{
  "best_global_step": 1600,
  "best_metric": 0.01271997,
  "best_model_checkpoint": "/home/serusr01/new/output/v48-20251225-153152/checkpoint-1600",
  "epoch": 1.0,
  "eval_steps": 100,
  "global_step": 2318,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "acc": 0.5208333333333334,
      "epoch": 0.0004314063848144953,
      "grad_norm": 144.0,
      "learning_rate": 5.999997244730563e-06,
      "loss": 1.0161079168319702,
      "step": 1
    },
    {
      "acc": 0.8512820512820513,
      "epoch": 0.0021570319240724763,
      "grad_norm": 12.3125,
      "learning_rate": 5.9999311185171385e-06,
      "loss": 0.40382882952690125,
      "step": 5
    },
    {
      "acc": 0.8622047244094488,
      "epoch": 0.004314063848144953,
      "grad_norm": 18.75,
      "learning_rate": 5.999724477231659e-06,
      "loss": 0.35493576526641846,
      "step": 10
    },
    {
      "acc": 0.8986083499005965,
      "epoch": 0.0064710957722174285,
      "grad_norm": 17.0,
      "learning_rate": 5.9993800856327355e-06,
      "loss": 0.29610111713409426,
      "step": 15
    },
    {
      "acc": 0.8972868217054264,
      "epoch": 0.008628127696289905,
      "grad_norm": 9.125,
      "learning_rate": 5.998897959535169e-06,
      "loss": 0.302515435218811,
      "step": 20
    },
    {
      "acc": 0.876953125,
      "epoch": 0.010785159620362382,
      "grad_norm": 14.9375,
      "learning_rate": 5.998278121078668e-06,
      "loss": 0.3312162160873413,
      "step": 25
    },
    {
      "acc": 0.8975903614457831,
      "epoch": 0.012942191544434857,
      "grad_norm": 13.375,
      "learning_rate": 5.997520598726825e-06,
      "loss": 0.30696473121643064,
      "step": 30
    },
    {
      "acc": 0.8742632612966601,
      "epoch": 0.015099223468507334,
      "grad_norm": 13.25,
      "learning_rate": 5.996625427265816e-06,
      "loss": 0.36511917114257814,
      "step": 35
    },
    {
      "acc": 0.8631790744466801,
      "epoch": 0.01725625539257981,
      "grad_norm": 9.3125,
      "learning_rate": 5.995592647802802e-06,
      "loss": 0.338372015953064,
      "step": 40
    },
    {
      "acc": 0.8964803312629399,
      "epoch": 0.019413287316652286,
      "grad_norm": 6.0,
      "learning_rate": 5.9944223077640325e-06,
      "loss": 0.32863070964813235,
      "step": 45
    },
    {
      "acc": 0.8957915831663327,
      "epoch": 0.021570319240724764,
      "grad_norm": 7.96875,
      "learning_rate": 5.993114460892682e-06,
      "loss": 0.27804901599884035,
      "step": 50
    },
    {
      "acc": 0.8773006134969326,
      "epoch": 0.02372735116479724,
      "grad_norm": 8.9375,
      "learning_rate": 5.9916691672463725e-06,
      "loss": 0.31299686431884766,
      "step": 55
    },
    {
      "acc": 0.9154929577464789,
      "epoch": 0.025884383088869714,
      "grad_norm": 9.1875,
      "learning_rate": 5.990086493194415e-06,
      "loss": 0.2637639045715332,
      "step": 60
    },
    {
      "acc": 0.8927125506072875,
      "epoch": 0.028041415012942193,
      "grad_norm": 9.5,
      "learning_rate": 5.988366511414766e-06,
      "loss": 0.30198609828948975,
      "step": 65
    },
    {
      "acc": 0.9156118143459916,
      "epoch": 0.030198446937014668,
      "grad_norm": 9.0625,
      "learning_rate": 5.986509300890691e-06,
      "loss": 0.24096782207489015,
      "step": 70
    },
    {
      "acc": 0.8798449612403101,
      "epoch": 0.032355478861087146,
      "grad_norm": 7.4375,
      "learning_rate": 5.984514946907133e-06,
      "loss": 0.2953530550003052,
      "step": 75
    },
    {
      "acc": 0.8818737270875764,
      "epoch": 0.03451251078515962,
      "grad_norm": 15.9375,
      "learning_rate": 5.982383541046798e-06,
      "loss": 0.3139867067337036,
      "step": 80
    },
    {
      "acc": 0.9300567107750473,
      "epoch": 0.036669542709232096,
      "grad_norm": 6.5,
      "learning_rate": 5.98011518118595e-06,
      "loss": 0.20404579639434814,
      "step": 85
    },
    {
      "acc": 0.9036885245901639,
      "epoch": 0.03882657463330457,
      "grad_norm": 9.75,
      "learning_rate": 5.977709971489917e-06,
      "loss": 0.2777871131896973,
      "step": 90
    },
    {
      "acc": 0.8767676767676768,
      "epoch": 0.040983606557377046,
      "grad_norm": 12.3125,
      "learning_rate": 5.975168022408304e-06,
      "loss": 0.31117370128631594,
      "step": 95
    },
    {
      "acc": 0.9202334630350194,
      "epoch": 0.04314063848144953,
      "grad_norm": 5.9375,
      "learning_rate": 5.972489450669929e-06,
      "loss": 0.23061840534210204,
      "step": 100
    },
    {
      "epoch": 0.04314063848144953,
      "eval_acc": 0.9116072164099185,
      "eval_loss": 0.031446851789951324,
      "eval_mrr": 0.8872438524590165,
      "eval_ndcg": 0.9145527118078347,
      "eval_runtime": 92.9513,
      "eval_samples_per_second": 21.0,
      "eval_steps_per_second": 10.5,
      "step": 100
    },
    {
      "acc": 0.9188118811881189,
      "epoch": 0.045297670405522,
      "grad_norm": 10.6875,
      "learning_rate": 5.969674379277452e-06,
      "loss": 0.253974175453186,
      "step": 105
    },
    {
      "acc": 0.916030534351145,
      "epoch": 0.04745470232959448,
      "grad_norm": 8.5,
      "learning_rate": 5.9667229375017345e-06,
      "loss": 0.21623244285583496,
      "step": 110
    },
    {
      "acc": 0.9288461538461539,
      "epoch": 0.04961173425366695,
      "grad_norm": 9.4375,
      "learning_rate": 5.9636352608759e-06,
      "loss": 0.2198498487472534,
      "step": 115
    },
    {
      "acc": 0.8913480885311871,
      "epoch": 0.05176876617773943,
      "grad_norm": 10.1875,
      "learning_rate": 5.960411491189113e-06,
      "loss": 0.3081125497817993,
      "step": 120
    },
    {
      "acc": 0.8951612903225806,
      "epoch": 0.05392579810181191,
      "grad_norm": 10.125,
      "learning_rate": 5.957051776480063e-06,
      "loss": 0.2795823335647583,
      "step": 125
    },
    {
      "acc": 0.91,
      "epoch": 0.056082830025884385,
      "grad_norm": 10.375,
      "learning_rate": 5.953556271030172e-06,
      "loss": 0.24584596157073973,
      "step": 130
    },
    {
      "acc": 0.9195402298850575,
      "epoch": 0.05823986194995686,
      "grad_norm": 10.0,
      "learning_rate": 5.949925135356506e-06,
      "loss": 0.20374622344970703,
      "step": 135
    },
    {
      "acc": 0.9190283400809717,
      "epoch": 0.060396893874029335,
      "grad_norm": 13.25,
      "learning_rate": 5.946158536204403e-06,
      "loss": 0.2638064384460449,
      "step": 140
    },
    {
      "acc": 0.9001956947162426,
      "epoch": 0.06255392579810182,
      "grad_norm": 7.21875,
      "learning_rate": 5.942256646539821e-06,
      "loss": 0.2462130069732666,
      "step": 145
    },
    {
      "acc": 0.9201520912547528,
      "epoch": 0.06471095772217429,
      "grad_norm": 7.21875,
      "learning_rate": 5.9382196455413914e-06,
      "loss": 0.25420570373535156,
      "step": 150
    },
    {
      "acc": 0.9288702928870293,
      "epoch": 0.06686798964624677,
      "grad_norm": 10.1875,
      "learning_rate": 5.9340477185921895e-06,
      "loss": 0.2146221160888672,
      "step": 155
    },
    {
      "acc": 0.904397705544933,
      "epoch": 0.06902502157031924,
      "grad_norm": 27.875,
      "learning_rate": 5.9297410572712245e-06,
      "loss": 0.2619572877883911,
      "step": 160
    },
    {
      "acc": 0.903041825095057,
      "epoch": 0.07118205349439172,
      "grad_norm": 8.1875,
      "learning_rate": 5.925299859344643e-06,
      "loss": 0.2647380352020264,
      "step": 165
    },
    {
      "acc": 0.9238095238095239,
      "epoch": 0.07333908541846419,
      "grad_norm": 9.75,
      "learning_rate": 5.9207243287566444e-06,
      "loss": 0.2059628963470459,
      "step": 170
    },
    {
      "acc": 0.9080234833659491,
      "epoch": 0.07549611734253667,
      "grad_norm": 19.0,
      "learning_rate": 5.916014675620117e-06,
      "loss": 0.22357699871063233,
      "step": 175
    },
    {
      "acc": 0.9107883817427386,
      "epoch": 0.07765314926660914,
      "grad_norm": 12.125,
      "learning_rate": 5.911171116206986e-06,
      "loss": 0.23458666801452638,
      "step": 180
    },
    {
      "acc": 0.9092664092664092,
      "epoch": 0.07981018119068162,
      "grad_norm": 10.3125,
      "learning_rate": 5.9061938729382915e-06,
      "loss": 0.2300126552581787,
      "step": 185
    },
    {
      "acc": 0.924901185770751,
      "epoch": 0.08196721311475409,
      "grad_norm": 12.25,
      "learning_rate": 5.901083174373961e-06,
      "loss": 0.21468648910522461,
      "step": 190
    },
    {
      "acc": 0.9180327868852459,
      "epoch": 0.08412424503882658,
      "grad_norm": 9.375,
      "learning_rate": 5.895839255202328e-06,
      "loss": 0.2179854154586792,
      "step": 195
    },
    {
      "acc": 0.935871743486974,
      "epoch": 0.08628127696289906,
      "grad_norm": 11.9375,
      "learning_rate": 5.8904623562293435e-06,
      "loss": 0.22542862892150878,
      "step": 200
    },
    {
      "epoch": 0.08628127696289906,
      "eval_acc": 0.9252821484471538,
      "eval_loss": 0.02712932601571083,
      "eval_mrr": 0.9001439305230289,
      "eval_ndcg": 0.9243608674894062,
      "eval_runtime": 85.9276,
      "eval_samples_per_second": 22.717,
      "eval_steps_per_second": 11.358,
      "step": 200
    },
    {
      "acc": 0.9358178053830227,
      "epoch": 0.08843830888697153,
      "grad_norm": 5.96875,
      "learning_rate": 5.884952724367524e-06,
      "loss": 0.20223827362060548,
      "step": 205
    },
    {
      "acc": 0.9375,
      "epoch": 0.090595340811044,
      "grad_norm": 7.1875,
      "learning_rate": 5.879310612624611e-06,
      "loss": 0.19913212060928345,
      "step": 210
    },
    {
      "acc": 0.9217557251908397,
      "epoch": 0.09275237273511648,
      "grad_norm": 18.5,
      "learning_rate": 5.873536280091955e-06,
      "loss": 0.21275293827056885,
      "step": 215
    },
    {
      "acc": 0.9144050104384134,
      "epoch": 0.09490940465918896,
      "grad_norm": 14.375,
      "learning_rate": 5.867629991932611e-06,
      "loss": 0.23634729385375977,
      "step": 220
    },
    {
      "acc": 0.9239766081871345,
      "epoch": 0.09706643658326143,
      "grad_norm": 13.125,
      "learning_rate": 5.861592019369172e-06,
      "loss": 0.2160120725631714,
      "step": 225
    },
    {
      "acc": 0.8967611336032388,
      "epoch": 0.0992234685073339,
      "grad_norm": 12.0625,
      "learning_rate": 5.855422639671309e-06,
      "loss": 0.25560975074768066,
      "step": 230
    },
    {
      "acc": 0.9405940594059405,
      "epoch": 0.10138050043140638,
      "grad_norm": 11.3125,
      "learning_rate": 5.849122136143034e-06,
      "loss": 0.20292723178863525,
      "step": 235
    },
    {
      "acc": 0.9289827255278311,
      "epoch": 0.10353753235547886,
      "grad_norm": 16.125,
      "learning_rate": 5.842690798109697e-06,
      "loss": 0.19165314435958863,
      "step": 240
    },
    {
      "acc": 0.9137931034482759,
      "epoch": 0.10569456427955133,
      "grad_norm": 14.375,
      "learning_rate": 5.8361289209047004e-06,
      "loss": 0.22526559829711915,
      "step": 245
    },
    {
      "acc": 0.9123505976095617,
      "epoch": 0.10785159620362382,
      "grad_norm": 11.375,
      "learning_rate": 5.829436805855932e-06,
      "loss": 0.1998949646949768,
      "step": 250
    },
    {
      "acc": 0.9298245614035088,
      "epoch": 0.1100086281276963,
      "grad_norm": 9.3125,
      "learning_rate": 5.82261476027193e-06,
      "loss": 0.211327862739563,
      "step": 255
    },
    {
      "acc": 0.9357429718875502,
      "epoch": 0.11216566005176877,
      "grad_norm": 23.375,
      "learning_rate": 5.8156630974277715e-06,
      "loss": 0.21703736782073973,
      "step": 260
    },
    {
      "acc": 0.9092783505154639,
      "epoch": 0.11432269197584125,
      "grad_norm": 17.875,
      "learning_rate": 5.808582136550686e-06,
      "loss": 0.2739748954772949,
      "step": 265
    },
    {
      "acc": 0.9468302658486708,
      "epoch": 0.11647972389991372,
      "grad_norm": 9.9375,
      "learning_rate": 5.8013722028053985e-06,
      "loss": 0.14410748481750488,
      "step": 270
    },
    {
      "acc": 0.9310344827586207,
      "epoch": 0.1186367558239862,
      "grad_norm": 8.5,
      "learning_rate": 5.794033627279193e-06,
      "loss": 0.1934072494506836,
      "step": 275
    },
    {
      "acc": 0.9155470249520153,
      "epoch": 0.12079378774805867,
      "grad_norm": 11.4375,
      "learning_rate": 5.786566746966714e-06,
      "loss": 0.20481424331665038,
      "step": 280
    },
    {
      "acc": 0.9216494845360824,
      "epoch": 0.12295081967213115,
      "grad_norm": 12.5,
      "learning_rate": 5.778971904754485e-06,
      "loss": 0.22041404247283936,
      "step": 285
    },
    {
      "acc": 0.917864476386037,
      "epoch": 0.12510785159620363,
      "grad_norm": 19.375,
      "learning_rate": 5.771249449405169e-06,
