277622491121292,
      "learning_rate": 2.2253749493311715e-05,
      "loss": 0.5101,
      "step": 17540
    },
    {
      "epoch": 3.556951763275233,
      "grad_norm": 0.005033069755882025,
      "learning_rate": 2.2152411836238347e-05,
      "loss": 0.0018,
      "step": 17550
    },
    {
      "epoch": 3.5589785164167003,
      "grad_norm": 0.005153517704457045,
      "learning_rate": 2.205107417916498e-05,
      "loss": 0.0039,
      "step": 17560
    },
    {
      "epoch": 3.5610052695581675,
      "grad_norm": 0.005008082836866379,
      "learning_rate": 2.194973652209161e-05,
      "loss": 0.0012,
      "step": 17570
    },
    {
      "epoch": 3.5630320226996353,
      "grad_norm": 0.005480339750647545,
      "learning_rate": 2.1848398865018242e-05,
      "loss": 0.0027,
      "step": 17580
    },
    {
      "epoch": 3.5650587758411025,
      "grad_norm": 0.004882423207163811,
      "learning_rate": 2.1747061207944875e-05,
      "loss": 0.0011,
      "step": 17590
    },
    {
      "epoch": 3.56708552898257,
      "grad_norm": 0.006747744046151638,
      "learning_rate": 2.1645723550871504e-05,
      "loss": 0.0012,
      "step": 17600
    },
    {
      "epoch": 3.56708552898257,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.0995689406991005,
      "eval_runtime": 46.2745,
      "eval_samples_per_second": 11.799,
      "eval_steps_per_second": 1.491,
      "step": 17600
    },
    {
      "epoch": 3.5691122821240375,
      "grad_norm": 0.0049659782089293,
      "learning_rate": 2.1544385893798137e-05,
      "loss": 0.0018,
      "step": 17610
    },
    {
      "epoch": 3.571139035265505,
      "grad_norm": 0.0050674243830144405,
      "learning_rate": 2.144304823672477e-05,
      "loss": 0.0011,
      "step": 17620
    },
    {
      "epoch": 3.573165788406972,
      "grad_norm": 0.048832111060619354,
      "learning_rate": 2.13417105796514e-05,
      "loss": 0.0035,
      "step": 17630
    },
    {
      "epoch": 3.5751925415484394,
      "grad_norm": 0.008065484464168549,
      "learning_rate": 2.124037292257803e-05,
      "loss": 0.0022,
      "step": 17640
    },
    {
      "epoch": 3.5772192946899066,
      "grad_norm": 0.004766256082803011,
      "learning_rate": 2.113903526550466e-05,
      "loss": 0.0009,
      "step": 17650
    },
    {
      "epoch": 3.579246047831374,
      "grad_norm": 0.005500113591551781,
      "learning_rate": 2.1037697608431296e-05,
      "loss": 0.0011,
      "step": 17660
    },
    {
      "epoch": 3.5812728009728416,
      "grad_norm": 0.004638390615582466,
      "learning_rate": 2.0936359951357926e-05,
      "loss": 0.0018,
      "step": 17670
    },
    {
      "epoch": 3.583299554114309,
      "grad_norm": 0.0049532013945281506,
      "learning_rate": 2.0835022294284555e-05,
      "loss": 0.7605,
      "step": 17680
    },
    {
      "epoch": 3.585326307255776,
      "grad_norm": 0.005197311285883188,
      "learning_rate": 2.073368463721119e-05,
      "loss": 0.0016,
      "step": 17690
    },
    {
      "epoch": 3.587353060397244,
      "grad_norm": 0.0053718313574790955,
      "learning_rate": 2.063234698013782e-05,
      "loss": 0.0016,
      "step": 17700
    },
    {
      "epoch": 3.587353060397244,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.09795909374952316,
      "eval_runtime": 50.1492,
      "eval_samples_per_second": 10.888,
      "eval_steps_per_second": 1.376,
      "step": 17700
    },
    {
      "epoch": 3.589379813538711,
      "grad_norm": 0.0051916828379035,
      "learning_rate": 2.0531009323064453e-05,
      "loss": 0.0016,
      "step": 17710
    },
    {
      "epoch": 3.5914065666801784,
      "grad_norm": 0.005293379537761211,
      "learning_rate": 2.0429671665991082e-05,
      "loss": 0.7548,
      "step": 17720
    },
    {
      "epoch": 3.5934333198216457,
      "grad_norm": 0.00580208096653223,
      "learning_rate": 2.0328334008917715e-05,
      "loss": 0.0006,
      "step": 17730
    },
    {
      "epoch": 3.595460072963113,
      "grad_norm": 0.005405276548117399,
      "learning_rate": 2.0226996351844347e-05,
      "loss": 0.0023,
      "step": 17740
    },
    {
      "epoch": 3.5974868261045803,
      "grad_norm": 0.0055334181524813175,
      "learning_rate": 2.0125658694770977e-05,
      "loss": 0.2852,
      "step": 17750
    },
    {
      "epoch": 3.599513579246048,
      "grad_norm": 0.00532138254493475,
      "learning_rate": 2.002432103769761e-05,
      "loss": 0.0022,
      "step": 17760
    },
    {
      "epoch": 3.6015403323875153,
      "grad_norm": 0.04558062180876732,
      "learning_rate": 1.9922983380624242e-05,
      "loss": 0.0011,
      "step": 17770
    },
    {
      "epoch": 3.6035670855289825,
      "grad_norm": 0.005479319021105766,
      "learning_rate": 1.982164572355087e-05,
      "loss": 0.0017,
      "step": 17780
    },
    {
      "epoch": 3.60559383867045,
      "grad_norm": 0.005388789344578981,
      "learning_rate": 1.9720308066477504e-05,
      "loss": 0.0012,
      "step": 17790
    },
    {
      "epoch": 3.6076205918119175,
      "grad_norm": 0.00526474229991436,
      "learning_rate": 1.9618970409404137e-05,
      "loss": 0.0015,
      "step": 17800
    },
    {
      "epoch": 3.6076205918119175,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10518113523721695,
      "eval_runtime": 50.1299,
      "eval_samples_per_second": 10.892,
      "eval_steps_per_second": 1.376,
      "step": 17800
    },
    {
      "epoch": 3.609647344953385,
      "grad_norm": 0.005050942301750183,
      "learning_rate": 1.9517632752330766e-05,
      "loss": 0.0039,
