{
  "best_metric": 0.8163072349117797,
  "best_model_checkpoint": "output/arabert-2stage/checkpoint-4000",
  "epoch": 0.6798096532970768,
  "eval_steps": 500,
  "global_step": 4000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.001699524133242692,
      "grad_norm": 18.25199317932129,
      "learning_rate": 1.3582342954159594e-07,
      "loss": 4.8813,
      "step": 10
    },
    {
      "epoch": 0.003399048266485384,
      "grad_norm": 22.529903411865234,
      "learning_rate": 3.0560271646859084e-07,
      "loss": 4.555,
      "step": 20
    },
    {
      "epoch": 0.005098572399728076,
      "grad_norm": 16.15692901611328,
      "learning_rate": 4.7538200339558575e-07,
      "loss": 4.5007,
      "step": 30
    },
    {
      "epoch": 0.006798096532970768,
      "grad_norm": 15.619988441467285,
      "learning_rate": 6.451612903225807e-07,
      "loss": 3.9173,
      "step": 40
    },
    {
      "epoch": 0.00849762066621346,
      "grad_norm": 16.881023406982422,
      "learning_rate": 8.149405772495757e-07,
      "loss": 4.3681,
      "step": 50
    },
    {
      "epoch": 0.010197144799456152,
      "grad_norm": 16.038394927978516,
      "learning_rate": 9.67741935483871e-07,
      "loss": 4.2357,
      "step": 60
    },
    {
      "epoch": 0.011896668932698844,
      "grad_norm": 12.993019104003906,
      "learning_rate": 1.137521222410866e-06,
      "loss": 4.1423,
      "step": 70
    },
    {
      "epoch": 0.013596193065941536,
      "grad_norm": 15.696857452392578,
      "learning_rate": 1.307300509337861e-06,
      "loss": 3.8028,
      "step": 80
    },
    {
      "epoch": 0.015295717199184228,
      "grad_norm": 10.482597351074219,
      "learning_rate": 1.477079796264856e-06,
      "loss": 3.8624,
      "step": 90
    },
    {
      "epoch": 0.01699524133242692,
      "grad_norm": 13.150389671325684,
      "learning_rate": 1.6468590831918508e-06,
      "loss": 3.5161,
      "step": 100
    },
    {
      "epoch": 0.018694765465669613,
      "grad_norm": 11.572787284851074,
      "learning_rate": 1.8166383701188457e-06,
      "loss": 3.4791,
      "step": 110
    },
    {
      "epoch": 0.020394289598912305,
      "grad_norm": 13.110374450683594,
      "learning_rate": 1.9864176570458403e-06,
      "loss": 3.6538,
      "step": 120
    },
    {
      "epoch": 0.022093813732154997,
      "grad_norm": 10.92455768585205,
      "learning_rate": 2.1561969439728354e-06,
      "loss": 2.8184,
      "step": 130
    },
    {
      "epoch": 0.02379333786539769,
      "grad_norm": 13.158881187438965,
      "learning_rate": 2.3259762308998304e-06,
      "loss": 3.321,
      "step": 140
    },
    {
      "epoch": 0.02549286199864038,
      "grad_norm": 7.724843502044678,
      "learning_rate": 2.4957555178268255e-06,
      "loss": 2.7731,
      "step": 150
    },
    {
      "epoch": 0.027192386131883073,
      "grad_norm": 9.516618728637695,
      "learning_rate": 2.6655348047538205e-06,
      "loss": 3.0967,
      "step": 160
    },
    {
      "epoch": 0.028891910265125765,
      "grad_norm": 8.883516311645508,
      "learning_rate": 2.835314091680815e-06,
      "loss": 2.8666,
      "step": 170
    },
    {
      "epoch": 0.030591434398368457,
      "grad_norm": 7.585137844085693,
      "learning_rate": 3.0050933786078102e-06,
      "loss": 2.2543,
      "step": 180
    },
    {
      "epoch": 0.032290958531611146,
      "grad_norm": 11.794560432434082,
      "learning_rate": 3.174872665534805e-06,
      "loss": 3.1811,
      "step": 190
    },
    {
      "epoch": 0.03399048266485384,
      "grad_norm": 8.631832122802734,
      "learning_rate": 3.3446519524618e-06,
      "loss": 2.5646,
      "step": 200
    },
    {
      "epoch": 0.03569000679809653,
      "grad_norm": 7.68147611618042,
      "learning_rate": 3.5144312393887946e-06,
      "loss": 2.8453,
      "step": 210
    },
    {
      "epoch": 0.037389530931339225,
      "grad_norm": 9.345367431640625,
      "learning_rate": 3.6842105263157896e-06,
      "loss": 3.0061,
      "step": 220
    },
    {
      "epoch": 0.039089055064581914,
      "grad_norm": 10.231043815612793,
      "learning_rate": 3.853989813242784e-06,
      "loss": 2.5211,
      "step": 230
    },
    {
      "epoch": 0.04078857919782461,
      "grad_norm": 6.374863147735596,
      "learning_rate": 4.02376910016978e-06,
      "loss": 2.6467,
      "step": 240
    },
    {
      "epoch": 0.0424881033310673,
      "grad_norm": 12.607449531555176,
      "learning_rate": 4.193548387096774e-06,
      "loss": 2.1417,
      "step": 250
    },
    {
      "epoch": 0.04418762746430999,
      "grad_norm": 12.550193786621094,
      "learning_rate": 4.36332767402377e-06,
      "loss": 2.7977,
      "step": 260
    },
    {
      "epoch": 0.04588715159755268,
      "grad_norm": 9.875368118286133,
      "learning_rate": 4.5331069609507645e-06,
      "loss": 2.8345,
      "step": 270
    },
    {
      "epoch": 0.04758667573079538,
      "grad_norm": 14.792099952697754,
      "learning_rate": 4.702886247877759e-06,
      "loss": 2.8812,
      "step": 280
    },
    {
      "epoch": 0.049286199864038066,
      "grad_norm": 11.13675308227539,
      "learning_rate": 4.872665534804754e-06,
      "loss": 2.493,
      "step": 290
    },
    {
      "epoch": 0.05098572399728076,
      "grad_norm": 15.685995101928711,
      "learning_rate": 5.042444821731749e-06,
      "loss": 2.5967,
      "step": 300
    },
    {
      "epoch": 0.05268524813052345,
      "grad_norm": 8.819021224975586,
      "learning_rate": 5.212224108658745e-06,
      "loss": 2.4955,
      "step": 310
    },
    {
      "epoch": 0.054384772263766146,
      "grad_norm": 12.058821678161621,
      "learning_rate": 5.3820033955857386e-06,
      "loss": 2.2387,
      "step": 320
    },
    {
      "epoch": 0.056084296397008834,
      "grad_norm": 9.35466194152832,
      "learning_rate": 5.551782682512734e-06,
      "loss": 2.6786,
      "step": 330
    },
    {
      "epoch": 0.05778382053025153,
      "grad_norm": 9.817248344421387,
      "learning_rate": 5.721561969439729e-06,
      "loss": 2.2292,
      "step": 340
    },
    {
      "epoch": 0.05948334466349422,
      "grad_norm": 7.38469934463501,
      "learning_rate": 5.891341256366724e-06,
      "loss": 1.8386,
      "step": 350
    },
    {
      "epoch": 0.061182868796736914,
      "grad_norm": 7.639986038208008,
      "learning_rate": 6.061120543293718e-06,
      "loss": 2.5317,
      "step": 360
    },
    {
      "epoch": 0.0628823929299796,
      "grad_norm": 14.127429962158203,
      "learning_rate": 6.2308998302207134e-06,
      "loss": 2.3225,
      "step": 370
    },
    {
      "epoch": 0.06458191706322229,
      "grad_norm": 6.480667591094971,
      "learning_rate": 6.400679117147709e-06,
      "loss": 2.5349,
      "step": 380
    },
    {
      "epoch": 0.066281441196465,
      "grad_norm": 12.269805908203125,
      "learning_rate": 6.5704584040747036e-06,
      "loss": 2.5192,
      "step": 390
    },
    {
      "epoch": 0.06798096532970768,
      "grad_norm": 10.89499568939209,
      "learning_rate": 6.740237691001699e-06,
      "loss": 2.6806,
      "step": 400
    },
    {
      "epoch": 0.06968048946295037,
      "grad_norm": 6.393769264221191,
      "learning_rate": 6.910016977928693e-06,
      "loss": 2.593,
      "step": 410
    },
    {
      "epoch": 0.07138001359619306,
      "grad_norm": 10.500598907470703,
      "learning_rate": 7.079796264855688e-06,
      "loss": 1.893,
      "step": 420
    },
    {
      "epoch": 0.07307953772943576,
      "grad_norm": 9.394804954528809,
      "learning_rate": 7.249575551782683e-06,
      "loss": 2.4957,
      "step": 430
    },
    {
      "epoch": 0.07477906186267845,
      "grad_norm": 8.060555458068848,
      "learning_rate": 7.4193548387096784e-06,
      "loss": 2.2887,
      "step": 440
    },
    {
      "epoch": 0.07647858599592114,
      "grad_norm": 7.361180305480957,
      "learning_rate": 7.589134125636672e-06,
      "loss": 1.8902,
      "step": 450
    },
    {
      "epoch": 0.07817811012916383,
      "grad_norm": 9.406976699829102,
      "learning_rate": 7.758913412563669e-06,
      "loss": 2.445,
      "step": 460
    },
    {
      "epoch": 0.07987763426240653,
      "grad_norm": 6.804340362548828,
      "learning_rate": 7.928692699490664e-06,
      "loss": 2.3188,
      "step": 470
    },
    {
      "epoch": 0.08157715839564922,
      "grad_norm": 8.055757522583008,
      "learning_rate": 8.098471986417658e-06,
      "loss": 1.7857,
      "step": 480
    },
    {
      "epoch": 0.08327668252889191,
      "grad_norm": 14.029518127441406,
      "learning_rate": 8.268251273344653e-06,
      "loss": 1.9323,
      "step": 490
    },
    {
      "epoch": 0.0849762066621346,
      "grad_norm": 7.238176345825195,
      "learning_rate": 8.438030560271647e-06,
      "loss": 2.2119,
      "step": 500
    },
    {
      "epoch": 0.0849762066621346,
      "eval_cosine_accuracy@1": 0.644,
      "eval_cosine_accuracy@10": 0.89,
      "eval_cosine_accuracy@3": 0.7905,
      "eval_cosine_accuracy@5": 0.842,
      "eval_cosine_map@100": 0.7319568991974511,
      "eval_cosine_mrr@10": 0.727909523809523,
      "eval_cosine_ndcg@10": 0.7672958038269853,
      "eval_cosine_precision@1": 0.644,
      "eval_cosine_precision@10": 0.089,
      "eval_cosine_precision@3": 0.2635,
      "eval_cosine_precision@5": 0.1684,
      "eval_cosine_recall@1": 0.644,
      "eval_cosine_recall@10": 0.89,
      "eval_cosine_recall@3": 0.7905,
      "eval_cosine_recall@5": 0.842,
      "eval_loss": 2.020264148712158,
      "eval_runtime": 2.7651,
      "eval_samples_per_second": 272.684,
      "eval_sequential_score": 0.7319568991974511,
      "eval_steps_per_second": 2.17,
      "eval_sts-dev_pearson_cosine": 0.7945015650907715,
      "eval_sts-dev_pearson_dot": 0.7892461139806375,
      "eval_sts-dev_pearson_euclidean": 0.7772212795710262,
      "eval_sts-dev_pearson_manhattan": 0.7758198419084321,
      "eval_sts-dev_pearson_max": 0.7945015650907715,
      "eval_sts-dev_spearman_cosine": 0.8029369749021982,
      "eval_sts-dev_spearman_dot": 0.7981715789995407,
      "eval_sts-dev_spearman_euclidean": 0.7974969539844081,