      "loss": 0.20896604061126708,
      "step": 290
    },
    {
      "acc": 0.950592885375494,
      "epoch": 0.1272648835202761,
      "grad_norm": 16.875,
      "learning_rate": 5.763399735541551e-06,
      "loss": 0.1629479169845581,
      "step": 295
    },
    {
      "acc": 0.9405737704918032,
      "epoch": 0.12942191544434858,
      "grad_norm": 13.25,
      "learning_rate": 5.755423123630251e-06,
      "loss": 0.19808459281921387,
      "step": 300
    },
    {
      "epoch": 0.12942191544434858,
      "eval_acc": 0.9337671966389324,
      "eval_loss": 0.024512404575943947,
      "eval_mrr": 0.9131787909836067,
      "eval_ndcg": 0.9342680843294187,
      "eval_runtime": 79.6383,
      "eval_samples_per_second": 24.511,
      "eval_steps_per_second": 12.255,
      "step": 300
    },
    {
      "acc": 0.9308943089430894,
      "epoch": 0.13157894736842105,
      "grad_norm": 14.125,
      "learning_rate": 5.747319979965173e-06,
      "loss": 0.20357720851898192,
      "step": 305
    },
    {
      "acc": 0.9433962264150944,
      "epoch": 0.13373597929249353,
      "grad_norm": 14.6875,
      "learning_rate": 5.739090676650683e-06,
      "loss": 0.15039076805114746,
      "step": 310
    },
    {
      "acc": 0.9392712550607287,
      "epoch": 0.135893011216566,
      "grad_norm": 8.625,
      "learning_rate": 5.730735591584529e-06,
      "loss": 0.16305069923400878,
      "step": 315
    },
    {
      "acc": 0.9192307692307692,
      "epoch": 0.13805004314063848,
      "grad_norm": 12.75,
      "learning_rate": 5.722255108440474e-06,
      "loss": 0.2407283067703247,
      "step": 320
    },
    {
      "acc": 0.9298597194388778,
      "epoch": 0.14020707506471095,
      "grad_norm": 12.125,
      "learning_rate": 5.713649616650687e-06,
      "loss": 0.2122425317764282,
      "step": 325
    },
    {
      "acc": 0.9352226720647774,
      "epoch": 0.14236410698878343,
      "grad_norm": 11.125,
      "learning_rate": 5.7049195113878585e-06,
      "loss": 0.19069076776504518,
      "step": 330
    },
    {
      "acc": 0.9189765458422174,
      "epoch": 0.14452113891285592,
      "grad_norm": 17.75,
      "learning_rate": 5.696065193547054e-06,
      "loss": 0.22681286334991455,
      "step": 335
    },
    {
      "acc": 0.943579766536965,
      "epoch": 0.14667817083692838,
      "grad_norm": 14.8125,
      "learning_rate": 5.6870870697273e-06,
      "loss": 0.1799636721611023,
      "step": 340
    },
    {
      "acc": 0.9392712550607287,
      "epoch": 0.14883520276100087,
      "grad_norm": 8.0625,
      "learning_rate": 5.677985552212919e-06,
      "loss": 0.21220760345458983,
      "step": 345
    },
    {
      "acc": 0.9386138613861386,
      "epoch": 0.15099223468507333,
      "grad_norm": 15.0625,
      "learning_rate": 5.668761058954594e-06,
      "loss": 0.19215874671936034,
      "step": 350
    },
    {
      "acc": 0.9487704918032787,
      "epoch": 0.15314926660914582,
      "grad_norm": 9.75,
      "learning_rate": 5.659414013550172e-06,
      "loss": 0.17874677181243898,
      "step": 355
    },
    {
      "acc": 0.9449901768172888,
      "epoch": 0.15530629853321828,
      "grad_norm": 15.1875,
      "learning_rate": 5.649944845225219e-06,
      "loss": 0.17652757167816163,
      "step": 360
    },
    {
      "acc": 0.9281314168377823,
      "epoch": 0.15746333045729077,
      "grad_norm": 28.25,
      "learning_rate": 5.6403539888133056e-06,
      "loss": 0.2142080545425415,
      "step": 365
    },
    {
      "acc": 0.950920245398773,
      "epoch": 0.15962036238136323,
      "grad_norm": 12.0625,
      "learning_rate": 5.63064188473604e-06,
      "loss": 0.17437468767166137,
      "step": 370
    },
    {
      "acc": 0.9422680412371134,
      "epoch": 0.16177739430543572,
      "grad_norm": 19.75,
      "learning_rate": 5.6208089789828435e-06,
      "loss": 0.18122278451919555,
      "step": 375
    },
    {
      "acc": 0.9439071566731141,
      "epoch": 0.16393442622950818,
      "grad_norm": 13.8125,
      "learning_rate": 5.610855723090466e-06,
      "loss": 0.18091570138931273,
      "step": 380
    },
    {
      "acc": 0.923728813559322,
      "epoch": 0.16609145815358067,
      "grad_norm": 29.875,
      "learning_rate": 5.600782574122259e-06,
      "loss": 0.24442174434661865,
      "step": 385
    },
    {
      "acc": 0.9310344827586207,
      "epoch": 0.16824849007765316,
      "grad_norm": 16.75,
      "learning_rate": 5.590589994647182e-06,
      "loss": 0.20708949565887452,
      "step": 390
    },
    {
      "acc": 0.9395161290322581,
      "epoch": 0.17040552200172562,
      "grad_norm": 8.0625,
      "learning_rate": 5.58027845271856e-06,
      "loss": 0.21145825386047362,
      "step": 395
    },
    {
      "acc": 0.9618473895582329,
      "epoch": 0.1725625539257981,
      "grad_norm": 8.5,
      "learning_rate": 5.569848421852592e-06,
      "loss": 0.12078192234039306,
      "step": 400
    },
    {
      "epoch": 0.1725625539257981,
      "eval_acc": 0.9410989373094983,
      "eval_loss": 0.02222530171275139,
      "eval_mrr": 0.9228056693989071,
      "eval_ndcg": 0.9416691570915218,
      "eval_runtime": 87.651,
      "eval_samples_per_second": 22.27,
      "eval_steps_per_second": 11.135,
      "step": 400
    },
    {
      "acc": 0.951310861423221,
      "epoch": 0.17471958584987057,
      "grad_norm": 9.25,
      "learning_rate": 5.559300381006607e-06,
      "loss": 0.1467392086982727,
      "step": 405
    },
    {
      "acc": 0.9362139917695473,
      "epoch": 0.17687661777394306,
      "grad_norm": 9.5625,
      "learning_rate": 5.548634814557066e-06,
      "loss": 0.16467268466949464,
      "step": 410
    },
    {
      "acc": 0.9474671669793621,
      "epoch": 0.17903364969801552,
      "grad_norm": 21.0,
      "learning_rate": 5.537852212277326e-06,
      "loss": 0.16810253858566285,
      "step": 415
    },
    {
      "acc": 0.9329388560157791,
      "epoch": 0.181190681622088,
      "grad_norm": 21.75,
      "learning_rate": 5.5269530693151425e-06,
      "loss": 0.15131351947784424,
      "step": 420
    },
    {
      "acc": 0.9265536723163842,
      "epoch": 0.18334771354616047,
      "grad_norm": 26.25,
      "learning_rate": 5.5159378861699356e-06,
      "loss": 0.2086423635482788,
      "step": 425
    },
    {
      "acc": 0.9393939393939394,
      "epoch": 0.18550474547023296,
      "grad_norm": 9.875,
      "learning_rate": 5.504807168669804e-06,
      "loss": 0.19425034523010254,
      "step": 430
    },
    {
      "acc": 0.9387351778656127,
      "epoch": 0.18766177739430542,
      "grad_norm": 40.0,
      "learning_rate": 5.4935614279482984e-06,
      "loss": 0.1988288640975952,
      "step": 435
    },
    {
      "acc": 0.9350912778904665,
      "epoch": 0.1898188093183779,
      "grad_norm": 29.125,
      "learning_rate": 5.482201180420952e-06,
      "loss": 0.1814996600151062,
      "step": 440
    },
    {
      "acc": 0.9528688524590164,
      "epoch": 0.1919758412424504,
      "grad_norm": 17.25,
      "learning_rate": 5.4707269477615584e-06,
      "loss": 0.1311618208885193,
      "step": 445
    },
    {
      "acc": 0.9548872180451128,
      "epoch": 0.19413287316652286,
      "grad_norm": 21.375,
      "learning_rate": 5.4591392568782275e-06,
      "loss": 0.15495063066482545,
      "step": 450
    },
    {
      "acc": 0.9633204633204633,
      "epoch": 0.19628990509059535,
      "grad_norm": 14.8125,
      "learning_rate": 5.447438639889178e-06,
      "loss": 0.12518519163131714,
      "step": 455
    },
    {
      "acc": 0.9411764705882353,
      "epoch": 0.1984469370146678,
      "grad_norm": 17.875,
      "learning_rate": 5.435625634098311e-06,
      "loss": 0.17223405838012695,
      "step": 460
    },
    {
      "acc": 0.9486166007905138,
      "epoch": 0.2006039689387403,
      "grad_norm": 19.0,
      "learning_rate": 5.423700781970527e-06,
      "loss": 0.14980016946792601,
      "step": 465
    },
    {
      "acc": 0.9354838709677419,
      "epoch": 0.20276100086281276,
      "grad_norm": 23.0,
      "learning_rate": 5.411664631106827e-06,
      "loss": 0.1699580192565918,
      "step": 470
    },
    {
      "acc": 0.9409368635437881,
      "epoch": 0.20491803278688525,
      "grad_norm": 11.6875,
      "learning_rate": 5.399517734219159e-06,
      "loss": 0.2162698268890381,
      "step": 475
    },
    {
      "acc": 0.9535353535353536,
      "epoch": 0.2070750647109577,
      "grad_norm": 16.625,
      "learning_rate": 5.387260649105032e-06,
      "loss": 0.15655677318572997,
      "step": 480
    },
    {
      "acc": 0.9541984732824428,
      "epoch": 0.2092320966350302,
      "grad_norm": 14.4375,
      "learning_rate": 5.374893938621913e-06,
      "loss": 0.15797934532165528,
      "step": 485
    },
    {
      "acc": 0.9343629343629344,
      "epoch": 0.21138912855910266,
      "grad_norm": 6.59375,
      "learning_rate": 5.362418170661375e-06,
      "loss": 0.2039581060409546,
      "step": 490
    },
    {
      "acc": 0.9559386973180076,
      "epoch": 0.21354616048317515,
      "grad_norm": 16.625,
      "learning_rate": 5.3498339181230125e-06,
      "loss": 0.1275785207748413,
      "step": 495
    },
    {
      "acc": 0.9521988527724665,
      "epoch": 0.21570319240724764,
      "grad_norm": 8.1875,
      "learning_rate": 5.3371417588881456e-06,
      "loss": 0.16547932624816894,
      "step": 500
    },
    {
      "epoch": 0.21570319240724764,
      "eval_acc": 0.9440645852211879,
      "eval_loss": 0.0200980044901371,
      "eval_mrr": 0.9341749609679937,
      "eval_ndcg": 0.9502093836313069,
      "eval_runtime": 89.2845,
      "eval_samples_per_second": 21.863,
      "eval_steps_per_second": 10.931,
      "step": 500
    },
    {
      "acc": 0.9288617886178862,
      "epoch": 0.2178602243313201,
      "grad_norm": 17.0,
      "learning_rate": 5.324342275793272e-06,
      "loss": 0.22141244411468505,
      "step": 505
    },
    {
      "acc": 0.9502982107355865,
      "epoch": 0.2200172562553926,
      "grad_norm": 11.0,
      "learning_rate": 5.3114360566033085e-06,
      "loss": 0.13880524635314942,
      "step": 510
    },
    {
      "acc": 0.9284253578732107,
      "epoch": 0.22217428817946505,
      "grad_norm": 10.125,
      "learning_rate": 5.298423693984598e-06,
      "loss": 0.22597496509552,
      "step": 515
    },
    {
      "acc": 0.9553398058252427,
      "epoch": 0.22433132010353754,
      "grad_norm": 20.75,
      "learning_rate": 5.285305785477699e-06,
      "loss": 0.16823456287384034,
      "step": 520
    },
    {
      "acc": 0.9405940594059405,
      "epoch": 0.22648835202761,
      "grad_norm": 9.875,
      "learning_rate": 5.272082933469936e-06,
      "loss": 0.18091850280761718,
      "step": 525
    },
    {
      "acc": 0.9634888438133874,
      "epoch": 0.2286453839516825,
      "grad_norm": 9.6875,
      "learning_rate": 5.258755745167744e-06,
      "loss": 0.13490262031555175,
      "step": 530
    },
    {
      "acc": 0.9320754716981132,
      "epoch": 0.23080241587575495,
      "grad_norm": 18.625,
      "learning_rate": 5.245324832568787e-06,
      "loss": 0.19664840698242186,
      "step": 535
    },
    {
      "acc": 0.9470588235294117,
      "epoch": 0.23295944779982744,
      "grad_norm": 10.3125,
      "learning_rate": 5.2317908124338475e-06,
      "loss": 0.14916815757751464,
      "step": 540
    },
    {
      "acc": 0.9350912778904665,
      "epoch": 0.2351164797238999,
      "grad_norm": 12.0625,
      "learning_rate": 5.21815430625851e-06,
      "loss": 0.1888782262802124,
      "step": 545
    },
    {
      "acc": 0.9365079365079365,
      "epoch": 0.2372735116479724,
      "grad_norm": 14.6875,
      "learning_rate": 5.204415940244618e-06,
      "loss": 0.18109892606735228,
      "step": 550
    },
    {
      "acc": 0.9479768786127167,
      "epoch": 0.23943054357204488,
      "grad_norm": 35.0,
      "learning_rate": 5.1905763452715215e-06,
      "loss": 0.150074303150177,
      "step": 555
    },
    {
      "acc": 0.9408284023668639,
      "epoch": 0.24158757549611734,
      "grad_norm": 10.0625,
      "learning_rate": 5.176636156867102e-06,
      "loss": 0.13501374721527098,
      "step": 560
    },
    {
      "acc": 0.9490196078431372,
      "epoch": 0.24374460742018983,
      "grad_norm": 11.5625,
      "learning_rate": 5.162596015178593e-06,
      "loss": 0.14429715871810914,
      "step": 565
    },
    {
      "acc": 0.961038961038961,
      "epoch": 0.2459016393442623,
      "grad_norm": 4.53125,
      "learning_rate": 5.14845656494318e-06,
      "loss": 0.10108070373535157,
      "step": 570
    },
    {
      "acc": 0.9427402862985685,
      "epoch": 0.24805867126833478,
      "grad_norm": 13.5,
      "learning_rate": 5.134218455458396e-06,
      "loss": 0.16440529823303224,
      "step": 575
    },
    {
      "acc": 0.9512670565302144,
      "epoch": 0.25021570319240727,
      "grad_norm": 23.0,
      "learning_rate": 5.119882340552303e-06,
      "loss": 0.15792548656463623,
      "step": 580
    },
    {
