{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.6040931545518702,
  "eval_steps": 27,
  "global_step": 107,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.016937191249117856,
      "grad_norm": 34.22002029418945,
      "learning_rate": 6.818181818181818e-07,
      "loss": 7.2372,
      "step": 3
    },
    {
      "epoch": 0.03387438249823571,
      "grad_norm": 21.76839828491211,
      "learning_rate": 1.3636363636363636e-06,
      "loss": 6.855,
      "step": 6
    },
    {
      "epoch": 0.05081157374735357,
      "grad_norm": 21.260774612426758,
      "learning_rate": 2.0454545454545457e-06,
      "loss": 7.4707,
      "step": 9
    },
    {
      "epoch": 0.06774876499647142,
      "grad_norm": 16.885921478271484,
      "learning_rate": 2.7272727272727272e-06,
      "loss": 7.0187,
      "step": 12
    },
    {
      "epoch": 0.08468595624558928,
      "grad_norm": 19.509899139404297,
      "learning_rate": 3.409090909090909e-06,
      "loss": 6.6756,
      "step": 15
    },
    {
      "epoch": 0.10162314749470713,
      "grad_norm": 7.9427289962768555,
      "learning_rate": 4.0909090909090915e-06,
      "loss": 6.0155,
      "step": 18
    },
    {
      "epoch": 0.11856033874382499,
      "grad_norm": 7.325345039367676,
      "learning_rate": 4.772727272727273e-06,
      "loss": 6.1644,
      "step": 21
    },
    {
      "epoch": 0.13549752999294284,
      "grad_norm": 7.544689655303955,
      "learning_rate": 5.4545454545454545e-06,
      "loss": 6.2158,
      "step": 24
    },
    {
      "epoch": 0.1524347212420607,
      "grad_norm": 5.141758918762207,
      "learning_rate": 6.136363636363637e-06,
      "loss": 6.1369,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_NLI-v2_cosine_accuracy": 1.0,
      "eval_NLI-v2_dot_accuracy": 0.109375,
      "eval_NLI-v2_euclidean_accuracy": 1.0,
      "eval_NLI-v2_manhattan_accuracy": 1.0,
      "eval_NLI-v2_max_accuracy": 1.0,
      "eval_VitaminC_cosine_accuracy": 0.5546875,
      "eval_VitaminC_cosine_accuracy_threshold": 0.9544724822044373,
      "eval_VitaminC_cosine_ap": 0.5356492030729136,
      "eval_VitaminC_cosine_f1": 0.6542553191489362,
      "eval_VitaminC_cosine_f1_threshold": 0.7148199081420898,
      "eval_VitaminC_cosine_precision": 0.48616600790513836,
      "eval_VitaminC_cosine_recall": 1.0,
      "eval_VitaminC_dot_accuracy": 0.55078125,
      "eval_VitaminC_dot_accuracy_threshold": 414.4264831542969,
      "eval_VitaminC_dot_ap": 0.5108219546857565,
      "eval_VitaminC_dot_f1": 0.6507936507936508,
      "eval_VitaminC_dot_f1_threshold": 271.6522521972656,
      "eval_VitaminC_dot_precision": 0.4823529411764706,
      "eval_VitaminC_dot_recall": 1.0,
      "eval_VitaminC_euclidean_accuracy": 0.55078125,
      "eval_VitaminC_euclidean_accuracy_threshold": 6.519885063171387,
      "eval_VitaminC_euclidean_ap": 0.5226419655984281,
      "eval_VitaminC_euclidean_f1": 0.6505376344086021,
      "eval_VitaminC_euclidean_f1_threshold": 15.194067001342773,
      "eval_VitaminC_euclidean_precision": 0.4859437751004016,
      "eval_VitaminC_euclidean_recall": 0.983739837398374,
      "eval_VitaminC_manhattan_accuracy": 0.546875,
      "eval_VitaminC_manhattan_accuracy_threshold": 149.20114135742188,
      "eval_VitaminC_manhattan_ap": 0.5237451656134715,
      "eval_VitaminC_manhattan_f1": 0.6542553191489362,
