hmBERT-CoNLL-cp3 / trainer_state.json
emilys's picture
End of training
95064b2
raw
history blame
20 kB
{
"best_metric": 0.9181643400484828,
"best_model_checkpoint": "hmBERT-CoNLL-cp3/checkpoint-1000",
"epoch": 3.0,
"global_step": 1317,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.06,
"eval_accuracy": 0.9007242708617266,
"eval_f1": 0.3684910996506405,
"eval_loss": 0.4115295708179474,
"eval_precision": 0.36430921052631576,
"eval_recall": 0.37277011107371255,
"eval_runtime": 21.3063,
"eval_samples_per_second": 152.537,
"eval_steps_per_second": 4.787,
"step": 25
},
{
"epoch": 0.11,
"eval_accuracy": 0.9460301390132783,
"eval_f1": 0.6640782981476988,
"eval_loss": 0.22425369918346405,
"eval_precision": 0.6393085189222862,
"eval_recall": 0.6908448333894311,
"eval_runtime": 21.4481,
"eval_samples_per_second": 151.529,
"eval_steps_per_second": 4.756,
"step": 50
},
{
"epoch": 0.17,
"eval_accuracy": 0.958023441454772,
"eval_f1": 0.7474880579805633,
"eval_loss": 0.16170601546764374,
"eval_precision": 0.7319354838709677,
"eval_recall": 0.7637159205654662,
"eval_runtime": 21.6197,
"eval_samples_per_second": 150.326,
"eval_steps_per_second": 4.718,
"step": 75
},
{
"epoch": 0.23,
"eval_accuracy": 0.9585101826252872,
"eval_f1": 0.7455232462625266,
"eval_loss": 0.15437664091587067,
"eval_precision": 0.7281771501925546,
"eval_recall": 0.7637159205654662,
"eval_runtime": 21.768,
"eval_samples_per_second": 149.302,
"eval_steps_per_second": 4.686,
"step": 100
},
{
"epoch": 0.28,
"eval_accuracy": 0.964409485611931,
"eval_f1": 0.784738041002278,
"eval_loss": 0.13407838344573975,
"eval_precision": 0.7595275590551182,
"eval_recall": 0.81167956916863,
"eval_runtime": 21.8222,
"eval_samples_per_second": 148.931,
"eval_steps_per_second": 4.674,
"step": 125
},
{
"epoch": 0.34,
"eval_accuracy": 0.9692963669639033,
"eval_f1": 0.811351977494622,
"eval_loss": 0.12208627909421921,
"eval_precision": 0.7980143229166666,
"eval_recall": 0.8251430494782901,
"eval_runtime": 21.2673,
"eval_samples_per_second": 152.817,
"eval_steps_per_second": 4.796,
"step": 150
},
{
"epoch": 0.4,
"eval_accuracy": 0.9718663603442234,
"eval_f1": 0.8151923709306149,
"eval_loss": 0.10127394646406174,
"eval_precision": 0.796849887495982,
"eval_recall": 0.8343991921911814,
"eval_runtime": 21.1083,
"eval_samples_per_second": 153.968,
"eval_steps_per_second": 4.832,
"step": 175
},
{
"epoch": 0.46,
"eval_accuracy": 0.9731513570343834,
"eval_f1": 0.8333472419260619,
"eval_loss": 0.10758877545595169,
"eval_precision": 0.8265187882800861,
"eval_recall": 0.8402894648266577,
"eval_runtime": 20.8538,
"eval_samples_per_second": 155.847,
"eval_steps_per_second": 4.891,
"step": 200
},
{
"epoch": 0.51,
"eval_accuracy": 0.9762859701725011,
"eval_f1": 0.8543123543123543,
"eval_loss": 0.08832722157239914,
"eval_precision": 0.8453047775947282,
"eval_recall": 0.8635139683608213,
"eval_runtime": 21.5789,