      "step": 17810
    },
    {
      "epoch": 3.611674098094852,
      "grad_norm": 0.005604945123195648,
      "learning_rate": 1.94162950952574e-05,
      "loss": 0.0021,
      "step": 17820
    },
    {
      "epoch": 3.6137008512363193,
      "grad_norm": 0.005479446146637201,
      "learning_rate": 1.931495743818403e-05,
      "loss": 0.0015,
      "step": 17830
    },
    {
      "epoch": 3.6157276043777866,
      "grad_norm": 0.004991782829165459,
      "learning_rate": 1.921361978111066e-05,
      "loss": 0.001,
      "step": 17840
    },
    {
      "epoch": 3.617754357519254,
      "grad_norm": 0.005143380258232355,
      "learning_rate": 1.9112282124037293e-05,
      "loss": 0.5258,
      "step": 17850
    },
    {
      "epoch": 3.6197811106607216,
      "grad_norm": 0.005482929293066263,
      "learning_rate": 1.9010944466963926e-05,
      "loss": 0.005,
      "step": 17860
    },
    {
      "epoch": 3.621807863802189,
      "grad_norm": 0.005034882575273514,
      "learning_rate": 1.8909606809890555e-05,
      "loss": 0.0011,
      "step": 17870
    },
    {
      "epoch": 3.623834616943656,
      "grad_norm": 0.005108018405735493,
      "learning_rate": 1.8808269152817188e-05,
      "loss": 0.0011,
      "step": 17880
    },
    {
      "epoch": 3.625861370085124,
      "grad_norm": 0.004975921008735895,
      "learning_rate": 1.8706931495743817e-05,
      "loss": 0.0015,
      "step": 17890
    },
    {
      "epoch": 3.627888123226591,
      "grad_norm": 0.005141813308000565,
      "learning_rate": 1.8605593838670453e-05,
      "loss": 0.0018,
      "step": 17900
    },
    {
      "epoch": 3.627888123226591,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10540201514959335,
      "eval_runtime": 46.0356,
      "eval_samples_per_second": 11.86,
      "eval_steps_per_second": 1.499,
      "step": 17900
    },
    {
      "epoch": 3.6299148763680584,
      "grad_norm": 0.04432378336787224,
      "learning_rate": 1.8504256181597082e-05,
      "loss": 0.0026,
      "step": 17910
    },
    {
      "epoch": 3.6319416295095257,
      "grad_norm": 0.005204243119806051,
      "learning_rate": 1.840291852452371e-05,
      "loss": 0.0021,
      "step": 17920
    },
    {
      "epoch": 3.633968382650993,
      "grad_norm": 0.005375333596020937,
      "learning_rate": 1.8301580867450347e-05,
      "loss": 0.7511,
      "step": 17930
    },
    {
      "epoch": 3.6359951357924603,
      "grad_norm": 0.042768221348524094,
      "learning_rate": 1.8200243210376977e-05,
      "loss": 0.0015,
      "step": 17940
    },
    {
      "epoch": 3.638021888933928,
      "grad_norm": 0.008297201246023178,
      "learning_rate": 1.809890555330361e-05,
      "loss": 0.0006,
      "step": 17950
    },
    {
      "epoch": 3.6400486420753952,
      "grad_norm": 0.005426391493529081,
      "learning_rate": 1.7997567896230242e-05,
      "loss": 0.7386,
      "step": 17960
    },
    {
      "epoch": 3.6420753952168625,
      "grad_norm": 0.005903474520891905,
      "learning_rate": 1.789623023915687e-05,
      "loss": 0.0011,
      "step": 17970
    },
    {
      "epoch": 3.64410214835833,
      "grad_norm": 0.006364389322698116,
      "learning_rate": 1.7794892582083504e-05,
      "loss": 0.0018,
      "step": 17980
    },
    {
      "epoch": 3.6461289014997975,
      "grad_norm": 0.04269855469465256,
      "learning_rate": 1.7693554925010133e-05,
      "loss": 0.0057,
      "step": 17990
    },
    {
      "epoch": 3.648155654641265,
      "grad_norm": 0.04199114069342613,
      "learning_rate": 1.7592217267936766e-05,
      "loss": 0.003,
      "step": 18000
    },
    {
      "epoch": 3.648155654641265,
      "eval_accuracy": 0.978021978021978,
      "eval_loss": 0.10523013025522232,
      "eval_runtime": 51.7239,
      "eval_samples_per_second": 10.556,
      "eval_steps_per_second": 1.334,
      "step": 18000
    },
    {
      "epoch": 3.650182407782732,
      "grad_norm": 0.006271640304476023,
      "learning_rate": 1.74908796108634e-05,
      "loss": 0.0032,
      "step": 18010
    },
    {
      "epoch": 3.6522091609241993,
      "grad_norm": 0.005405428819358349,
      "learning_rate": 1.7389541953790028e-05,
      "loss": 0.0016,
      "step": 18020
    },
    {
      "epoch": 3.6542359140656666,
      "grad_norm": 0.005375401582568884,
      "learning_rate": 1.728820429671666e-05,
      "loss": 0.001,
      "step": 18030
    },
    {
      "epoch": 3.6562626672071343,
      "grad_norm": 0.00598566560074687,
      "learning_rate": 1.7186866639643293e-05,
      "loss": 0.0015,
      "step": 18040
    },
    {
      "epoch": 3.6582894203486016,
      "grad_norm": 0.005265843588858843,
      "learning_rate": 1.7085528982569922e-05,
      "loss": 0.0011,
      "step": 18050
    },
    {
      "epoch": 3.660316173490069,
      "grad_norm": 0.040170665830373764,
      "learning_rate": 1.6984191325496555e-05,
      "loss": 0.3142,
      "step": 18060
    },
    {
      "epoch": 3.662342926631536,
      "grad_norm": 0.040453970432281494,
      "learning_rate": 1.6882853668423188e-05,
      "loss": 0.0015,
      "step": 18070
    },
    {
      "epoch": 3.664369679773004,
      "grad_norm": 0.09230988472700119,
      "learning_rate": 1.6781516011349817e-05,
      "loss": 0.0015,
      "step": 18080
    },
    {
      "epoch": 3.666396432914471,
      "grad_norm": 0.006269339006394148,
      "learning_rate": 1.668017835427645e-05,
      "loss": 0.0015,
      "step": 18090
    },
    {
      "epoch": 3.6684231860559384,
      "grad_norm": 0.005188298877328634,
      "learning_rate": 1.6578840697203082e-05,