      "eval_VitaminC_manhattan_f1_threshold": 259.007080078125,
      "eval_VitaminC_manhattan_precision": 0.48616600790513836,
      "eval_VitaminC_manhattan_recall": 1.0,
      "eval_VitaminC_max_accuracy": 0.5546875,
      "eval_VitaminC_max_accuracy_threshold": 414.4264831542969,
      "eval_VitaminC_max_ap": 0.5356492030729136,
      "eval_VitaminC_max_f1": 0.6542553191489362,
      "eval_VitaminC_max_f1_threshold": 271.6522521972656,
      "eval_VitaminC_max_precision": 0.48616600790513836,
      "eval_VitaminC_max_recall": 1.0,
      "eval_sequential_score": 0.5356492030729136,
      "eval_sts-test_pearson_cosine": 0.056062031998983373,
      "eval_sts-test_pearson_dot": 0.2979259445723872,
      "eval_sts-test_pearson_euclidean": 0.0498319208592713,
      "eval_sts-test_pearson_manhattan": 0.07381429239121526,
      "eval_sts-test_pearson_max": 0.2979259445723872,
      "eval_sts-test_spearman_cosine": 0.1066788491614481,
      "eval_sts-test_spearman_dot": 0.315952670306405,
      "eval_sts-test_spearman_euclidean": 0.07303394554435191,
      "eval_sts-test_spearman_manhattan": 0.09039525717692232,
      "eval_sts-test_spearman_max": 0.315952670306405,
      "eval_vitaminc-pairs_loss": 2.698580741882324,
      "eval_vitaminc-pairs_runtime": 1.4747,
      "eval_vitaminc-pairs_samples_per_second": 73.236,
      "eval_vitaminc-pairs_steps_per_second": 1.356,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_negation-triplets_loss": 5.142906665802002,
      "eval_negation-triplets_runtime": 0.2993,
      "eval_negation-triplets_samples_per_second": 213.865,
      "eval_negation-triplets_steps_per_second": 3.342,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_scitail-pairs-pos_loss": 1.9216958284378052,
      "eval_scitail-pairs-pos_runtime": 0.3834,
      "eval_scitail-pairs-pos_samples_per_second": 140.842,
      "eval_scitail-pairs-pos_steps_per_second": 2.608,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_xsum-pairs_loss": 6.073049545288086,
      "eval_xsum-pairs_runtime": 3.1587,
      "eval_xsum-pairs_samples_per_second": 40.523,
      "eval_xsum-pairs_steps_per_second": 0.633,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_sciq_pairs_loss": 0.3449864387512207,
      "eval_sciq_pairs_runtime": 3.3747,
      "eval_sciq_pairs_samples_per_second": 37.93,
      "eval_sciq_pairs_steps_per_second": 0.593,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_qasc_pairs_loss": 3.2267842292785645,
      "eval_qasc_pairs_runtime": 0.6576,
      "eval_qasc_pairs_samples_per_second": 194.646,
      "eval_qasc_pairs_steps_per_second": 3.041,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_openbookqa_pairs_loss": 4.405983924865723,
      "eval_openbookqa_pairs_runtime": 0.6107,
      "eval_openbookqa_pairs_samples_per_second": 209.594,
      "eval_openbookqa_pairs_steps_per_second": 3.275,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_msmarco_pairs_loss": 6.937691688537598,
      "eval_msmarco_pairs_runtime": 1.3091,
      "eval_msmarco_pairs_samples_per_second": 97.779,
      "eval_msmarco_pairs_steps_per_second": 1.528,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_nq_pairs_loss": 6.794108867645264,
      "eval_nq_pairs_runtime": 2.3968,
      "eval_nq_pairs_samples_per_second": 53.404,
      "eval_nq_pairs_steps_per_second": 0.834,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_trivia_pairs_loss": 6.3355631828308105,