      "eval_sts-dev_spearman_manhattan": 0.793442700340276,
      "eval_sts-dev_spearman_max": 0.8029369749021982,
      "step": 500
    },
    {
      "epoch": 0.0866757307953773,
      "grad_norm": 8.096364974975586,
      "learning_rate": 8.607809847198643e-06,
      "loss": 2.4202,
      "step": 510
    },
    {
      "epoch": 0.08837525492861999,
      "grad_norm": 5.928977012634277,
      "learning_rate": 8.777589134125636e-06,
      "loss": 1.9747,
      "step": 520
    },
    {
      "epoch": 0.09007477906186268,
      "grad_norm": 5.913745880126953,
      "learning_rate": 8.947368421052632e-06,
      "loss": 2.727,
      "step": 530
    },
    {
      "epoch": 0.09177430319510536,
      "grad_norm": 11.565766334533691,
      "learning_rate": 9.117147707979627e-06,
      "loss": 2.2645,
      "step": 540
    },
    {
      "epoch": 0.09347382732834807,
      "grad_norm": 5.834039688110352,
      "learning_rate": 9.286926994906623e-06,
      "loss": 2.5484,
      "step": 550
    },
    {
      "epoch": 0.09517335146159075,
      "grad_norm": 9.030844688415527,
      "learning_rate": 9.456706281833618e-06,
      "loss": 2.0806,
      "step": 560
    },
    {
      "epoch": 0.09687287559483344,
      "grad_norm": 8.642946243286133,
      "learning_rate": 9.626485568760612e-06,
      "loss": 2.4332,
      "step": 570
    },
    {
      "epoch": 0.09857239972807613,
      "grad_norm": 8.626252174377441,
      "learning_rate": 9.796264855687608e-06,
      "loss": 1.866,
      "step": 580
    },
    {
      "epoch": 0.10027192386131883,
      "grad_norm": 7.73045539855957,
      "learning_rate": 9.966044142614601e-06,
      "loss": 2.1285,
      "step": 590
    },
    {
      "epoch": 0.10197144799456152,
      "grad_norm": 8.957172393798828,
      "learning_rate": 9.999943676843767e-06,
      "loss": 2.3799,
      "step": 600
    },
    {
      "epoch": 0.10367097212780421,
      "grad_norm": 7.423664093017578,
      "learning_rate": 9.999714866196319e-06,
      "loss": 2.6725,
      "step": 610
    },
    {
      "epoch": 0.1053704962610469,
      "grad_norm": 9.436025619506836,
      "learning_rate": 9.999310055908776e-06,
      "loss": 2.4767,
      "step": 620
    },
    {
      "epoch": 0.1070700203942896,
      "grad_norm": 8.621382713317871,
      "learning_rate": 9.998729260231243e-06,
      "loss": 1.9994,
      "step": 630
    },
    {
      "epoch": 0.10876954452753229,
      "grad_norm": 8.81654167175293,
      "learning_rate": 9.997972499608852e-06,
      "loss": 2.6699,
      "step": 640
    },
    {
      "epoch": 0.11046906866077498,
      "grad_norm": 6.4313154220581055,
      "learning_rate": 9.997039800681044e-06,
      "loss": 2.2264,
      "step": 650
    },
    {
      "epoch": 0.11216859279401767,
      "grad_norm": 8.586142539978027,
      "learning_rate": 9.995931196280622e-06,
      "loss": 2.0675,
      "step": 660
    },
    {
      "epoch": 0.11386811692726037,
      "grad_norm": 6.31871223449707,
      "learning_rate": 9.994646725432611e-06,
      "loss": 1.7862,
      "step": 670
    },
    {
      "epoch": 0.11556764106050306,
      "grad_norm": 9.907938957214355,
      "learning_rate": 9.993186433352867e-06,
      "loss": 2.5078,
      "step": 680
    },
    {
      "epoch": 0.11726716519374575,
      "grad_norm": 5.77314567565918,
      "learning_rate": 9.9915503714465e-06,
      "loss": 2.135,
      "step": 690
    },
    {
      "epoch": 0.11896668932698844,
      "grad_norm": 15.68774700164795,
      "learning_rate": 9.989738597306053e-06,
      "loss": 2.5356,
      "step": 700
    },
    {
      "epoch": 0.12066621346023114,
      "grad_norm": 13.062637329101562,
      "learning_rate": 9.987751174709489e-06,
      "loss": 2.4131,
      "step": 710
    },
    {
      "epoch": 0.12236573759347383,
      "grad_norm": 6.619427680969238,
      "learning_rate": 9.98558817361792e-06,
      "loss": 2.0465,
      "step": 720
    },
    {
      "epoch": 0.12406526172671652,
      "grad_norm": 10.573486328125,
      "learning_rate": 9.98324967017318e-06,
      "loss": 2.5195,
      "step": 730
    },
    {
      "epoch": 0.1257647858599592,
      "grad_norm": 10.98731803894043,
      "learning_rate": 9.98073574669511e-06,
      "loss": 2.3977,
      "step": 740
    },
    {
      "epoch": 0.1274643099932019,
      "grad_norm": 7.568258285522461,
      "learning_rate": 9.978046491678686e-06,
      "loss": 1.8389,
      "step": 750
    },
    {
      "epoch": 0.12916383412644458,
      "grad_norm": 6.886653423309326,
      "learning_rate": 9.975181999790888e-06,
      "loss": 2.1653,
      "step": 760
    },
    {
      "epoch": 0.13086335825968728,
      "grad_norm": 8.51340103149414,
      "learning_rate": 9.972142371867375e-06,
      "loss": 2.3171,
      "step": 770
    },
    {
      "epoch": 0.13256288239293,
      "grad_norm": 13.55966567993164,
      "learning_rate": 9.968927714908934e-06,
      "loss": 2.3126,
      "step": 780
    },
    {
      "epoch": 0.13426240652617266,
      "grad_norm": 6.183629035949707,
      "learning_rate": 9.965538142077709e-06,
      "loss": 1.8593,
      "step": 790
    },
    {
      "epoch": 0.13596193065941536,
      "grad_norm": 8.71070671081543,
      "learning_rate": 9.961973772693226e-06,
      "loss": 2.3248,
      "step": 800
    },
    {
      "epoch": 0.13766145479265807,
      "grad_norm": 6.578171253204346,
      "learning_rate": 9.958234732228184e-06,
      "loss": 1.4786,
      "step": 810
    },
    {
      "epoch": 0.13936097892590074,
      "grad_norm": 8.017528533935547,
      "learning_rate": 9.954321152304049e-06,
      "loss": 2.5924,
      "step": 820
    },
    {
      "epoch": 0.14106050305914344,
      "grad_norm": 9.630891799926758,
      "learning_rate": 9.950233170686404e-06,
      "loss": 2.7052,
      "step": 830
    },
    {
      "epoch": 0.14276002719238612,
      "grad_norm": 9.43925952911377,
      "learning_rate": 9.945970931280117e-06,
      "loss": 2.9104,
      "step": 840
    },
    {
      "epoch": 0.14445955132562882,
      "grad_norm": 7.2036967277526855,
      "learning_rate": 9.941534584124262e-06,
      "loss": 1.905,
      "step": 850
    },
    {
      "epoch": 0.14615907545887152,
      "grad_norm": 8.345267295837402,
      "learning_rate": 9.936924285386849e-06,
      "loss": 1.4475,
      "step": 860
    },
    {
      "epoch": 0.1478585995921142,
      "grad_norm": 6.2061614990234375,
      "learning_rate": 9.932140197359312e-06,
      "loss": 1.9715,
      "step": 870
    },
    {
      "epoch": 0.1495581237253569,
      "grad_norm": 8.147466659545898,
      "learning_rate": 9.927182488450813e-06,
      "loss": 1.922,
      "step": 880
    },
    {
      "epoch": 0.1512576478585996,
      "grad_norm": 12.269794464111328,
      "learning_rate": 9.9220513331823e-06,
      "loss": 2.5347,
      "step": 890
    },
    {
      "epoch": 0.15295717199184228,
      "grad_norm": 4.720547199249268,
      "learning_rate": 9.916746912180369e-06,
      "loss": 1.9091,
      "step": 900
    },
    {
      "epoch": 0.15465669612508498,
      "grad_norm": 10.795013427734375,
      "learning_rate": 9.911269412170906e-06,
      "loss": 2.127,
      "step": 910
    },
    {
      "epoch": 0.15635622025832766,
      "grad_norm": 9.331660270690918,
      "learning_rate": 9.905619025972513e-06,
      "loss": 2.0569,
      "step": 920
    },
    {
      "epoch": 0.15805574439157036,
      "grad_norm": 8.150259017944336,
      "learning_rate": 9.89979595248972e-06,
      "loss": 1.5199,
      "step": 930
    },
    {
      "epoch": 0.15975526852481306,
      "grad_norm": 5.694140434265137,
      "learning_rate": 9.893800396705985e-06,
      "loss": 2.1115,
      "step": 940
    },
    {
      "epoch": 0.16145479265805573,
      "grad_norm": 6.206644535064697,
      "learning_rate": 9.887632569676475e-06,
      "loss": 2.3499,
      "step": 950
    },
    {
      "epoch": 0.16315431679129844,
      "grad_norm": 4.6663408279418945,
      "learning_rate": 9.881292688520638e-06,
      "loss": 2.2428,
      "step": 960
    },
    {
      "epoch": 0.16485384092454114,
      "grad_norm": 7.22851037979126,
      "learning_rate": 9.874780976414565e-06,
      "loss": 2.3808,
      "step": 970
    },
    {
      "epoch": 0.16655336505778381,
      "grad_norm": 5.55798864364624,
      "learning_rate": 9.86809766258312e-06,
      "loss": 2.0025,
      "step": 980
    },
    {
      "epoch": 0.16825288919102652,
      "grad_norm": 5.861523628234863,
      "learning_rate": 9.86124298229189e-06,
      "loss": 2.0427,
      "step": 990
    },
    {
      "epoch": 0.1699524133242692,
      "grad_norm": 6.302640438079834,
      "learning_rate": 9.854217176838886e-06,
      "loss": 1.8858,
      "step": 1000
    },
    {
      "epoch": 0.1699524133242692,
      "eval_cosine_accuracy@1": 0.682,
      "eval_cosine_accuracy@10": 0.906,
      "eval_cosine_accuracy@3": 0.8185,
      "eval_cosine_accuracy@5": 0.858,
      "eval_cosine_map@100": 0.762636433194321,
      "eval_cosine_mrr@10": 0.7592498015873005,
      "eval_cosine_ndcg@10": 0.7948932488118442,
      "eval_cosine_precision@1": 0.682,
      "eval_cosine_precision@10": 0.09060000000000001,
      "eval_cosine_precision@3": 0.2728333333333333,
      "eval_cosine_precision@5": 0.1716,
      "eval_cosine_recall@1": 0.682,
      "eval_cosine_recall@10": 0.906,
      "eval_cosine_recall@3": 0.8185,
      "eval_cosine_recall@5": 0.858,
      "eval_loss": 1.8376002311706543,
      "eval_runtime": 2.7558,
      "eval_samples_per_second": 273.604,
      "eval_sequential_score": 0.762636433194321,
      "eval_steps_per_second": 2.177,
      "eval_sts-dev_pearson_cosine": 0.7985157354895591,
      "eval_sts-dev_pearson_dot": 0.7942162925984169,
      "eval_sts-dev_pearson_euclidean": 0.7855985286403331,
      "eval_sts-dev_pearson_manhattan": 0.7854261466194248,
      "eval_sts-dev_pearson_max": 0.7985157354895591,
      "eval_sts-dev_spearman_cosine": 0.8055793147203388,
      "eval_sts-dev_spearman_dot": 0.8010661505883895,
      "eval_sts-dev_spearman_euclidean": 0.8034068438279502,
      "eval_sts-dev_spearman_manhattan": 0.8018568915859646,