      "loss": 0.002,
      "step": 18100
    },
    {
      "epoch": 3.6684231860559384,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10631264001131058,
      "eval_runtime": 52.7075,
      "eval_samples_per_second": 10.359,
      "eval_steps_per_second": 1.309,
      "step": 18100
    },
    {
      "epoch": 3.6704499391974057,
      "grad_norm": 0.005731545854359865,
      "learning_rate": 1.647750304012971e-05,
      "loss": 0.0024,
      "step": 18110
    },
    {
      "epoch": 3.672476692338873,
      "grad_norm": 0.005191850941628218,
      "learning_rate": 1.6376165383056344e-05,
      "loss": 0.0013,
      "step": 18120
    },
    {
      "epoch": 3.6745034454803402,
      "grad_norm": 0.005665656179189682,
      "learning_rate": 1.6274827725982977e-05,
      "loss": 0.0011,
      "step": 18130
    },
    {
      "epoch": 3.676530198621808,
      "grad_norm": 0.03839230164885521,
      "learning_rate": 1.617349006890961e-05,
      "loss": 0.0014,
      "step": 18140
    },
    {
      "epoch": 3.6785569517632752,
      "grad_norm": 0.005275613162666559,
      "learning_rate": 1.607215241183624e-05,
      "loss": 0.7439,
      "step": 18150
    },
    {
      "epoch": 3.6805837049047425,
      "grad_norm": 0.005713291000574827,
      "learning_rate": 1.5970814754762868e-05,
      "loss": 0.3076,
      "step": 18160
    },
    {
      "epoch": 3.6826104580462102,
      "grad_norm": 0.005449062213301659,
      "learning_rate": 1.5869477097689504e-05,
      "loss": 0.0014,
      "step": 18170
    },
    {
      "epoch": 3.6846372111876775,
      "grad_norm": 0.0056668976321816444,
      "learning_rate": 1.5768139440616133e-05,
      "loss": 0.0006,
      "step": 18180
    },
    {
      "epoch": 3.686663964329145,
      "grad_norm": 0.006036930251866579,
      "learning_rate": 1.5666801783542766e-05,
      "loss": 0.0022,
      "step": 18190
    },
    {
      "epoch": 3.688690717470612,
      "grad_norm": 0.005712085869163275,
      "learning_rate": 1.55654641264694e-05,
      "loss": 0.0011,
      "step": 18200
    },
    {
      "epoch": 3.688690717470612,
      "eval_accuracy": 0.9761904761904762,
      "eval_loss": 0.11952827125787735,
      "eval_runtime": 46.1217,
      "eval_samples_per_second": 11.838,
      "eval_steps_per_second": 1.496,
      "step": 18200
    },
    {
      "epoch": 3.6907174706120793,
      "grad_norm": 0.037588682025671005,
      "learning_rate": 1.5464126469396028e-05,
      "loss": 0.0047,
      "step": 18210
    },
    {
      "epoch": 3.6927442237535466,
      "grad_norm": 0.0060033174231648445,
      "learning_rate": 1.536278881232266e-05,
      "loss": 0.2324,
      "step": 18220
    },
    {
      "epoch": 3.6947709768950143,
      "grad_norm": 0.005877983290702105,
      "learning_rate": 1.526145115524929e-05,
      "loss": 0.0008,
      "step": 18230
    },
    {
      "epoch": 3.6967977300364816,
      "grad_norm": 0.06802316009998322,
      "learning_rate": 1.5160113498175924e-05,
      "loss": 0.0017,
      "step": 18240
    },
    {
      "epoch": 3.698824483177949,
      "grad_norm": 0.0809393897652626,
      "learning_rate": 1.5058775841102555e-05,
      "loss": 0.0014,
      "step": 18250
    },
    {
      "epoch": 3.700851236319416,
      "grad_norm": 0.0054187779314816,
      "learning_rate": 1.4957438184029184e-05,
      "loss": 0.001,
      "step": 18260
    },
    {
      "epoch": 3.702877989460884,
      "grad_norm": 0.005456519313156605,
      "learning_rate": 1.4856100526955819e-05,
      "loss": 0.0016,
      "step": 18270
    },
    {
      "epoch": 3.704904742602351,
      "grad_norm": 0.005753775592893362,
      "learning_rate": 1.475476286988245e-05,
      "loss": 0.0015,
      "step": 18280
    },
    {
      "epoch": 3.7069314957438184,
      "grad_norm": 0.006126856431365013,
      "learning_rate": 1.465342521280908e-05,
      "loss": 0.7524,
      "step": 18290
    },
    {
      "epoch": 3.7089582488852857,
      "grad_norm": 0.03605639562010765,
      "learning_rate": 1.4552087555735713e-05,
      "loss": 0.4766,
      "step": 18300
    },
    {
      "epoch": 3.7089582488852857,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.08734480291604996,
      "eval_runtime": 50.1542,
      "eval_samples_per_second": 10.886,
      "eval_steps_per_second": 1.376,
      "step": 18300
    },
    {
      "epoch": 3.710985002026753,
      "grad_norm": 0.0060221729800105095,
      "learning_rate": 1.4450749898662344e-05,
      "loss": 0.001,
      "step": 18310
    },
    {
      "epoch": 3.7130117551682202,
      "grad_norm": 0.08299941569566727,
      "learning_rate": 1.4349412241588975e-05,
      "loss": 0.0013,
      "step": 18320
    },
    {
      "epoch": 3.715038508309688,
      "grad_norm": 0.006154602859169245,
      "learning_rate": 1.4248074584515606e-05,
      "loss": 0.0011,
      "step": 18330
    },
    {
      "epoch": 3.7170652614511552,
      "grad_norm": 0.03808287903666496,
      "learning_rate": 1.4146736927442239e-05,
      "loss": 0.0023,
      "step": 18340
    },
    {
      "epoch": 3.7190920145926225,
      "grad_norm": 0.005551365204155445,
      "learning_rate": 1.404539927036887e-05,
      "loss": 0.0014,
      "step": 18350
    },
    {
      "epoch": 3.7211187677340902,
      "grad_norm": 0.005745531991124153,
      "learning_rate": 1.39440616132955e-05,
      "loss": 0.0014,
      "step": 18360
    },
    {
      "epoch": 3.7231455208755575,
      "grad_norm": 0.006904330104589462,
      "learning_rate": 1.3842723956222133e-05,
      "loss": 0.001,
      "step": 18370
    },