      "eval_trivia_pairs_runtime": 4.4974,
      "eval_trivia_pairs_samples_per_second": 28.461,
      "eval_trivia_pairs_steps_per_second": 0.445,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_gooaq_pairs_loss": 6.405998706817627,
      "eval_gooaq_pairs_runtime": 0.8745,
      "eval_gooaq_pairs_samples_per_second": 146.37,
      "eval_gooaq_pairs_steps_per_second": 2.287,
      "step": 27
    },
    {
      "epoch": 0.1524347212420607,
      "eval_paws-pos_loss": 2.2308223247528076,
      "eval_paws-pos_runtime": 0.6998,
      "eval_paws-pos_samples_per_second": 182.908,
      "eval_paws-pos_steps_per_second": 2.858,
      "step": 27
    },
    {
      "epoch": 0.16937191249117856,
      "grad_norm": 5.885251522064209,
      "learning_rate": 6.818181818181818e-06,
      "loss": 5.7653,
      "step": 30
    },
    {
      "epoch": 0.1863091037402964,
      "grad_norm": 7.357480049133301,
      "learning_rate": 7.500000000000001e-06,
      "loss": 6.1259,
      "step": 33
    },
    {
      "epoch": 0.20324629498941427,
      "grad_norm": 7.321795463562012,
      "learning_rate": 8.181818181818183e-06,
      "loss": 5.7539,
      "step": 36
    },
    {
      "epoch": 0.22018348623853212,
      "grad_norm": 4.239792346954346,
      "learning_rate": 8.863636363636365e-06,
      "loss": 6.0131,
      "step": 39
    },
    {
      "epoch": 0.23712067748764998,
      "grad_norm": 3.9554407596588135,
      "learning_rate": 9.545454545454547e-06,
      "loss": 6.0074,
      "step": 42
    },
    {
      "epoch": 0.25405786873676783,
      "grad_norm": 4.406026840209961,
      "learning_rate": 1.0227272727272729e-05,
      "loss": 5.7125,
      "step": 45
    },
    {
      "epoch": 0.2709950599858857,
      "grad_norm": 7.235893249511719,
      "learning_rate": 1.0909090909090909e-05,
      "loss": 5.5634,
      "step": 48
    },
    {
      "epoch": 0.28793225123500354,
      "grad_norm": 5.330288410186768,
      "learning_rate": 1.1590909090909093e-05,
      "loss": 5.2924,
      "step": 51
    },
    {
      "epoch": 0.3048694424841214,
      "grad_norm": 7.216403961181641,
      "learning_rate": 1.2272727272727274e-05,
      "loss": 5.2286,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_NLI-v2_cosine_accuracy": 1.0,
      "eval_NLI-v2_dot_accuracy": 0.046875,
      "eval_NLI-v2_euclidean_accuracy": 1.0,
      "eval_NLI-v2_manhattan_accuracy": 1.0,
      "eval_NLI-v2_max_accuracy": 1.0,
      "eval_VitaminC_cosine_accuracy": 0.54296875,
      "eval_VitaminC_cosine_accuracy_threshold": 0.9328227043151855,
      "eval_VitaminC_cosine_ap": 0.5212059026196154,
      "eval_VitaminC_cosine_f1": 0.6576819407008085,
      "eval_VitaminC_cosine_f1_threshold": 0.7373804450035095,
      "eval_VitaminC_cosine_precision": 0.49193548387096775,
      "eval_VitaminC_cosine_recall": 0.991869918699187,
      "eval_VitaminC_dot_accuracy": 0.55078125,
      "eval_VitaminC_dot_accuracy_threshold": 418.2774658203125,
      "eval_VitaminC_dot_ap": 0.5160594099493883,
      "eval_VitaminC_dot_f1": 0.6521739130434782,
      "eval_VitaminC_dot_f1_threshold": 291.5081481933594,
      "eval_VitaminC_dot_precision": 0.4897959183673469,
      "eval_VitaminC_dot_recall": 0.975609756097561,
      "eval_VitaminC_euclidean_accuracy": 0.5390625,
      "eval_VitaminC_euclidean_accuracy_threshold": 8.120429039001465,