      "acc": 0.9310344827586207,
      "epoch": 0.25237273511647973,
      "grad_norm": 20.375,
      "learning_rate": 5.105448878553472e-06,
      "loss": 0.17942277193069459,
      "step": 585
    },
    {
      "acc": 0.9586614173228346,
      "epoch": 0.2545297670405522,
      "grad_norm": 6.21875,
      "learning_rate": 5.0909187322607434e-06,
      "loss": 0.1267695426940918,
      "step": 590
    },
    {
      "acc": 0.9421157684630739,
      "epoch": 0.25668679896462465,
      "grad_norm": 10.8125,
      "learning_rate": 5.076292568912801e-06,
      "loss": 0.17552590370178223,
      "step": 595
    },
    {
      "acc": 0.9418837675350702,
      "epoch": 0.25884383088869717,
      "grad_norm": 15.5625,
      "learning_rate": 5.061571060157525e-06,
      "loss": 0.1582822322845459,
      "step": 600
    },
    {
      "epoch": 0.25884383088869717,
      "eval_acc": 0.9523848751956504,
      "eval_loss": 0.01805831491947174,
      "eval_mrr": 0.9341017759562842,
      "eval_ndcg": 0.9502097307474537,
      "eval_runtime": 80.1164,
      "eval_samples_per_second": 24.365,
      "eval_steps_per_second": 12.182,
      "step": 600
    },
    {
      "acc": 0.9516806722689075,
      "epoch": 0.26100086281276963,
      "grad_norm": 9.5625,
      "learning_rate": 5.04675488202115e-06,
      "loss": 0.1585369348526001,
      "step": 605
    },
    {
      "acc": 0.967479674796748,
      "epoch": 0.2631578947368421,
      "grad_norm": 14.375,
      "learning_rate": 5.031844714877224e-06,
      "loss": 0.14685235023498536,
      "step": 610
    },
    {
      "acc": 0.9669902912621359,
      "epoch": 0.2653149266609146,
      "grad_norm": 11.875,
      "learning_rate": 5.016841243415359e-06,
      "loss": 0.09848676323890686,
      "step": 615
    },
    {
      "acc": 0.9692622950819673,
      "epoch": 0.26747195858498707,
      "grad_norm": 11.875,
      "learning_rate": 5.001745156609801e-06,
      "loss": 0.09488856792449951,
      "step": 620
    },
    {
      "acc": 0.962,
      "epoch": 0.26962899050905953,
      "grad_norm": 8.625,
      "learning_rate": 4.9865571476877775e-06,
      "loss": 0.10182394981384277,
      "step": 625
    },
    {
      "acc": 0.9538152610441767,
      "epoch": 0.271786022433132,
      "grad_norm": 13.0,
      "learning_rate": 4.9712779140976725e-06,
      "loss": 0.14965349435806274,
      "step": 630
    },
    {
      "acc": 0.9563567362428842,
      "epoch": 0.2739430543572045,
      "grad_norm": 15.5625,
      "learning_rate": 4.9559081574769965e-06,
      "loss": 0.14634023904800414,
      "step": 635
    },
    {
      "acc": 0.9402390438247012,
      "epoch": 0.27610008628127697,
      "grad_norm": 19.875,
      "learning_rate": 4.9404485836201695e-06,
      "loss": 0.15552909374237062,
      "step": 640
    },
    {
      "acc": 0.9487666034155597,
      "epoch": 0.27825711820534943,
      "grad_norm": 38.25,
      "learning_rate": 4.924899902446105e-06,
      "loss": 0.16740797758102416,
      "step": 645
    },
    {
      "acc": 0.9620253164556962,
      "epoch": 0.2804141501294219,
      "grad_norm": 8.9375,
      "learning_rate": 4.909262827965613e-06,
      "loss": 0.13096400499343872,
      "step": 650
    },
    {
      "acc": 0.9373737373737374,
      "epoch": 0.2825711820534944,
      "grad_norm": 23.375,
      "learning_rate": 4.893538078248613e-06,
      "loss": 0.19330756664276122,
      "step": 655
    },
    {
      "acc": 0.9327902240325866,
      "epoch": 0.28472821397756687,
      "grad_norm": 31.25,
      "learning_rate": 4.877726375391156e-06,
      "loss": 0.17176212072372438,
      "step": 660
    },
    {
      "acc": 0.9481037924151696,
      "epoch": 0.28688524590163933,
      "grad_norm": 17.125,
      "learning_rate": 4.861828445482269e-06,
      "loss": 0.15666646957397462,
      "step": 665
    },
    {
      "acc": 0.9557344064386318,
      "epoch": 0.28904227782571185,
      "grad_norm": 15.875,
      "learning_rate": 4.8458450185706095e-06,
      "loss": 0.12006416320800781,
      "step": 670
    },
    {
      "acc": 0.9530612244897959,
      "epoch": 0.2911993097497843,
      "grad_norm": 13.375,
      "learning_rate": 4.829776828630942e-06,
      "loss": 0.12216727733612061,
      "step": 675
    },
    {
      "acc": 0.9454545454545454,
      "epoch": 0.29335634167385677,
      "grad_norm": 12.8125,
      "learning_rate": 4.813624613530434e-06,
      "loss": 0.1579872488975525,
      "step": 680
    },
    {
      "acc": 0.94831013916501,
      "epoch": 0.29551337359792923,
      "grad_norm": 10.75,
      "learning_rate": 4.79738911499477e-06,
      "loss": 0.14849933385848998,
      "step": 685
    },
    {
      "acc": 0.958904109589041,
      "epoch": 0.29767040552200175,
      "grad_norm": 9.125,
      "learning_rate": 4.781071078574092e-06,
      "loss": 0.12549154758453368,
      "step": 690
    },
    {
      "acc": 0.943359375,
      "epoch": 0.2998274374460742,
      "grad_norm": 19.25,
      "learning_rate": 4.764671253608765e-06,
      "loss": 0.11965863704681397,
      "step": 695
    },
    {
      "acc": 0.9474747474747475,
      "epoch": 0.30198446937014667,
      "grad_norm": 24.5,
      "learning_rate": 4.748190393194964e-06,
      "loss": 0.15644127130508423,
      "step": 700
    },
    {
      "epoch": 0.30198446937014667,
      "eval_acc": 0.9549386275640498,
      "eval_loss": 0.017001556232571602,
      "eval_mrr": 0.9429937548790007,
      "eval_ndcg": 0.9569733975457528,
      "eval_runtime": 87.9286,
      "eval_samples_per_second": 22.2,
      "eval_steps_per_second": 11.1,
      "step": 700
    },
    {
      "acc": 0.9433962264150944,
      "epoch": 0.30414150129421913,
      "grad_norm": 14.875,
      "learning_rate": 4.731629254150091e-06,
      "loss": 0.1656881332397461,
      "step": 705
    },
    {
      "acc": 0.9484536082474226,
      "epoch": 0.30629853321829165,
      "grad_norm": 16.875,
      "learning_rate": 4.714988596978023e-06,
      "loss": 0.16551544666290283,
      "step": 710
    },
    {
      "acc": 0.9601593625498008,
      "epoch": 0.3084555651423641,
      "grad_norm": 13.0625,
      "learning_rate": 4.698269185834188e-06,
      "loss": 0.13289109468460084,
      "step": 715
    },
    {
      "acc": 0.9621928166351607,
      "epoch": 0.31061259706643657,
      "grad_norm": 6.625,
      "learning_rate": 4.681471788490473e-06,
      "loss": 0.13107165098190307,
      "step": 720
    },
    {
      "acc": 0.9474708171206225,
      "epoch": 0.3127696289905091,
      "grad_norm": 17.125,
      "learning_rate": 4.664597176299972e-06,
      "loss": 0.15164241790771485,
      "step": 725
    },
    {
      "acc": 0.9386138613861386,
      "epoch": 0.31492666091458155,
      "grad_norm": 13.25,
      "learning_rate": 4.647646124161557e-06,
      "loss": 0.14998191595077515,
      "step": 730
    },
    {
      "acc": 0.9588014981273408,
      "epoch": 0.317083692838654,
      "grad_norm": 17.0,
      "learning_rate": 4.6306194104843005e-06,
      "loss": 0.1452179431915283,
      "step": 735
    },
    {
      "acc": 0.9697542533081286,
      "epoch": 0.31924072476272647,
      "grad_norm": 21.75,
      "learning_rate": 4.613517817151725e-06,
      "loss": 0.0863348662853241,
      "step": 740
    },
    {
      "acc": 0.9271653543307087,
      "epoch": 0.321397756686799,
      "grad_norm": 22.25,
      "learning_rate": 4.596342129485904e-06,
      "loss": 0.2219313383102417,
      "step": 745
    },
    {
      "acc": 0.948559670781893,
      "epoch": 0.32355478861087145,
      "grad_norm": 17.25,
      "learning_rate": 4.579093136211394e-06,
      "loss": 0.14628000259399415,
      "step": 750
    },
    {
      "acc": 0.9550321199143469,
      "epoch": 0.3257118205349439,
      "grad_norm": 19.375,
      "learning_rate": 4.561771629419018e-06,
      "loss": 0.14832402467727662,
      "step": 755
    },
    {
      "acc": 0.9540918163672655,
      "epoch": 0.32786885245901637,
      "grad_norm": 24.875,
      "learning_rate": 4.544378404529493e-06,
      "loss": 0.16552494764328002,
      "step": 760
    },
    {
      "acc": 0.9468302658486708,
      "epoch": 0.3300258843830889,
      "grad_norm": 11.75,
      "learning_rate": 4.526914260256897e-06,
      "loss": 0.18186469078063966,
      "step": 765
    },
    {
      "acc": 0.9342915811088296,
      "epoch": 0.33218291630716135,
      "grad_norm": 29.0,
      "learning_rate": 4.509379998572003e-06,
      "loss": 0.189409601688385,
      "step": 770
    },
    {
      "acc": 0.9652509652509652,
      "epoch": 0.3343399482312338,
      "grad_norm": 8.125,
      "learning_rate": 4.491776424665441e-06,
      "loss": 0.11067872047424317,
      "step": 775
    },
    {
      "acc": 0.9634408602150538,
      "epoch": 0.3364969801553063,
      "grad_norm": 19.125,
      "learning_rate": 4.474104346910724e-06,
      "loss": 0.12605880498886107,
      "step": 780
    },
    {
      "acc": 0.9620493358633776,
      "epoch": 0.3386540120793788,
      "grad_norm": 16.625,
      "learning_rate": 4.4563645768271375e-06,
      "loss": 0.1279581904411316,
      "step": 785
    },
    {
      "acc": 0.9608610567514677,
      "epoch": 0.34081104400345125,
      "grad_norm": 26.125,
      "learning_rate": 4.438557929042457e-06,
      "loss": 0.13559631109237671,
      "step": 790
    },
    {
      "acc": 0.963265306122449,
      "epoch": 0.3429680759275237,
      "grad_norm": 16.75,
      "learning_rate": 4.4206852212555544e-06,
      "loss": 0.11860382556915283,
      "step": 795
    },
    {
      "acc": 0.9424184261036468,
      "epoch": 0.3451251078515962,
      "grad_norm": 27.25,
      "learning_rate": 4.402747274198838e-06,
      "loss": 0.1454222559928894,
      "step": 800
    },
    {
      "epoch": 0.3451251078515962,
      "eval_acc": 0.9581514128017135,
      "eval_loss": 0.01614278368651867,
      "eval_mrr": 0.9435499609679938,
      "eval_ndcg": 0.9573436789903933,
      "eval_runtime": 106.3602,
      "eval_samples_per_second": 18.353,
      "eval_steps_per_second": 9.176,
      "step": 800
    },
    {
      "acc": 0.9533468559837728,
      "epoch": 0.3472821397756687,
      "grad_norm": 23.125,
      "learning_rate": 4.384744911600571e-06,
      "loss": 0.15178160667419432,
      "step": 805
    },
    {
      "acc": 0.9653767820773931,
      "epoch": 0.34943917169974115,
      "grad_norm": 16.625,
      "learning_rate": 4.36667896014704e-06,
      "loss": 0.12023615837097168,
      "step": 810
    },
    {
      "acc": 0.9449715370018975,
      "epoch": 0.3515962036238136,
      "grad_norm": 15.0625,
      "learning_rate": 4.3485502494445945e-06,
      "loss": 0.1382935166358948,
      "step": 815
    },
    {
      "acc": 0.9700598802395209,
      "epoch": 0.3537532355478861,
      "grad_norm": 4.75,
      "learning_rate": 4.330359611981552e-06,
      "loss": 0.09947049617767334,
      "step": 820
    },
    {
      "acc": 0.9504132231404959,
      "epoch": 0.3559102674719586,
      "grad_norm": 11.4375,
      "learning_rate": 4.312107883089969e-06,
      "loss": 0.15309990644454957,
      "step": 825
    },
    {
      "acc": 0.9606625258799172,
      "epoch": 0.35806729939603105,
      "grad_norm": 13.375,
      "learning_rate": 4.293795900907278e-06,
      "loss": 0.1290311336517334,
      "step": 830
    },
    {
      "acc": 0.9569471624266145,
      "epoch": 0.36022433132010356,
      "grad_norm": 13.875,
      "learning_rate": 4.275424506337804e-06,
      "loss": 0.1386420726776123,
      "step": 835
    },
    {
      "acc": 0.9567779960707269,
      "epoch": 0.362381363244176,
      "grad_norm": 15.0,
      "learning_rate": 4.256994543014147e-06,
      "loss": 0.13683913946151732,
      "step": 840
    },
    {
      "acc": 0.978515625,
      "epoch": 0.3645383951682485,
      "grad_norm": 10.875,
      "learning_rate": 4.2385068572584416e-06,
      "loss": 0.07557401657104493,
      "step": 845
    },
    {
      "acc": 0.9506903353057199,
      "epoch": 0.36669542709232095,
      "grad_norm": 21.875,
      "learning_rate": 4.2199622980434965e-06,
      "loss": 0.15320024490356446,
      "step": 850
    },
    {
      "acc": 0.9474708171206225,
      "epoch": 0.36885245901639346,
      "grad_norm": 30.875,
      "learning_rate": 4.2013617169537995e-06,
      "loss": 0.15146974325180054,
      "step": 855
    },
    {
      "acc": 0.936127744510978,
      "epoch": 0.3710094909404659,
      "grad_norm": 13.9375,
      "learning_rate": 4.182705968146426e-06,
      "loss": 0.18005824089050293,
      "step": 860
    },
    {
      "acc": 0.9661354581673307,
      "epoch": 0.3731665228645384,
      "grad_norm": 11.0625,
      "learning_rate": 4.163995908311802e-06,
      "loss": 0.10441750288009644,
      "step": 865
    },
    {
      "acc": 0.9671179883945842,
      "epoch": 0.37532355478861085,
      "grad_norm": 11.625,
      "learning_rate": 4.145232396634372e-06,
      "loss": 0.11835912466049195,
      "step": 870
    },
    {
      "acc": 0.9686888454011742,
      "epoch": 0.37748058671268336,
