{
  "best_metric": 2.003035545349121,
  "best_model_checkpoint": "/mnt/default/projects/sca-xiaoke-v3/amlt-results/7301932201.25563-cd1e6021-6ea9-4835-8578-ba26f723a708/checkpoint-100000",
  "epoch": 3.673229503379371,
  "global_step": 100000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "_do_backward_in_ms": 13833.612655987963,
      "_prepare_inputs_in_ms": 9.499168023467064,
      "compute_loss_in_ms": 1344.5582571439445,
      "epoch": 0.0,
      "learning_rate/full": 0.0,
      "loss": 9.9783,
      "step": 1,
      "training_step_in_ms": 15198.054818203673
    },
    {
      "epoch": 0.0,
      "eval_objects365-local-v2-validation_loss": 9.392258644104004,
      "eval_objects365-local-v2-validation_runtime": 16.1869,
      "eval_objects365-local-v2-validation_samples_per_second": 49.423,
      "eval_objects365-local-v2-validation_steps_per_second": 0.803,
      "step": 1
    },
    {
      "_do_backward_in_ms": 2550.187955102608,
      "_prepare_inputs_in_ms": 12.46627290174549,
      "compute_loss_in_ms": 289.9478549773164,
      "epoch": 0.04,
      "learning_rate/full": 9.998519814211118e-05,
      "loss": 3.8861,
      "step": 1000,
      "training_step_in_ms": 2849.577255269468
    },
    {
      "_do_backward_in_ms": 2556.9965168242343,
      "_prepare_inputs_in_ms": 3.3681516400538385,
      "compute_loss_in_ms": 289.68224829342216,
      "epoch": 0.07,
      "learning_rate/full": 9.992214555651498e-05,
      "loss": 2.6909,
      "step": 2000,
      "training_step_in_ms": 2856.0172235199716
    },
    {
      "_do_backward_in_ms": 2544.9560291268863,
      "_prepare_inputs_in_ms": 3.326028081588447,
      "compute_loss_in_ms": 289.6684127182234,
      "epoch": 0.11,
      "learning_rate/full": 9.980962819544264e-05,
      "loss": 2.4899,
      "step": 3000,
      "training_step_in_ms": 2843.9263667755295
    },
    {
      "_do_backward_in_ms": 2553.390086729545,
      "_prepare_inputs_in_ms": 3.3251649560406804,
      "compute_loss_in_ms": 289.7164152129553,
      "epoch": 0.15,
      "learning_rate/full": 9.964775754540861e-05,
      "loss": 2.364,
      "step": 4000,
      "training_step_in_ms": 2852.3637848414946
    },
    {
      "_do_backward_in_ms": 2533.654973218916,
      "_prepare_inputs_in_ms": 3.3234398325439543,
      "compute_loss_in_ms": 289.49879492027685,
      "epoch": 0.18,
      "learning_rate/full": 9.94366939940579e-05,
      "loss": 2.2671,
      "step": 5000,
      "training_step_in_ms": 2832.4309401281644
    },
    {
      "epoch": 0.18,
      "eval_objects365-local-v2-validation_loss": 3.2006750106811523,
      "eval_objects365-local-v2-validation_runtime": 15.8807,
      "eval_objects365-local-v2-validation_samples_per_second": 50.376,
      "eval_objects365-local-v2-validation_steps_per_second": 0.819,
      "step": 5000
    },
    {
      "_do_backward_in_ms": 2536.1245670726057,
      "_prepare_inputs_in_ms": 13.602690789379688,
      "compute_loss_in_ms": 290.3555030166838,
      "epoch": 0.22,
      "learning_rate/full": 9.917664667124788e-05,
      "loss": 2.2035,
      "step": 6000,
      "training_step_in_ms": 2835.8918179108296
    },
    {
      "_do_backward_in_ms": 2521.147600293625,
      "_prepare_inputs_in_ms": 3.3214590828865767,
      "compute_loss_in_ms": 290.27019435027614,
      "epoch": 0.26,
      "learning_rate/full": 9.886787324183434e-05,
      "loss": 2.1482,
      "step": 7000,
      "training_step_in_ms": 2820.6920346897095
    },
    {
      "_do_backward_in_ms": 2548.451570129022,
      "_prepare_inputs_in_ms": 3.319471804657951,
      "compute_loss_in_ms": 289.1897811254021,
      "epoch": 0.29,
      "learning_rate/full": 9.851067965036731e-05,
      "loss": 2.0894,
      "step": 8000,
      "training_step_in_ms": 2846.943020476494
    },
    {
      "_do_backward_in_ms": 2545.683652488282,
      "_prepare_inputs_in_ms": 3.316459863912314,
      "compute_loss_in_ms": 289.3646022947505,
      "epoch": 0.33,
      "learning_rate/full": 9.81058489571687e-05,
      "loss": 2.0449,
      "step": 9000,
      "training_step_in_ms": 2844.350300100632
    },
    {
      "_do_backward_in_ms": 2587.985424251994,
      "_prepare_inputs_in_ms": 3.3290419806726277,
      "compute_loss_in_ms": 289.5182610661723,
      "epoch": 0.37,
      "learning_rate/full": 9.765344841950912e-05,
      "loss": 2.0159,
      "step": 10000,
      "training_step_in_ms": 2886.8384181782603
    },
    {
      "epoch": 0.37,
      "eval_objects365-local-v2-validation_loss": 2.7690136432647705,
      "eval_objects365-local-v2-validation_runtime": 9.6182,
      "eval_objects365-local-v2-validation_samples_per_second": 83.175,
      "eval_objects365-local-v2-validation_steps_per_second": 1.352,
      "step": 10000
    },
    {
      "_do_backward_in_ms": 2580.9500538480934,
      "_prepare_inputs_in_ms": 7.9110309222817925,
      "compute_loss_in_ms": 289.0381096436588,
      "epoch": 0.4,