      "eval_sts-dev_spearman_max": 0.8055793147203388,
      "step": 1000
    },
    {
      "epoch": 0.1716519374575119,
      "grad_norm": 7.927140235900879,
      "learning_rate": 9.847020493546058e-06,
      "loss": 1.9554,
      "step": 1010
    },
    {
      "epoch": 0.1733514615907546,
      "grad_norm": 8.299236297607422,
      "learning_rate": 9.839653185750588e-06,
      "loss": 2.2341,
      "step": 1020
    },
    {
      "epoch": 0.17505098572399727,
      "grad_norm": 8.708532333374023,
      "learning_rate": 9.83211551279597e-06,
      "loss": 2.4421,
      "step": 1030
    },
    {
      "epoch": 0.17675050985723997,
      "grad_norm": 9.657612800598145,
      "learning_rate": 9.824407740022878e-06,
      "loss": 1.8227,
      "step": 1040
    },
    {
      "epoch": 0.17845003399048268,
      "grad_norm": 6.149839401245117,
      "learning_rate": 9.816530138759837e-06,
      "loss": 2.1157,
      "step": 1050
    },
    {
      "epoch": 0.18014955812372535,
      "grad_norm": 5.039497375488281,
      "learning_rate": 9.808482986313653e-06,
      "loss": 1.7899,
      "step": 1060
    },
    {
      "epoch": 0.18184908225696805,
      "grad_norm": 8.380406379699707,
      "learning_rate": 9.80026656595967e-06,
      "loss": 1.8153,
      "step": 1070
    },
    {
      "epoch": 0.18354860639021073,
      "grad_norm": 5.239027500152588,
      "learning_rate": 9.791881166931788e-06,
      "loss": 2.0095,
      "step": 1080
    },
    {
      "epoch": 0.18524813052345343,
      "grad_norm": 6.085160732269287,
      "learning_rate": 9.783327084412277e-06,
      "loss": 1.5087,
      "step": 1090
    },
    {
      "epoch": 0.18694765465669613,
      "grad_norm": 5.803791522979736,
      "learning_rate": 9.774604619521404e-06,
      "loss": 1.9106,
      "step": 1100
    },
    {
      "epoch": 0.1886471787899388,
      "grad_norm": 5.130899906158447,
      "learning_rate": 9.765714079306814e-06,
      "loss": 2.0189,
      "step": 1110
    },
    {
      "epoch": 0.1903467029231815,
      "grad_norm": 13.715828895568848,
      "learning_rate": 9.756655776732727e-06,
      "loss": 2.3282,
      "step": 1120
    },
    {
      "epoch": 0.1920462270564242,
      "grad_norm": 9.30579662322998,
      "learning_rate": 9.74743003066893e-06,
      "loss": 2.147,
      "step": 1130
    },
    {
      "epoch": 0.1937457511896669,
      "grad_norm": 6.4321675300598145,
      "learning_rate": 9.738037165879538e-06,
      "loss": 1.9017,
      "step": 1140
    },
    {
      "epoch": 0.1954452753229096,
      "grad_norm": 5.368739128112793,
      "learning_rate": 9.72847751301157e-06,
      "loss": 2.102,
      "step": 1150
    },
    {
      "epoch": 0.19714479945615226,
      "grad_norm": 5.458316802978516,
      "learning_rate": 9.718751408583312e-06,
      "loss": 2.1275,
      "step": 1160
    },
    {
      "epoch": 0.19884432358939497,
      "grad_norm": 8.618191719055176,
      "learning_rate": 9.708859194972462e-06,
      "loss": 2.6718,
      "step": 1170
    },
    {
      "epoch": 0.20054384772263767,
      "grad_norm": 7.549973487854004,
      "learning_rate": 9.698801220404086e-06,
      "loss": 1.8549,
      "step": 1180
    },
    {
      "epoch": 0.20224337185588034,
      "grad_norm": 5.044092655181885,
      "learning_rate": 9.688577838938358e-06,
      "loss": 2.1802,
      "step": 1190
    },
    {
      "epoch": 0.20394289598912305,
      "grad_norm": 5.195682048797607,
      "learning_rate": 9.67818941045809e-06,
      "loss": 2.2905,
      "step": 1200
    },
    {
      "epoch": 0.20564242012236575,
      "grad_norm": 11.789019584655762,
      "learning_rate": 9.667636300656073e-06,
      "loss": 2.283,
      "step": 1210
    },
    {
      "epoch": 0.20734194425560842,
      "grad_norm": 7.097940921783447,
      "learning_rate": 9.656918881022196e-06,
      "loss": 2.1976,
      "step": 1220
    },
    {
      "epoch": 0.20904146838885113,
      "grad_norm": 10.327863693237305,
      "learning_rate": 9.646037528830374e-06,
      "loss": 2.5894,
      "step": 1230
    },
    {
      "epoch": 0.2107409925220938,
      "grad_norm": 6.603916645050049,
      "learning_rate": 9.634992627125264e-06,
      "loss": 1.7659,
      "step": 1240
    },
    {
      "epoch": 0.2124405166553365,
      "grad_norm": 4.55158805847168,
      "learning_rate": 9.623784564708782e-06,
      "loss": 1.9581,
      "step": 1250
    },
    {
      "epoch": 0.2141400407885792,
      "grad_norm": 6.308363437652588,
      "learning_rate": 9.612413736126421e-06,
      "loss": 2.1419,
      "step": 1260
    },
    {
      "epoch": 0.21583956492182188,
      "grad_norm": 5.769819259643555,
      "learning_rate": 9.600880541653352e-06,
      "loss": 1.7988,
      "step": 1270
    },
    {
      "epoch": 0.21753908905506458,
      "grad_norm": 9.009718894958496,
      "learning_rate": 9.589185387280343e-06,
      "loss": 2.3537,
      "step": 1280
    },
    {
      "epoch": 0.21923861318830729,
      "grad_norm": 10.911920547485352,
      "learning_rate": 9.577328684699468e-06,
      "loss": 2.2776,
      "step": 1290
    },
    {
      "epoch": 0.22093813732154996,
      "grad_norm": 7.810164928436279,
      "learning_rate": 9.565310851289602e-06,
      "loss": 2.0633,
      "step": 1300
    },
    {
      "epoch": 0.22263766145479266,
      "grad_norm": 7.256144046783447,
      "learning_rate": 9.553132310101741e-06,
      "loss": 2.247,
      "step": 1310
    },
    {
      "epoch": 0.22433718558803534,
      "grad_norm": 8.476386070251465,
      "learning_rate": 9.540793489844106e-06,
      "loss": 1.6592,
      "step": 1320
    },
    {
      "epoch": 0.22603670972127804,
      "grad_norm": 7.557313919067383,
      "learning_rate": 9.528294824867054e-06,
      "loss": 1.8965,
      "step": 1330
    },
    {
      "epoch": 0.22773623385452074,
      "grad_norm": 5.973213195800781,
      "learning_rate": 9.515636755147781e-06,
      "loss": 2.3415,
      "step": 1340
    },
    {
      "epoch": 0.22943575798776342,
      "grad_norm": 6.3328471183776855,
      "learning_rate": 9.50281972627484e-06,
      "loss": 1.7645,
      "step": 1350
    },
    {
      "epoch": 0.23113528212100612,
      "grad_norm": 12.455780982971191,
      "learning_rate": 9.489844189432456e-06,
      "loss": 1.9076,
      "step": 1360
    },
    {
      "epoch": 0.23283480625424882,
      "grad_norm": 6.662944316864014,
      "learning_rate": 9.476710601384639e-06,
      "loss": 2.0724,
      "step": 1370
    },
    {
      "epoch": 0.2345343303874915,
      "grad_norm": 8.458195686340332,
      "learning_rate": 9.463419424459108e-06,
      "loss": 2.0766,
      "step": 1380
    },
    {
      "epoch": 0.2362338545207342,
      "grad_norm": 5.073389530181885,
      "learning_rate": 9.449971126531015e-06,
      "loss": 1.6035,
      "step": 1390
    },
    {
      "epoch": 0.23793337865397687,
      "grad_norm": 9.805305480957031,
      "learning_rate": 9.436366181006476e-06,
      "loss": 2.1389,
      "step": 1400
    },
    {
      "epoch": 0.23963290278721958,
      "grad_norm": 5.322582244873047,
      "learning_rate": 9.422605066805906e-06,
      "loss": 1.8332,
      "step": 1410
    },
    {
      "epoch": 0.24133242692046228,
      "grad_norm": 6.252927780151367,
      "learning_rate": 9.408688268347157e-06,
      "loss": 1.8316,
      "step": 1420
    },
    {
      "epoch": 0.24303195105370495,
      "grad_norm": 6.532680034637451,
      "learning_rate": 9.394616275528475e-06,
      "loss": 1.754,
      "step": 1430
    },
    {
      "epoch": 0.24473147518694766,
      "grad_norm": 10.015703201293945,
      "learning_rate": 9.380389583711238e-06,
      "loss": 2.0633,
      "step": 1440
    },
    {
      "epoch": 0.24643099932019036,
      "grad_norm": 4.93398904800415,
      "learning_rate": 9.366008693702535e-06,
      "loss": 1.9549,
      "step": 1450
    },
    {
      "epoch": 0.24813052345343303,
      "grad_norm": 10.338335037231445,
      "learning_rate": 9.351474111737531e-06,
      "loss": 2.1118,
      "step": 1460
    },
    {
      "epoch": 0.24983004758667574,
      "grad_norm": 6.729907512664795,
      "learning_rate": 9.336786349461637e-06,
      "loss": 2.1022,
      "step": 1470
    },
    {
      "epoch": 0.2515295717199184,
      "grad_norm": 6.920368194580078,
      "learning_rate": 9.321945923912518e-06,
      "loss": 1.7638,
      "step": 1480
    },
    {
      "epoch": 0.2532290958531611,
      "grad_norm": 8.145246505737305,
      "learning_rate": 9.306953357501874e-06,
      "loss": 2.3552,
      "step": 1490
    },
    {
      "epoch": 0.2549286199864038,
      "grad_norm": 7.566394805908203,
      "learning_rate": 9.291809177997061e-06,
      "loss": 2.3227,
      "step": 1500
    },
    {
      "epoch": 0.2549286199864038,
      "eval_cosine_accuracy@1": 0.6785,
      "eval_cosine_accuracy@10": 0.909,
      "eval_cosine_accuracy@3": 0.8225,
      "eval_cosine_accuracy@5": 0.863,
      "eval_cosine_map@100": 0.7639697235001636,
      "eval_cosine_mrr@10": 0.7603734126984125,
      "eval_cosine_ndcg@10": 0.7966821763190639,
      "eval_cosine_precision@1": 0.6785,
      "eval_cosine_precision@10": 0.09090000000000001,
      "eval_cosine_precision@3": 0.2741666666666666,
      "eval_cosine_precision@5": 0.17260000000000003,
      "eval_cosine_recall@1": 0.6785,
      "eval_cosine_recall@10": 0.909,
      "eval_cosine_recall@3": 0.8225,
      "eval_cosine_recall@5": 0.863,
      "eval_loss": 1.7280288934707642,
      "eval_runtime": 2.966,
      "eval_samples_per_second": 254.217,
      "eval_sequential_score": 0.7639697235001636,
      "eval_steps_per_second": 2.023,
      "eval_sts-dev_pearson_cosine": 0.7956596138775444,
      "eval_sts-dev_pearson_dot": 0.7898276993726723,
      "eval_sts-dev_pearson_euclidean": 0.7849266959832228,
      "eval_sts-dev_pearson_manhattan": 0.7850471763202945,
      "eval_sts-dev_pearson_max": 0.7956596138775444,
      "eval_sts-dev_spearman_cosine": 0.8035475738008022,
      "eval_sts-dev_spearman_dot": 0.7954777202652852,
      "eval_sts-dev_spearman_euclidean": 0.7982728913757071,
      "eval_sts-dev_spearman_manhattan": 0.8004630496417331,