"eval_samples_per_second": 150.61,
"eval_steps_per_second": 4.727,
"step": 225
},
{
"epoch": 0.57,
"eval_accuracy": 0.9762665005256804,
"eval_f1": 0.8535063638632394,
"eval_loss": 0.0972680076956749,
"eval_precision": 0.8438887974995888,
"eval_recall": 0.8633456748569506,
"eval_runtime": 21.3825,
"eval_samples_per_second": 151.993,
"eval_steps_per_second": 4.77,
"step": 250
},
{
"epoch": 0.63,
"eval_accuracy": 0.9765390755811689,
"eval_f1": 0.8575239683201333,
"eval_loss": 0.0882844626903534,
"eval_precision": 0.8496613249628283,
"eval_recall": 0.8655334904072702,
"eval_runtime": 21.4561,
"eval_samples_per_second": 151.472,
"eval_steps_per_second": 4.754,
"step": 275
},
{
"epoch": 0.68,
"eval_accuracy": 0.9766364238152719,
"eval_f1": 0.8551207327227309,
"eval_loss": 0.08785738795995712,
"eval_precision": 0.8462425840474621,
"eval_recall": 0.8641871423763042,
"eval_runtime": 21.5171,
"eval_samples_per_second": 151.043,
"eval_steps_per_second": 4.74,
"step": 300
},
{
"epoch": 0.74,
"eval_accuracy": 0.9786612670846151,
"eval_f1": 0.8711310264708322,
"eval_loss": 0.07805178314447403,
"eval_precision": 0.8592240955966607,
"eval_recall": 0.8833726018175698,
"eval_runtime": 21.5676,
"eval_samples_per_second": 150.689,
"eval_steps_per_second": 4.729,
"step": 325
},
{
"epoch": 0.8,
"eval_accuracy": 0.9803161870643666,
"eval_f1": 0.8810828765985717,
"eval_loss": 0.07252655178308487,
"eval_precision": 0.8696721311475409,
"eval_recall": 0.8927970380343319,
"eval_runtime": 21.8654,
"eval_samples_per_second": 148.637,
"eval_steps_per_second": 4.665,
"step": 350
},
{
"epoch": 0.85,
"eval_accuracy": 0.9807250496475994,
"eval_f1": 0.881333443900821,
"eval_loss": 0.07547631114721298,
"eval_precision": 0.8687264999182606,
"eval_recall": 0.8943116795691687,
"eval_runtime": 21.5439,
"eval_samples_per_second": 150.855,
"eval_steps_per_second": 4.735,
"step": 375
},
{
"epoch": 0.91,
"eval_accuracy": 0.9822242124527861,
"eval_f1": 0.8890735355213959,
"eval_loss": 0.06660941988229752,
"eval_precision": 0.8780567864762843,
"eval_recall": 0.9003702457085156,
"eval_runtime": 21.2792,
"eval_samples_per_second": 152.731,
"eval_steps_per_second": 4.793,
"step": 400
},
{
"epoch": 0.97,
"eval_accuracy": 0.982282621393248,
"eval_f1": 0.8935885647412856,
"eval_loss": 0.06582659482955933,
"eval_precision": 0.8877262913137353,
"eval_recall": 0.8995287781891619,
"eval_runtime": 21.0258,
"eval_samples_per_second": 154.572,
"eval_steps_per_second": 4.851,
"step": 425
},
{
"epoch": 1.03,
"eval_accuracy": 0.983664966317511,
"eval_f1": 0.8993299832495812,
"eval_loss": 0.0645442008972168,
"eval_precision": 0.89513171057019,
"eval_recall": 0.9035678222820599,
"eval_runtime": 21.8717,
"eval_samples_per_second": 148.594,
"eval_steps_per_second": 4.664,
"step": 450
},
{
"epoch": 1.08,
"eval_accuracy": 0.9831003465597135,
"eval_f1": 0.8950920756603616,
"eval_loss": 0.06972972303628922,
"eval_precision": 0.8864499092259449,