    {
      "epoch": 3.7251722740170248,
      "grad_norm": 0.006295742932707071,
      "learning_rate": 1.3741386299148764e-05,
      "loss": 0.001,
      "step": 18380
    },
    {
      "epoch": 3.727199027158492,
      "grad_norm": 0.024064790457487106,
      "learning_rate": 1.3640048642075395e-05,
      "loss": 0.0018,
      "step": 18390
    },
    {
      "epoch": 3.7292257802999593,
      "grad_norm": 0.005999110173434019,
      "learning_rate": 1.3538710985002026e-05,
      "loss": 0.0026,
      "step": 18400
    },
    {
      "epoch": 3.7292257802999593,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.08764051645994186,
      "eval_runtime": 49.6519,
      "eval_samples_per_second": 10.997,
      "eval_steps_per_second": 1.39,
      "step": 18400
    },
    {
      "epoch": 3.7312525334414266,
      "grad_norm": 0.005963246803730726,
      "learning_rate": 1.3437373327928659e-05,
      "loss": 0.0006,
      "step": 18410
    },
    {
      "epoch": 3.7332792865828943,
      "grad_norm": 0.00517231086269021,
      "learning_rate": 1.333603567085529e-05,
      "loss": 0.001,
      "step": 18420
    },
    {
      "epoch": 3.7353060397243616,
      "grad_norm": 0.034096866846084595,
      "learning_rate": 1.323469801378192e-05,
      "loss": 0.0017,
      "step": 18430
    },
    {
      "epoch": 3.737332792865829,
      "grad_norm": 0.005253325682133436,
      "learning_rate": 1.3133360356708555e-05,
      "loss": 0.001,
      "step": 18440
    },
    {
      "epoch": 3.7393595460072966,
      "grad_norm": 0.034667935222387314,
      "learning_rate": 1.3032022699635186e-05,
      "loss": 0.0009,
      "step": 18450
    },
    {
      "epoch": 3.741386299148764,
      "grad_norm": 0.0051004644483327866,
      "learning_rate": 1.2930685042561815e-05,
      "loss": 0.0006,
      "step": 18460
    },
    {
      "epoch": 3.743413052290231,
      "grad_norm": 0.005032215733081102,
      "learning_rate": 1.282934738548845e-05,
      "loss": 0.0015,
      "step": 18470
    },
    {
      "epoch": 3.7454398054316984,
      "grad_norm": 0.005197310354560614,
      "learning_rate": 1.272800972841508e-05,
      "loss": 0.5533,
      "step": 18480
    },
    {
      "epoch": 3.7474665585731657,
      "grad_norm": 0.005205828696489334,
      "learning_rate": 1.2626672071341711e-05,
      "loss": 0.0021,
      "step": 18490
    },
    {
      "epoch": 3.749493311714633,
      "grad_norm": 0.005383576266467571,
      "learning_rate": 1.252533441426834e-05,
      "loss": 0.0006,
      "step": 18500
    },
    {
      "epoch": 3.749493311714633,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.09415023028850555,
      "eval_runtime": 47.0485,
      "eval_samples_per_second": 11.605,
      "eval_steps_per_second": 1.467,
      "step": 18500
    },
    {
      "epoch": 3.7515200648561007,
      "grad_norm": 0.034614093601703644,
      "learning_rate": 1.2423996757194973e-05,
      "loss": 0.0014,
      "step": 18510
    },
    {
      "epoch": 3.753546817997568,
      "grad_norm": 0.004989073611795902,
      "learning_rate": 1.2322659100121606e-05,
      "loss": 0.0012,
      "step": 18520
    },
    {
      "epoch": 3.7555735711390352,
      "grad_norm": 0.005143389571458101,
      "learning_rate": 1.2221321443048237e-05,
      "loss": 0.0007,
      "step": 18530
    },
    {
      "epoch": 3.7576003242805025,
      "grad_norm": 0.005387092940509319,
      "learning_rate": 1.2119983785974868e-05,
      "loss": 0.0012,
      "step": 18540
    },
    {
      "epoch": 3.75962707742197,
      "grad_norm": 0.005190431606024504,
      "learning_rate": 1.20186461289015e-05,
      "loss": 0.006,
      "step": 18550
    },
    {
      "epoch": 3.7616538305634375,
      "grad_norm": 0.005023384466767311,
      "learning_rate": 1.1917308471828132e-05,
      "loss": 0.001,
      "step": 18560
    },
    {
      "epoch": 3.7636805837049048,
      "grad_norm": 0.005527614150196314,
      "learning_rate": 1.1815970814754764e-05,
      "loss": 0.545,
      "step": 18570
    },
    {
      "epoch": 3.765707336846372,
      "grad_norm": 0.005237712990492582,
      "learning_rate": 1.1714633157681395e-05,
      "loss": 0.0006,
      "step": 18580
    },
    {
      "epoch": 3.7677340899878393,
      "grad_norm": 0.005363814067095518,
      "learning_rate": 1.1613295500608026e-05,
      "loss": 0.0013,
      "step": 18590
    },
    {
      "epoch": 3.7697608431293066,
      "grad_norm": 0.036191727966070175,
      "learning_rate": 1.1511957843534659e-05,
      "loss": 0.0014,
      "step": 18600
    },
    {
      "epoch": 3.7697608431293066,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.09435998648405075,
      "eval_runtime": 49.8122,
      "eval_samples_per_second": 10.961,
      "eval_steps_per_second": 1.385,
      "step": 18600
    },
    {
      "epoch": 3.7717875962707743,
      "grad_norm": 0.0053384387865662575,
      "learning_rate": 1.141062018646129e-05,
      "loss": 0.001,
      "step": 18610
    },
    {
      "epoch": 3.7738143494122416,
      "grad_norm": 0.0050865658558905125,
      "learning_rate": 1.130928252938792e-05,
      "loss": 0.0073,
      "step": 18620
    },
    {
      "epoch": 3.775841102553709,
      "grad_norm": 0.004879043437540531,
      "learning_rate": 1.1207944872314553e-05,
      "loss": 0.0009,
      "step": 18630
    },
    {
      "epoch": 3.7778678556951766,
      "grad_norm": 0.00479856925085187,
      "learning_rate": 1.1106607215241184e-05,
      "loss": 0.0017,
      "step": 18640
    },
    {
      "epoch": 3.779894608836644,
      "grad_norm": 0.004984683822840452,