      "eval_sts-dev_spearman_max": 0.8035475738008022,
      "step": 1500
    },
    {
      "epoch": 0.2566281441196465,
      "grad_norm": 8.947423934936523,
      "learning_rate": 9.276513918502508e-06,
      "loss": 2.3639,
      "step": 1510
    },
    {
      "epoch": 0.25832766825288916,
      "grad_norm": 7.069028377532959,
      "learning_rate": 9.261068117440955e-06,
      "loss": 1.717,
      "step": 1520
    },
    {
      "epoch": 0.26002719238613187,
      "grad_norm": 9.209027290344238,
      "learning_rate": 9.24547231853449e-06,
      "loss": 2.0703,
      "step": 1530
    },
    {
      "epoch": 0.26172671651937457,
      "grad_norm": 6.722319602966309,
      "learning_rate": 9.229727070785423e-06,
      "loss": 1.9918,
      "step": 1540
    },
    {
      "epoch": 0.26342624065261727,
      "grad_norm": 7.064426422119141,
      "learning_rate": 9.21383292845695e-06,
      "loss": 2.0636,
      "step": 1550
    },
    {
      "epoch": 0.26512576478586,
      "grad_norm": 5.03215217590332,
      "learning_rate": 9.197790451053641e-06,
      "loss": 1.7062,
      "step": 1560
    },
    {
      "epoch": 0.2668252889191027,
      "grad_norm": 8.875913619995117,
      "learning_rate": 9.18160020330175e-06,
      "loss": 2.2969,
      "step": 1570
    },
    {
      "epoch": 0.2685248130523453,
      "grad_norm": 8.441971778869629,
      "learning_rate": 9.165262755129337e-06,
      "loss": 1.9633,
      "step": 1580
    },
    {
      "epoch": 0.270224337185588,
      "grad_norm": 6.123201847076416,
      "learning_rate": 9.148778681646196e-06,
      "loss": 1.8068,
      "step": 1590
    },
    {
      "epoch": 0.27192386131883073,
      "grad_norm": 4.153637886047363,
      "learning_rate": 9.132148563123617e-06,
      "loss": 1.6155,
      "step": 1600
    },
    {
      "epoch": 0.27362338545207343,
      "grad_norm": 5.999012470245361,
      "learning_rate": 9.115372984973963e-06,
      "loss": 1.6978,
      "step": 1610
    },
    {
      "epoch": 0.27532290958531613,
      "grad_norm": 7.118943214416504,
      "learning_rate": 9.098452537730049e-06,
      "loss": 2.1703,
      "step": 1620
    },
    {
      "epoch": 0.2770224337185588,
      "grad_norm": 7.293047904968262,
      "learning_rate": 9.081387817024371e-06,
      "loss": 2.5395,
      "step": 1630
    },
    {
      "epoch": 0.2787219578518015,
      "grad_norm": 4.569684982299805,
      "learning_rate": 9.064179423568122e-06,
      "loss": 1.7744,
      "step": 1640
    },
    {
      "epoch": 0.2804214819850442,
      "grad_norm": 5.9083404541015625,
      "learning_rate": 9.046827963130063e-06,
      "loss": 1.7637,
      "step": 1650
    },
    {
      "epoch": 0.2821210061182869,
      "grad_norm": 7.332223892211914,
      "learning_rate": 9.02933404651518e-06,
      "loss": 1.9622,
      "step": 1660
    },
    {
      "epoch": 0.2838205302515296,
      "grad_norm": 6.511028289794922,
      "learning_rate": 9.011698289543199e-06,
      "loss": 1.3597,
      "step": 1670
    },
    {
      "epoch": 0.28552005438477224,
      "grad_norm": 6.4721832275390625,
      "learning_rate": 8.9939213130269e-06,
      "loss": 1.6605,
      "step": 1680
    },
    {
      "epoch": 0.28721957851801494,
      "grad_norm": 7.247612953186035,
      "learning_rate": 8.97600374275026e-06,
      "loss": 2.166,
      "step": 1690
    },
    {
      "epoch": 0.28891910265125764,
      "grad_norm": 10.479696273803711,
      "learning_rate": 8.957946209446435e-06,
      "loss": 1.959,
      "step": 1700
    },
    {
      "epoch": 0.29061862678450034,
      "grad_norm": 6.533411026000977,
      "learning_rate": 8.939749348775544e-06,
      "loss": 2.0912,
      "step": 1710
    },
    {
      "epoch": 0.29231815091774305,
      "grad_norm": 8.847211837768555,
      "learning_rate": 8.921413801302305e-06,
      "loss": 2.0446,
      "step": 1720
    },
    {
      "epoch": 0.29401767505098575,
      "grad_norm": 6.280109882354736,
      "learning_rate": 8.902940212473477e-06,
      "loss": 2.1172,
      "step": 1730
    },
    {
      "epoch": 0.2957171991842284,
      "grad_norm": 6.145227909088135,
      "learning_rate": 8.884329232595139e-06,
      "loss": 1.8379,
      "step": 1740
    },
    {
      "epoch": 0.2974167233174711,
      "grad_norm": 6.113027095794678,
      "learning_rate": 8.865581516809806e-06,
      "loss": 2.0789,
      "step": 1750
    },
    {
      "epoch": 0.2991162474507138,
      "grad_norm": 8.188186645507812,
      "learning_rate": 8.846697725073359e-06,
      "loss": 1.8014,
      "step": 1760
    },
    {
      "epoch": 0.3008157715839565,
      "grad_norm": 7.898641109466553,
      "learning_rate": 8.827678522131815e-06,
      "loss": 1.7158,
      "step": 1770
    },
    {
      "epoch": 0.3025152957171992,
      "grad_norm": 8.399277687072754,
      "learning_rate": 8.80852457749793e-06,
      "loss": 1.785,
      "step": 1780
    },
    {
      "epoch": 0.30421481985044185,
      "grad_norm": 4.910240650177002,
      "learning_rate": 8.789236565427627e-06,
      "loss": 1.9891,
      "step": 1790
    },
    {
      "epoch": 0.30591434398368456,
      "grad_norm": 10.41693115234375,
      "learning_rate": 8.769815164896262e-06,
      "loss": 2.134,
      "step": 1800
    },
    {
      "epoch": 0.30761386811692726,
      "grad_norm": 8.666768074035645,
      "learning_rate": 8.750261059574722e-06,
      "loss": 1.8226,
      "step": 1810
    },
    {
      "epoch": 0.30931339225016996,
      "grad_norm": 4.843673229217529,
      "learning_rate": 8.730574937805363e-06,
      "loss": 1.9502,
      "step": 1820
    },
    {
      "epoch": 0.31101291638341266,
      "grad_norm": 6.370514392852783,
      "learning_rate": 8.71075749257777e-06,
      "loss": 2.0422,
      "step": 1830
    },
    {
      "epoch": 0.3127124405166553,
      "grad_norm": 4.965259552001953,
      "learning_rate": 8.690809421504375e-06,
      "loss": 1.7759,
      "step": 1840
    },
    {
      "epoch": 0.314411964649898,
      "grad_norm": 6.794595241546631,
      "learning_rate": 8.670731426795888e-06,
      "loss": 2.0413,
      "step": 1850
    },
    {
      "epoch": 0.3161114887831407,
      "grad_norm": 4.695109844207764,
      "learning_rate": 8.65052421523658e-06,
      "loss": 1.9104,
      "step": 1860
    },
    {
      "epoch": 0.3178110129163834,
      "grad_norm": 6.058196544647217,
      "learning_rate": 8.63018849815941e-06,
      "loss": 2.3626,
      "step": 1870
    },
    {
      "epoch": 0.3195105370496261,
      "grad_norm": 4.891868591308594,
      "learning_rate": 8.609724991420981e-06,
      "loss": 1.8832,
      "step": 1880
    },
    {
      "epoch": 0.3212100611828688,
      "grad_norm": 5.250426292419434,
      "learning_rate": 8.589134415376333e-06,
      "loss": 2.0618,
      "step": 1890
    },
    {
      "epoch": 0.32290958531611147,
      "grad_norm": 4.848729133605957,
      "learning_rate": 8.568417494853598e-06,
      "loss": 2.0059,
      "step": 1900
    },
    {
      "epoch": 0.32460910944935417,
      "grad_norm": 8.6367769241333,
      "learning_rate": 8.547574959128472e-06,
      "loss": 1.9333,
      "step": 1910
    },
    {
      "epoch": 0.3263086335825969,
      "grad_norm": 5.955835342407227,
      "learning_rate": 8.526607541898555e-06,
      "loss": 2.3307,
      "step": 1920
    },
    {
      "epoch": 0.3280081577158396,
      "grad_norm": 7.700503349304199,
      "learning_rate": 8.505515981257515e-06,
      "loss": 2.3562,
      "step": 1930
    },
    {
      "epoch": 0.3297076818490823,
      "grad_norm": 6.47861909866333,
      "learning_rate": 8.48430101966911e-06,
      "loss": 1.7028,
      "step": 1940
    },
    {
      "epoch": 0.3314072059823249,
      "grad_norm": 6.167905807495117,
      "learning_rate": 8.462963403941046e-06,
      "loss": 1.6633,
      "step": 1950
    },
    {
      "epoch": 0.33310673011556763,
      "grad_norm": 5.260265827178955,
      "learning_rate": 8.441503885198699e-06,
      "loss": 1.6875,
      "step": 1960
    },
    {
      "epoch": 0.33480625424881033,
      "grad_norm": 5.841998100280762,
      "learning_rate": 8.419923218858658e-06,
      "loss": 1.9344,
      "step": 1970
    },
    {
      "epoch": 0.33650577838205303,
      "grad_norm": 9.941591262817383,
      "learning_rate": 8.398222164602151e-06,
      "loss": 2.1813,
      "step": 1980
    },
    {
      "epoch": 0.33820530251529574,
      "grad_norm": 8.419736862182617,
      "learning_rate": 8.376401486348286e-06,
      "loss": 1.9077,
      "step": 1990
    },
    {
      "epoch": 0.3399048266485384,
      "grad_norm": 6.201753616333008,
      "learning_rate": 8.354461952227175e-06,
      "loss": 1.9249,
      "step": 2000
    },
    {
      "epoch": 0.3399048266485384,
      "eval_cosine_accuracy@1": 0.695,
      "eval_cosine_accuracy@10": 0.9095,
      "eval_cosine_accuracy@3": 0.831,
      "eval_cosine_accuracy@5": 0.871,
      "eval_cosine_map@100": 0.7742690226366353,
      "eval_cosine_mrr@10": 0.7707448412698402,
      "eval_cosine_ndcg@10": 0.8045879491595074,
      "eval_cosine_precision@1": 0.695,
      "eval_cosine_precision@10": 0.09095000000000002,
      "eval_cosine_precision@3": 0.277,
      "eval_cosine_precision@5": 0.17420000000000002,
      "eval_cosine_recall@1": 0.695,
      "eval_cosine_recall@10": 0.9095,
      "eval_cosine_recall@3": 0.831,
      "eval_cosine_recall@5": 0.871,
      "eval_loss": 1.7007904052734375,
      "eval_runtime": 2.7948,
      "eval_samples_per_second": 269.79,
      "eval_sequential_score": 0.7742690226366353,
      "eval_steps_per_second": 2.147,
      "eval_sts-dev_pearson_cosine": 0.7952822056751716,
      "eval_sts-dev_pearson_dot": 0.7915034361455775,
      "eval_sts-dev_pearson_euclidean": 0.7874083556892966,
      "eval_sts-dev_pearson_manhattan": 0.7873169677574228,
      "eval_sts-dev_pearson_max": 0.7952822056751716,
      "eval_sts-dev_spearman_cosine": 0.8004780454585122,
      "eval_sts-dev_spearman_dot": 0.7972326119671326,
      "eval_sts-dev_spearman_euclidean": 0.7979533651258709,
      "eval_sts-dev_spearman_manhattan": 0.7990561344213346,