      "eval_VitaminC_euclidean_ap": 0.5224837623095228,
      "eval_VitaminC_euclidean_f1": 0.6576819407008085,
      "eval_VitaminC_euclidean_f1_threshold": 14.879999160766602,
      "eval_VitaminC_euclidean_precision": 0.49193548387096775,
      "eval_VitaminC_euclidean_recall": 0.991869918699187,
      "eval_VitaminC_manhattan_accuracy": 0.53515625,
      "eval_VitaminC_manhattan_accuracy_threshold": 137.40658569335938,
      "eval_VitaminC_manhattan_ap": 0.5186382518671783,
      "eval_VitaminC_manhattan_f1": 0.6576086956521738,
      "eval_VitaminC_manhattan_f1_threshold": 263.32452392578125,
      "eval_VitaminC_manhattan_precision": 0.49387755102040815,
      "eval_VitaminC_manhattan_recall": 0.983739837398374,
      "eval_VitaminC_max_accuracy": 0.55078125,
      "eval_VitaminC_max_accuracy_threshold": 418.2774658203125,
      "eval_VitaminC_max_ap": 0.5224837623095228,
      "eval_VitaminC_max_f1": 0.6576819407008085,
      "eval_VitaminC_max_f1_threshold": 291.5081481933594,
      "eval_VitaminC_max_precision": 0.49387755102040815,
      "eval_VitaminC_max_recall": 0.991869918699187,
      "eval_sequential_score": 0.5224837623095228,
      "eval_sts-test_pearson_cosine": 0.14377091128453176,
      "eval_sts-test_pearson_dot": 0.24728387094758872,
      "eval_sts-test_pearson_euclidean": 0.14604155960515372,
      "eval_sts-test_pearson_manhattan": 0.1446467532231986,
      "eval_sts-test_pearson_max": 0.24728387094758872,
      "eval_sts-test_spearman_cosine": 0.1968510434344728,
      "eval_sts-test_spearman_dot": 0.29467218283745694,
      "eval_sts-test_spearman_euclidean": 0.17218164683969664,
      "eval_sts-test_spearman_manhattan": 0.17741843340856742,
      "eval_sts-test_spearman_max": 0.29467218283745694,
      "eval_vitaminc-pairs_loss": 2.664700746536255,
      "eval_vitaminc-pairs_runtime": 1.4487,
      "eval_vitaminc-pairs_samples_per_second": 74.551,
      "eval_vitaminc-pairs_steps_per_second": 1.381,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_negation-triplets_loss": 4.6218037605285645,
      "eval_negation-triplets_runtime": 0.2971,
      "eval_negation-triplets_samples_per_second": 215.438,
      "eval_negation-triplets_steps_per_second": 3.366,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_scitail-pairs-pos_loss": 1.2413936853408813,
      "eval_scitail-pairs-pos_runtime": 0.372,
      "eval_scitail-pairs-pos_samples_per_second": 145.175,
      "eval_scitail-pairs-pos_steps_per_second": 2.688,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_xsum-pairs_loss": 5.249766826629639,
      "eval_xsum-pairs_runtime": 3.1506,
      "eval_xsum-pairs_samples_per_second": 40.627,
      "eval_xsum-pairs_steps_per_second": 0.635,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_sciq_pairs_loss": 0.2961578667163849,
      "eval_sciq_pairs_runtime": 3.2909,
      "eval_sciq_pairs_samples_per_second": 38.895,
      "eval_sciq_pairs_steps_per_second": 0.608,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_qasc_pairs_loss": 2.530872344970703,
      "eval_qasc_pairs_runtime": 0.6255,
      "eval_qasc_pairs_samples_per_second": 204.63,
      "eval_qasc_pairs_steps_per_second": 3.197,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_openbookqa_pairs_loss": 3.8855104446411133,
      "eval_openbookqa_pairs_runtime": 0.5742,
      "eval_openbookqa_pairs_samples_per_second": 222.914,