      "learning_rate/full": 9.715392539719363e-05,
      "loss": 1.9885,
      "step": 11000,
      "training_step_in_ms": 2879.298744171858
    },
    {
      "_do_backward_in_ms": 2542.390609878581,
      "_prepare_inputs_in_ms": 3.3009646027348936,
      "compute_loss_in_ms": 289.6751387268305,
      "epoch": 0.44,
      "learning_rate/full": 9.660720396490479e-05,
      "loss": 1.9603,
      "step": 12000,
      "training_step_in_ms": 2841.272174295038
    },
    {
      "_do_backward_in_ms": 2512.176790253725,
      "_prepare_inputs_in_ms": 3.362945111002773,
      "compute_loss_in_ms": 289.53350385534577,
      "epoch": 0.48,
      "learning_rate/full": 9.601491810524181e-05,
      "loss": 1.9403,
      "step": 13000,
      "training_step_in_ms": 2811.0723278834485
    },
    {
      "_do_backward_in_ms": 2531.5591839374974,
      "_prepare_inputs_in_ms": 3.3155137847643346,
      "compute_loss_in_ms": 290.233249894809,
      "epoch": 0.51,
      "learning_rate/full": 9.537646901827713e-05,
      "loss": 1.9229,
      "step": 14000,
      "training_step_in_ms": 2831.124616945861
    },
    {
      "_do_backward_in_ms": 2558.8225115824025,
      "_prepare_inputs_in_ms": 3.351889422861859,
      "compute_loss_in_ms": 289.55780304642394,
      "epoch": 0.55,
      "learning_rate/full": 9.469376483242311e-05,
      "loss": 1.9078,
      "step": 15000,
      "training_step_in_ms": 2857.7440589836333
    },
    {
      "epoch": 0.55,
      "eval_objects365-local-v2-validation_loss": 2.566831588745117,
      "eval_objects365-local-v2-validation_runtime": 12.0469,
      "eval_objects365-local-v2-validation_samples_per_second": 66.407,
      "eval_objects365-local-v2-validation_steps_per_second": 1.079,
      "step": 15000
    },
    {
      "_do_backward_in_ms": 2569.1758075398393,
      "_prepare_inputs_in_ms": 9.903900120278632,
      "compute_loss_in_ms": 289.8735368660185,
      "epoch": 0.59,
      "learning_rate/full": 9.396611533498878e-05,
      "loss": 1.8953,
      "step": 16000,
      "training_step_in_ms": 2868.428954576142
    },
    {
      "_do_backward_in_ms": 2524.4527391090523,
      "_prepare_inputs_in_ms": 3.3152836377266794,
      "compute_loss_in_ms": 289.7500243561808,
      "epoch": 0.62,
      "learning_rate/full": 9.319569524704217e-05,
      "loss": 1.8846,
      "step": 17000,
      "training_step_in_ms": 2823.4884984181263
    },
    {
      "_do_backward_in_ms": 2517.142957175616,
      "_prepare_inputs_in_ms": 3.341707782819867,
      "compute_loss_in_ms": 289.59233529726043,
      "epoch": 0.66,
      "learning_rate/full": 9.238172567527468e-05,
      "loss": 1.8726,
      "step": 18000,
      "training_step_in_ms": 2816.0720933733974
    },
    {
      "_do_backward_in_ms": 2532.7779561954085,
      "_prepare_inputs_in_ms": 3.3180122550111264,
      "compute_loss_in_ms": 289.8389355558902,
      "epoch": 0.7,
      "learning_rate/full": 9.152751594369358e-05,
      "loss": 1.864,
      "step": 19000,
      "training_step_in_ms": 2831.926313831471
    },
    {
      "_do_backward_in_ms": 2530.6621135415044,
      "_prepare_inputs_in_ms": 3.3357447627931833,
      "compute_loss_in_ms": 289.85987694049254,
      "epoch": 0.73,
      "learning_rate/full": 9.063048903303299e-05,
      "loss": 1.8557,
      "step": 20000,
      "training_step_in_ms": 2829.814629596658
    },
    {
      "epoch": 0.73,
      "eval_objects365-local-v2-validation_loss": 2.4183120727539062,
      "eval_objects365-local-v2-validation_runtime": 19.7661,
      "eval_objects365-local-v2-validation_samples_per_second": 40.473,
      "eval_objects365-local-v2-validation_steps_per_second": 0.658,
      "step": 20000
    },
    {
      "_do_backward_in_ms": 2476.3556728472468,
      "_prepare_inputs_in_ms": 17.271267849380134,
      "compute_loss_in_ms": 289.4574518314115,
      "epoch": 0.77,
      "learning_rate/full": 8.969416096468137e-05,
      "loss": 1.8426,
      "step": 21000,
      "training_step_in_ms": 2775.230575547088
    },
    {
      "_do_backward_in_ms": 2468.883111936506,
      "_prepare_inputs_in_ms": 3.332945456728339,
      "compute_loss_in_ms": 289.88871885929257,
      "epoch": 0.81,
      "learning_rate/full": 8.871858101523774e-05,
      "loss": 1.8415,
      "step": 22000,
      "training_step_in_ms": 2768.1600090207066
    },
    {
      "_do_backward_in_ms": 2506.8567285528407,
      "_prepare_inputs_in_ms": 3.3408744835760444,
      "compute_loss_in_ms": 289.73892450495623,
      "epoch": 0.84,
      "learning_rate/full": 8.770471389455464e-05,
      "loss": 1.8293,
      "step": 23000,
      "training_step_in_ms": 2805.954835511511
    },
    {
      "_do_backward_in_ms": 2541.85831053718,
      "_prepare_inputs_in_ms": 3.357209531823173,
      "compute_loss_in_ms": 290.00427257712,
      "epoch": 0.88,
      "learning_rate/full": 8.665356217305291e-05,
      "loss": 1.8266,
      "step": 24000,
      "training_step_in_ms": 2841.2785381632857
    },