      "eval_sts-dev_spearman_max": 0.8004780454585122,
      "step": 2000
    },
    {
      "epoch": 0.3416043507817811,
      "grad_norm": 5.67038106918335,
      "learning_rate": 8.332404334552882e-06,
      "loss": 1.7941,
      "step": 2010
    },
    {
      "epoch": 0.3433038749150238,
      "grad_norm": 5.878876686096191,
      "learning_rate": 8.310229409796235e-06,
      "loss": 2.0737,
      "step": 2020
    },
    {
      "epoch": 0.3450033990482665,
      "grad_norm": 5.597275733947754,
      "learning_rate": 8.287937958557513e-06,
      "loss": 1.559,
      "step": 2030
    },
    {
      "epoch": 0.3467029231815092,
      "grad_norm": 6.382761478424072,
      "learning_rate": 8.265530765538938e-06,
      "loss": 1.3173,
      "step": 2040
    },
    {
      "epoch": 0.3484024473147519,
      "grad_norm": 7.027074813842773,
      "learning_rate": 8.243008619517074e-06,
      "loss": 2.1037,
      "step": 2050
    },
    {
      "epoch": 0.35010197144799454,
      "grad_norm": 6.432947158813477,
      "learning_rate": 8.220372313315053e-06,
      "loss": 2.27,
      "step": 2060
    },
    {
      "epoch": 0.35180149558123724,
      "grad_norm": 6.3885416984558105,
      "learning_rate": 8.197622643774663e-06,
      "loss": 1.7745,
      "step": 2070
    },
    {
      "epoch": 0.35350101971447995,
      "grad_norm": 6.885595798492432,
      "learning_rate": 8.174760411728308e-06,
      "loss": 1.3386,
      "step": 2080
    },
    {
      "epoch": 0.35520054384772265,
      "grad_norm": 7.397289752960205,
      "learning_rate": 8.1517864219708e-06,
      "loss": 1.9689,
      "step": 2090
    },
    {
      "epoch": 0.35690006798096535,
      "grad_norm": 4.5459885597229,
      "learning_rate": 8.128701483231048e-06,
      "loss": 1.699,
      "step": 2100
    },
    {
      "epoch": 0.358599592114208,
      "grad_norm": 4.440462589263916,
      "learning_rate": 8.105506408143571e-06,
      "loss": 1.8324,
      "step": 2110
    },
    {
      "epoch": 0.3602991162474507,
      "grad_norm": 5.693051338195801,
      "learning_rate": 8.082202013219912e-06,
      "loss": 1.8712,
      "step": 2120
    },
    {
      "epoch": 0.3619986403806934,
      "grad_norm": 7.692087650299072,
      "learning_rate": 8.058789118819872e-06,
      "loss": 2.4458,
      "step": 2130
    },
    {
      "epoch": 0.3636981645139361,
      "grad_norm": 5.480040073394775,
      "learning_rate": 8.03526854912265e-06,
      "loss": 1.7687,
      "step": 2140
    },
    {
      "epoch": 0.3653976886471788,
      "grad_norm": 7.727152347564697,
      "learning_rate": 8.011641132097822e-06,
      "loss": 1.9936,
      "step": 2150
    },
    {
      "epoch": 0.36709721278042146,
      "grad_norm": 5.268275260925293,
      "learning_rate": 7.987907699476202e-06,
      "loss": 1.7563,
      "step": 2160
    },
    {
      "epoch": 0.36879673691366416,
      "grad_norm": 12.53956413269043,
      "learning_rate": 7.96406908672055e-06,
      "loss": 2.0094,
      "step": 2170
    },
    {
      "epoch": 0.37049626104690686,
      "grad_norm": 6.193202495574951,
      "learning_rate": 7.940126132996172e-06,
      "loss": 2.0172,
      "step": 2180
    },
    {
      "epoch": 0.37219578518014956,
      "grad_norm": 8.85962200164795,
      "learning_rate": 7.916079681141385e-06,
      "loss": 1.9722,
      "step": 2190
    },
    {
      "epoch": 0.37389530931339227,
      "grad_norm": 8.260863304138184,
      "learning_rate": 7.891930577637832e-06,
      "loss": 1.8102,
      "step": 2200
    },
    {
      "epoch": 0.37559483344663497,
      "grad_norm": 7.650066375732422,
      "learning_rate": 7.867679672580696e-06,
      "loss": 2.1561,
      "step": 2210
    },
    {
      "epoch": 0.3772943575798776,
      "grad_norm": 9.542856216430664,
      "learning_rate": 7.843327819648774e-06,
      "loss": 2.1181,
      "step": 2220
    },
    {
      "epoch": 0.3789938817131203,
      "grad_norm": 9.622944831848145,
      "learning_rate": 7.818875876074421e-06,
      "loss": 1.9387,
      "step": 2230
    },
    {
      "epoch": 0.380693405846363,
      "grad_norm": 5.241047382354736,
      "learning_rate": 7.794324702613376e-06,
      "loss": 1.7689,
      "step": 2240
    },
    {
      "epoch": 0.3823929299796057,
      "grad_norm": 10.481565475463867,
      "learning_rate": 7.769675163514463e-06,
      "loss": 1.9585,
      "step": 2250
    },
    {
      "epoch": 0.3840924541128484,
      "grad_norm": 8.667862892150879,
      "learning_rate": 7.744928126489169e-06,
      "loss": 1.5681,
      "step": 2260
    },
    {
      "epoch": 0.38579197824609107,
      "grad_norm": 5.036750793457031,
      "learning_rate": 7.720084462681092e-06,
      "loss": 2.0402,
      "step": 2270
    },
    {
      "epoch": 0.3874915023793338,
      "grad_norm": 5.431954383850098,
      "learning_rate": 7.695145046635281e-06,
      "loss": 1.78,
      "step": 2280
    },
    {
      "epoch": 0.3891910265125765,
      "grad_norm": 7.302848815917969,
      "learning_rate": 7.670110756267455e-06,
      "loss": 1.9524,
      "step": 2290
    },
    {
      "epoch": 0.3908905506458192,
      "grad_norm": 8.023348808288574,
      "learning_rate": 7.644982472833083e-06,
      "loss": 1.9678,
      "step": 2300
    },
    {
      "epoch": 0.3925900747790619,
      "grad_norm": 7.136281967163086,
      "learning_rate": 7.61976108089638e-06,
      "loss": 1.6553,
      "step": 2310
    },
    {
      "epoch": 0.39428959891230453,
      "grad_norm": 8.916802406311035,
      "learning_rate": 7.594447468299156e-06,
      "loss": 1.8378,
      "step": 2320
    },
    {
      "epoch": 0.39598912304554723,
      "grad_norm": 8.570812225341797,
      "learning_rate": 7.56904252612957e-06,
      "loss": 1.8066,
      "step": 2330
    },
    {
      "epoch": 0.39768864717878993,
      "grad_norm": 10.376533508300781,
      "learning_rate": 7.543547148690757e-06,
      "loss": 1.8585,
      "step": 2340
    },
    {
      "epoch": 0.39938817131203264,
      "grad_norm": 5.894679546356201,
      "learning_rate": 7.517962233469345e-06,
      "loss": 1.4762,
      "step": 2350
    },
    {
      "epoch": 0.40108769544527534,
      "grad_norm": 6.78580904006958,
      "learning_rate": 7.492288681103872e-06,
      "loss": 1.6828,
      "step": 2360
    },
    {
      "epoch": 0.40278721957851804,
      "grad_norm": 6.131538391113281,
      "learning_rate": 7.4665273953530695e-06,
      "loss": 1.6288,
      "step": 2370
    },
    {
      "epoch": 0.4044867437117607,
      "grad_norm": 5.721993923187256,
      "learning_rate": 7.440679283064059e-06,
      "loss": 1.7679,
      "step": 2380
    },
    {
      "epoch": 0.4061862678450034,
      "grad_norm": 6.973949909210205,
      "learning_rate": 7.41474525414042e-06,
      "loss": 2.128,
      "step": 2390
    },
    {
      "epoch": 0.4078857919782461,
      "grad_norm": 6.6178202629089355,
      "learning_rate": 7.388726221510163e-06,
      "loss": 1.3543,
      "step": 2400
    },
    {
      "epoch": 0.4095853161114888,
      "grad_norm": 5.370213985443115,
      "learning_rate": 7.3626231010935974e-06,
      "loss": 1.806,
      "step": 2410
    },
    {
      "epoch": 0.4112848402447315,
      "grad_norm": 8.617044448852539,
      "learning_rate": 7.336436811771085e-06,
      "loss": 1.8597,
      "step": 2420
    },
    {
      "epoch": 0.41298436437797414,
      "grad_norm": 12.804397583007812,
      "learning_rate": 7.310168275350692e-06,
      "loss": 2.2319,
      "step": 2430
    },
    {
      "epoch": 0.41468388851121685,
      "grad_norm": 8.611510276794434,
      "learning_rate": 7.28381841653574e-06,
      "loss": 1.9433,
      "step": 2440
    },
    {
      "epoch": 0.41638341264445955,
      "grad_norm": 5.746379375457764,
      "learning_rate": 7.257388162892261e-06,
      "loss": 1.7124,
      "step": 2450
    },
    {
      "epoch": 0.41808293677770225,
      "grad_norm": 9.15091609954834,
      "learning_rate": 7.23087844481634e-06,
      "loss": 1.702,
      "step": 2460
    },
    {
      "epoch": 0.41978246091094495,
      "grad_norm": 5.277801513671875,
      "learning_rate": 7.204290195501358e-06,
      "loss": 1.5501,
      "step": 2470
    },
    {
      "epoch": 0.4214819850441876,
      "grad_norm": 9.658906936645508,
      "learning_rate": 7.177624350905155e-06,
      "loss": 2.0232,
      "step": 2480
    },
    {
      "epoch": 0.4231815091774303,
      "grad_norm": 5.400667667388916,
      "learning_rate": 7.153559522573959e-06,
      "loss": 1.9549,
      "step": 2490
    },
    {
      "epoch": 0.424881033310673,
      "grad_norm": 6.810690402984619,
      "learning_rate": 7.126748835263528e-06,
      "loss": 1.9216,
      "step": 2500
    },
    {
      "epoch": 0.424881033310673,
      "eval_cosine_accuracy@1": 0.706,
      "eval_cosine_accuracy@10": 0.912,
      "eval_cosine_accuracy@3": 0.8375,
      "eval_cosine_accuracy@5": 0.87,
      "eval_cosine_map@100": 0.782229618428024,
      "eval_cosine_mrr@10": 0.7788738095238087,
      "eval_cosine_ndcg@10": 0.811319534720964,
      "eval_cosine_precision@1": 0.706,
      "eval_cosine_precision@10": 0.09120000000000002,
      "eval_cosine_precision@3": 0.2791666666666666,
      "eval_cosine_precision@5": 0.174,
      "eval_cosine_recall@1": 0.706,
      "eval_cosine_recall@10": 0.912,
      "eval_cosine_recall@3": 0.8375,
      "eval_cosine_recall@5": 0.87,
      "eval_loss": 1.6223056316375732,
      "eval_runtime": 2.7895,
      "eval_samples_per_second": 270.299,
      "eval_sequential_score": 0.782229618428024,
      "eval_steps_per_second": 2.151,
      "eval_sts-dev_pearson_cosine": 0.7982554363739512,
      "eval_sts-dev_pearson_dot": 0.7952587919868024,
      "eval_sts-dev_pearson_euclidean": 0.7869234232815028,
      "eval_sts-dev_pearson_manhattan": 0.7872366700437463,
      "eval_sts-dev_pearson_max": 0.7982554363739512,
      "eval_sts-dev_spearman_cosine": 0.8022561650534226,
      "eval_sts-dev_spearman_dot": 0.7985926145853256,
      "eval_sts-dev_spearman_euclidean": 0.8008740888247353,
      "eval_sts-dev_spearman_manhattan": 0.801728465873034,
      "eval_sts-dev_spearman_max": 0.8022561650534226,
      "step": 2500