      "eval_openbookqa_pairs_steps_per_second": 3.483,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_msmarco_pairs_loss": 5.246406555175781,
      "eval_msmarco_pairs_runtime": 1.2872,
      "eval_msmarco_pairs_samples_per_second": 99.442,
      "eval_msmarco_pairs_steps_per_second": 1.554,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_nq_pairs_loss": 5.332630157470703,
      "eval_nq_pairs_runtime": 2.3739,
      "eval_nq_pairs_samples_per_second": 53.92,
      "eval_nq_pairs_steps_per_second": 0.843,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_trivia_pairs_loss": 5.647429943084717,
      "eval_trivia_pairs_runtime": 4.4729,
      "eval_trivia_pairs_samples_per_second": 28.617,
      "eval_trivia_pairs_steps_per_second": 0.447,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_gooaq_pairs_loss": 5.225871562957764,
      "eval_gooaq_pairs_runtime": 0.8715,
      "eval_gooaq_pairs_samples_per_second": 146.868,
      "eval_gooaq_pairs_steps_per_second": 2.295,
      "step": 54
    },
    {
      "epoch": 0.3048694424841214,
      "eval_paws-pos_loss": 0.8335962891578674,
      "eval_paws-pos_runtime": 0.6844,
      "eval_paws-pos_samples_per_second": 187.036,
      "eval_paws-pos_steps_per_second": 2.922,
      "step": 54
    },
    {
      "epoch": 0.32180663373323926,
      "grad_norm": 6.847682952880859,
      "learning_rate": 1.2954545454545455e-05,
      "loss": 4.4811,
      "step": 57
    },
    {
      "epoch": 0.3387438249823571,
      "grad_norm": 8.383002281188965,
      "learning_rate": 1.3636363636363637e-05,
      "loss": 4.4239,
      "step": 60
    },
    {
      "epoch": 0.35568101623147497,
      "grad_norm": 7.014843463897705,
      "learning_rate": 1.431818181818182e-05,
      "loss": 4.0273,
      "step": 63
    },
    {
      "epoch": 0.3726182074805928,
      "grad_norm": 5.9739885330200195,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 3.4508,
      "step": 66
    },
    {
      "epoch": 0.3895553987297107,
      "grad_norm": 11.202752113342285,
      "learning_rate": 1.5681818181818182e-05,
      "loss": 3.9702,
      "step": 69
    },
    {
      "epoch": 0.40649258997882853,
      "grad_norm": 7.064818859100342,
      "learning_rate": 1.6363636363636366e-05,
      "loss": 3.5295,
      "step": 72
    },
    {
      "epoch": 0.4234297812279464,
      "grad_norm": 5.912719249725342,
      "learning_rate": 1.7045454545454546e-05,
      "loss": 3.6395,
      "step": 75
    },
    {
      "epoch": 0.44036697247706424,
      "grad_norm": 5.033207893371582,
      "learning_rate": 1.772727272727273e-05,
      "loss": 3.2398,
      "step": 78
    },
    {
      "epoch": 0.4573041637261821,
      "grad_norm": 5.218384265899658,
      "learning_rate": 1.840909090909091e-05,
      "loss": 3.116,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_NLI-v2_cosine_accuracy": 1.0,
      "eval_NLI-v2_dot_accuracy": 0.0,
      "eval_NLI-v2_euclidean_accuracy": 1.0,
      "eval_NLI-v2_manhattan_accuracy": 1.0,
      "eval_NLI-v2_max_accuracy": 1.0,
      "eval_VitaminC_cosine_accuracy": 0.5546875,
      "eval_VitaminC_cosine_accuracy_threshold": 0.9041332006454468,
      "eval_VitaminC_cosine_ap": 0.5292859731465609,
      "eval_VitaminC_cosine_f1": 0.6542553191489362,
      "eval_VitaminC_cosine_f1_threshold": 0.452939510345459,
      "eval_VitaminC_cosine_precision": 0.48616600790513836,