      "grad_norm": 8.1875,
      "learning_rate": 4.126416294753141e-06,
      "loss": 0.08334695696830749,
      "step": 875
    },
    {
      "acc": 0.9533898305084746,
      "epoch": 0.3796376186367558,
      "grad_norm": 15.0625,
      "learning_rate": 4.1075484667221095e-06,
      "loss": 0.1570521354675293,
      "step": 880
    },
    {
      "acc": 0.9628865979381444,
      "epoch": 0.3817946505608283,
      "grad_norm": 11.875,
      "learning_rate": 4.088629778970591e-06,
      "loss": 0.10125092267990113,
      "step": 885
    },
    {
      "acc": 0.9517102615694165,
      "epoch": 0.3839516824849008,
      "grad_norm": 10.3125,
      "learning_rate": 4.06966110026343e-06,
      "loss": 0.15835769176483155,
      "step": 890
    },
    {
      "acc": 0.9445506692160612,
      "epoch": 0.38610871440897326,
      "grad_norm": 51.0,
      "learning_rate": 4.050643301661107e-06,
      "loss": 0.17581632137298583,
      "step": 895
    },
    {
      "acc": 0.9544554455445544,
      "epoch": 0.3882657463330457,
      "grad_norm": 17.75,
      "learning_rate": 4.0315772564797325e-06,
      "loss": 0.16853252649307252,
      "step": 900
    },
    {
      "epoch": 0.3882657463330457,
      "eval_acc": 0.9614465771480353,
      "eval_loss": 0.015336191281676292,
      "eval_mrr": 0.9436902322404371,
      "eval_ndcg": 0.9575083647474146,
      "eval_runtime": 79.2201,
      "eval_samples_per_second": 24.64,
      "eval_steps_per_second": 12.32,
      "step": 900
    },
    {
      "acc": 0.9625984251968503,
      "epoch": 0.3904227782571182,
      "grad_norm": 9.5625,
      "learning_rate": 4.012463840250949e-06,
      "loss": 0.11725049018859864,
      "step": 905
    },
    {
      "acc": 0.9578313253012049,
      "epoch": 0.3925798101811907,
      "grad_norm": 26.125,
      "learning_rate": 3.993303930681726e-06,
      "loss": 0.11894035339355469,
      "step": 910
    },
    {
      "acc": 0.952,
      "epoch": 0.39473684210526316,
      "grad_norm": 26.375,
      "learning_rate": 3.974098407614051e-06,
      "loss": 0.14351377487182618,
      "step": 915
    },
    {
      "acc": 0.9588477366255144,
      "epoch": 0.3968938740293356,
      "grad_norm": 33.75,
      "learning_rate": 3.954848152984529e-06,
      "loss": 0.12398861646652222,
      "step": 920
    },
    {
      "acc": 0.9679358717434869,
      "epoch": 0.3990509059534081,
      "grad_norm": 10.4375,
      "learning_rate": 3.935554050783885e-06,
      "loss": 0.10695126056671142,
      "step": 925
    },
    {
      "acc": 0.9646365422396856,
      "epoch": 0.4012079378774806,
      "grad_norm": 9.5,
      "learning_rate": 3.916216987016363e-06,
      "loss": 0.11947808265686036,
      "step": 930
    },
    {
      "acc": 0.9588235294117647,
      "epoch": 0.40336496980155306,
      "grad_norm": 19.375,
      "learning_rate": 3.8968378496590485e-06,
      "loss": 0.12318050861358643,
      "step": 935
    },
    {
      "acc": 0.9699812382739212,
      "epoch": 0.4055220017256255,
      "grad_norm": 9.5625,
      "learning_rate": 3.877417528621087e-06,
      "loss": 0.08819143176078796,
      "step": 940
    },
    {
      "acc": 0.9529411764705882,
      "epoch": 0.40767903364969804,
      "grad_norm": 25.0,
      "learning_rate": 3.8579569157028195e-06,
      "loss": 0.1318346619606018,
      "step": 945
    },
    {
      "acc": 0.9395085066162571,
      "epoch": 0.4098360655737705,
      "grad_norm": 14.875,
      "learning_rate": 3.838456904554829e-06,
      "loss": 0.14923615455627443,
      "step": 950
    },
    {
      "acc": 0.9635974304068522,
      "epoch": 0.41199309749784296,
      "grad_norm": 11.25,
      "learning_rate": 3.8189183906369035e-06,
      "loss": 0.10555909872055054,
      "step": 955
    },
    {
      "acc": 0.9612244897959183,
      "epoch": 0.4141501294219154,
      "grad_norm": 27.0,
      "learning_rate": 3.799342271176918e-06,
      "loss": 0.11741663217544555,
      "step": 960
    },
    {
      "acc": 0.9591397849462365,
      "epoch": 0.41630716134598794,
      "grad_norm": 11.875,
      "learning_rate": 3.7797294451296307e-06,
      "loss": 0.1104978084564209,
      "step": 965
    },
    {
      "acc": 0.9538461538461539,
      "epoch": 0.4184641932700604,
      "grad_norm": 32.0,
      "learning_rate": 3.7600808131354033e-06,
      "loss": 0.12546448707580565,
      "step": 970
    },
    {
      "acc": 0.9642147117296223,
      "epoch": 0.42062122519413286,
      "grad_norm": 17.375,
      "learning_rate": 3.740397277478841e-06,
      "loss": 0.15062413215637208,
      "step": 975
    },
    {
      "acc": 0.9447731755424064,
      "epoch": 0.4227782571182053,
      "grad_norm": 20.25,
      "learning_rate": 3.720679742047358e-06,
      "loss": 0.19598883390426636,
      "step": 980
    },
    {
      "acc": 0.9592668024439919,
      "epoch": 0.42493528904227784,
      "grad_norm": 8.4375,
      "learning_rate": 3.700929112289674e-06,
      "loss": 0.11623775959014893,
      "step": 985
    },
    {
      "acc": 0.9693486590038314,
      "epoch": 0.4270923209663503,
      "grad_norm": 17.625,
      "learning_rate": 3.681146295174234e-06,
      "loss": 0.1038408637046814,
      "step": 990
    },
    {
      "acc": 0.9513184584178499,
      "epoch": 0.42924935289042276,
      "grad_norm": 6.40625,
      "learning_rate": 3.6613321991475553e-06,
      "loss": 0.16103934049606322,
      "step": 995
    },
    {
      "acc": 0.9710982658959537,
      "epoch": 0.4314063848144953,
      "grad_norm": 35.5,
      "learning_rate": 3.6414877340925163e-06,
      "loss": 0.1310266375541687,
      "step": 1000
    },
    {
      "epoch": 0.4314063848144953,
      "eval_acc": 0.9601285114095065,
      "eval_loss": 0.01538186427205801,
      "eval_mrr": 0.9503037177985949,
      "eval_ndcg": 0.9625558802620603,
      "eval_runtime": 84.007,
      "eval_samples_per_second": 23.236,
      "eval_steps_per_second": 11.618,
      "step": 1000
    },
    {
      "acc": 0.9592668024439919,
      "epoch": 0.43356341673856774,
      "grad_norm": 14.5625,
      "learning_rate": 3.6216138112865695e-06,
      "loss": 0.11078298091888428,
      "step": 1005
    },
    {
      "acc": 0.9509433962264151,
      "epoch": 0.4357204486626402,
      "grad_norm": 31.75,
      "learning_rate": 3.601711343359897e-06,
      "loss": 0.12666620016098024,
      "step": 1010
    },
    {
      "acc": 0.958984375,
      "epoch": 0.43787748058671266,
      "grad_norm": 7.84375,
      "learning_rate": 3.5817812442535008e-06,
      "loss": 0.10231037139892578,
      "step": 1015
    },
    {
      "acc": 0.9519038076152304,
      "epoch": 0.4400345125107852,
      "grad_norm": 34.75,
      "learning_rate": 3.561824429177234e-06,
      "loss": 0.14617985486984253,
      "step": 1020
    },
    {
      "acc": 0.9489603024574669,
      "epoch": 0.44219154443485764,
      "grad_norm": 24.75,
      "learning_rate": 3.541841814567774e-06,
      "loss": 0.15405564308166503,
      "step": 1025
    },
    {
      "acc": 0.9458413926499033,
      "epoch": 0.4443485763589301,
      "grad_norm": 25.875,
      "learning_rate": 3.521834318046539e-06,
      "loss": 0.1635340690612793,
      "step": 1030
    },
    {
      "acc": 0.964509394572025,
      "epoch": 0.44650560828300256,
      "grad_norm": 10.9375,
      "learning_rate": 3.5018028583775472e-06,
      "loss": 0.10313594341278076,
      "step": 1035
    },
    {
      "acc": 0.9542857142857143,
      "epoch": 0.4486626402070751,
      "grad_norm": 23.375,
      "learning_rate": 3.481748355425229e-06,
      "loss": 0.13693466186523437,
      "step": 1040
    },
    {
      "acc": 0.9563567362428842,
      "epoch": 0.45081967213114754,
      "grad_norm": 22.875,
      "learning_rate": 3.4616717301121857e-06,
      "loss": 0.123587965965271,
      "step": 1045
    },
    {
      "acc": 0.9590163934426229,
      "epoch": 0.45297670405522,
      "grad_norm": 12.5,
      "learning_rate": 3.441573904376899e-06,
      "loss": 0.13192965984344482,
      "step": 1050
    },
    {
      "acc": 0.9732824427480916,
      "epoch": 0.4551337359792925,
      "grad_norm": 20.875,
      "learning_rate": 3.4214558011313937e-06,
      "loss": 0.09249483346939087,
      "step": 1055
    },
    {
      "acc": 0.9580838323353293,
      "epoch": 0.457290767903365,
      "grad_norm": 26.75,
      "learning_rate": 3.4013183442188606e-06,
      "loss": 0.11342880725860596,
      "step": 1060
    },
    {
      "acc": 0.9686274509803922,
      "epoch": 0.45944779982743744,
      "grad_norm": 22.375,
      "learning_rate": 3.381162458371229e-06,
      "loss": 0.08873859643936158,
      "step": 1065
    },
    {
      "acc": 0.969758064516129,
      "epoch": 0.4616048317515099,
      "grad_norm": 39.0,
      "learning_rate": 3.3609890691667005e-06,
      "loss": 0.0837608277797699,
      "step": 1070
    },
    {
      "acc": 0.962671905697446,
      "epoch": 0.4637618636755824,
      "grad_norm": 16.375,
      "learning_rate": 3.340799102987251e-06,
      "loss": 0.11619801521301269,
      "step": 1075
    },
    {
      "acc": 0.9606003752345216,
      "epoch": 0.4659188955996549,
      "grad_norm": 15.0625,
      "learning_rate": 3.320593486976088e-06,
      "loss": 0.10996119976043701,
      "step": 1080
    },
    {
      "acc": 0.955193482688391,
      "epoch": 0.46807592752372734,
      "grad_norm": 24.5,
      "learning_rate": 3.300373148995072e-06,
      "loss": 0.12069922685623169,
      "step": 1085
    },
    {
      "acc": 0.9691991786447639,
      "epoch": 0.4702329594477998,
      "grad_norm": 7.90625,
      "learning_rate": 3.280139017582113e-06,
      "loss": 0.14299554824829103,
      "step": 1090
    },
    {
      "acc": 0.9743589743589743,
      "epoch": 0.4723899913718723,
      "grad_norm": 10.875,
      "learning_rate": 3.2598920219085285e-06,
      "loss": 0.08163526654243469,
      "step": 1095
    },
    {
      "acc": 0.9682242990654205,
      "epoch": 0.4745470232959448,
      "grad_norm": 9.25,
      "learning_rate": 3.2396330917363754e-06,
      "loss": 0.108160400390625,
      "step": 1100
    },
    {
      "epoch": 0.4745470232959448,
      "eval_acc": 0.964494604168383,
      "eval_loss": 0.014143780805170536,
      "eval_mrr": 0.9522601971116316,
      "eval_ndcg": 0.963948349172127,
      "eval_runtime": 88.1934,
      "eval_samples_per_second": 22.133,
      "eval_steps_per_second": 11.067,
      "step": 1100
    },
    {
      "acc": 0.9725490196078431,
      "epoch": 0.47670405522001724,
      "grad_norm": 11.875,
      "learning_rate": 3.219363157375755e-06,
      "loss": 0.08103461265563965,
      "step": 1105
    },
    {
      "acc": 0.97678916827853,
      "epoch": 0.47886108714408976,
      "grad_norm": 26.5,
      "learning_rate": 3.1990831496420897e-06,
      "loss": 0.08858168721199036,
      "step": 1110
    },
    {
      "acc": 0.9602385685884692,
      "epoch": 0.4810181190681622,
      "grad_norm": 15.6875,
      "learning_rate": 3.1787939998133853e-06,
      "loss": 0.11988803148269653,
      "step": 1115
    },
    {
      "acc": 0.9718875502008032,
      "epoch": 0.4831751509922347,
      "grad_norm": 19.0,
      "learning_rate": 3.158496639587459e-06,
      "loss": 0.09709318280220032,
      "step": 1120
    },
    {
      "acc": 0.948559670781893,
      "epoch": 0.48533218291630714,
      "grad_norm": 17.75,
      "learning_rate": 3.1381920010391566e-06,
      "loss": 0.1310401439666748,
      "step": 1125
    },
    {
      "acc": 0.9658444022770398,
      "epoch": 0.48748921484037966,
      "grad_norm": 27.875,
      "learning_rate": 3.1178810165775532e-06,
      "loss": 0.11905833482742309,
      "step": 1130
    },
    {
      "acc": 0.967680608365019,
      "epoch": 0.4896462467644521,
      "grad_norm": 14.9375,
      "learning_rate": 3.0975646189031345e-06,
      "loss": 0.09188791513442993,
      "step": 1135
    },
    {
      "acc": 0.9701195219123506,
      "epoch": 0.4918032786885246,
      "grad_norm": 14.3125,
      "learning_rate": 3.0772437409649664e-06,
      "loss": 0.09571850299835205,
      "step": 1140
    },
    {
      "acc": 0.9606741573033708,
      "epoch": 0.49396031061259704,
      "grad_norm": 16.875,
      "learning_rate": 3.056919315917851e-06,
      "loss": 0.1439572811126709,
      "step": 1145
    },
    {
      "acc": 0.9613152804642167,
      "epoch": 0.49611734253666956,
      "grad_norm": 23.625,
      "learning_rate": 3.0365922770794798e-06,
      "loss": 0.11582168340682983,
      "step": 1150
    },
    {
      "acc": 0.980276134122288,
      "epoch": 0.498274374460742,
      "grad_norm": 18.0,
      "learning_rate": 3.016263557887571e-06,
      "loss": 0.0658139169216156,
      "step": 1155
    },
    {
      "acc": 0.9581673306772909,
      "epoch": 0.5004314063848145,
      "grad_norm": 9.1875,
      "learning_rate": 2.995934091857007e-06,
      "loss": 0.11524436473846436,
      "step": 1160
    },
    {
      "acc": 0.9642105263157895,
      "epoch": 0.502588438308887,
      "grad_norm": 28.125,
      "learning_rate": 2.975604812536964e-06,