    },
    {
      "epoch": 0.4265805574439157,
      "grad_norm": 6.482539653778076,
      "learning_rate": 7.099863282277834e-06,
      "loss": 1.5548,
      "step": 2510
    },
    {
      "epoch": 0.4282800815771584,
      "grad_norm": 7.853833198547363,
      "learning_rate": 7.072903810040381e-06,
      "loss": 1.6629,
      "step": 2520
    },
    {
      "epoch": 0.4299796057104011,
      "grad_norm": 6.526813507080078,
      "learning_rate": 7.0458713675767665e-06,
      "loss": 1.9463,
      "step": 2530
    },
    {
      "epoch": 0.43167912984364376,
      "grad_norm": 9.038359642028809,
      "learning_rate": 7.018766906481288e-06,
      "loss": 2.16,
      "step": 2540
    },
    {
      "epoch": 0.43337865397688646,
      "grad_norm": 5.049062728881836,
      "learning_rate": 6.991591380883435e-06,
      "loss": 1.6674,
      "step": 2550
    },
    {
      "epoch": 0.43507817811012917,
      "grad_norm": 6.659489154815674,
      "learning_rate": 6.964345747414307e-06,
      "loss": 1.7055,
      "step": 2560
    },
    {
      "epoch": 0.43677770224337187,
      "grad_norm": 8.091437339782715,
      "learning_rate": 6.937030965172935e-06,
      "loss": 1.8062,
      "step": 2570
    },
    {
      "epoch": 0.43847722637661457,
      "grad_norm": 7.644534587860107,
      "learning_rate": 6.909647995692522e-06,
      "loss": 2.1972,
      "step": 2580
    },
    {
      "epoch": 0.4401767505098572,
      "grad_norm": 7.349899768829346,
      "learning_rate": 6.88219780290659e-06,
      "loss": 1.9069,
      "step": 2590
    },
    {
      "epoch": 0.4418762746430999,
      "grad_norm": 8.312362670898438,
      "learning_rate": 6.854681353115056e-06,
      "loss": 2.6629,
      "step": 2600
    },
    {
      "epoch": 0.4435757987763426,
      "grad_norm": 8.937472343444824,
      "learning_rate": 6.82709961495021e-06,
      "loss": 2.4573,
      "step": 2610
    },
    {
      "epoch": 0.4452753229095853,
      "grad_norm": 6.268832683563232,
      "learning_rate": 6.799453559342619e-06,
      "loss": 2.0894,
      "step": 2620
    },
    {
      "epoch": 0.446974847042828,
      "grad_norm": 9.547843933105469,
      "learning_rate": 6.771744159486947e-06,
      "loss": 1.5755,
      "step": 2630
    },
    {
      "epoch": 0.4486743711760707,
      "grad_norm": 8.299346923828125,
      "learning_rate": 6.7439723908077e-06,
      "loss": 2.3673,
      "step": 2640
    },
    {
      "epoch": 0.4503738953093134,
      "grad_norm": 6.921226978302002,
      "learning_rate": 6.716139230924891e-06,
      "loss": 1.8997,
      "step": 2650
    },
    {
      "epoch": 0.4520734194425561,
      "grad_norm": 6.874094486236572,
      "learning_rate": 6.6882456596196125e-06,
      "loss": 2.0955,
      "step": 2660
    },
    {
      "epoch": 0.4537729435757988,
      "grad_norm": 7.336355686187744,
      "learning_rate": 6.660292658799565e-06,
      "loss": 2.0553,
      "step": 2670
    },
    {
      "epoch": 0.4554724677090415,
      "grad_norm": 6.311689376831055,
      "learning_rate": 6.632281212464479e-06,
      "loss": 1.9258,
      "step": 2680
    },
    {
      "epoch": 0.4571719918422842,
      "grad_norm": 5.280473232269287,
      "learning_rate": 6.60421230667148e-06,
      "loss": 1.7009,
      "step": 2690
    },
    {
      "epoch": 0.45887151597552683,
      "grad_norm": 5.1102986335754395,
      "learning_rate": 6.5760869295003796e-06,
      "loss": 1.9853,
      "step": 2700
    },
    {
      "epoch": 0.46057104010876954,
      "grad_norm": 6.345133304595947,
      "learning_rate": 6.54790607101889e-06,
      "loss": 1.8553,
      "step": 2710
    },
    {
      "epoch": 0.46227056424201224,
      "grad_norm": 9.678619384765625,
      "learning_rate": 6.519670723247773e-06,
      "loss": 1.7589,
      "step": 2720
    },
    {
      "epoch": 0.46397008837525494,
      "grad_norm": 4.833879470825195,
      "learning_rate": 6.491381880125916e-06,
      "loss": 1.827,
      "step": 2730
    },
    {
      "epoch": 0.46566961250849764,
      "grad_norm": 6.266608238220215,
      "learning_rate": 6.463040537475354e-06,
      "loss": 1.7476,
      "step": 2740
    },
    {
      "epoch": 0.4673691366417403,
      "grad_norm": 8.750253677368164,
      "learning_rate": 6.434647692966199e-06,
      "loss": 2.1204,
      "step": 2750
    },
    {
      "epoch": 0.469068660774983,
      "grad_norm": 5.197884559631348,
      "learning_rate": 6.406204346081531e-06,
      "loss": 1.5477,
      "step": 2760
    },
    {
      "epoch": 0.4707681849082257,
      "grad_norm": 7.584977626800537,
      "learning_rate": 6.377711498082213e-06,
      "loss": 1.9078,
      "step": 2770
    },
    {
      "epoch": 0.4724677090414684,
      "grad_norm": 5.226884365081787,
      "learning_rate": 6.349170151971641e-06,
      "loss": 1.53,
      "step": 2780
    },
    {
      "epoch": 0.4741672331747111,
      "grad_norm": 9.499238967895508,
      "learning_rate": 6.320581312460439e-06,
      "loss": 1.7868,
      "step": 2790
    },
    {
      "epoch": 0.47586675730795375,
      "grad_norm": 6.44546365737915,
      "learning_rate": 6.291945985931093e-06,
      "loss": 1.7781,
      "step": 2800
    },
    {
      "epoch": 0.47756628144119645,
      "grad_norm": 5.152192115783691,
      "learning_rate": 6.263265180402517e-06,
      "loss": 2.0043,
      "step": 2810
    },
    {
      "epoch": 0.47926580557443915,
      "grad_norm": 7.129688262939453,
      "learning_rate": 6.234539905494576e-06,
      "loss": 1.9584,
      "step": 2820
    },
    {
      "epoch": 0.48096532970768185,
      "grad_norm": 6.289784908294678,
      "learning_rate": 6.2057711723925455e-06,
      "loss": 1.9584,
      "step": 2830
    },
    {
      "epoch": 0.48266485384092456,
      "grad_norm": 7.36097526550293,
      "learning_rate": 6.176959993811511e-06,
      "loss": 1.9952,
      "step": 2840
    },
    {
      "epoch": 0.48436437797416726,
      "grad_norm": 4.622279644012451,
      "learning_rate": 6.148107383960723e-06,
      "loss": 1.9764,
      "step": 2850
    },
    {
      "epoch": 0.4860639021074099,
      "grad_norm": 8.651782989501953,
      "learning_rate": 6.119214358507889e-06,
      "loss": 1.8804,
      "step": 2860
    },
    {
      "epoch": 0.4877634262406526,
      "grad_norm": 6.295717716217041,
      "learning_rate": 6.090281934543429e-06,
      "loss": 1.8634,
      "step": 2870
    },
    {
      "epoch": 0.4894629503738953,
      "grad_norm": 5.505663871765137,
      "learning_rate": 6.061311130544664e-06,
      "loss": 1.8853,
      "step": 2880
    },
    {
      "epoch": 0.491162474507138,
      "grad_norm": 10.88358211517334,
      "learning_rate": 6.032302966339965e-06,
      "loss": 2.2094,
      "step": 2890
    },
    {
      "epoch": 0.4928619986403807,
      "grad_norm": 6.576531410217285,
      "learning_rate": 6.003258463072859e-06,
      "loss": 1.5955,
      "step": 2900
    },
    {
      "epoch": 0.49456152277362336,
      "grad_norm": 5.74630880355835,
      "learning_rate": 5.974178643166075e-06,
      "loss": 1.8756,
      "step": 2910
    },
    {
      "epoch": 0.49626104690686607,
      "grad_norm": 8.021421432495117,
      "learning_rate": 5.945064530285556e-06,
      "loss": 1.9114,
      "step": 2920
    },
    {
      "epoch": 0.49796057104010877,
      "grad_norm": 4.470407009124756,
      "learning_rate": 5.915917149304424e-06,
      "loss": 2.2359,
      "step": 2930
    },
    {
      "epoch": 0.49966009517335147,
      "grad_norm": 8.264505386352539,
      "learning_rate": 5.8867375262669076e-06,
      "loss": 1.7543,
      "step": 2940
    },
    {
      "epoch": 0.5013596193065941,
      "grad_norm": 6.8826069831848145,
      "learning_rate": 5.857526688352211e-06,
      "loss": 1.7754,
      "step": 2950
    },
    {
      "epoch": 0.5030591434398368,
      "grad_norm": 7.148299694061279,
      "learning_rate": 5.828285663838365e-06,
      "loss": 2.1194,
      "step": 2960
    },
    {
      "epoch": 0.5047586675730795,
      "grad_norm": 6.472208499908447,
      "learning_rate": 5.7990154820660295e-06,
      "loss": 1.7127,
      "step": 2970
    },
    {
      "epoch": 0.5064581917063222,
      "grad_norm": 8.591309547424316,
      "learning_rate": 5.769717173402253e-06,
      "loss": 1.6715,
      "step": 2980
    },
    {
      "epoch": 0.5081577158395649,
      "grad_norm": 5.60316801071167,
      "learning_rate": 5.740391769204209e-06,
      "loss": 1.4632,
      "step": 2990
    },
    {
      "epoch": 0.5098572399728076,
      "grad_norm": 5.79714822769165,
      "learning_rate": 5.71104030178288e-06,
      "loss": 1.7171,
      "step": 3000
    },
    {
      "epoch": 0.5098572399728076,
      "eval_cosine_accuracy@1": 0.707,
      "eval_cosine_accuracy@10": 0.915,
      "eval_cosine_accuracy@3": 0.839,
      "eval_cosine_accuracy@5": 0.874,
      "eval_cosine_map@100": 0.7842834572194356,
      "eval_cosine_mrr@10": 0.7809134920634915,
      "eval_cosine_ndcg@10": 0.8136383995155022,
      "eval_cosine_precision@1": 0.707,
      "eval_cosine_precision@10": 0.0915,
      "eval_cosine_precision@3": 0.2796666666666666,
      "eval_cosine_precision@5": 0.1748,
      "eval_cosine_recall@1": 0.707,
      "eval_cosine_recall@10": 0.915,
      "eval_cosine_recall@3": 0.839,
      "eval_cosine_recall@5": 0.874,
      "eval_loss": 1.6132309436798096,
      "eval_runtime": 2.7954,
      "eval_samples_per_second": 269.727,
      "eval_sequential_score": 0.7842834572194356,
      "eval_steps_per_second": 2.146,
      "eval_sts-dev_pearson_cosine": 0.8056038655154354,
      "eval_sts-dev_pearson_dot": 0.8019913183334593,
      "eval_sts-dev_pearson_euclidean": 0.7922504417572196,
      "eval_sts-dev_pearson_manhattan": 0.7937333279050955,
      "eval_sts-dev_pearson_max": 0.8056038655154354,
      "eval_sts-dev_spearman_cosine": 0.8083570013976158,
      "eval_sts-dev_spearman_dot": 0.8047381961082496,
      "eval_sts-dev_spearman_euclidean": 0.8071350345841625,
      "eval_sts-dev_spearman_manhattan": 0.8069596988802814,
      "eval_sts-dev_spearman_max": 0.8083570013976158,
      "step": 3000
    },
    {
      "epoch": 0.5115567641060503,