      "eval_VitaminC_cosine_recall": 1.0,
      "eval_VitaminC_dot_accuracy": 0.5546875,
      "eval_VitaminC_dot_accuracy_threshold": 414.42559814453125,
      "eval_VitaminC_dot_ap": 0.5222732504955002,
      "eval_VitaminC_dot_f1": 0.6542553191489362,
      "eval_VitaminC_dot_f1_threshold": 212.6934814453125,
      "eval_VitaminC_dot_precision": 0.48616600790513836,
      "eval_VitaminC_dot_recall": 1.0,
      "eval_VitaminC_euclidean_accuracy": 0.5546875,
      "eval_VitaminC_euclidean_accuracy_threshold": 9.18377685546875,
      "eval_VitaminC_euclidean_ap": 0.5291787221346742,
      "eval_VitaminC_euclidean_f1": 0.6542553191489362,
      "eval_VitaminC_euclidean_f1_threshold": 22.683509826660156,
      "eval_VitaminC_euclidean_precision": 0.48616600790513836,
      "eval_VitaminC_euclidean_recall": 1.0,
      "eval_VitaminC_manhattan_accuracy": 0.55859375,
      "eval_VitaminC_manhattan_accuracy_threshold": 173.8212127685547,
      "eval_VitaminC_manhattan_ap": 0.5305698453165033,
      "eval_VitaminC_manhattan_f1": 0.6542553191489362,
      "eval_VitaminC_manhattan_f1_threshold": 415.5366516113281,
      "eval_VitaminC_manhattan_precision": 0.48616600790513836,
      "eval_VitaminC_manhattan_recall": 1.0,
      "eval_VitaminC_max_accuracy": 0.55859375,
      "eval_VitaminC_max_accuracy_threshold": 414.42559814453125,
      "eval_VitaminC_max_ap": 0.5305698453165033,
      "eval_VitaminC_max_f1": 0.6542553191489362,
      "eval_VitaminC_max_f1_threshold": 415.5366516113281,
      "eval_VitaminC_max_precision": 0.48616600790513836,
      "eval_VitaminC_max_recall": 1.0,
      "eval_sequential_score": 0.5305698453165033,
      "eval_sts-test_pearson_cosine": 0.45494716382349193,
      "eval_sts-test_pearson_dot": 0.44837123659858896,
      "eval_sts-test_pearson_euclidean": 0.4480861256491879,
      "eval_sts-test_pearson_manhattan": 0.4417008219313264,
      "eval_sts-test_pearson_max": 0.45494716382349193,
      "eval_sts-test_spearman_cosine": 0.48921418507251446,
      "eval_sts-test_spearman_dot": 0.46707725062744593,
      "eval_sts-test_spearman_euclidean": 0.4610824798409968,
      "eval_sts-test_spearman_manhattan": 0.46068648052845956,
      "eval_sts-test_spearman_max": 0.48921418507251446,
      "eval_vitaminc-pairs_loss": 2.5043575763702393,
      "eval_vitaminc-pairs_runtime": 1.4778,
      "eval_vitaminc-pairs_samples_per_second": 73.079,
      "eval_vitaminc-pairs_steps_per_second": 1.353,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_negation-triplets_loss": 3.4229447841644287,
      "eval_negation-triplets_runtime": 0.2991,
      "eval_negation-triplets_samples_per_second": 213.954,
      "eval_negation-triplets_steps_per_second": 3.343,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_scitail-pairs-pos_loss": 0.2784869372844696,
      "eval_scitail-pairs-pos_runtime": 0.3633,
      "eval_scitail-pairs-pos_samples_per_second": 148.649,
      "eval_scitail-pairs-pos_steps_per_second": 2.753,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_xsum-pairs_loss": 2.428964614868164,
      "eval_xsum-pairs_runtime": 3.1548,
      "eval_xsum-pairs_samples_per_second": 40.573,
      "eval_xsum-pairs_steps_per_second": 0.634,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_sciq_pairs_loss": 0.15256048738956451,
      "eval_sciq_pairs_runtime": 3.2432,