    {
      "_do_backward_in_ms": 2533.210183262825,
      "_prepare_inputs_in_ms": 3.364060287596658,
      "compute_loss_in_ms": 289.9935355405323,
      "epoch": 0.92,
      "learning_rate/full": 8.556616529032215e-05,
      "loss": 1.8162,
      "step": 25000,
      "training_step_in_ms": 2832.5927242138423
    },
    {
      "epoch": 0.92,
      "eval_objects365-local-v2-validation_loss": 2.3316762447357178,
      "eval_objects365-local-v2-validation_runtime": 10.2829,
      "eval_objects365-local-v2-validation_samples_per_second": 77.799,
      "eval_objects365-local-v2-validation_steps_per_second": 1.264,
      "step": 25000
    },
    {
      "_do_backward_in_ms": 2552.305114510702,
      "_prepare_inputs_in_ms": 8.565855091612377,
      "compute_loss_in_ms": 289.73671219767664,
      "epoch": 0.96,
      "learning_rate/full": 8.444359852726274e-05,
      "loss": 1.8117,
      "step": 26000,
      "training_step_in_ms": 2851.485752185108
    },
    {
      "_do_backward_in_ms": 2533.5027522125747,
      "_prepare_inputs_in_ms": 3.3632643420714885,
      "compute_loss_in_ms": 289.90648507210426,
      "epoch": 0.99,
      "learning_rate/full": 8.328579747384175e-05,
      "loss": 1.806,
      "step": 27000,
      "training_step_in_ms": 2832.807456281269
    },
    {
      "_do_backward_in_ms": 2546.2225131660234,
      "_prepare_inputs_in_ms": 3.3682647191453725,
      "compute_loss_in_ms": 291.1998807019554,
      "epoch": 1.03,
      "learning_rate/full": 8.209501557634378e-05,
      "loss": 1.7955,
      "step": 28000,
      "training_step_in_ms": 2846.875886055641
    },
    {
      "_do_backward_in_ms": 2583.1195299711544,
      "_prepare_inputs_in_ms": 3.3649198710918427,
      "compute_loss_in_ms": 290.97770567121916,
      "epoch": 1.07,
      "learning_rate/full": 8.087367077395005e-05,
      "loss": 1.7877,
      "step": 29000,
      "training_step_in_ms": 2883.5424125664867
    },
    {
      "_do_backward_in_ms": 2570.092649807455,
      "_prepare_inputs_in_ms": 3.4090082803741097,
      "compute_loss_in_ms": 291.62430305662565,
      "epoch": 1.1,
      "learning_rate/full": 7.962179630107982e-05,
      "loss": 1.786,
      "step": 30000,
      "training_step_in_ms": 2871.2795688530896
    },
    {
      "epoch": 1.1,
      "eval_objects365-local-v2-validation_loss": 2.2865772247314453,
      "eval_objects365-local-v2-validation_runtime": 8.6589,
      "eval_objects365-local-v2-validation_samples_per_second": 92.391,
      "eval_objects365-local-v2-validation_steps_per_second": 1.501,
      "step": 30000
    },
    {
      "_do_backward_in_ms": 721.1331692694221,
      "_prepare_inputs_in_ms": 3.2962444906588644,
      "compute_loss_in_ms": 299.71373338252306,
      "epoch": 1.14,
      "learning_rate/full": 9.998418572322853e-05,
      "loss": 1.7871,
      "step": 31000,
      "training_step_in_ms": 1029.60611718148
    },
    {
      "_do_backward_in_ms": 729.436703273328,
      "_prepare_inputs_in_ms": 3.258270466234535,
      "compute_loss_in_ms": 298.0247277948074,
      "epoch": 1.18,
      "learning_rate/full": 9.991993511089866e-05,
      "loss": 1.7845,
      "step": 32000,
      "training_step_in_ms": 1036.1627226730343
    },
    {
      "_do_backward_in_ms": 703.6630942693446,
      "_prepare_inputs_in_ms": 3.2456943639554083,
      "compute_loss_in_ms": 297.64328660978936,
      "epoch": 1.21,
      "learning_rate/full": 9.98061823549655e-05,
      "loss": 1.7817,
      "step": 33000,
      "training_step_in_ms": 1009.9646131193731
    },
    {
      "_do_backward_in_ms": 676.1822207763325,
      "_prepare_inputs_in_ms": 3.2518609322141856,
      "compute_loss_in_ms": 297.877519285772,
      "epoch": 1.25,
      "learning_rate/full": 9.964326742751142e-05,
      "loss": 1.7795,
      "step": 34000,
      "training_step_in_ms": 982.7506944458
    },
    {
      "_do_backward_in_ms": 624.6346801738255,
      "_prepare_inputs_in_ms": 3.2795886190142483,
      "compute_loss_in_ms": 298.1166247774381,
      "epoch": 1.29,
      "learning_rate/full": 9.943126236733435e-05,
      "loss": 1.7769,
      "step": 35000,
      "training_step_in_ms": 931.5135621828958
    },
    {
      "epoch": 1.29,
      "eval_objects365-local-v2-validation_loss": 2.2724013328552246,
      "eval_objects365-local-v2-validation_runtime": 12.4784,
      "eval_objects365-local-v2-validation_samples_per_second": 64.111,
      "eval_objects365-local-v2-validation_steps_per_second": 1.042,
      "step": 35000
    },
    {
      "_do_backward_in_ms": 655.5177808874287,
      "_prepare_inputs_in_ms": 7.8633947225300425,
      "compute_loss_in_ms": 297.8193518089228,
      "epoch": 1.32,
      "learning_rate/full": 9.917037681729384e-05,
      "loss": 1.7745,
      "step": 36000,
      "training_step_in_ms": 962.1334383783396
    },
    {
      "_do_backward_in_ms": 699.8656730484217,
      "_prepare_inputs_in_ms": 3.240072426153347,
      "compute_loss_in_ms": 298.73781585809775,