      "learning_rate": 1.1005269558167817e-05,
      "loss": 0.0013,
      "step": 18650
    },
    {
      "epoch": 3.781921361978111,
      "grad_norm": 0.004968602210283279,
      "learning_rate": 1.0903931901094446e-05,
      "loss": 0.001,
      "step": 18660
    },
    {
      "epoch": 3.7839481151195784,
      "grad_norm": 0.005061035510152578,
      "learning_rate": 1.0802594244021079e-05,
      "loss": 0.0009,
      "step": 18670
    },
    {
      "epoch": 3.7859748682610457,
      "grad_norm": 0.00481655215844512,
      "learning_rate": 1.070125658694771e-05,
      "loss": 0.0017,
      "step": 18680
    },
    {
      "epoch": 3.788001621402513,
      "grad_norm": 0.005656828172504902,
      "learning_rate": 1.0599918929874342e-05,
      "loss": 0.0013,
      "step": 18690
    },
    {
      "epoch": 3.7900283745439807,
      "grad_norm": 0.004866285249590874,
      "learning_rate": 1.0498581272800973e-05,
      "loss": 0.0013,
      "step": 18700
    },
    {
      "epoch": 3.7900283745439807,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.09718893468379974,
      "eval_runtime": 50.1654,
      "eval_samples_per_second": 10.884,
      "eval_steps_per_second": 1.375,
      "step": 18700
    },
    {
      "epoch": 3.792055127685448,
      "grad_norm": 0.010259117931127548,
      "learning_rate": 1.0397243615727604e-05,
      "loss": 0.0006,
      "step": 18710
    },
    {
      "epoch": 3.794081880826915,
      "grad_norm": 0.004835317377001047,
      "learning_rate": 1.0295905958654237e-05,
      "loss": 0.0009,
      "step": 18720
    },
    {
      "epoch": 3.7961086339683825,
      "grad_norm": 0.005045641213655472,
      "learning_rate": 1.0194568301580868e-05,
      "loss": 0.0011,
      "step": 18730
    },
    {
      "epoch": 3.79813538710985,
      "grad_norm": 0.004773715045303106,
      "learning_rate": 1.0093230644507499e-05,
      "loss": 0.001,
      "step": 18740
    },
    {
      "epoch": 3.8001621402513175,
      "grad_norm": 0.03401784598827362,
      "learning_rate": 9.991892987434132e-06,
      "loss": 0.002,
      "step": 18750
    },
    {
      "epoch": 3.8021888933927848,
      "grad_norm": 0.004623086657375097,
      "learning_rate": 9.890555330360763e-06,
      "loss": 0.0017,
      "step": 18760
    },
    {
      "epoch": 3.804215646534252,
      "grad_norm": 0.004645565524697304,
      "learning_rate": 9.789217673287395e-06,
      "loss": 0.001,
      "step": 18770
    },
    {
      "epoch": 3.8062423996757193,
      "grad_norm": 0.03281537815928459,
      "learning_rate": 9.687880016214024e-06,
      "loss": 0.0147,
      "step": 18780
    },
    {
      "epoch": 3.808269152817187,
      "grad_norm": 0.03286786004900932,
      "learning_rate": 9.586542359140657e-06,
      "loss": 0.005,
      "step": 18790
    },
    {
      "epoch": 3.8102959059586543,
      "grad_norm": 0.02435663342475891,
      "learning_rate": 9.48520470206729e-06,
      "loss": 0.0016,
      "step": 18800
    },
    {
      "epoch": 3.8102959059586543,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.10444357991218567,
      "eval_runtime": 48.0273,
      "eval_samples_per_second": 11.369,
      "eval_steps_per_second": 1.437,
      "step": 18800
    },
    {
      "epoch": 3.8123226591001216,
      "grad_norm": 0.004839787259697914,
      "learning_rate": 9.38386704499392e-06,
      "loss": 0.0024,
      "step": 18810
    },
    {
      "epoch": 3.814349412241589,
      "grad_norm": 0.20349133014678955,
      "learning_rate": 9.282529387920552e-06,
      "loss": 0.0021,
      "step": 18820
    },
    {
      "epoch": 3.8163761653830566,
      "grad_norm": 0.004515075124800205,
      "learning_rate": 9.181191730847183e-06,
      "loss": 0.0034,
      "step": 18830
    },
    {
      "epoch": 3.818402918524524,
      "grad_norm": 0.03329044207930565,
      "learning_rate": 9.079854073773815e-06,
      "loss": 0.0012,
      "step": 18840
    },
    {
      "epoch": 3.820429671665991,
      "grad_norm": 0.004499832168221474,
      "learning_rate": 8.978516416700446e-06,
      "loss": 0.0005,
      "step": 18850
    },
    {
      "epoch": 3.8224564248074584,
      "grad_norm": 0.004518959205597639,
      "learning_rate": 8.877178759627077e-06,
      "loss": 0.0018,
      "step": 18860
    },
    {
      "epoch": 3.8244831779489257,
      "grad_norm": 0.004625269211828709,
      "learning_rate": 8.77584110255371e-06,
      "loss": 0.0009,
      "step": 18870
    },
    {
      "epoch": 3.826509931090393,
      "grad_norm": 0.00611333129927516,
      "learning_rate": 8.67450344548034e-06,
      "loss": 0.0012,
      "step": 18880
    },
    {
      "epoch": 3.8285366842318607,
      "grad_norm": 0.004493276588618755,
      "learning_rate": 8.573165788406973e-06,
      "loss": 1.116,
      "step": 18890
    },
    {
      "epoch": 3.830563437373328,
      "grad_norm": 0.004905337002128363,
      "learning_rate": 8.471828131333603e-06,
      "loss": 0.0009,
      "step": 18900
    },
    {
      "epoch": 3.830563437373328,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10386329889297485,
      "eval_runtime": 48.9261,
      "eval_samples_per_second": 11.16,
      "eval_steps_per_second": 1.41,
      "step": 18900
    },
    {
      "epoch": 3.832590190514795,
      "grad_norm": 0.004614606965333223,
      "learning_rate": 8.370490474260235e-06,
      "loss": 0.0012,
      "step": 18910
    },
    {
      "epoch": 3.834616943656263,
      "grad_norm": 0.006155083421617746,
      "learning_rate": 8.269152817186868e-06,
      "loss": 0.0009,
      "step": 18920