      "eval_sciq_pairs_samples_per_second": 39.467,
      "eval_sciq_pairs_steps_per_second": 0.617,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_qasc_pairs_loss": 1.2902077436447144,
      "eval_qasc_pairs_runtime": 0.6211,
      "eval_qasc_pairs_samples_per_second": 206.085,
      "eval_qasc_pairs_steps_per_second": 3.22,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_openbookqa_pairs_loss": 2.4784862995147705,
      "eval_openbookqa_pairs_runtime": 0.5758,
      "eval_openbookqa_pairs_samples_per_second": 222.308,
      "eval_openbookqa_pairs_steps_per_second": 3.474,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_msmarco_pairs_loss": 2.967724084854126,
      "eval_msmarco_pairs_runtime": 1.2944,
      "eval_msmarco_pairs_samples_per_second": 98.885,
      "eval_msmarco_pairs_steps_per_second": 1.545,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_nq_pairs_loss": 3.358661413192749,
      "eval_nq_pairs_runtime": 2.3827,
      "eval_nq_pairs_samples_per_second": 53.722,
      "eval_nq_pairs_steps_per_second": 0.839,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_trivia_pairs_loss": 3.1391680240631104,
      "eval_trivia_pairs_runtime": 4.4155,
      "eval_trivia_pairs_samples_per_second": 28.989,
      "eval_trivia_pairs_steps_per_second": 0.453,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_gooaq_pairs_loss": 2.8774912357330322,
      "eval_gooaq_pairs_runtime": 0.8746,
      "eval_gooaq_pairs_samples_per_second": 146.346,
      "eval_gooaq_pairs_steps_per_second": 2.287,
      "step": 81
    },
    {
      "epoch": 0.4573041637261821,
      "eval_paws-pos_loss": 0.19754411280155182,
      "eval_paws-pos_runtime": 0.684,
      "eval_paws-pos_samples_per_second": 187.141,
      "eval_paws-pos_steps_per_second": 2.924,
      "step": 81
    },
    {
      "epoch": 0.47424135497529996,
      "grad_norm": 5.149569988250732,
      "learning_rate": 1.9090909090909094e-05,
      "loss": 2.6049,
      "step": 84
    },
    {
      "epoch": 0.4911785462244178,
      "grad_norm": 5.012928009033203,
      "learning_rate": 1.9772727272727274e-05,
      "loss": 2.7738,
      "step": 87
    },
    {
      "epoch": 0.5081157374735357,
      "grad_norm": 4.880725383758545,
      "learning_rate": 2.0454545454545457e-05,
      "loss": 2.5416,
      "step": 90
    },
    {
      "epoch": 0.5250529287226535,
      "grad_norm": 5.618528366088867,
      "learning_rate": 2.113636363636364e-05,
      "loss": 2.3913,
      "step": 93
    },
    {
      "epoch": 0.5419901199717714,
      "grad_norm": 5.020515441894531,
      "learning_rate": 2.1818181818181818e-05,
      "loss": 2.3144,
      "step": 96
    },
    {
      "epoch": 0.5589273112208892,
      "grad_norm": 4.818451404571533,
      "learning_rate": 2.25e-05,
      "loss": 2.1857,
      "step": 99
    },
    {
      "epoch": 0.5758645024700071,
      "grad_norm": 5.094771385192871,
      "learning_rate": 2.3181818181818185e-05,
      "loss": 1.8881,
      "step": 102
    },
    {
      "epoch": 0.592801693719125,
      "grad_norm": 3.795962333679199,
      "learning_rate": 2.3863636363636365e-05,
      "loss": 2.2699,
      "step": 105
    }
  ],
  "logging_steps": 3,
  "max_steps": 531,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 107,
  "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": 160,
  "trial_name": null,
  "trial_params": null
}