      "epoch": 1.36,
      "learning_rate/full": 9.886053467655043e-05,
      "loss": 1.7686,
      "step": 37000,
      "training_step_in_ms": 1007.3340494644362
    },
    {
      "_do_backward_in_ms": 698.6605496255215,
      "_prepare_inputs_in_ms": 3.2106620045378804,
      "compute_loss_in_ms": 299.0653761769645,
      "epoch": 1.4,
      "learning_rate/full": 9.850266196818751e-05,
      "loss": 1.7686,
      "step": 38000,
      "training_step_in_ms": 1006.4037391303573
    },
    {
      "_do_backward_in_ms": 724.4765353850089,
      "_prepare_inputs_in_ms": 3.3089412197005004,
      "compute_loss_in_ms": 298.82350925635546,
      "epoch": 1.43,
      "learning_rate/full": 9.80968270237185e-05,
      "loss": 1.763,
      "step": 39000,
      "training_step_in_ms": 1032.1099421884865
    },
    {
      "_do_backward_in_ms": 681.9796452447772,
      "_prepare_inputs_in_ms": 3.263611613307148,
      "compute_loss_in_ms": 298.8235752664041,
      "epoch": 1.47,
      "learning_rate/full": 9.764343115619788e-05,
      "loss": 1.7614,
      "step": 40000,
      "training_step_in_ms": 989.5278342715465
    },
    {
      "epoch": 1.47,
      "eval_objects365-local-v2-validation_loss": 2.228322982788086,
      "eval_objects365-local-v2-validation_runtime": 16.6211,
      "eval_objects365-local-v2-validation_samples_per_second": 48.132,
      "eval_objects365-local-v2-validation_steps_per_second": 0.782,
      "step": 40000
    },
    {
      "_do_backward_in_ms": 720.4739852924831,
      "_prepare_inputs_in_ms": 12.448386072667512,
      "compute_loss_in_ms": 298.166242549251,
      "epoch": 1.51,
      "learning_rate/full": 9.714292270967042e-05,
      "loss": 1.7603,
      "step": 41000,
      "training_step_in_ms": 1027.4099870913196
    },
    {
      "_do_backward_in_ms": 658.6267548131291,
      "_prepare_inputs_in_ms": 3.201066299341619,
      "compute_loss_in_ms": 298.838454146171,
      "epoch": 1.54,
      "learning_rate/full": 9.659579661582255e-05,
      "loss": 1.7573,
      "step": 42000,
      "training_step_in_ms": 966.1794079060201
    },
    {
      "_do_backward_in_ms": 676.473127261037,
      "_prepare_inputs_in_ms": 3.228976390324533,
      "compute_loss_in_ms": 298.2980008148588,
      "epoch": 1.58,
      "learning_rate/full": 9.600197721584953e-05,
      "loss": 1.7545,
      "step": 43000,
      "training_step_in_ms": 983.4983903854154
    },
    {
      "_do_backward_in_ms": 697.463578726165,
      "_prepare_inputs_in_ms": 3.195912489667535,
      "compute_loss_in_ms": 299.1437525388319,
      "epoch": 1.62,
      "learning_rate/full": 9.536323925372398e-05,
      "loss": 1.7526,
      "step": 44000,
      "training_step_in_ms": 1005.2843026786577
    },
    {
      "_do_backward_in_ms": 813.7108269445598,
      "_prepare_inputs_in_ms": 3.206419989466667,
      "compute_loss_in_ms": 298.0525588088203,
      "epoch": 1.65,
      "learning_rate/full": 9.467964349816328e-05,
      "loss": 1.7477,
      "step": 45000,
      "training_step_in_ms": 1120.4696311727166
    },
    {
      "epoch": 1.65,
      "eval_objects365-local-v2-validation_loss": 2.1873245239257812,
      "eval_objects365-local-v2-validation_runtime": 12.2979,
      "eval_objects365-local-v2-validation_samples_per_second": 65.052,
      "eval_objects365-local-v2-validation_steps_per_second": 1.057,
      "step": 45000
    },
    {
      "_do_backward_in_ms": 708.7151438989677,
      "_prepare_inputs_in_ms": 8.210845838180447,
      "compute_loss_in_ms": 297.9545111299988,
      "epoch": 1.69,
      "learning_rate/full": 9.395186592816932e-05,
      "loss": 1.7454,
      "step": 46000,
      "training_step_in_ms": 1015.4890209392179
    },
    {
      "_do_backward_in_ms": 677.8432183256373,
      "_prepare_inputs_in_ms": 3.263900319347158,
      "compute_loss_in_ms": 297.8792646545917,
      "epoch": 1.73,
      "learning_rate/full": 9.31806262122764e-05,
      "loss": 1.7443,
      "step": 47000,
      "training_step_in_ms": 984.4030563381966
    },
    {
      "_do_backward_in_ms": 720.129483740544,
      "_prepare_inputs_in_ms": 3.2109336624853313,
      "compute_loss_in_ms": 297.79731379216537,
      "epoch": 1.76,
      "learning_rate/full": 9.23666869969011e-05,
      "loss": 1.7429,
      "step": 48000,
      "training_step_in_ms": 1026.6044916820247
    },
    {
      "_do_backward_in_ms": 655.5070220401976,
      "_prepare_inputs_in_ms": 3.2323538628406823,
      "compute_loss_in_ms": 297.94803192745894,
      "epoch": 1.8,
      "learning_rate/full": 9.15108531521937e-05,
      "loss": 1.7414,
      "step": 49000,
      "training_step_in_ms": 962.1004992513917
    },
    {
      "_do_backward_in_ms": 672.1013942162972,
      "_prepare_inputs_in_ms": 3.2201343055348843,
      "compute_loss_in_ms": 298.3921029092744,
      "epoch": 1.84,
      "learning_rate/full": 9.061305292392976e-05,
      "loss": 1.7364,
      "step": 50000,