    },
    {
      "epoch": 3.83664369679773,
      "grad_norm": 0.03494027629494667,
      "learning_rate": 8.167815160113499e-06,
      "loss": 0.0013,
      "step": 18930
    },
    {
      "epoch": 3.8386704499391975,
      "grad_norm": 0.004626502748578787,
      "learning_rate": 8.06647750304013e-06,
      "loss": 0.0005,
      "step": 18940
    },
    {
      "epoch": 3.8406972030806648,
      "grad_norm": 0.004527637269347906,
      "learning_rate": 7.96513984596676e-06,
      "loss": 0.0009,
      "step": 18950
    },
    {
      "epoch": 3.842723956222132,
      "grad_norm": 0.03365058824419975,
      "learning_rate": 7.863802188893393e-06,
      "loss": 0.003,
      "step": 18960
    },
    {
      "epoch": 3.8447507093635993,
      "grad_norm": 0.004913891199976206,
      "learning_rate": 7.762464531820026e-06,
      "loss": 0.0012,
      "step": 18970
    },
    {
      "epoch": 3.846777462505067,
      "grad_norm": 0.004651523195207119,
      "learning_rate": 7.661126874746655e-06,
      "loss": 0.0012,
      "step": 18980
    },
    {
      "epoch": 3.8488042156465343,
      "grad_norm": 0.004609841853380203,
      "learning_rate": 7.559789217673288e-06,
      "loss": 0.0005,
      "step": 18990
    },
    {
      "epoch": 3.8508309687880016,
      "grad_norm": 0.004604669287800789,
      "learning_rate": 7.458451560599919e-06,
      "loss": 0.0008,
      "step": 19000
    },
    {
      "epoch": 3.8508309687880016,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.09755395352840424,
      "eval_runtime": 50.2984,
      "eval_samples_per_second": 10.855,
      "eval_steps_per_second": 1.372,
      "step": 19000
    },
    {
      "epoch": 3.852857721929469,
      "grad_norm": 0.004643167834728956,
      "learning_rate": 7.357113903526551e-06,
      "loss": 0.0008,
      "step": 19010
    },
    {
      "epoch": 3.8548844750709366,
      "grad_norm": 0.0045625087805092335,
      "learning_rate": 7.255776246453182e-06,
      "loss": 0.0008,
      "step": 19020
    },
    {
      "epoch": 3.856911228212404,
      "grad_norm": 0.009439327754080296,
      "learning_rate": 7.1544385893798136e-06,
      "loss": 0.0012,
      "step": 19030
    },
    {
      "epoch": 3.858937981353871,
      "grad_norm": 0.11327062547206879,
      "learning_rate": 7.053100932306446e-06,
      "loss": 0.0021,
      "step": 19040
    },
    {
      "epoch": 3.8609647344953384,
      "grad_norm": 0.004486795514822006,
      "learning_rate": 6.951763275233076e-06,
      "loss": 0.0012,
      "step": 19050
    },
    {
      "epoch": 3.8629914876368057,
      "grad_norm": 0.004348400980234146,
      "learning_rate": 6.850425618159709e-06,
      "loss": 0.0009,
      "step": 19060
    },
    {
      "epoch": 3.865018240778273,
      "grad_norm": 0.004538117442280054,
      "learning_rate": 6.749087961086339e-06,
      "loss": 0.5545,
      "step": 19070
    },
    {
      "epoch": 3.8670449939197407,
      "grad_norm": 0.004487019032239914,
      "learning_rate": 6.647750304012972e-06,
      "loss": 0.0023,
      "step": 19080
    },
    {
      "epoch": 3.869071747061208,
      "grad_norm": 0.004544503055512905,
      "learning_rate": 6.5464126469396035e-06,
      "loss": 0.0005,
      "step": 19090
    },
    {
      "epoch": 3.871098500202675,
      "grad_norm": 0.004624968860298395,
      "learning_rate": 6.4450749898662345e-06,
      "loss": 0.0005,
      "step": 19100
    },
    {
      "epoch": 3.871098500202675,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.09687044471502304,
      "eval_runtime": 49.0878,
      "eval_samples_per_second": 11.123,
      "eval_steps_per_second": 1.406,
      "step": 19100
    },
    {
      "epoch": 3.873125253344143,
      "grad_norm": 0.004524781834334135,
      "learning_rate": 6.343737332792866e-06,
      "loss": 0.001,
      "step": 19110
    },
    {
      "epoch": 3.87515200648561,
      "grad_norm": 0.004462063778191805,
      "learning_rate": 6.242399675719498e-06,
      "loss": 0.0012,
      "step": 19120
    },
    {
      "epoch": 3.8771787596270775,
      "grad_norm": 0.10787799954414368,
      "learning_rate": 6.141062018646129e-06,
      "loss": 0.0019,
      "step": 19130
    },
    {
      "epoch": 3.8792055127685448,
      "grad_norm": 0.004407837055623531,
      "learning_rate": 6.039724361572761e-06,
      "loss": 0.5658,
      "step": 19140
    },
    {
      "epoch": 3.881232265910012,
      "grad_norm": 0.006913017015904188,
      "learning_rate": 5.938386704499392e-06,
      "loss": 0.0009,
      "step": 19150
    },
    {
      "epoch": 3.8832590190514793,
      "grad_norm": 0.0045287227258086205,
      "learning_rate": 5.837049047426024e-06,
      "loss": 0.534,
      "step": 19160
    },
    {
      "epoch": 3.885285772192947,
      "grad_norm": 0.0046244170516729355,
      "learning_rate": 5.735711390352655e-06,
      "loss": 0.0054,
      "step": 19170
    },
    {
      "epoch": 3.8873125253344143,
      "grad_norm": 0.004677692428231239,
      "learning_rate": 5.634373733279287e-06,
      "loss": 0.0017,
      "step": 19180
    },
    {
      "epoch": 3.8893392784758816,
      "grad_norm": 0.03539438173174858,
      "learning_rate": 5.533036076205918e-06,
      "loss": 0.7679,
      "step": 19190
    },
    {
      "epoch": 3.891366031617349,
      "grad_norm": 0.004666884895414114,
      "learning_rate": 5.43169841913255e-06,
      "loss": 0.0009,
      "step": 19200
    },
    {
      "epoch": 3.891366031617349,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.09635677188634872,
      "eval_runtime": 46.483,