      "loss": 0.10523393154144287,
      "step": 1165
    },
    {
      "acc": 0.9673076923076923,
      "epoch": 0.5047454702329595,
      "grad_norm": 14.75,
      "learning_rate": 2.9552766534680456e-06,
      "loss": 0.10329176187515259,
      "step": 1170
    },
    {
      "acc": 0.9471544715447154,
      "epoch": 0.5069025021570319,
      "grad_norm": 22.5,
      "learning_rate": 2.9349505481394128e-06,
      "loss": 0.15873119831085206,
      "step": 1175
    },
    {
      "acc": 0.9680638722554891,
      "epoch": 0.5090595340811044,
      "grad_norm": 14.25,
      "learning_rate": 2.914627429945915e-06,
      "loss": 0.10048485994338989,
      "step": 1180
    },
    {
      "acc": 0.9631067961165048,
      "epoch": 0.5112165660051768,
      "grad_norm": 38.25,
      "learning_rate": 2.894308232145232e-06,
      "loss": 0.10857952833175659,
      "step": 1185
    },
    {
      "acc": 0.9545454545454546,
      "epoch": 0.5133735979292493,
      "grad_norm": 23.125,
      "learning_rate": 2.8739938878150124e-06,
      "loss": 0.11859817504882812,
      "step": 1190
    },
    {
      "acc": 0.9759036144578314,
      "epoch": 0.5155306298533219,
      "grad_norm": 14.8125,
      "learning_rate": 2.8536853298100302e-06,
      "loss": 0.06508604288101197,
      "step": 1195
    },
    {
      "acc": 0.9620253164556962,
      "epoch": 0.5176876617773943,
      "grad_norm": 10.6875,
      "learning_rate": 2.833383490719347e-06,
      "loss": 0.13556112051010133,
      "step": 1200
    },
    {
      "epoch": 0.5176876617773943,
      "eval_acc": 0.964824120603015,
      "eval_loss": 0.013948180712759495,
      "eval_mrr": 0.9515539617486339,
      "eval_ndcg": 0.9634276661273489,
      "eval_runtime": 79.2148,
      "eval_samples_per_second": 24.642,
      "eval_steps_per_second": 12.321,
      "step": 1200
    },
    {
      "acc": 0.948559670781893,
      "epoch": 0.5198446937014668,
      "grad_norm": 13.125,
      "learning_rate": 2.8130893028234826e-06,
      "loss": 0.16991710662841797,
      "step": 1205
    },
    {
      "acc": 0.9775967413441955,
      "epoch": 0.5220017256255393,
      "grad_norm": 16.5,
      "learning_rate": 2.7928036980516074e-06,
      "loss": 0.0850608766078949,
      "step": 1210
    },
    {
      "acc": 0.9673704414587332,
      "epoch": 0.5241587575496117,
      "grad_norm": 7.8125,
      "learning_rate": 2.7725276079387467e-06,
      "loss": 0.10886862277984619,
      "step": 1215
    },
    {
      "acc": 0.9599198396793587,
      "epoch": 0.5263157894736842,
      "grad_norm": 15.4375,
      "learning_rate": 2.752261963583003e-06,
      "loss": 0.10846264362335205,
      "step": 1220
    },
    {
      "acc": 0.9634888438133874,
      "epoch": 0.5284728213977566,
      "grad_norm": 44.75,
      "learning_rate": 2.7320076956028e-06,
      "loss": 0.137969446182251,
      "step": 1225
    },
    {
      "acc": 0.9713114754098361,
      "epoch": 0.5306298533218292,
      "grad_norm": 14.4375,
      "learning_rate": 2.711765734094143e-06,
      "loss": 0.11962342262268066,
      "step": 1230
    },
    {
      "acc": 0.9619238476953907,
      "epoch": 0.5327868852459017,
      "grad_norm": 27.75,
      "learning_rate": 2.6915370085879188e-06,
      "loss": 0.11013026237487793,
      "step": 1235
    },
    {
      "acc": 0.955193482688391,
      "epoch": 0.5349439171699741,
      "grad_norm": 34.0,
      "learning_rate": 2.6713224480071984e-06,
      "loss": 0.142191481590271,
      "step": 1240
    },
    {
      "acc": 0.9458874458874459,
      "epoch": 0.5371009490940466,
      "grad_norm": 9.0,
      "learning_rate": 2.651122980624588e-06,
      "loss": 0.16410914659500123,
      "step": 1245
    },
    {
      "acc": 0.9612403100775194,
      "epoch": 0.5392579810181191,
      "grad_norm": 18.875,
      "learning_rate": 2.630939534019599e-06,
      "loss": 0.11810561418533325,
      "step": 1250
    },
    {
      "acc": 0.9647058823529412,
      "epoch": 0.5414150129421915,
      "grad_norm": 24.25,
      "learning_rate": 2.6107730350360508e-06,
      "loss": 0.09320048093795777,
      "step": 1255
    },
    {
      "acc": 0.9604743083003953,
      "epoch": 0.543572044866264,
      "grad_norm": 12.5,
      "learning_rate": 2.5906244097395137e-06,
      "loss": 0.13027727603912354,
      "step": 1260
    },
    {
      "acc": 0.9547244094488189,
      "epoch": 0.5457290767903364,
      "grad_norm": 20.5,
      "learning_rate": 2.570494583374779e-06,
      "loss": 0.144273042678833,
      "step": 1265
    },
    {
      "acc": 0.9634615384615385,
      "epoch": 0.547886108714409,
      "grad_norm": 9.3125,
      "learning_rate": 2.5503844803233732e-06,
      "loss": 0.11550074815750122,
      "step": 1270
    },
    {
      "acc": 0.96,
      "epoch": 0.5500431406384815,
      "grad_norm": 18.875,
      "learning_rate": 2.53029502406111e-06,
      "loss": 0.13190865516662598,
      "step": 1275
    },
    {
      "acc": 0.9737903225806451,
      "epoch": 0.5522001725625539,
      "grad_norm": 11.5,
      "learning_rate": 2.510227137115681e-06,
      "loss": 0.07091407179832458,
      "step": 1280
    },
    {
      "acc": 0.952191235059761,
      "epoch": 0.5543572044866264,
      "grad_norm": 26.5,
      "learning_rate": 2.490181741024292e-06,
      "loss": 0.14382394552230834,
      "step": 1285
    },
    {
      "acc": 0.9730848861283644,
      "epoch": 0.5565142364106989,
      "grad_norm": 49.0,
      "learning_rate": 2.4701597562913486e-06,
      "loss": 0.09190328121185302,
      "step": 1290
    },
    {
      "acc": 0.9789674952198852,
      "epoch": 0.5586712683347713,
      "grad_norm": 9.0,
      "learning_rate": 2.4501621023461826e-06,
      "loss": 0.07015281319618225,
      "step": 1295
    },
    {
      "acc": 0.9668615984405458,
      "epoch": 0.5608283002588438,
      "grad_norm": 35.5,
      "learning_rate": 2.4301896975008326e-06,
      "loss": 0.14583213329315187,
      "step": 1300
    },
    {
      "epoch": 0.5608283002588438,
      "eval_acc": 0.9630941593211961,
      "eval_loss": 0.013961615972220898,
      "eval_mrr": 0.9501536885245901,
      "eval_ndcg": 0.962423945763765,
      "eval_runtime": 79.8407,
      "eval_samples_per_second": 24.449,
      "eval_steps_per_second": 12.224,
      "step": 1300
    },
    {
      "acc": 0.9674329501915708,
      "epoch": 0.5629853321829164,
      "grad_norm": 18.75,
      "learning_rate": 2.410243458907872e-06,
      "loss": 0.11050969362258911,
      "step": 1305
    },
    {
      "acc": 0.950381679389313,
      "epoch": 0.5651423641069888,
      "grad_norm": 18.25,
      "learning_rate": 2.390324302518298e-06,
      "loss": 0.1483738660812378,
      "step": 1310
    },
    {
      "acc": 0.9652351738241309,
      "epoch": 0.5672993960310613,
      "grad_norm": 15.6875,
      "learning_rate": 2.370433143039462e-06,
      "loss": 0.12723606824874878,
      "step": 1315
    },
    {
      "acc": 0.9678638941398866,
      "epoch": 0.5694564279551337,
      "grad_norm": 14.25,
      "learning_rate": 2.3505708938930706e-06,
      "loss": 0.10463025569915771,
      "step": 1320
    },
    {
      "acc": 0.9717171717171718,
      "epoch": 0.5716134598792062,
      "grad_norm": 22.5,
      "learning_rate": 2.3307384671732414e-06,
      "loss": 0.09857140183448791,
      "step": 1325
    },
    {
      "acc": 0.9667359667359667,
      "epoch": 0.5737704918032787,
      "grad_norm": 14.1875,
      "learning_rate": 2.310936773604614e-06,
      "loss": 0.11535818576812744,
      "step": 1330
    },
    {
      "acc": 0.9669421487603306,
      "epoch": 0.5759275237273511,
      "grad_norm": 8.4375,
      "learning_rate": 2.291166722500533e-06,
      "loss": 0.10772225856781006,
      "step": 1335
    },
    {
      "acc": 0.9666666666666667,
      "epoch": 0.5780845556514237,
      "grad_norm": 11.375,
      "learning_rate": 2.2714292217212883e-06,
      "loss": 0.10940985679626465,
      "step": 1340
    },
    {
      "acc": 0.9622266401590457,
      "epoch": 0.5802415875754962,
      "grad_norm": 17.125,
      "learning_rate": 2.2517251776324297e-06,
      "loss": 0.106083083152771,
      "step": 1345
    },
    {
      "acc": 0.9638095238095238,
      "epoch": 0.5823986194995686,
      "grad_norm": 18.375,
      "learning_rate": 2.232055495063139e-06,
      "loss": 0.1073039174079895,
      "step": 1350
    },
    {
      "acc": 0.9717741935483871,
      "epoch": 0.5845556514236411,
      "grad_norm": 15.875,
      "learning_rate": 2.212421077264685e-06,
      "loss": 0.08703945875167847,
      "step": 1355
    },
    {
      "acc": 0.9564356435643564,
      "epoch": 0.5867126833477135,
      "grad_norm": 55.5,
      "learning_rate": 2.192822825868944e-06,
      "loss": 0.16233315467834472,
      "step": 1360
    },
    {
      "acc": 0.9764705882352941,
      "epoch": 0.588869715271786,
      "grad_norm": 20.125,
      "learning_rate": 2.1732616408469933e-06,
      "loss": 0.08231468200683593,
      "step": 1365
    },
    {
      "acc": 0.9704724409448819,
      "epoch": 0.5910267471958585,
      "grad_norm": 16.625,
      "learning_rate": 2.1537384204677878e-06,
      "loss": 0.10110975503921509,
      "step": 1370
    },
    {
      "acc": 0.9421157684630739,
      "epoch": 0.5931837791199309,
      "grad_norm": 20.625,
      "learning_rate": 2.1342540612569065e-06,
      "loss": 0.1582808256149292,
      "step": 1375
    },
    {
      "acc": 0.9518304431599229,
      "epoch": 0.5953408110440035,
      "grad_norm": 24.125,
      "learning_rate": 2.1148094579553858e-06,
      "loss": 0.13189674615859986,
      "step": 1380
    },
    {
      "acc": 0.9643564356435643,
      "epoch": 0.597497842968076,
      "grad_norm": 14.5,
      "learning_rate": 2.0954055034786334e-06,
      "loss": 0.11206245422363281,
      "step": 1385
    },
    {
      "acc": 0.9660678642714571,
      "epoch": 0.5996548748921484,
      "grad_norm": 23.375,
      "learning_rate": 2.07604308887542e-06,
      "loss": 0.10885739326477051,
      "step": 1390
    },
    {
      "acc": 0.9610136452241715,
      "epoch": 0.6018119068162209,
      "grad_norm": 29.125,
      "learning_rate": 2.0567231032869656e-06,
      "loss": 0.11518199443817138,
      "step": 1395
    },
    {
      "acc": 0.9661733615221987,
      "epoch": 0.6039689387402933,
      "grad_norm": 49.0,
      "learning_rate": 2.0374464339061075e-06,
      "loss": 0.1330866575241089,
      "step": 1400
    },
    {
      "epoch": 0.6039689387402933,
      "eval_acc": 0.9657302907982536,
      "eval_loss": 0.013689501211047173,
      "eval_mrr": 0.9512752488290397,
      "eval_ndcg": 0.9632696768901375,
      "eval_runtime": 79.2378,
      "eval_samples_per_second": 24.635,
      "eval_steps_per_second": 12.317,
      "step": 1400
    },
    {
      "acc": 0.9659574468085106,
      "epoch": 0.6061259706643658,
      "grad_norm": 18.0,
      "learning_rate": 2.0182139659365603e-06,
      "loss": 0.13202260732650756,
      "step": 1405
    },
    {
      "acc": 0.9670103092783505,
      "epoch": 0.6082830025884383,
      "grad_norm": 15.1875,
      "learning_rate": 1.9990265825522662e-06,
      "loss": 0.13020153045654298,
      "step": 1410
    },
    {
      "acc": 0.9691991786447639,
      "epoch": 0.6104400345125108,
      "grad_norm": 11.3125,
      "learning_rate": 1.9798851648568368e-06,
      "loss": 0.10563652515411377,
      "step": 1415
    },
    {
      "acc": 0.9745098039215686,
      "epoch": 0.6125970664365833,
      "grad_norm": 16.875,
      "learning_rate": 1.960790591843097e-06,
      "loss": 0.10446355342864991,
      "step": 1420
    },
    {
      "acc": 0.967479674796748,
      "epoch": 0.6147540983606558,
      "grad_norm": 16.125,
      "learning_rate": 1.9417437403527154e-06,
      "loss": 0.09148574471473694,
      "step": 1425
    },
    {
      "acc": 0.9619238476953907,
      "epoch": 0.6169111302847282,
      "grad_norm": 24.25,
      "learning_rate": 1.9227454850359425e-06,
      "loss": 0.10498310327529907,
      "step": 1430
    },
    {
      "acc": 0.9658444022770398,
      "epoch": 0.6190681622088007,
      "grad_norm": 17.375,
      "learning_rate": 1.903796698311443e-06,
      "loss": 0.09558966159820556,
      "step": 1435
    },
    {
      "acc": 0.9481327800829875,
      "epoch": 0.6212251941328731,
      "grad_norm": 19.5,
      "learning_rate": 1.8848982503262351e-06,
      "loss": 0.16058825254440307,
      "step": 1440
    },
    {
      "acc": 0.9735349716446124,
      "epoch": 0.6233822260569456,
      "grad_norm": 21.0,
      "learning_rate": 1.8660510089157334e-06,
      "loss": 0.10411704778671264,
      "step": 1445
    },
    {
      "acc": 0.9736842105263158,
      "epoch": 0.6255392579810182,
      "grad_norm": 77.0,
      "learning_rate": 1.8472558395638938e-06,
      "loss": 0.09424603581428528,
      "step": 1450
    },
    {
      "acc": 0.9693251533742331,
      "epoch": 0.6276962899050906,
      "grad_norm": 9.4375,
      "learning_rate": 1.8285136053634757e-06,
      "loss": 0.12103959321975707,