      "training_step_in_ms": 979.1467305382248
    },
    {
      "epoch": 1.84,
      "eval_objects365-local-v2-validation_loss": 2.164118766784668,
      "eval_objects365-local-v2-validation_runtime": 10.1507,
      "eval_objects365-local-v2-validation_samples_per_second": 78.812,
      "eval_objects365-local-v2-validation_steps_per_second": 1.281,
      "step": 50000
    },
    {
      "_do_backward_in_ms": 674.3136331241112,
      "_prepare_inputs_in_ms": 6.509397196586208,
      "compute_loss_in_ms": 298.9732424640313,
      "epoch": 1.87,
      "learning_rate/full": 8.967692735767203e-05,
      "loss": 1.7366,
      "step": 51000,
      "training_step_in_ms": 982.0552003385965
    },
    {
      "_do_backward_in_ms": 697.6172431716695,
      "_prepare_inputs_in_ms": 3.218969340668991,
      "compute_loss_in_ms": 298.4035959227476,
      "epoch": 1.91,
      "learning_rate/full": 8.86996523066913e-05,
      "loss": 1.735,
      "step": 52000,
      "training_step_in_ms": 1004.6795647891704
    },
    {
      "_do_backward_in_ms": 704.6787912775762,
      "_prepare_inputs_in_ms": 3.23072279850021,
      "compute_loss_in_ms": 298.14702042611316,
      "epoch": 1.95,
      "learning_rate/full": 8.768506659844343e-05,
      "loss": 1.7307,
      "step": 53000,
      "training_step_in_ms": 1011.5234109486919
    },
    {
      "_do_backward_in_ms": 769.2823166255839,
      "_prepare_inputs_in_ms": 3.204921918688342,
      "compute_loss_in_ms": 298.67282255436294,
      "epoch": 1.98,
      "learning_rate/full": 8.663321571775915e-05,
      "loss": 1.7305,
      "step": 54000,
      "training_step_in_ms": 1076.5878808272537
    },
    {
      "_do_backward_in_ms": 696.9331407416612,
      "_prepare_inputs_in_ms": 3.244183993898332,
      "compute_loss_in_ms": 298.70368046709336,
      "epoch": 2.02,
      "learning_rate/full": 8.554513979559709e-05,
      "loss": 1.7214,
      "step": 55000,
      "training_step_in_ms": 1004.3571789248381
    },
    {
      "epoch": 2.02,
      "eval_objects365-local-v2-validation_loss": 2.1291964054107666,
      "eval_objects365-local-v2-validation_runtime": 9.885,
      "eval_objects365-local-v2-validation_samples_per_second": 80.931,
      "eval_objects365-local-v2-validation_steps_per_second": 1.315,
      "step": 55000
    },
    {
      "_do_backward_in_ms": 653.2630679495633,
      "_prepare_inputs_in_ms": 6.026035463639699,
      "compute_loss_in_ms": 299.06775847170223,
      "epoch": 2.06,
      "learning_rate/full": 8.442077319354145e-05,
      "loss": 1.7136,
      "step": 56000,
      "training_step_in_ms": 961.0975561500527
    },
    {
      "_do_backward_in_ms": 745.036297386745,
      "_prepare_inputs_in_ms": 3.249741542385891,
      "compute_loss_in_ms": 297.8976241340861,
      "epoch": 2.09,
      "learning_rate/full": 8.326347629835318e-05,
      "loss": 1.7138,
      "step": 57000,
      "training_step_in_ms": 1051.66180648515
    },
    {
      "_do_backward_in_ms": 626.0586955258623,
      "_prepare_inputs_in_ms": 3.230042038485408,
      "compute_loss_in_ms": 298.4531378012616,
      "epoch": 2.13,
      "learning_rate/full": 8.207328655483055e-05,
      "loss": 1.7121,
      "step": 58000,
      "training_step_in_ms": 933.1826650444418
    },
    {
      "_do_backward_in_ms": 704.3030783196446,
      "_prepare_inputs_in_ms": 3.246616828488186,
      "compute_loss_in_ms": 297.92309659463353,
      "epoch": 2.17,
      "learning_rate/full": 8.085138089139716e-05,
      "loss": 1.712,
      "step": 59000,
      "training_step_in_ms": 1010.959731190931
    },
    {
      "_do_backward_in_ms": 695.9465066853445,
      "_prepare_inputs_in_ms": 3.237966085318476,
      "compute_loss_in_ms": 298.90500363637693,
      "epoch": 2.2,
      "learning_rate/full": 7.95989675990117e-05,
      "loss": 1.7112,
      "step": 60000,
      "training_step_in_ms": 1003.5448978319764
    },
    {
      "epoch": 2.2,
      "eval_objects365-local-v2-validation_loss": 2.1127543449401855,
      "eval_objects365-local-v2-validation_runtime": 8.4515,
      "eval_objects365-local-v2-validation_samples_per_second": 94.657,
      "eval_objects365-local-v2-validation_steps_per_second": 1.538,
      "step": 60000
    },
    {
      "_do_backward_in_ms": 716.5119342987891,
      "_prepare_inputs_in_ms": 5.995148892154933,
      "compute_loss_in_ms": 299.30640732260883,
      "epoch": 2.24,
      "learning_rate/full": 7.831598792818578e-05,
      "loss": 1.7064,
      "step": 61000,
      "training_step_in_ms": 1024.6346847999375
    },
    {
      "_do_backward_in_ms": 679.5819691745564,
      "_prepare_inputs_in_ms": 3.2997133519966155,
      "compute_loss_in_ms": 298.4492077725008,
      "epoch": 2.28,
      "learning_rate/full": 7.700627631147224e-05,
      "loss": 1.7071,
      "step": 62000,
      "training_step_in_ms": 986.8024183101952
    },
    {
      "_do_backward_in_ms": 696.5671245567501,
      "_prepare_inputs_in_ms": 3.2379625719040632,