      "grad_norm": 6.9394049644470215,
      "learning_rate": 5.6816638043667294e-06,
      "loss": 1.4526,
      "step": 3010
    },
    {
      "epoch": 0.513256288239293,
      "grad_norm": 10.348281860351562,
      "learning_rate": 5.652263311065325e-06,
      "loss": 1.8309,
      "step": 3020
    },
    {
      "epoch": 0.5149558123725357,
      "grad_norm": 8.404902458190918,
      "learning_rate": 5.6228398568329344e-06,
      "loss": 1.9336,
      "step": 3030
    },
    {
      "epoch": 0.5166553365057783,
      "grad_norm": 6.444024562835693,
      "learning_rate": 5.593394477432096e-06,
      "loss": 1.7704,
      "step": 3040
    },
    {
      "epoch": 0.518354860639021,
      "grad_norm": 9.629279136657715,
      "learning_rate": 5.563928209397151e-06,
      "loss": 1.943,
      "step": 3050
    },
    {
      "epoch": 0.5200543847722637,
      "grad_norm": 6.889398097991943,
      "learning_rate": 5.534442089997771e-06,
      "loss": 1.8689,
      "step": 3060
    },
    {
      "epoch": 0.5217539089055064,
      "grad_norm": 7.757649898529053,
      "learning_rate": 5.5049371572024235e-06,
      "loss": 1.8908,
      "step": 3070
    },
    {
      "epoch": 0.5234534330387491,
      "grad_norm": 6.982544422149658,
      "learning_rate": 5.475414449641852e-06,
      "loss": 1.7924,
      "step": 3080
    },
    {
      "epoch": 0.5251529571719918,
      "grad_norm": 7.109350204467773,
      "learning_rate": 5.445875006572502e-06,
      "loss": 1.7891,
      "step": 3090
    },
    {
      "epoch": 0.5268524813052345,
      "grad_norm": 7.507669448852539,
      "learning_rate": 5.416319867839943e-06,
      "loss": 2.0208,
      "step": 3100
    },
    {
      "epoch": 0.5285520054384772,
      "grad_norm": 4.527754783630371,
      "learning_rate": 5.386750073842259e-06,
      "loss": 1.5862,
      "step": 3110
    },
    {
      "epoch": 0.53025152957172,
      "grad_norm": 8.16139030456543,
      "learning_rate": 5.357166665493434e-06,
      "loss": 2.2195,
      "step": 3120
    },
    {
      "epoch": 0.5319510537049627,
      "grad_norm": 7.385548114776611,
      "learning_rate": 5.327570684186697e-06,
      "loss": 1.6103,
      "step": 3130
    },
    {
      "epoch": 0.5336505778382054,
      "grad_norm": 7.784885406494141,
      "learning_rate": 5.297963171757874e-06,
      "loss": 2.5839,
      "step": 3140
    },
    {
      "epoch": 0.535350101971448,
      "grad_norm": 7.986569404602051,
      "learning_rate": 5.268345170448704e-06,
      "loss": 2.4304,
      "step": 3150
    },
    {
      "epoch": 0.5370496261046906,
      "grad_norm": 7.29852294921875,
      "learning_rate": 5.23871772287016e-06,
      "loss": 1.9009,
      "step": 3160
    },
    {
      "epoch": 0.5387491502379333,
      "grad_norm": 8.884086608886719,
      "learning_rate": 5.209081871965741e-06,
      "loss": 2.2161,
      "step": 3170
    },
    {
      "epoch": 0.540448674371176,
      "grad_norm": 6.930916786193848,
      "learning_rate": 5.179438660974756e-06,
      "loss": 1.9054,
      "step": 3180
    },
    {
      "epoch": 0.5421481985044188,
      "grad_norm": 6.724717617034912,
      "learning_rate": 5.149789133395606e-06,
      "loss": 1.3603,
      "step": 3190
    },
    {
      "epoch": 0.5438477226376615,
      "grad_norm": 11.375778198242188,
      "learning_rate": 5.120134332949049e-06,
      "loss": 2.1651,
      "step": 3200
    },
    {
      "epoch": 0.5455472467709042,
      "grad_norm": 9.330193519592285,
      "learning_rate": 5.0904753035414545e-06,
      "loss": 2.0636,
      "step": 3210
    },
    {
      "epoch": 0.5472467709041469,
      "grad_norm": 7.2718000411987305,
      "learning_rate": 5.060813089228064e-06,
      "loss": 2.0036,
      "step": 3220
    },
    {
      "epoch": 0.5489462950373896,
      "grad_norm": 5.489546775817871,
      "learning_rate": 5.0311487341762285e-06,
      "loss": 1.7724,
      "step": 3230
    },
    {
      "epoch": 0.5506458191706323,
      "grad_norm": 7.590505123138428,
      "learning_rate": 5.001483282628664e-06,
      "loss": 1.2798,
      "step": 3240
    },
    {
      "epoch": 0.552345343303875,
      "grad_norm": 5.193188190460205,
      "learning_rate": 4.971817778866681e-06,
      "loss": 1.8975,
      "step": 3250
    },
    {
      "epoch": 0.5540448674371176,
      "grad_norm": 4.518863677978516,
      "learning_rate": 4.9421532671734265e-06,
      "loss": 1.8153,
      "step": 3260
    },
    {
      "epoch": 0.5557443915703603,
      "grad_norm": 7.619812488555908,
      "learning_rate": 4.912490791797129e-06,
      "loss": 1.6361,
      "step": 3270
    },
    {
      "epoch": 0.557443915703603,
      "grad_norm": 8.381867408752441,
      "learning_rate": 4.882831396914332e-06,
      "loss": 2.2208,
      "step": 3280
    },
    {
      "epoch": 0.5591434398368457,
      "grad_norm": 4.878629684448242,
      "learning_rate": 4.853176126593139e-06,
      "loss": 1.7356,
      "step": 3290
    },
    {
      "epoch": 0.5608429639700884,
      "grad_norm": 6.435664653778076,
      "learning_rate": 4.823526024756464e-06,
      "loss": 1.8563,
      "step": 3300
    },
    {
      "epoch": 0.5625424881033311,
      "grad_norm": 6.569293022155762,
      "learning_rate": 4.793882135145278e-06,
      "loss": 1.8743,
      "step": 3310
    },
    {
      "epoch": 0.5642420122365738,
      "grad_norm": 6.728095054626465,
      "learning_rate": 4.764245501281869e-06,
      "loss": 1.7494,
      "step": 3320
    },
    {
      "epoch": 0.5659415363698165,
      "grad_norm": 6.320137977600098,
      "learning_rate": 4.734617166433113e-06,
      "loss": 1.5363,
      "step": 3330
    },
    {
      "epoch": 0.5676410605030592,
      "grad_norm": 5.589328765869141,
      "learning_rate": 4.7049981735737385e-06,
      "loss": 2.0732,
      "step": 3340
    },
    {
      "epoch": 0.5693405846363019,
      "grad_norm": 6.616926193237305,
      "learning_rate": 4.67538956534962e-06,
      "loss": 1.7303,
      "step": 3350
    },
    {
      "epoch": 0.5710401087695445,
      "grad_norm": 4.177309036254883,
      "learning_rate": 4.645792384041078e-06,
      "loss": 1.532,
      "step": 3360
    },
    {
      "epoch": 0.5727396329027872,
      "grad_norm": 6.281280517578125,
      "learning_rate": 4.616207671526175e-06,
      "loss": 2.2504,
      "step": 3370
    },
    {
      "epoch": 0.5744391570360299,
      "grad_norm": 6.781736373901367,
      "learning_rate": 4.586636469244054e-06,
      "loss": 1.6377,
      "step": 3380
    },
    {
      "epoch": 0.5761386811692726,
      "grad_norm": 8.028448104858398,
      "learning_rate": 4.557079818158269e-06,
      "loss": 2.1332,
      "step": 3390
    },
    {
      "epoch": 0.5778382053025153,
      "grad_norm": 5.485851764678955,
      "learning_rate": 4.527538758720144e-06,
      "loss": 1.8658,
      "step": 3400
    },
    {
      "epoch": 0.579537729435758,
      "grad_norm": 5.404677867889404,
      "learning_rate": 4.498014330832148e-06,
      "loss": 1.8165,
      "step": 3410
    },
    {
      "epoch": 0.5812372535690007,
      "grad_norm": 5.938612461090088,
      "learning_rate": 4.468507573811285e-06,
      "loss": 1.7847,
      "step": 3420
    },
    {
      "epoch": 0.5829367777022434,
      "grad_norm": 6.639806270599365,
      "learning_rate": 4.439019526352511e-06,
      "loss": 1.747,
      "step": 3430
    },
    {
      "epoch": 0.5846363018354861,
      "grad_norm": 10.548473358154297,
      "learning_rate": 4.409551226492173e-06,
      "loss": 1.8863,
      "step": 3440
    },
    {
      "epoch": 0.5863358259687288,
      "grad_norm": 6.586738109588623,
      "learning_rate": 4.380103711571459e-06,
      "loss": 2.0465,
      "step": 3450
    },
    {
      "epoch": 0.5880353501019715,
      "grad_norm": 6.210246562957764,
      "learning_rate": 4.350678018199887e-06,
      "loss": 1.6285,
      "step": 3460
    },
    {
      "epoch": 0.5897348742352141,
      "grad_norm": 8.666604995727539,
      "learning_rate": 4.321275182218821e-06,
      "loss": 2.278,
      "step": 3470
    },
    {
      "epoch": 0.5914343983684568,
      "grad_norm": 5.970696926116943,
      "learning_rate": 4.291896238664991e-06,
      "loss": 1.7391,
      "step": 3480
    },
    {
      "epoch": 0.5931339225016995,
      "grad_norm": 6.197153091430664,
      "learning_rate": 4.262542221734077e-06,
      "loss": 1.8116,
      "step": 3490
    },
    {
      "epoch": 0.5948334466349422,
      "grad_norm": 5.297378063201904,
      "learning_rate": 4.233214164744286e-06,
      "loss": 1.5104,
      "step": 3500
    },
    {
      "epoch": 0.5948334466349422,
      "eval_cosine_accuracy@1": 0.7,
      "eval_cosine_accuracy@10": 0.9105,
      "eval_cosine_accuracy@3": 0.8315,
      "eval_cosine_accuracy@5": 0.871,
      "eval_cosine_map@100": 0.7789008444240532,
      "eval_cosine_mrr@10": 0.775367857142857,
      "eval_cosine_ndcg@10": 0.8083837978692977,
      "eval_cosine_precision@1": 0.7,
      "eval_cosine_precision@10": 0.09105,
      "eval_cosine_precision@3": 0.2771666666666666,
      "eval_cosine_precision@5": 0.17420000000000005,
      "eval_cosine_recall@1": 0.7,
      "eval_cosine_recall@10": 0.9105,
      "eval_cosine_recall@3": 0.8315,
      "eval_cosine_recall@5": 0.871,
      "eval_loss": 1.5698647499084473,
      "eval_runtime": 2.7632,
      "eval_samples_per_second": 272.875,
      "eval_sequential_score": 0.7789008444240532,
      "eval_steps_per_second": 2.171,
      "eval_sts-dev_pearson_cosine": 0.8017937499678631,
      "eval_sts-dev_pearson_dot": 0.7968971369444564,
      "eval_sts-dev_pearson_euclidean": 0.7889399312152248,
      "eval_sts-dev_pearson_manhattan": 0.7898347176162472,
      "eval_sts-dev_pearson_max": 0.8017937499678631,
      "eval_sts-dev_spearman_cosine": 0.8044304044501694,
      "eval_sts-dev_spearman_dot": 0.8013449184419915,
      "eval_sts-dev_spearman_euclidean": 0.8042950575910329,
      "eval_sts-dev_spearman_manhattan": 0.8042523771894302,
      "eval_sts-dev_spearman_max": 0.8044304044501694,
      "step": 3500
    },
    {
      "epoch": 0.5965329707681849,
      "grad_norm": 5.6484575271606445,