      "step": 1455
    },
    {
      "acc": 0.9722222222222222,
      "epoch": 0.6298533218291631,
      "grad_norm": 15.625,
      "learning_rate": 1.8098251669764012e-06,
      "loss": 0.07807242274284362,
      "step": 1460
    },
    {
      "acc": 0.9619238476953907,
      "epoch": 0.6320103537532356,
      "grad_norm": 12.5625,
      "learning_rate": 1.7911913825942357e-06,
      "loss": 0.1451573133468628,
      "step": 1465
    },
    {
      "acc": 0.9725738396624473,
      "epoch": 0.634167385677308,
      "grad_norm": 9.0,
      "learning_rate": 1.7726131078987784e-06,
      "loss": 0.10025861263275146,
      "step": 1470
    },
    {
      "acc": 0.9704724409448819,
      "epoch": 0.6363244176013805,
      "grad_norm": 14.0625,
      "learning_rate": 1.7540911960227694e-06,
      "loss": 0.08995563387870789,
      "step": 1475
    },
    {
      "acc": 0.9669724770642202,
      "epoch": 0.6384814495254529,
      "grad_norm": 40.5,
      "learning_rate": 1.7356264975107124e-06,
      "loss": 0.10488073825836182,
      "step": 1480
    },
    {
      "acc": 0.9637404580152672,
      "epoch": 0.6406384814495254,
      "grad_norm": 45.25,
      "learning_rate": 1.7172198602798172e-06,
      "loss": 0.1231122612953186,
      "step": 1485
    },
    {
      "acc": 0.9597701149425287,
      "epoch": 0.642795513373598,
      "grad_norm": 13.4375,
      "learning_rate": 1.6988721295810601e-06,
      "loss": 0.13030797243118286,
      "step": 1490
    },
    {
      "acc": 0.9408163265306122,
      "epoch": 0.6449525452976704,
      "grad_norm": 12.9375,
      "learning_rate": 1.680584147960375e-06,
      "loss": 0.15982915163040162,
      "step": 1495
    },
    {
      "acc": 0.958984375,
      "epoch": 0.6471095772217429,
      "grad_norm": 44.0,
      "learning_rate": 1.6623567552199572e-06,
      "loss": 0.14942604303359985,
      "step": 1500
    },
    {
      "epoch": 0.6471095772217429,
      "eval_acc": 0.966554081884834,
      "eval_loss": 0.013300775550305843,
      "eval_mrr": 0.955933474824356,
      "eval_ndcg": 0.9667304856555736,
      "eval_runtime": 151.3502,
      "eval_samples_per_second": 12.897,
      "eval_steps_per_second": 6.449,
      "step": 1500
    },
    {
      "acc": 0.9682242990654205,
      "epoch": 0.6492666091458154,
      "grad_norm": 21.0,
      "learning_rate": 1.644190788379699e-06,
      "loss": 0.10443732738494874,
      "step": 1505
    },
    {
      "acc": 0.9662921348314607,
      "epoch": 0.6514236410698878,
      "grad_norm": 19.75,
      "learning_rate": 1.6260870816387567e-06,
      "loss": 0.1014707326889038,
      "step": 1510
    },
    {
      "acc": 0.9601518026565465,
      "epoch": 0.6535806729939603,
      "grad_norm": 19.625,
      "learning_rate": 1.6080464663372418e-06,
      "loss": 0.13788410425186157,
      "step": 1515
    },
    {
      "acc": 0.9372623574144486,
      "epoch": 0.6557377049180327,
      "grad_norm": 16.875,
      "learning_rate": 1.5900697709180413e-06,
      "loss": 0.15338197946548462,
      "step": 1520
    },
    {
      "acc": 0.9784735812133072,
      "epoch": 0.6578947368421053,
      "grad_norm": 17.0,
      "learning_rate": 1.5721578208887794e-06,
      "loss": 0.07349593639373779,
      "step": 1525
    },
    {
      "acc": 0.9643564356435643,
      "epoch": 0.6600517687661778,
      "grad_norm": 25.875,
      "learning_rate": 1.5543114387839121e-06,
      "loss": 0.12870337963104247,
      "step": 1530
    },
    {
      "acc": 0.9527720739219713,
      "epoch": 0.6622088006902502,
      "grad_norm": 44.5,
      "learning_rate": 1.5365314441269463e-06,
      "loss": 0.12610948085784912,
      "step": 1535
    },
    {
      "acc": 0.9777365491651205,
      "epoch": 0.6643658326143227,
      "grad_norm": 17.375,
      "learning_rate": 1.5188186533928136e-06,
      "loss": 0.07375502586364746,
      "step": 1540
    },
    {
      "acc": 0.9482758620689655,
      "epoch": 0.6665228645383952,
      "grad_norm": 21.25,
      "learning_rate": 1.501173879970375e-06,
      "loss": 0.14943064451217652,
      "step": 1545
    },
    {
      "acc": 0.9661654135338346,
      "epoch": 0.6686798964624676,
      "grad_norm": 2.265625,
      "learning_rate": 1.4835979341250695e-06,
      "loss": 0.08926547765731811,
      "step": 1550
    },
    {
      "acc": 0.9563409563409564,
      "epoch": 0.6708369283865401,
      "grad_norm": 38.0,
      "learning_rate": 1.4660916229617048e-06,
      "loss": 0.1205858826637268,
      "step": 1555
    },
    {
      "acc": 0.9730290456431535,
      "epoch": 0.6729939603106126,
      "grad_norm": 8.1875,
      "learning_rate": 1.448655750387395e-06,
      "loss": 0.09265510439872741,
      "step": 1560
    },
    {
      "acc": 0.969939879759519,
      "epoch": 0.6751509922346851,
      "grad_norm": 18.375,
      "learning_rate": 1.431291117074648e-06,
      "loss": 0.08199839591979981,
      "step": 1565
    },
    {
      "acc": 0.9661654135338346,
      "epoch": 0.6773080241587576,
      "grad_norm": 12.9375,
      "learning_rate": 1.4139985204245914e-06,
      "loss": 0.11682524681091308,
      "step": 1570
    },
    {
      "acc": 0.9781746031746031,
      "epoch": 0.67946505608283,
      "grad_norm": 5.8125,
      "learning_rate": 1.3967787545303558e-06,
      "loss": 0.06525716781616211,
      "step": 1575
    },
    {
      "acc": 0.9621052631578947,
      "epoch": 0.6816220880069025,
      "grad_norm": 9.1875,
      "learning_rate": 1.3796326101406157e-06,
      "loss": 0.1133004069328308,
      "step": 1580
    },
    {
      "acc": 0.977319587628866,
      "epoch": 0.683779119930975,
      "grad_norm": 21.375,
      "learning_rate": 1.362560874623272e-06,
      "loss": 0.09791160225868226,
      "step": 1585
    },
    {
      "acc": 0.9688715953307393,
      "epoch": 0.6859361518550474,
      "grad_norm": 11.3125,
      "learning_rate": 1.3455643319292957e-06,
      "loss": 0.09991573691368102,
      "step": 1590
    },
    {
      "acc": 0.946611909650924,
      "epoch": 0.6880931837791199,
      "grad_norm": 18.125,
      "learning_rate": 1.3286437625567298e-06,
      "loss": 0.149417245388031,
      "step": 1595
    },
    {
      "acc": 0.9684418145956607,
      "epoch": 0.6902502157031924,
      "grad_norm": 16.25,
      "learning_rate": 1.3117999435148463e-06,
      "loss": 0.10367587804794312,
      "step": 1600
    },
    {
      "epoch": 0.6902502157031924,
      "eval_acc": 0.9691902133618914,
      "eval_loss": 0.012719967402517796,
      "eval_mrr": 0.9549814597970335,
      "eval_ndcg": 0.9660630214705206,
      "eval_runtime": 79.2818,
      "eval_samples_per_second": 24.621,
      "eval_steps_per_second": 12.311,
      "step": 1600
    },
    {
      "acc": 0.9466950959488273,
      "epoch": 0.6924072476272649,
      "grad_norm": 40.0,
      "learning_rate": 1.2950336482884685e-06,
      "loss": 0.1607167363166809,
      "step": 1605
    },
    {
      "acc": 0.9825581395348837,
      "epoch": 0.6945642795513374,
      "grad_norm": 17.25,
      "learning_rate": 1.2783456468024478e-06,
      "loss": 0.05497826337814331,
      "step": 1610
    },
    {
      "acc": 0.9757575757575757,
      "epoch": 0.6967213114754098,
      "grad_norm": 6.1875,
      "learning_rate": 1.261736705386309e-06,
      "loss": 0.07923665642738342,
      "step": 1615
    },
    {
      "acc": 0.9713114754098361,
      "epoch": 0.6988783433994823,
      "grad_norm": 18.625,
      "learning_rate": 1.2452075867390614e-06,
      "loss": 0.1029202938079834,
      "step": 1620
    },
    {
      "acc": 0.9633401221995926,
      "epoch": 0.7010353753235548,
      "grad_norm": 26.25,
      "learning_rate": 1.2287590498941737e-06,
      "loss": 0.09318680763244629,
      "step": 1625
    },
    {
      "acc": 0.9791666666666666,
      "epoch": 0.7031924072476272,
      "grad_norm": 15.125,
      "learning_rate": 1.212391850184718e-06,
      "loss": 0.06999911665916443,
      "step": 1630
    },
    {
      "acc": 0.9701789264413518,
      "epoch": 0.7053494391716998,
      "grad_norm": 5.875,
      "learning_rate": 1.1961067392086857e-06,
      "loss": 0.0950109601020813,
      "step": 1635
    },
    {
      "acc": 0.9804305283757339,
      "epoch": 0.7075064710957722,
      "grad_norm": 16.25,
      "learning_rate": 1.1799044647944712e-06,
      "loss": 0.07231690287590027,
      "step": 1640
    },
    {
      "acc": 0.9720558882235529,
      "epoch": 0.7096635030198447,
      "grad_norm": 7.78125,
      "learning_rate": 1.1637857709665332e-06,
      "loss": 0.09113675355911255,
      "step": 1645
    },
    {
      "acc": 0.9636015325670498,
      "epoch": 0.7118205349439172,
      "grad_norm": 13.5625,
      "learning_rate": 1.147751397911227e-06,
      "loss": 0.10403518676757813,
      "step": 1650
    },
    {
      "acc": 0.9708171206225681,
      "epoch": 0.7139775668679896,
      "grad_norm": 7.53125,
      "learning_rate": 1.1318020819428137e-06,
      "loss": 0.09691992402076721,
      "step": 1655
    },
    {
      "acc": 0.9642105263157895,
      "epoch": 0.7161345987920621,
      "grad_norm": 9.6875,
      "learning_rate": 1.1159385554696505e-06,
      "loss": 0.10156847238540649,
      "step": 1660
    },
    {
      "acc": 0.9689922480620154,
      "epoch": 0.7182916307161346,
      "grad_norm": 17.625,
      "learning_rate": 1.1001615469605548e-06,
      "loss": 0.11043925285339355,
      "step": 1665
    },
    {
      "acc": 0.9627450980392157,
      "epoch": 0.7204486626402071,
      "grad_norm": 19.75,
      "learning_rate": 1.0844717809113533e-06,
      "loss": 0.10573551654815674,
      "step": 1670
    },
    {
      "acc": 0.9693251533742331,
      "epoch": 0.7226056945642796,
      "grad_norm": 9.0,
      "learning_rate": 1.0688699778116152e-06,
      "loss": 0.10334875583648681,
      "step": 1675
    },
    {
      "acc": 0.9562624254473161,
      "epoch": 0.724762726488352,
      "grad_norm": 25.25,
      "learning_rate": 1.053356854111562e-06,
      "loss": 0.12973486185073851,
      "step": 1680
    },
    {
      "acc": 0.9851380042462845,
      "epoch": 0.7269197584124245,
      "grad_norm": 10.75,
      "learning_rate": 1.0379331221891693e-06,
      "loss": 0.060935062170028684,
      "step": 1685
    },
    {
      "acc": 0.9683794466403162,
      "epoch": 0.729076790336497,
      "grad_norm": 19.5,
      "learning_rate": 1.0225994903174548e-06,
      "loss": 0.10458909273147583,
      "step": 1690
    },
    {
      "acc": 0.9680638722554891,
      "epoch": 0.7312338222605694,
      "grad_norm": 11.75,
      "learning_rate": 1.0073566626319522e-06,
      "loss": 0.10490895509719848,
      "step": 1695
    },
    {
      "acc": 0.9766536964980544,
      "epoch": 0.7333908541846419,
      "grad_norm": 14.4375,
      "learning_rate": 9.922053390983784e-07,
      "loss": 0.07817023992538452,
      "step": 1700
    },
    {
      "epoch": 0.7333908541846419,
      "eval_acc": 0.9688606969272593,
      "eval_loss": 0.012777810916304588,
      "eval_mrr": 0.9589456479313037,
      "eval_ndcg": 0.9690763069755148,
      "eval_runtime": 78.9281,
      "eval_samples_per_second": 24.731,
      "eval_steps_per_second": 12.366,
      "step": 1700
    },
    {
      "acc": 0.9959514170040485,
      "epoch": 0.7355478861087144,
      "grad_norm": 8.5625,
      "learning_rate": 9.771462154804868e-07,
      "loss": 0.03793552815914154,
      "step": 1705
    },
    {
      "acc": 0.9590643274853801,
      "epoch": 0.7377049180327869,
      "grad_norm": 12.25,
      "learning_rate": 9.621799833081207e-07,
      "loss": 0.10620832443237305,
      "step": 1710
    },
    {
      "acc": 0.9735234215885947,
      "epoch": 0.7398619499568594,
      "grad_norm": 32.75,
      "learning_rate": 9.473073298454589e-07,
      "loss": 0.10178903341293336,
      "step": 1715
    },
    {
      "acc": 0.9735234215885947,
      "epoch": 0.7420189818809318,
      "grad_norm": 19.875,
      "learning_rate": 9.325289380594497e-07,
      "loss": 0.10075312852859497,
      "step": 1720
    },
    {
      "acc": 0.972,
      "epoch": 0.7441760138050043,
      "grad_norm": 65.0,
      "learning_rate": 9.178454865884533e-07,
      "loss": 0.13943676948547362,
      "step": 1725
    },
    {
      "acc": 0.9418837675350702,
      "epoch": 0.7463330457290768,
      "grad_norm": 19.25,
      "learning_rate": 9.03257649711078e-07,
      "loss": 0.19055825471878052,
      "step": 1730
    },
    {
      "acc": 0.9684418145956607,
      "epoch": 0.7484900776531492,
      "grad_norm": 21.625,
      "learning_rate": 8.887660973152132e-07,
      "loss": 0.0985870122909546,
      "step": 1735
    },
    {
      "acc": 0.9464285714285714,
      "epoch": 0.7506471095772217,
      "grad_norm": 17.125,
      "learning_rate": 8.743714948672717e-07,
      "loss": 0.1542567014694214,
      "step": 1740
    },
    {
      "acc": 0.9735772357723578,
      "epoch": 0.7528041415012943,
      "grad_norm": 4.3125,
      "learning_rate": 8.600745033816266e-07,
      "loss": 0.08904030919075012,