      "compute_loss_in_ms": 298.1275162412785,
      "epoch": 2.31,
      "learning_rate/full": 7.566850863280712e-05,
      "loss": 1.7073,
      "step": 63000,
      "training_step_in_ms": 1003.3855868254323
    },
    {
      "_do_backward_in_ms": 788.7977120715659,
      "_prepare_inputs_in_ms": 3.261386409169063,
      "compute_loss_in_ms": 297.86948832380585,
      "epoch": 2.35,
      "learning_rate/full": 7.430805850264685e-05,
      "loss": 1.7044,
      "step": 64000,
      "training_step_in_ms": 1095.416541912593
    },
    {
      "_do_backward_in_ms": 679.518492219504,
      "_prepare_inputs_in_ms": 3.206374272936955,
      "compute_loss_in_ms": 298.958141958341,
      "epoch": 2.39,
      "learning_rate/full": 7.292361925349194e-05,
      "loss": 1.7045,
      "step": 65000,
      "training_step_in_ms": 987.1717654119711
    },
    {
      "epoch": 2.39,
      "eval_objects365-local-v2-validation_loss": 2.0909905433654785,
      "eval_objects365-local-v2-validation_runtime": 9.9304,
      "eval_objects365-local-v2-validation_samples_per_second": 80.561,
      "eval_objects365-local-v2-validation_steps_per_second": 1.309,
      "step": 65000
    },
    {
      "_do_backward_in_ms": 673.5720948528033,
      "_prepare_inputs_in_ms": 6.18732621827092,
      "compute_loss_in_ms": 297.60415547758134,
      "epoch": 2.42,
      "learning_rate/full": 7.151371560919644e-05,
      "loss": 1.702,
      "step": 66000,
      "training_step_in_ms": 980.0054668276571
    },
    {
      "_do_backward_in_ms": 654.3511845318135,
      "_prepare_inputs_in_ms": 3.2777874602470547,
      "compute_loss_in_ms": 298.2891470948234,
      "epoch": 2.46,
      "learning_rate/full": 7.008393674839574e-05,
      "loss": 1.7017,
      "step": 67000,
      "training_step_in_ms": 961.3662255166564
    },
    {
      "_do_backward_in_ms": 674.1237894105725,
      "_prepare_inputs_in_ms": 3.2963655965868384,
      "compute_loss_in_ms": 298.1315166691784,
      "epoch": 2.5,
      "learning_rate/full": 6.863429772988044e-05,
      "loss": 1.7001,
      "step": 68000,
      "training_step_in_ms": 980.9758842557203
    },
    {
      "_do_backward_in_ms": 651.5303289373405,
      "_prepare_inputs_in_ms": 3.2912338944151998,
      "compute_loss_in_ms": 297.70729713048786,
      "epoch": 2.53,
      "learning_rate/full": 6.716475375716726e-05,
      "loss": 1.6992,
      "step": 69000,
      "training_step_in_ms": 957.9476293225307
    },
    {
      "_do_backward_in_ms": 701.2599963427056,
      "_prepare_inputs_in_ms": 3.2801425319630653,
      "compute_loss_in_ms": 297.9513049093075,
      "epoch": 2.57,
      "learning_rate/full": 6.568119138852548e-05,
      "loss": 1.6991,
      "step": 70000,
      "training_step_in_ms": 1007.9343056466896
    },
    {
      "epoch": 2.57,
      "eval_objects365-local-v2-validation_loss": 2.0831971168518066,
      "eval_objects365-local-v2-validation_runtime": 13.482,
      "eval_objects365-local-v2-validation_samples_per_second": 59.338,
      "eval_objects365-local-v2-validation_steps_per_second": 0.964,
      "step": 70000
    },
    {
      "_do_backward_in_ms": 684.7173160158563,
      "_prepare_inputs_in_ms": 9.682496656596602,
      "compute_loss_in_ms": 298.04157538694375,
      "epoch": 2.61,
      "learning_rate/full": 6.41791349446638e-05,
      "loss": 1.6959,
      "step": 71000,
      "training_step_in_ms": 991.6520508083049
    },
    {
      "_do_backward_in_ms": 719.4974615401588,
      "_prepare_inputs_in_ms": 3.2725699762813747,
      "compute_loss_in_ms": 298.8009250371251,
      "epoch": 2.64,
      "learning_rate/full": 6.266455189473463e-05,
      "loss": 1.6968,
      "step": 72000,
      "training_step_in_ms": 1027.078579184832
    },
    {
      "_do_backward_in_ms": 630.4531340918038,
      "_prepare_inputs_in_ms": 3.2720213141292334,
      "compute_loss_in_ms": 298.3485946042929,
      "epoch": 2.68,
      "learning_rate/full": 6.11374454037508e-05,
      "loss": 1.6951,
      "step": 73000,
      "training_step_in_ms": 937.6128895445727
    },
    {
      "_do_backward_in_ms": 673.5870531778783,
      "_prepare_inputs_in_ms": 3.220968232722953,
      "compute_loss_in_ms": 298.4312469626311,
      "epoch": 2.72,
      "learning_rate/full": 5.9599325562893006e-05,
      "loss": 1.6948,
      "step": 74000,
      "training_step_in_ms": 980.7399763246067
    },
    {
      "_do_backward_in_ms": 670.7334653646685,
      "_prepare_inputs_in_ms": 3.225922678830102,
      "compute_loss_in_ms": 298.11172814434394,
      "epoch": 2.75,
      "learning_rate/full": 5.8050159947610774e-05,
      "loss": 1.6937,
      "step": 75000,
      "training_step_in_ms": 977.5889939961489
    },
    {
      "epoch": 2.75,
      "eval_objects365-local-v2-validation_loss": 2.060175895690918,
      "eval_objects365-local-v2-validation_runtime": 10.0939,
      "eval_objects365-local-v2-validation_samples_per_second": 79.255,