      "eval_samples_per_second": 11.746,
      "eval_steps_per_second": 1.484,
      "step": 19200
    },
    {
      "epoch": 3.8933927847588166,
      "grad_norm": 0.004748017527163029,
      "learning_rate": 5.330360762059181e-06,
      "loss": 0.0012,
      "step": 19210
    },
    {
      "epoch": 3.895419537900284,
      "grad_norm": 0.15614542365074158,
      "learning_rate": 5.229023104985813e-06,
      "loss": 0.0016,
      "step": 19220
    },
    {
      "epoch": 3.897446291041751,
      "grad_norm": 0.00450799660757184,
      "learning_rate": 5.1276854479124445e-06,
      "loss": 0.2846,
      "step": 19230
    },
    {
      "epoch": 3.8994730441832184,
      "grad_norm": 0.004577334504574537,
      "learning_rate": 5.026347790839076e-06,
      "loss": 0.0015,
      "step": 19240
    },
    {
      "epoch": 3.9014997973246857,
      "grad_norm": 0.004710934590548277,
      "learning_rate": 4.925010133765707e-06,
      "loss": 0.0013,
      "step": 19250
    },
    {
      "epoch": 3.9035265504661534,
      "grad_norm": 0.03456436097621918,
      "learning_rate": 4.823672476692339e-06,
      "loss": 0.0014,
      "step": 19260
    },
    {
      "epoch": 3.9055533036076207,
      "grad_norm": 0.01188341248780489,
      "learning_rate": 4.72233481961897e-06,
      "loss": 0.2853,
      "step": 19270
    },
    {
      "epoch": 3.907580056749088,
      "grad_norm": 0.03473988547921181,
      "learning_rate": 4.620997162545603e-06,
      "loss": 0.5524,
      "step": 19280
    },
    {
      "epoch": 3.909606809890555,
      "grad_norm": 0.0046269637532532215,
      "learning_rate": 4.519659505472234e-06,
      "loss": 0.0013,
      "step": 19290
    },
    {
      "epoch": 3.911633563032023,
      "grad_norm": 0.0044867959804832935,
      "learning_rate": 4.4183218483988654e-06,
      "loss": 0.0005,
      "step": 19300
    },
    {
      "epoch": 3.911633563032023,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10195985436439514,
      "eval_runtime": 50.1764,
      "eval_samples_per_second": 10.882,
      "eval_steps_per_second": 1.375,
      "step": 19300
    },
    {
      "epoch": 3.91366031617349,
      "grad_norm": 0.004721532575786114,
      "learning_rate": 4.316984191325496e-06,
      "loss": 0.0005,
      "step": 19310
    },
    {
      "epoch": 3.9156870693149575,
      "grad_norm": 0.0053703757002949715,
      "learning_rate": 4.215646534252128e-06,
      "loss": 0.0009,
      "step": 19320
    },
    {
      "epoch": 3.9177138224564247,
      "grad_norm": 0.03514038398861885,
      "learning_rate": 4.114308877178759e-06,
      "loss": 0.4019,
      "step": 19330
    },
    {
      "epoch": 3.919740575597892,
      "grad_norm": 0.00470903841778636,
      "learning_rate": 4.012971220105392e-06,
      "loss": 0.0019,
      "step": 19340
    },
    {
      "epoch": 3.9217673287393593,
      "grad_norm": 0.005483035929501057,
      "learning_rate": 3.911633563032023e-06,
      "loss": 0.0005,
      "step": 19350
    },
    {
      "epoch": 3.923794081880827,
      "grad_norm": 0.0046259555965662,
      "learning_rate": 3.8102959059586546e-06,
      "loss": 0.002,
      "step": 19360
    },
    {
      "epoch": 3.9258208350222943,
      "grad_norm": 0.0044891200959682465,
      "learning_rate": 3.708958248885286e-06,
      "loss": 0.0005,
      "step": 19370
    },
    {
      "epoch": 3.9278475881637616,
      "grad_norm": 0.004647380672395229,
      "learning_rate": 3.6076205918119173e-06,
      "loss": 0.0009,
      "step": 19380
    },
    {
      "epoch": 3.9298743413052293,
      "grad_norm": 0.005284647457301617,
      "learning_rate": 3.5062829347385487e-06,
      "loss": 0.0014,
      "step": 19390
    },
    {
      "epoch": 3.9319010944466966,
      "grad_norm": 0.46924781799316406,
      "learning_rate": 3.404945277665181e-06,
      "loss": 0.5488,
      "step": 19400
    },
    {
      "epoch": 3.9319010944466966,
      "eval_accuracy": 0.9816849816849816,
      "eval_loss": 0.09855089336633682,
      "eval_runtime": 49.0745,
      "eval_samples_per_second": 11.126,
      "eval_steps_per_second": 1.406,
      "step": 19400
    },
    {
      "epoch": 3.933927847588164,
      "grad_norm": 0.004715560935437679,
      "learning_rate": 3.3036076205918123e-06,
      "loss": 0.002,
      "step": 19410
    },
    {
      "epoch": 3.935954600729631,
      "grad_norm": 0.03552348166704178,
      "learning_rate": 3.2022699635184437e-06,
      "loss": 0.5464,
      "step": 19420
    },
    {
      "epoch": 3.9379813538710984,
      "grad_norm": 0.6693766713142395,
      "learning_rate": 3.100932306445075e-06,
      "loss": 0.0054,
      "step": 19430
    },
    {
      "epoch": 3.9400081070125657,
      "grad_norm": 0.03712967783212662,
      "learning_rate": 2.999594649371707e-06,
      "loss": 0.2938,
      "step": 19440
    },
    {
      "epoch": 3.9420348601540334,
      "grad_norm": 0.004706588573753834,
      "learning_rate": 2.8982569922983382e-06,
      "loss": 0.0013,
      "step": 19450
    },
    {
      "epoch": 3.9440616132955006,
      "grad_norm": 0.004487375728785992,
      "learning_rate": 2.7969193352249696e-06,
      "loss": 0.0005,
      "step": 19460
    },
    {
      "epoch": 3.946088366436968,
      "grad_norm": 0.004523433744907379,
      "learning_rate": 2.6955816781516014e-06,
      "loss": 0.002,
      "step": 19470
    },
    {
      "epoch": 3.948115119578435,
      "grad_norm": 0.004606922157108784,
      "learning_rate": 2.594244021078233e-06,
      "loss": 0.0006,
      "step": 19480
    },
    {
      "epoch": 3.950141872719903,
      "grad_norm": 0.006367456633597612,