      "step": 1745
    },
    {
      "acc": 0.9654471544715447,
      "epoch": 0.7549611734253667,
      "grad_norm": 19.75,
      "learning_rate": 8.458757793902626e-07,
      "loss": 0.12530235052108765,
      "step": 1750
    },
    {
      "acc": 0.9553398058252427,
      "epoch": 0.7571182053494392,
      "grad_norm": 16.25,
      "learning_rate": 8.317759749126208e-07,
      "loss": 0.11530464887619019,
      "step": 1755
    },
    {
      "acc": 0.9684418145956607,
      "epoch": 0.7592752372735116,
      "grad_norm": 15.375,
      "learning_rate": 8.17775737425663e-07,
      "loss": 0.09011921286582947,
      "step": 1760
    },
    {
      "acc": 0.959349593495935,
      "epoch": 0.7614322691975841,
      "grad_norm": 19.0,
      "learning_rate": 8.038757098341323e-07,
      "loss": 0.1243793249130249,
      "step": 1765
    },
    {
      "acc": 0.9701492537313433,
      "epoch": 0.7635893011216566,
      "grad_norm": 23.75,
      "learning_rate": 7.900765304410384e-07,
      "loss": 0.10980342626571656,
      "step": 1770
    },
    {
      "acc": 0.972,
      "epoch": 0.765746333045729,
      "grad_norm": 12.1875,
      "learning_rate": 7.763788329183404e-07,
      "loss": 0.11545180082321167,
      "step": 1775
    },
    {
      "acc": 0.9709864603481625,
      "epoch": 0.7679033649698016,
      "grad_norm": 54.0,
      "learning_rate": 7.627832462778483e-07,
      "loss": 0.08927801847457886,
      "step": 1780
    },
    {
      "acc": 0.9572649572649573,
      "epoch": 0.7700603968938741,
      "grad_norm": 34.0,
      "learning_rate": 7.492903948423435e-07,
      "loss": 0.11710215806961059,
      "step": 1785
    },
    {
      "acc": 0.980276134122288,
      "epoch": 0.7722174288179465,
      "grad_norm": 17.5,
      "learning_rate": 7.359008982169011e-07,
      "loss": 0.06560088992118836,
      "step": 1790
    },
    {
      "acc": 0.9651162790697675,
      "epoch": 0.774374460742019,
      "grad_norm": 18.75,
      "learning_rate": 7.226153712604432e-07,
      "loss": 0.11870806217193604,
      "step": 1795
    },
    {
      "acc": 0.967479674796748,
      "epoch": 0.7765314926660914,
      "grad_norm": 7.78125,
      "learning_rate": 7.094344240575007e-07,
      "loss": 0.10688618421554566,
      "step": 1800
    },
    {
      "epoch": 0.7765314926660914,
      "eval_acc": 0.9681192849493369,
      "eval_loss": 0.012869248166680336,
      "eval_mrr": 0.9557627097970336,
      "eval_ndcg": 0.9666167539180269,
      "eval_runtime": 104.0332,
      "eval_samples_per_second": 18.763,
      "eval_steps_per_second": 9.382,
      "step": 1800
    },
    {
      "acc": 0.9689922480620154,
      "epoch": 0.7786885245901639,
      "grad_norm": 25.875,
      "learning_rate": 6.963586618902001e-07,
      "loss": 0.10927298069000244,
      "step": 1805
    },
    {
      "acc": 0.9309664694280079,
      "epoch": 0.7808455565142364,
      "grad_norm": 42.25,
      "learning_rate": 6.83388685210466e-07,
      "loss": 0.19510157108306886,
      "step": 1810
    },
    {
      "acc": 0.9536290322580645,
      "epoch": 0.7830025884383088,
      "grad_norm": 20.5,
      "learning_rate": 6.705250896124497e-07,
      "loss": 0.14289662837982178,
      "step": 1815
    },
    {
      "acc": 0.9706498951781971,
      "epoch": 0.7851596203623814,
      "grad_norm": 11.4375,
      "learning_rate": 6.577684658051778e-07,
      "loss": 0.08139024376869201,
      "step": 1820
    },
    {
      "acc": 0.9577464788732394,
      "epoch": 0.7873166522864539,
      "grad_norm": 15.5,
      "learning_rate": 6.45119399585426e-07,
      "loss": 0.13177180290222168,
      "step": 1825
    },
    {
      "acc": 0.9609375,
      "epoch": 0.7894736842105263,
      "grad_norm": 12.9375,
      "learning_rate": 6.325784718108197e-07,
      "loss": 0.12202094793319702,
      "step": 1830
    },
    {
      "acc": 0.9866156787762906,
      "epoch": 0.7916307161345988,
      "grad_norm": 3.3125,
      "learning_rate": 6.201462583731599e-07,
      "loss": 0.06428595781326293,
      "step": 1835
    },
    {
      "acc": 0.959915611814346,
      "epoch": 0.7937877480586712,
      "grad_norm": 42.25,
      "learning_rate": 6.078233301719783e-07,
      "loss": 0.15991504192352296,
      "step": 1840
    },
    {
      "acc": 0.9642857142857143,
      "epoch": 0.7959447799827437,
      "grad_norm": 18.125,
      "learning_rate": 5.956102530883194e-07,
      "loss": 0.09823997020721435,
      "step": 1845
    },
    {
      "acc": 0.9693877551020408,
      "epoch": 0.7981018119068162,
      "grad_norm": 18.875,
      "learning_rate": 5.83507587958757e-07,
      "loss": 0.09222990274429321,
      "step": 1850
    },
    {
      "acc": 0.9811320754716981,
      "epoch": 0.8002588438308887,
      "grad_norm": 14.5,
      "learning_rate": 5.715158905496371e-07,
      "loss": 0.07574660181999207,
      "step": 1855
    },
    {
      "acc": 0.9813278008298755,
      "epoch": 0.8024158757549612,
      "grad_norm": 21.375,
      "learning_rate": 5.596357115315609e-07,
      "loss": 0.06097576022148132,
      "step": 1860
    },
    {
      "acc": 0.9708333333333333,
      "epoch": 0.8045729076790337,
      "grad_norm": 12.875,
      "learning_rate": 5.478675964540932e-07,
      "loss": 0.09323927164077758,
      "step": 1865
    },
    {
      "acc": 0.959830866807611,
      "epoch": 0.8067299396031061,
      "grad_norm": 15.375,
      "learning_rate": 5.362120857207107e-07,
      "loss": 0.1452409267425537,
      "step": 1870
    },
    {
      "acc": 0.9698795180722891,
      "epoch": 0.8088869715271786,
      "grad_norm": 21.375,
      "learning_rate": 5.246697145639902e-07,
      "loss": 0.09878731966018676,
      "step": 1875
    },
    {
      "acc": 0.95703125,
      "epoch": 0.811044003451251,
      "grad_norm": 19.75,
      "learning_rate": 5.132410130210248e-07,
      "loss": 0.11701784133911133,
      "step": 1880
    },
    {
      "acc": 0.9629629629629629,
      "epoch": 0.8132010353753235,
      "grad_norm": 45.0,
      "learning_rate": 5.019265059090872e-07,
      "loss": 0.10859339237213135,
      "step": 1885
    },
    {
      "acc": 0.9723865877712031,
      "epoch": 0.8153580672993961,
      "grad_norm": 11.375,
      "learning_rate": 4.907267128015301e-07,
      "loss": 0.12401903867721557,
      "step": 1890
    },
    {
      "acc": 0.9619771863117871,
      "epoch": 0.8175150992234685,
      "grad_norm": 22.25,
      "learning_rate": 4.796421480039241e-07,
      "loss": 0.15579386949539184,
      "step": 1895
    },
    {
      "acc": 0.9611650485436893,
      "epoch": 0.819672131147541,
      "grad_norm": 36.75,
      "learning_rate": 4.686733205304431e-07,
      "loss": 0.1097487449645996,
      "step": 1900
    },
    {
      "epoch": 0.819672131147541,
      "eval_acc": 0.9676250102973886,
      "eval_loss": 0.013201375491917133,
      "eval_mrr": 0.9557108704137391,
      "eval_ndcg": 0.966545438225074,
      "eval_runtime": 79.4758,
      "eval_samples_per_second": 24.561,
      "eval_steps_per_second": 12.28,
      "step": 1900
    },
    {
      "acc": 0.9821073558648111,
      "epoch": 0.8218291630716135,
      "grad_norm": 10.3125,
      "learning_rate": 4.5782073408048806e-07,
      "loss": 0.06990676522254943,
      "step": 1905
    },
    {
      "acc": 0.9524752475247524,
      "epoch": 0.8239861949956859,
      "grad_norm": 31.25,
      "learning_rate": 4.470848870155572e-07,
      "loss": 0.13890234231948853,
      "step": 1910
    },
    {
      "acc": 0.9555984555984556,
      "epoch": 0.8261432269197584,
      "grad_norm": 18.0,
      "learning_rate": 4.3646627233636093e-07,
      "loss": 0.14572898149490357,
      "step": 1915
    },
    {
      "acc": 0.9524714828897338,
      "epoch": 0.8283002588438308,
      "grad_norm": 37.25,
      "learning_rate": 4.259653776601826e-07,
      "loss": 0.15419791936874389,
      "step": 1920
    },
    {
      "acc": 0.9623015873015873,
      "epoch": 0.8304572907679033,
      "grad_norm": 29.375,
      "learning_rate": 4.155826851984866e-07,
      "loss": 0.11790182590484619,
      "step": 1925
    },
    {
      "acc": 0.9747081712062257,
      "epoch": 0.8326143226919759,
      "grad_norm": 30.625,
      "learning_rate": 4.0531867173477464e-07,
      "loss": 0.07656689286231995,
      "step": 1930
    },
    {
      "acc": 0.9800796812749004,
      "epoch": 0.8347713546160483,
      "grad_norm": 24.875,
      "learning_rate": 3.951738086026925e-07,
      "loss": 0.08629719018936158,
      "step": 1935
    },
    {
      "acc": 0.9488752556237219,
      "epoch": 0.8369283865401208,
      "grad_norm": 15.9375,
      "learning_rate": 3.8514856166438363e-07,
      "loss": 0.10850644111633301,
      "step": 1940
    },
    {
      "acc": 0.9452332657200812,
      "epoch": 0.8390854184641933,
      "grad_norm": 8.375,
      "learning_rate": 3.752433912890991e-07,
      "loss": 0.2388686418533325,
      "step": 1945
    },
    {
      "acc": 0.9788461538461538,
      "epoch": 0.8412424503882657,
      "grad_norm": 11.4375,
      "learning_rate": 3.654587523320528e-07,
      "loss": 0.07610541582107544,
      "step": 1950
    },
    {
      "acc": 0.9603174603174603,
      "epoch": 0.8433994823123382,
      "grad_norm": 11.625,
      "learning_rate": 3.557950941135394e-07,
      "loss": 0.13932143449783324,
      "step": 1955
    },
    {
      "acc": 0.9722772277227723,
      "epoch": 0.8455565142364107,
      "grad_norm": 14.875,
      "learning_rate": 3.462528603982974e-07,
      "loss": 0.07739153504371643,
      "step": 1960
    },
    {
      "acc": 0.9672131147540983,
      "epoch": 0.8477135461604832,
      "grad_norm": 19.0,
      "learning_rate": 3.3683248937513147e-07,
      "loss": 0.10050212144851685,
      "step": 1965
    },
    {
      "acc": 0.9665271966527197,
      "epoch": 0.8498705780845557,
      "grad_norm": 20.0,
      "learning_rate": 3.275344136367935e-07,
      "loss": 0.11014137268066407,
      "step": 1970
    },
    {
      "acc": 0.9486652977412731,
      "epoch": 0.8520276100086281,
      "grad_norm": 14.125,
      "learning_rate": 3.183590601601121e-07,
      "loss": 0.12993695735931396,
      "step": 1975
    },
    {
      "acc": 0.9609053497942387,
      "epoch": 0.8541846419327006,
      "grad_norm": 10.875,
      "learning_rate": 3.0930685028638974e-07,
      "loss": 0.12024986743927002,
      "step": 1980
    },
    {
      "acc": 0.9685534591194969,
      "epoch": 0.8563416738567731,
      "grad_norm": 34.75,
      "learning_rate": 3.003781997020524e-07,
      "loss": 0.12062886953353882,
      "step": 1985
    },
    {
      "acc": 0.9519038076152304,
      "epoch": 0.8584987057808455,
      "grad_norm": 31.25,
      "learning_rate": 2.9157351841956137e-07,
      "loss": 0.1291287899017334,
      "step": 1990
    },
    {
      "acc": 0.964,
      "epoch": 0.860655737704918,
      "grad_norm": 43.75,
      "learning_rate": 2.828932107585858e-07,
      "loss": 0.09483218193054199,
      "step": 1995
    },
    {
      "acc": 0.974,
      "epoch": 0.8628127696289906,
      "grad_norm": 27.125,
      "learning_rate": 2.743376753274339e-07,
      "loss": 0.10480066537857055,
      "step": 2000
    },
    {
      "epoch": 0.8628127696289906,
      "eval_acc": 0.9682016640579949,
      "eval_loss": 0.012947115115821362,
      "eval_mrr": 0.9556596409055426,
      "eval_ndcg": 0.9665805610309751,
      "eval_runtime": 79.6091,
      "eval_samples_per_second": 24.52,
      "eval_steps_per_second": 12.26,
      "step": 2000
    },
    {
      "acc": 0.9676113360323887,
      "epoch": 0.864969801553063,
      "grad_norm": 37.75,
      "learning_rate": 2.659073050047511e-07,
      "loss": 0.10173193216323853,
      "step": 2005
    },
    {
      "acc": 0.9330628803245437,
      "epoch": 0.8671268334771355,
      "grad_norm": 43.5,
      "learning_rate": 2.576024869214767e-07,
      "loss": 0.18904460668563844,
      "step": 2010
    },
    {
      "acc": 0.9647058823529412,
      "epoch": 0.869283865401208,
      "grad_norm": 25.75,
      "learning_rate": 2.4942360244306616e-07,
      "loss": 0.10903534889221192,
      "step": 2015
    },
    {
      "acc": 0.950920245398773,
      "epoch": 0.8714408973252804,
      "grad_norm": 15.5625,
      "learning_rate": 2.41371027151981e-07,
      "loss": 0.1522771120071411,
      "step": 2020
    },
    {
      "acc": 0.9818913480885312,
      "epoch": 0.8735979292493529,
      "grad_norm": 13.1875,
      "learning_rate": 2.3344513083043905e-07,
      "loss": 0.04949050545692444,
      "step": 2025
    },
    {
      "acc": 0.984251968503937,
      "epoch": 0.8757549611734253,
      "grad_norm": 9.5,
      "learning_rate": 2.2564627744343414e-07,
      "loss": 0.08176502585411072,
      "step": 2030
    },
    {
      "acc": 0.9659318637274549,
      "epoch": 0.8779119930974978,
      "grad_norm": 29.0,
      "learning_rate": 2.1797482512202438e-07,
      "loss": 0.10490905046463013,
      "step": 2035
    },
    {
      "acc": 0.9746588693957114,
      "epoch": 0.8800690250215704,