      "eval_objects365-local-v2-validation_steps_per_second": 1.288,
      "step": 75000
    },
    {
      "_do_backward_in_ms": 707.8241047970951,
      "_prepare_inputs_in_ms": 7.805172383402789,
      "compute_loss_in_ms": 298.5676535603343,
      "epoch": 2.79,
      "learning_rate/full": 5.649457853856564e-05,
      "loss": 1.692,
      "step": 76000,
      "training_step_in_ms": 1015.2529031389859
    },
    {
      "_do_backward_in_ms": 679.5182852572761,
      "_prepare_inputs_in_ms": 3.2479509462136775,
      "compute_loss_in_ms": 298.08279935712926,
      "epoch": 2.83,
      "learning_rate/full": 5.4932574914808247e-05,
      "loss": 1.6897,
      "step": 77000,
      "training_step_in_ms": 986.3572093644179
    },
    {
      "_do_backward_in_ms": 624.3718270168174,
      "_prepare_inputs_in_ms": 3.2478769938461483,
      "compute_loss_in_ms": 298.39406253327616,
      "epoch": 2.87,
      "learning_rate/full": 5.336569367582159e-05,
      "loss": 1.6903,
      "step": 78000,
      "training_step_in_ms": 931.5286971002351
    },
    {
      "_do_backward_in_ms": 699.6423877081834,
      "_prepare_inputs_in_ms": 3.2958002001978457,
      "compute_loss_in_ms": 298.47152298805304,
      "epoch": 2.9,
      "learning_rate/full": 5.179548424435664e-05,
      "loss": 1.6903,
      "step": 79000,
      "training_step_in_ms": 1006.8662925218232
    },
    {
      "_do_backward_in_ms": 721.8450685073622,
      "_prepare_inputs_in_ms": 3.2630610479973257,
      "compute_loss_in_ms": 298.69656310253777,
      "epoch": 2.94,
      "learning_rate/full": 5.0223499334273436e-05,
      "loss": 1.6861,
      "step": 80000,
      "training_step_in_ms": 1029.3176082074642
    },
    {
      "epoch": 2.94,
      "eval_objects365-local-v2-validation_loss": 2.0467376708984375,
      "eval_objects365-local-v2-validation_runtime": 12.3724,
      "eval_objects365-local-v2-validation_samples_per_second": 64.66,
      "eval_objects365-local-v2-validation_steps_per_second": 1.051,
      "step": 80000
    },
    {
      "_do_backward_in_ms": 745.3300231467001,
      "_prepare_inputs_in_ms": 8.417346246764874,
      "compute_loss_in_ms": 297.6633286998154,
      "epoch": 2.98,
      "learning_rate/full": 4.8651293415127954e-05,
      "loss": 1.6859,
      "step": 81000,
      "training_step_in_ms": 1051.9083714645822
    },
    {
      "_do_backward_in_ms": 713.2887438628823,
      "_prepare_inputs_in_ms": 3.309611749369651,
      "compute_loss_in_ms": 299.24693407770246,
      "epoch": 3.01,
      "learning_rate/full": 4.7080421175022657e-05,
      "loss": 1.6808,
      "step": 82000,
      "training_step_in_ms": 1021.3982793183532
    },
    {
      "_do_backward_in_ms": 683.561601414578,
      "_prepare_inputs_in_ms": 3.340075889835134,
      "compute_loss_in_ms": 299.84076196700335,
      "epoch": 3.05,
      "learning_rate/full": 4.551243598324116e-05,
      "loss": 1.6701,
      "step": 83000,
      "training_step_in_ms": 992.2671003735159
    },
    {
      "_do_backward_in_ms": 691.1516734741163,
      "_prepare_inputs_in_ms": 3.3659417459275573,
      "compute_loss_in_ms": 298.7288506310433,
      "epoch": 3.09,
      "learning_rate/full": 4.394888835418673e-05,
      "loss": 1.6682,
      "step": 84000,
      "training_step_in_ms": 998.8259057179093
    },
    {
      "_do_backward_in_ms": 671.9453688920476,
      "_prepare_inputs_in_ms": 3.3872235738672316,
      "compute_loss_in_ms": 299.7595350977499,
      "epoch": 3.12,
      "learning_rate/full": 4.238976880417727e-05,
      "loss": 1.6682,
      "step": 85000,
      "training_step_in_ms": 980.6325940783136
    },
    {
      "epoch": 3.12,
      "eval_objects365-local-v2-validation_loss": 2.037609815597534,
      "eval_objects365-local-v2-validation_runtime": 8.3572,
      "eval_objects365-local-v2-validation_samples_per_second": 95.726,
      "eval_objects365-local-v2-validation_steps_per_second": 1.556,
      "step": 85000
    },
    {
      "_do_backward_in_ms": 682.5501747601666,
      "_prepare_inputs_in_ms": 6.052993975997133,
      "compute_loss_in_ms": 298.15280754882235,
      "epoch": 3.16,
      "learning_rate/full": 4.083818976295859e-05,
      "loss": 1.6684,
      "step": 86000,
      "training_step_in_ms": 989.5919032730162
    },
    {
      "_do_backward_in_ms": 702.3105279654264,
      "_prepare_inputs_in_ms": 3.27472277241759,
      "compute_loss_in_ms": 298.05750323599204,
      "epoch": 3.2,
      "learning_rate/full": 3.929722605177466e-05,
      "loss": 1.6664,
      "step": 87000,
      "training_step_in_ms": 1009.1526904681232
    },
    {
      "_do_backward_in_ms": 647.5646377876401,
      "_prepare_inputs_in_ms": 3.2817639869172126,
      "compute_loss_in_ms": 299.0744258032646,
      "epoch": 3.23,
      "learning_rate/full": 3.776684586220099e-05,
      "loss": 1.6671,
      "step": 88000,
      "training_step_in_ms": 955.4352292607073
    },
    {