      "learning_rate": 4.2039131000999865e-06,
      "loss": 1.8306,
      "step": 3510
    },
    {
      "epoch": 0.5982324949014276,
      "grad_norm": 7.109194278717041,
      "learning_rate": 4.174640059255369e-06,
      "loss": 1.6575,
      "step": 3520
    },
    {
      "epoch": 0.5999320190346703,
      "grad_norm": 8.67902660369873,
      "learning_rate": 4.145396072678126e-06,
      "loss": 1.56,
      "step": 3530
    },
    {
      "epoch": 0.601631543167913,
      "grad_norm": 5.615638732910156,
      "learning_rate": 4.116182169813185e-06,
      "loss": 1.8674,
      "step": 3540
    },
    {
      "epoch": 0.6033310673011557,
      "grad_norm": 6.5344085693359375,
      "learning_rate": 4.086999379046472e-06,
      "loss": 1.4909,
      "step": 3550
    },
    {
      "epoch": 0.6050305914343984,
      "grad_norm": 7.410928726196289,
      "learning_rate": 4.057848727668701e-06,
      "loss": 1.6653,
      "step": 3560
    },
    {
      "epoch": 0.6067301155676411,
      "grad_norm": 5.162845134735107,
      "learning_rate": 4.028731241839224e-06,
      "loss": 2.2836,
      "step": 3570
    },
    {
      "epoch": 0.6084296397008837,
      "grad_norm": 9.384202003479004,
      "learning_rate": 3.999647946549898e-06,
      "loss": 2.1572,
      "step": 3580
    },
    {
      "epoch": 0.6101291638341264,
      "grad_norm": 9.560714721679688,
      "learning_rate": 3.970599865589006e-06,
      "loss": 1.6969,
      "step": 3590
    },
    {
      "epoch": 0.6118286879673691,
      "grad_norm": 8.857442855834961,
      "learning_rate": 3.94158802150522e-06,
      "loss": 2.0679,
      "step": 3600
    },
    {
      "epoch": 0.6135282121006118,
      "grad_norm": 4.5523200035095215,
      "learning_rate": 3.912613435571603e-06,
      "loss": 1.8555,
      "step": 3610
    },
    {
      "epoch": 0.6152277362338545,
      "grad_norm": 4.791132926940918,
      "learning_rate": 3.8836771277496575e-06,
      "loss": 1.8506,
      "step": 3620
    },
    {
      "epoch": 0.6169272603670972,
      "grad_norm": 5.0394673347473145,
      "learning_rate": 3.854780116653429e-06,
      "loss": 1.4441,
      "step": 3630
    },
    {
      "epoch": 0.6186267845003399,
      "grad_norm": 8.9247465133667,
      "learning_rate": 3.8259234195136344e-06,
      "loss": 1.6772,
      "step": 3640
    },
    {
      "epoch": 0.6203263086335826,
      "grad_norm": 7.963625907897949,
      "learning_rate": 3.7971080521418636e-06,
      "loss": 1.6228,
      "step": 3650
    },
    {
      "epoch": 0.6220258327668253,
      "grad_norm": 5.572691917419434,
      "learning_rate": 3.7683350288948204e-06,
      "loss": 1.7809,
      "step": 3660
    },
    {
      "epoch": 0.623725356900068,
      "grad_norm": 4.029000282287598,
      "learning_rate": 3.7396053626386115e-06,
      "loss": 1.5777,
      "step": 3670
    },
    {
      "epoch": 0.6254248810333106,
      "grad_norm": 5.965753555297852,
      "learning_rate": 3.7109200647130983e-06,
      "loss": 2.0558,
      "step": 3680
    },
    {
      "epoch": 0.6271244051665533,
      "grad_norm": 5.478359222412109,
      "learning_rate": 3.6822801448962835e-06,
      "loss": 1.9044,
      "step": 3690
    },
    {
      "epoch": 0.628823929299796,
      "grad_norm": 6.156702518463135,
      "learning_rate": 3.6536866113687754e-06,
      "loss": 1.9202,
      "step": 3700
    },
    {
      "epoch": 0.6305234534330387,
      "grad_norm": 5.119917392730713,
      "learning_rate": 3.6251404706783005e-06,
      "loss": 1.4821,
      "step": 3710
    },
    {
      "epoch": 0.6322229775662814,
      "grad_norm": 5.908693790435791,
      "learning_rate": 3.5966427277042593e-06,
      "loss": 1.5492,
      "step": 3720
    },
    {
      "epoch": 0.6339225016995241,
      "grad_norm": 6.6029181480407715,
      "learning_rate": 3.568194385622359e-06,
      "loss": 1.7337,
      "step": 3730
    },
    {
      "epoch": 0.6356220258327668,
      "grad_norm": 6.767298221588135,
      "learning_rate": 3.5397964458693045e-06,
      "loss": 2.0108,
      "step": 3740
    },
    {
      "epoch": 0.6373215499660095,
      "grad_norm": 7.079708576202393,
      "learning_rate": 3.5114499081075386e-06,
      "loss": 1.7365,
      "step": 3750
    },
    {
      "epoch": 0.6390210740992522,
      "grad_norm": 5.657531261444092,
      "learning_rate": 3.4831557701900507e-06,
      "loss": 1.5291,
      "step": 3760
    },
    {
      "epoch": 0.6407205982324949,
      "grad_norm": 9.403407096862793,
      "learning_rate": 3.4549150281252635e-06,
      "loss": 1.8147,
      "step": 3770
    },
    {
      "epoch": 0.6424201223657376,
      "grad_norm": 4.905513763427734,
      "learning_rate": 3.4267286760419527e-06,
      "loss": 1.6704,
      "step": 3780
    },
    {
      "epoch": 0.6441196464989802,
      "grad_norm": 5.188987731933594,
      "learning_rate": 3.398597706154268e-06,
      "loss": 1.727,
      "step": 3790
    },
    {
      "epoch": 0.6458191706322229,
      "grad_norm": 9.689862251281738,
      "learning_rate": 3.3705231087267955e-06,
      "loss": 1.741,
      "step": 3800
    },
    {
      "epoch": 0.6475186947654656,
      "grad_norm": 5.145129203796387,
      "learning_rate": 3.3425058720396986e-06,
      "loss": 1.6929,
      "step": 3810
    },
    {
      "epoch": 0.6492182188987083,
      "grad_norm": 5.635613918304443,
      "learning_rate": 3.314546982353941e-06,
      "loss": 1.7855,
      "step": 3820
    },
    {
      "epoch": 0.650917743031951,
      "grad_norm": 5.333996295928955,
      "learning_rate": 3.2866474238765522e-06,
      "loss": 1.4379,
      "step": 3830
    },
    {
      "epoch": 0.6526172671651937,
      "grad_norm": 7.929337501525879,
      "learning_rate": 3.258808178725984e-06,
      "loss": 1.8255,
      "step": 3840
    },
    {
      "epoch": 0.6543167912984365,
      "grad_norm": 7.606388092041016,
      "learning_rate": 3.2310302268975545e-06,
      "loss": 1.8072,
      "step": 3850
    },
    {
      "epoch": 0.6560163154316792,
      "grad_norm": 5.5722126960754395,
      "learning_rate": 3.2033145462289284e-06,
      "loss": 1.4806,
      "step": 3860
    },
    {
      "epoch": 0.6577158395649219,
      "grad_norm": 4.687691688537598,
      "learning_rate": 3.175662112365707e-06,
      "loss": 1.7177,
      "step": 3870
    },
    {
      "epoch": 0.6594153636981646,
      "grad_norm": 5.950764179229736,
      "learning_rate": 3.1480738987270854e-06,
      "loss": 1.3493,
      "step": 3880
    },
    {
      "epoch": 0.6611148878314073,
      "grad_norm": 6.248993396759033,
      "learning_rate": 3.120550876471576e-06,
      "loss": 1.7455,
      "step": 3890
    },
    {
      "epoch": 0.6628144119646499,
      "grad_norm": 6.485055446624756,
      "learning_rate": 3.0930940144628385e-06,
      "loss": 1.5822,
      "step": 3900
    },
    {
      "epoch": 0.6645139360978926,
      "grad_norm": 6.7886881828308105,
      "learning_rate": 3.0657042792355528e-06,
      "loss": 2.4897,
      "step": 3910
    },
    {
      "epoch": 0.6662134602311353,
      "grad_norm": 4.51090669631958,
      "learning_rate": 3.0383826349614115e-06,
      "loss": 1.8648,
      "step": 3920
    },
    {
      "epoch": 0.667912984364378,
      "grad_norm": 7.129496097564697,
      "learning_rate": 3.0111300434151746e-06,
      "loss": 1.8095,
      "step": 3930
    },
    {
      "epoch": 0.6696125084976207,
      "grad_norm": 8.079377174377441,
      "learning_rate": 2.9839474639408108e-06,
      "loss": 2.0793,
      "step": 3940
    },
    {
      "epoch": 0.6713120326308634,
      "grad_norm": 6.9486284255981445,
      "learning_rate": 2.956835853417726e-06,
      "loss": 1.9273,
      "step": 3950
    },
    {
      "epoch": 0.6730115567641061,
      "grad_norm": 6.231419563293457,
      "learning_rate": 2.9297961662270893e-06,
      "loss": 1.6705,
      "step": 3960
    },
    {
      "epoch": 0.6747110808973488,
      "grad_norm": 9.692667007446289,
      "learning_rate": 2.902829354218219e-06,
      "loss": 1.8497,
      "step": 3970
    },
    {
      "epoch": 0.6764106050305915,
      "grad_norm": 6.941633701324463,
      "learning_rate": 2.875936366675096e-06,
      "loss": 2.1609,
      "step": 3980
    },
    {
      "epoch": 0.6781101291638342,
      "grad_norm": 4.512592315673828,
      "learning_rate": 2.8491181502829335e-06,
      "loss": 1.719,
      "step": 3990
    },
    {
      "epoch": 0.6798096532970768,
      "grad_norm": 13.518653869628906,
      "learning_rate": 2.822375649094846e-06,
      "loss": 1.7953,
      "step": 4000
    },
    {
      "epoch": 0.6798096532970768,
      "eval_cosine_accuracy@1": 0.7145,
      "eval_cosine_accuracy@10": 0.914,
      "eval_cosine_accuracy@3": 0.839,
      "eval_cosine_accuracy@5": 0.8745,
      "eval_cosine_map@100": 0.7881479093038395,
      "eval_cosine_mrr@10": 0.7848144841269835,
      "eval_cosine_ndcg@10": 0.8163072349117797,
      "eval_cosine_precision@1": 0.7145,
      "eval_cosine_precision@10": 0.09140000000000001,
      "eval_cosine_precision@3": 0.2796666666666666,
      "eval_cosine_precision@5": 0.1749,
      "eval_cosine_recall@1": 0.7145,
      "eval_cosine_recall@10": 0.914,
      "eval_cosine_recall@3": 0.839,
      "eval_cosine_recall@5": 0.8745,
      "eval_loss": 1.5573265552520752,
      "eval_runtime": 2.7757,
      "eval_samples_per_second": 271.645,
      "eval_sequential_score": 0.7881479093038395,
      "eval_steps_per_second": 2.162,
      "eval_sts-dev_pearson_cosine": 0.8053811920422538,
      "eval_sts-dev_pearson_dot": 0.8004660633078914,
      "eval_sts-dev_pearson_euclidean": 0.7928635065778744,
      "eval_sts-dev_pearson_manhattan": 0.7941416034985971,
      "eval_sts-dev_pearson_max": 0.8053811920422538,
      "eval_sts-dev_spearman_cosine": 0.8086109320148909,
      "eval_sts-dev_spearman_dot": 0.804234875942229,
      "eval_sts-dev_spearman_euclidean": 0.8084185226972233,
      "eval_sts-dev_spearman_manhattan": 0.8088695507249705,
      "eval_sts-dev_spearman_max": 0.8088695507249705,
      "step": 4000
    }
  ],
  "logging_steps": 10,
  "max_steps": 5884,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": null
}