      "learning_rate": 2.492906364004864e-06,
      "loss": 0.0006,
      "step": 19490
    },
    {
      "epoch": 3.95216862586137,
      "grad_norm": 0.004643953405320644,
      "learning_rate": 2.391568706931496e-06,
      "loss": 0.0014,
      "step": 19500
    },
    {
      "epoch": 3.95216862586137,
      "eval_accuracy": 0.9835164835164835,
      "eval_loss": 0.0963364690542221,
      "eval_runtime": 46.4967,
      "eval_samples_per_second": 11.743,
      "eval_steps_per_second": 1.484,
      "step": 19500
    },
    {
      "epoch": 3.9541953790028375,
      "grad_norm": 0.00468785222619772,
      "learning_rate": 2.2902310498581274e-06,
      "loss": 0.0009,
      "step": 19510
    },
    {
      "epoch": 3.9562221321443047,
      "grad_norm": 2.83689546585083,
      "learning_rate": 2.188893392784759e-06,
      "loss": 0.0027,
      "step": 19520
    },
    {
      "epoch": 3.958248885285772,
      "grad_norm": 0.0045558735728263855,
      "learning_rate": 2.0875557357113905e-06,
      "loss": 0.0009,
      "step": 19530
    },
    {
      "epoch": 3.9602756384272393,
      "grad_norm": 0.004627866670489311,
      "learning_rate": 1.986218078638022e-06,
      "loss": 0.0009,
      "step": 19540
    },
    {
      "epoch": 3.962302391568707,
      "grad_norm": 0.035609010607004166,
      "learning_rate": 1.8848804215646537e-06,
      "loss": 0.0013,
      "step": 19550
    },
    {
      "epoch": 3.9643291447101743,
      "grad_norm": 0.004579546395689249,
      "learning_rate": 1.7835427644912851e-06,
      "loss": 0.0014,
      "step": 19560
    },
    {
      "epoch": 3.9663558978516416,
      "grad_norm": 0.03501616045832634,
      "learning_rate": 1.6822051074179165e-06,
      "loss": 0.0016,
      "step": 19570
    },
    {
      "epoch": 3.9683826509931093,
      "grad_norm": 0.00457020616158843,
      "learning_rate": 1.5808674503445483e-06,
      "loss": 0.001,
      "step": 19580
    },
    {
      "epoch": 3.9704094041345765,
      "grad_norm": 0.004598743282258511,
      "learning_rate": 1.4795297932711797e-06,
      "loss": 0.0011,
      "step": 19590
    },
    {
      "epoch": 3.972436157276044,
      "grad_norm": 0.004656678065657616,
      "learning_rate": 1.3781921361978113e-06,
      "loss": 0.001,
      "step": 19600
    },
    {
      "epoch": 3.972436157276044,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10372201353311539,
      "eval_runtime": 49.7746,
      "eval_samples_per_second": 10.969,
      "eval_steps_per_second": 1.386,
      "step": 19600
    },
    {
      "epoch": 3.974462910417511,
      "grad_norm": 0.03887145593762398,
      "learning_rate": 1.2768544791244426e-06,
      "loss": 0.001,
      "step": 19610
    },
    {
      "epoch": 3.9764896635589784,
      "grad_norm": 0.03490873798727989,
      "learning_rate": 1.1755168220510742e-06,
      "loss": 0.0013,
      "step": 19620
    },
    {
      "epoch": 3.9785164167004456,
      "grad_norm": 0.00451950216665864,
      "learning_rate": 1.0741791649777058e-06,
      "loss": 0.002,
      "step": 19630
    },
    {
      "epoch": 3.9805431698419134,
      "grad_norm": 0.00451238127425313,
      "learning_rate": 9.728415079043374e-07,
      "loss": 0.0019,
      "step": 19640
    },
    {
      "epoch": 3.9825699229833806,
      "grad_norm": 0.004648827016353607,
      "learning_rate": 8.715038508309688e-07,
      "loss": 0.0005,
      "step": 19650
    },
    {
      "epoch": 3.984596676124848,
      "grad_norm": 0.004489494021981955,
      "learning_rate": 7.701661937576004e-07,
      "loss": 0.0013,
      "step": 19660
    },
    {
      "epoch": 3.9866234292663156,
      "grad_norm": 0.004483028780668974,
      "learning_rate": 6.688285366842319e-07,
      "loss": 0.5184,
      "step": 19670
    },
    {
      "epoch": 3.988650182407783,
      "grad_norm": 0.004603186156600714,
      "learning_rate": 5.674908796108635e-07,
      "loss": 0.0027,
      "step": 19680
    },
    {
      "epoch": 3.99067693554925,
      "grad_norm": 0.00456126919016242,
      "learning_rate": 4.6615322253749494e-07,
      "loss": 0.3085,
      "step": 19690
    },
    {
      "epoch": 3.9927036886907175,
      "grad_norm": 0.004487769678235054,
      "learning_rate": 3.648155654641265e-07,
      "loss": 0.0009,
      "step": 19700
    },
    {
      "epoch": 3.9927036886907175,
      "eval_accuracy": 0.9798534798534798,
      "eval_loss": 0.10453352332115173,
      "eval_runtime": 50.4335,
      "eval_samples_per_second": 10.826,
      "eval_steps_per_second": 1.368,
      "step": 19700
    },
    {
      "epoch": 3.9947304418321847,
      "grad_norm": 0.004525844007730484,
      "learning_rate": 2.63477908390758e-07,
      "loss": 0.0013,
      "step": 19710
    },
    {
      "epoch": 3.996757194973652,
      "grad_norm": 0.004443651530891657,
      "learning_rate": 1.6214025131738955e-07,
      "loss": 0.0009,
      "step": 19720
    },
    {
      "epoch": 3.9987839481151197,
      "grad_norm": 0.004505722783505917,
      "learning_rate": 6.080259424402109e-08,
      "loss": 0.0013,
      "step": 19730
    },
    {
      "epoch": 4.0,
      "step": 19736,
      "total_flos": 1.5293819070335877e+18,
      "train_loss": 0.20495193709398876,
      "train_runtime": 28441.8571,
      "train_samples_per_second": 0.694,
      "train_steps_per_second": 0.694
    }
  ],
  "logging_steps": 10,
  "max_steps": 19736,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "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": 1.5293819070335877e+18,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}