      "_do_backward_in_ms": 1546.6698298668489,
      "_prepare_inputs_in_ms": 3.2690847222693264,
      "compute_loss_in_ms": 298.1887877949048,
      "epoch": 3.27,
      "learning_rate/full": 3.6248562522640714e-05,
      "loss": 1.6665,
      "step": 89000,
      "training_step_in_ms": 1853.662267722888
    },
    {
      "_do_backward_in_ms": 1196.733351110248,
      "_prepare_inputs_in_ms": 3.2732989322394133,
      "compute_loss_in_ms": 298.05953590921126,
      "epoch": 3.31,
      "learning_rate/full": 3.4743877399432354e-05,
      "loss": 1.6673,
      "step": 90000,
      "training_step_in_ms": 1503.5614393500146
    },
    {
      "epoch": 3.31,
      "eval_objects365-local-v2-validation_loss": 2.015079975128174,
      "eval_objects365-local-v2-validation_runtime": 21.4395,
      "eval_objects365-local-v2-validation_samples_per_second": 37.314,
      "eval_objects365-local-v2-validation_steps_per_second": 0.606,
      "step": 90000
    },
    {
      "_do_backward_in_ms": 844.9975665507372,
      "_prepare_inputs_in_ms": 19.031029790805363,
      "compute_loss_in_ms": 298.18785645280065,
      "epoch": 3.34,
      "learning_rate/full": 3.325427841221202e-05,
      "loss": 1.6653,
      "step": 91000,
      "training_step_in_ms": 1152.0402662665583
    },
    {
      "_do_backward_in_ms": 1347.8918932015076,
      "_prepare_inputs_in_ms": 3.3262957674451172,
      "compute_loss_in_ms": 298.0921304386575,
      "epoch": 3.38,
      "learning_rate/full": 3.178123856257234e-05,
      "loss": 1.6655,
      "step": 92000,
      "training_step_in_ms": 1654.7804647334851
    },
    {
      "_do_backward_in_ms": 1063.0321979811415,
      "_prepare_inputs_in_ms": 3.2982348231598735,
      "compute_loss_in_ms": 298.2737292505335,
      "epoch": 3.42,
      "learning_rate/full": 3.0326214477473512e-05,
      "loss": 1.6623,
      "step": 93000,
      "training_step_in_ms": 1370.0481948212255
    },
    {
      "_do_backward_in_ms": 675.3600625551771,
      "_prepare_inputs_in_ms": 3.2861924229655415,
      "compute_loss_in_ms": 297.8690059813671,
      "epoch": 3.45,
      "learning_rate/full": 2.8890644968846193e-05,
      "loss": 1.6633,
      "step": 94000,
      "training_step_in_ms": 981.9639976194594
    },
    {
      "_do_backward_in_ms": 685.7196875785012,
      "_prepare_inputs_in_ms": 3.2590834801085293,
      "compute_loss_in_ms": 298.93962507206015,
      "epoch": 3.49,
      "learning_rate/full": 2.74745444265936e-05,
      "loss": 1.6637,
      "step": 95000,
      "training_step_in_ms": 993.3962683330756
    },
    {
      "epoch": 3.49,
      "eval_objects365-local-v2-validation_loss": 2.0223588943481445,
      "eval_objects365-local-v2-validation_runtime": 13.6125,
      "eval_objects365-local-v2-validation_samples_per_second": 58.769,
      "eval_objects365-local-v2-validation_steps_per_second": 0.955,
      "step": 95000
    },
    {
      "_do_backward_in_ms": 663.6111688169185,
      "_prepare_inputs_in_ms": 11.124133992699312,
      "compute_loss_in_ms": 298.8600466307988,
      "epoch": 3.53,
      "learning_rate/full": 2.608214514070504e-05,
      "loss": 1.6625,
      "step": 96000,
      "training_step_in_ms": 971.3277539745905
    },
    {
      "_do_backward_in_ms": 862.6617220155895,
      "_prepare_inputs_in_ms": 3.263724277028814,
      "compute_loss_in_ms": 298.15222160494886,
      "epoch": 3.56,
      "learning_rate/full": 2.471339721239901e-05,
      "loss": 1.6617,
      "step": 97000,
      "training_step_in_ms": 1169.5883399837185
    },
    {
      "_do_backward_in_ms": 792.2627358706668,
      "_prepare_inputs_in_ms": 3.239450325258076,
      "compute_loss_in_ms": 298.1890718040522,
      "epoch": 3.6,
      "learning_rate/full": 2.3370986279672524e-05,
      "loss": 1.6611,
      "step": 98000,
      "training_step_in_ms": 1099.1772187727038
    },
    {
      "_do_backward_in_ms": 821.3568878679071,
      "_prepare_inputs_in_ms": 3.2639728772919625,
      "compute_loss_in_ms": 298.9591815781314,
      "epoch": 3.64,
      "learning_rate/full": 2.2053549815720452e-05,
      "loss": 1.6597,
      "step": 99000,
      "training_step_in_ms": 1129.0956585616805
    },
    {
      "_do_backward_in_ms": 841.5605684022885,
      "_prepare_inputs_in_ms": 3.277285093674436,
      "compute_loss_in_ms": 298.11495484854095,
      "epoch": 3.67,
      "learning_rate/full": 2.0762471598573356e-05,
      "loss": 1.6591,
      "step": 100000,
      "training_step_in_ms": 1148.4819840774871
    },
    {
      "epoch": 3.67,
      "eval_objects365-local-v2-validation_loss": 2.003035545349121,
      "eval_objects365-local-v2-validation_runtime": 15.3967,
      "eval_objects365-local-v2-validation_samples_per_second": 51.959,
      "eval_objects365-local-v2-validation_steps_per_second": 0.844,
      "step": 100000
    }
  ],
  "max_steps": 100000,
  "num_train_epochs": 4,
  "total_flos": 1.6535680982674692e+23,
  "trial_name": null,
  "trial_params": null
}