      "grad_norm": 17.875,
      "learning_rate": 2.1043112614688276e-07,
      "loss": 0.09062132239341736,
      "step": 2040
    },
    {
      "acc": 0.9331941544885177,
      "epoch": 0.8822260569456428,
      "grad_norm": 57.0,
      "learning_rate": 2.0301552693212566e-07,
      "loss": 0.14883551597595215,
      "step": 2045
    },
    {
      "acc": 0.9767441860465116,
      "epoch": 0.8843830888697153,
      "grad_norm": 15.375,
      "learning_rate": 1.9572836800939897e-07,
      "loss": 0.08334760069847107,
      "step": 2050
    },
    {
      "acc": 0.9729166666666667,
      "epoch": 0.8865401207937877,
      "grad_norm": 8.1875,
      "learning_rate": 1.885699840122439e-07,
      "loss": 0.08837355971336365,
      "step": 2055
    },
    {
      "acc": 0.9670103092783505,
      "epoch": 0.8886971527178602,
      "grad_norm": 15.0625,
      "learning_rate": 1.8154070366073094e-07,
      "loss": 0.10173630714416504,
      "step": 2060
    },
    {
      "acc": 0.9784735812133072,
      "epoch": 0.8908541846419327,
      "grad_norm": 14.5625,
      "learning_rate": 1.746408497463623e-07,
      "loss": 0.0801218330860138,
      "step": 2065
    },
    {
      "acc": 0.9614604462474645,
      "epoch": 0.8930112165660051,
      "grad_norm": 23.25,
      "learning_rate": 1.6787073911725103e-07,
      "loss": 0.11090768575668335,
      "step": 2070
    },
    {
      "acc": 0.9774127310061602,
      "epoch": 0.8951682484900777,
      "grad_norm": 16.25,
      "learning_rate": 1.6123068266356977e-07,
      "loss": 0.0828168511390686,
      "step": 2075
    },
    {
      "acc": 0.9747081712062257,
      "epoch": 0.8973252804141502,
      "grad_norm": 9.75,
      "learning_rate": 1.5472098530327506e-07,
      "loss": 0.07558534741401672,
      "step": 2080
    },
    {
      "acc": 0.9762376237623762,
      "epoch": 0.8994823123382226,
      "grad_norm": 25.375,
      "learning_rate": 1.4834194596810525e-07,
      "loss": 0.1000247836112976,
      "step": 2085
    },
    {
      "acc": 0.9703557312252964,
      "epoch": 0.9016393442622951,
      "grad_norm": 30.5,
      "learning_rate": 1.420938575898525e-07,
      "loss": 0.09505946040153504,
      "step": 2090
    },
    {
      "acc": 0.975,
      "epoch": 0.9037963761863675,
      "grad_norm": 25.875,
      "learning_rate": 1.3597700708691185e-07,
      "loss": 0.06200051307678223,
      "step": 2095
    },
    {
      "acc": 0.9657258064516129,
      "epoch": 0.90595340811044,
      "grad_norm": 23.625,
      "learning_rate": 1.299916753511058e-07,
      "loss": 0.11226943731307984,
      "step": 2100
    },
    {
      "epoch": 0.90595340811044,
      "eval_acc": 0.9686959387099432,
      "eval_loss": 0.013119321316480637,
      "eval_mrr": 0.9532567330210773,
      "eval_ndcg": 0.9646967453417445,
      "eval_runtime": 80.4978,
      "eval_samples_per_second": 24.249,
      "eval_steps_per_second": 12.125,
      "step": 2100
    },
    {
      "acc": 0.9623762376237623,
      "epoch": 0.9081104400345125,
      "grad_norm": 25.25,
      "learning_rate": 1.2413813723478384e-07,
      "loss": 0.1284480571746826,
      "step": 2105
    },
    {
      "acc": 0.974155069582505,
      "epoch": 0.910267471958585,
      "grad_norm": 20.375,
      "learning_rate": 1.1841666153820308e-07,
      "loss": 0.10379769802093505,
      "step": 2110
    },
    {
      "acc": 0.9459459459459459,
      "epoch": 0.9124245038826575,
      "grad_norm": 44.0,
      "learning_rate": 1.1282751099718425e-07,
      "loss": 0.11733911037445069,
      "step": 2115
    },
    {
      "acc": 0.9717171717171718,
      "epoch": 0.91458153580673,
      "grad_norm": 26.125,
      "learning_rate": 1.0737094227104516e-07,
      "loss": 0.09395996928215027,
      "step": 2120
    },
    {
      "acc": 0.9621513944223108,
      "epoch": 0.9167385677308024,
      "grad_norm": 8.75,
      "learning_rate": 1.0204720593081606e-07,
      "loss": 0.12494601011276245,
      "step": 2125
    },
    {
      "acc": 0.9595375722543352,
      "epoch": 0.9188955996548749,
      "grad_norm": 11.25,
      "learning_rate": 9.685654644773256e-08,
      "loss": 0.11165077686309814,
      "step": 2130
    },
    {
      "acc": 0.96875,
      "epoch": 0.9210526315789473,
      "grad_norm": 22.375,
      "learning_rate": 9.179920218200888e-08,
      "loss": 0.09169023633003234,
      "step": 2135
    },
    {
      "acc": 0.9607843137254902,
      "epoch": 0.9232096635030198,
      "grad_norm": 12.875,
      "learning_rate": 8.687540537189397e-08,
      "loss": 0.09511052966117858,
      "step": 2140
    },
    {
      "acc": 0.9585798816568047,
      "epoch": 0.9253666954270923,
      "grad_norm": 12.0,
      "learning_rate": 8.208538212300442e-08,
      "loss": 0.13230088949203492,
      "step": 2145
    },
    {
      "acc": 0.9676767676767677,
      "epoch": 0.9275237273511648,
      "grad_norm": 21.5,
      "learning_rate": 7.742935239794335e-08,
      "loss": 0.09337798357009888,
      "step": 2150
    },
    {
      "acc": 0.9746588693957114,
      "epoch": 0.9296807592752373,
      "grad_norm": 22.875,
      "learning_rate": 7.290753000619854e-08,
      "loss": 0.08780012726783752,
      "step": 2155
    },
    {
      "acc": 0.9766081871345029,
      "epoch": 0.9318377911993098,
      "grad_norm": 9.75,
      "learning_rate": 6.852012259432461e-08,
      "loss": 0.08427742719650269,
      "step": 2160
    },
    {
      "acc": 0.9631067961165048,
      "epoch": 0.9339948231233822,
      "grad_norm": 19.375,
      "learning_rate": 6.426733163640697e-08,
      "loss": 0.11798138618469238,
      "step": 2165
    },
    {
      "acc": 0.9481037924151696,
      "epoch": 0.9361518550474547,
      "grad_norm": 23.875,
      "learning_rate": 6.014935242481057e-08,
      "loss": 0.15075309276580812,
      "step": 2170
    },
    {
      "acc": 0.9627450980392157,
      "epoch": 0.9383088869715271,
      "grad_norm": 46.5,
      "learning_rate": 5.6166374061211724e-08,
      "loss": 0.11194332838058471,
      "step": 2175
    },
    {
      "acc": 0.9747368421052631,
      "epoch": 0.9404659188955996,
      "grad_norm": 21.375,
      "learning_rate": 5.2318579447914406e-08,
      "loss": 0.10739229917526245,
      "step": 2180
    },
    {
      "acc": 0.9782178217821782,
      "epoch": 0.9426229508196722,
      "grad_norm": 8.9375,
      "learning_rate": 4.8606145279450955e-08,
      "loss": 0.07264227271080018,
      "step": 2185
    },
    {
      "acc": 0.9506172839506173,
      "epoch": 0.9447799827437446,
      "grad_norm": 12.125,
      "learning_rate": 4.50292420344689e-08,
      "loss": 0.13095579147338868,
      "step": 2190
    },
    {
      "acc": 0.9532520325203252,
      "epoch": 0.9469370146678171,
      "grad_norm": 17.125,
      "learning_rate": 4.158803396790056e-08,
      "loss": 0.12362562417984009,
      "step": 2195
    },
    {
      "acc": 0.9645669291338582,
      "epoch": 0.9490940465918896,
      "grad_norm": 12.0,
      "learning_rate": 3.8282679103422756e-08,
      "loss": 0.11409047842025757,
      "step": 2200
    },
    {
      "epoch": 0.9490940465918896,
      "eval_acc": 0.9687783178186012,
      "eval_loss": 0.01316050160676241,
      "eval_mrr": 0.9534153005464481,
      "eval_ndcg": 0.9648771549963384,
      "eval_runtime": 79.4962,
      "eval_samples_per_second": 24.555,
      "eval_steps_per_second": 12.277,
      "step": 2200
    },
    {
      "acc": 0.9576271186440678,
      "epoch": 0.951251078515962,
      "grad_norm": 11.8125,
      "learning_rate": 3.5113329226198256e-08,
      "loss": 0.15816335678100585,
      "step": 2205
    },
    {
      "acc": 0.970954356846473,
      "epoch": 0.9534081104400345,
      "grad_norm": 14.0,
      "learning_rate": 3.20801298759067e-08,
      "loss": 0.09749370217323303,
      "step": 2210
    },
    {
      "acc": 0.9624505928853755,
      "epoch": 0.955565142364107,
      "grad_norm": 8.75,
      "learning_rate": 2.9183220340060936e-08,
      "loss": 0.12024798393249511,
      "step": 2215
    },
    {
      "acc": 0.9827213822894169,
      "epoch": 0.9577221742881795,
      "grad_norm": 8.4375,
      "learning_rate": 2.642273364760983e-08,
      "loss": 0.07396645545959472,
      "step": 2220
    },
    {
      "acc": 0.9569672131147541,
      "epoch": 0.959879206212252,
      "grad_norm": 35.75,
      "learning_rate": 2.3798796562832126e-08,
      "loss": 0.12084379196166992,
      "step": 2225
    },
    {
      "acc": 0.9570552147239264,
      "epoch": 0.9620362381363244,
      "grad_norm": 26.875,
      "learning_rate": 2.1311529579512766e-08,
      "loss": 0.1314695119857788,
      "step": 2230
    },
    {
      "acc": 0.9333333333333333,
      "epoch": 0.9641932700603969,
      "grad_norm": 25.75,
      "learning_rate": 1.8961046915409674e-08,
      "loss": 0.17206907272338867,
      "step": 2235
    },
    {
      "acc": 0.955193482688391,
      "epoch": 0.9663503019844694,
      "grad_norm": 12.625,
      "learning_rate": 1.674745650701126e-08,
      "loss": 0.14425760507583618,
      "step": 2240
    },
    {
      "acc": 0.9656488549618321,
      "epoch": 0.9685073339085418,
      "grad_norm": 18.125,
      "learning_rate": 1.4670860004576735e-08,
      "loss": 0.10901893377304077,
      "step": 2245
    },
    {
      "acc": 0.9776422764227642,
      "epoch": 0.9706643658326143,
      "grad_norm": 19.125,
      "learning_rate": 1.2731352767470838e-08,
      "loss": 0.09050199985504151,
      "step": 2250
    },
    {
      "acc": 0.9449715370018975,
      "epoch": 0.9728213977566867,
      "grad_norm": 25.625,
      "learning_rate": 1.0929023859783005e-08,
      "loss": 0.15088040828704835,
      "step": 2255
    },
    {
      "acc": 0.9625984251968503,
      "epoch": 0.9749784296807593,
      "grad_norm": 23.25,
      "learning_rate": 9.26395604623831e-09,
      "loss": 0.09685000777244568,
      "step": 2260
    },
    {
      "acc": 0.9695817490494296,
      "epoch": 0.9771354616048318,
      "grad_norm": 14.4375,
      "learning_rate": 7.736225788396833e-09,
      "loss": 0.09270592331886292,
      "step": 2265
    },
    {
      "acc": 0.9692622950819673,
      "epoch": 0.9792924935289042,
      "grad_norm": 8.75,
      "learning_rate": 6.345903241142481e-09,
      "loss": 0.09486258029937744,
      "step": 2270
    },
    {
      "acc": 0.9613034623217923,
      "epoch": 0.9814495254529767,
      "grad_norm": 7.375,
      "learning_rate": 5.0930522494612165e-09,
      "loss": 0.12018944025039673,
      "step": 2275
    },
    {
      "acc": 0.9679358717434869,
      "epoch": 0.9836065573770492,
      "grad_norm": 41.25,
      "learning_rate": 3.977730345508413e-09,
      "loss": 0.11004929542541504,
      "step": 2280
    },
    {
      "acc": 0.9537223340040242,
      "epoch": 0.9857635893011216,
      "grad_norm": 14.1875,
      "learning_rate": 2.9999887459692954e-09,
      "loss": 0.1255163073539734,
      "step": 2285
    },
    {
      "acc": 0.9727095516569201,
      "epoch": 0.9879206212251941,
      "grad_norm": 29.25,
      "learning_rate": 2.1598723497041616e-09,
      "loss": 0.09740299582481385,
      "step": 2290
    },
    {
      "acc": 0.978,
      "epoch": 0.9900776531492667,
      "grad_norm": 17.875,
      "learning_rate": 1.457419735688359e-09,
      "loss": 0.08217963576316833,
      "step": 2295
    },
    {
      "acc": 0.9664031620553359,
      "epoch": 0.9922346850733391,
      "grad_norm": 21.75,
      "learning_rate": 8.926631612410364e-10,
      "loss": 0.106975257396698,
      "step": 2300
    },
    {
      "epoch": 0.9922346850733391,
      "eval_acc": 0.9682016640579949,
      "eval_loss": 0.012847819365561008,
      "eval_mrr": 0.9552498048399688,
      "eval_ndcg": 0.9662674769692897,
      "eval_runtime": 79.8344,
      "eval_samples_per_second": 24.451,
      "eval_steps_per_second": 12.225,
      "step": 2300
    },
    {
      "acc": 0.944558521560575,
      "epoch": 0.9943917169974116,
      "grad_norm": 23.5,
      "learning_rate": 4.6562856054166436e-10,
      "loss": 0.17086119651794435,
      "step": 2305
    },
    {
      "acc": 0.9473684210526315,
      "epoch": 0.996548748921484,
      "grad_norm": 22.25,
      "learning_rate": 1.763355434416525e-10,
      "loss": 0.13650988340377807,
      "step": 2310
    },
    {
      "acc": 0.9756592292089249,
      "epoch": 0.9987057808455565,
      "grad_norm": 6.40625,
      "learning_rate": 2.479739456207053e-11,
      "loss": 0.08215545415878296,
      "step": 2315
    },
    {
      "epoch": 1.0,
      "eval_acc": 0.9688606969272593,
      "eval_loss": 0.0128474785014987,
      "eval_mrr": 0.954596018735363,
      "eval_ndcg": 0.9658596079307621,
      "eval_runtime": 78.5065,
      "eval_samples_per_second": 24.864,
      "eval_steps_per_second": 12.432,
      "step": 2318
    }
  ],
  "logging_steps": 5,
  "max_steps": 2318,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 400,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.864146106092093e+17,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}