"eval_recall": 0.9039044092898014,
"eval_runtime": 22.0283,
"eval_samples_per_second": 147.537,
"eval_steps_per_second": 4.63,
"step": 475
},
{
"epoch": 1.14,
"learning_rate": 3.1017463933181475e-05,
"loss": 0.1392,
"step": 500
},
{
"epoch": 1.14,
"eval_accuracy": 0.9824189089209922,
"eval_f1": 0.8908151358559759,
"eval_loss": 0.06883182376623154,
"eval_precision": 0.8824306472919419,
"eval_recall": 0.8993604846852912,
"eval_runtime": 21.1436,
"eval_samples_per_second": 153.711,
"eval_steps_per_second": 4.824,
"step": 500
},
{
"epoch": 1.2,
"eval_accuracy": 0.9827498929169425,
"eval_f1": 0.8999163179916319,
"eval_loss": 0.06805375963449478,
"eval_precision": 0.8949733688415447,
"eval_recall": 0.9049141703130259,
"eval_runtime": 21.618,
"eval_samples_per_second": 150.337,
"eval_steps_per_second": 4.718,
"step": 525
},
{
"epoch": 1.25,
"eval_accuracy": 0.9823215606868891,
"eval_f1": 0.8915259903058667,
"eval_loss": 0.06759364902973175,
"eval_precision": 0.8854581673306773,
"eval_recall": 0.8976775496465836,
"eval_runtime": 21.6174,
"eval_samples_per_second": 150.342,
"eval_steps_per_second": 4.718,
"step": 550
},
{
"epoch": 1.31,
"eval_accuracy": 0.9841906467816673,
"eval_f1": 0.9013520280420632,
"eval_loss": 0.06176742911338806,
"eval_precision": 0.8940397350993378,
"eval_recall": 0.9087849209020532,
"eval_runtime": 21.5727,
"eval_samples_per_second": 150.653,
"eval_steps_per_second": 4.728,
"step": 575
},
{
"epoch": 1.37,
"eval_accuracy": 0.984015419960282,
"eval_f1": 0.9010107760421018,
"eval_loss": 0.0644294023513794,
"eval_precision": 0.8945098689666612,
"eval_recall": 0.9076068663749579,
"eval_runtime": 22.0692,
"eval_samples_per_second": 147.264,
"eval_steps_per_second": 4.622,
"step": 600
},
{
"epoch": 1.42,
"eval_accuracy": 0.9837233752579728,
"eval_f1": 0.9010347129506008,
"eval_loss": 0.0641368106007576,
"eval_precision": 0.8935782853359815,
"eval_recall": 0.9086166273981824,
"eval_runtime": 21.6306,
"eval_samples_per_second": 150.25,
"eval_steps_per_second": 4.716,
"step": 625
},
{
"epoch": 1.48,
"eval_accuracy": 0.9845800397180795,
"eval_f1": 0.9043724966622163,
"eval_loss": 0.06189671903848648,
"eval_precision": 0.8968884475339292,
"eval_recall": 0.9119824974755975,
"eval_runtime": 20.932,
"eval_samples_per_second": 155.265,
"eval_steps_per_second": 4.873,
"step": 650
},
{
"epoch": 1.54,
"eval_accuracy": 0.9848136754799268,
"eval_f1": 0.9074897257401661,
"eval_loss": 0.06076198071241379,
"eval_precision": 0.9045310148804547,
"eval_recall": 0.9104678559407607,
"eval_runtime": 21.5144,
"eval_samples_per_second": 151.062,
"eval_steps_per_second": 4.741,
"step": 675
},
{
"epoch": 1.59,
"eval_accuracy": 0.9851057201822359,
"eval_f1": 0.9090604869070525,
"eval_loss": 0.06238234415650368,
"eval_precision": 0.903842954583264,
"eval_recall": 0.9143386065297879,
"eval_runtime": 21.3845,
"eval_samples_per_second": 151.979,
"eval_steps_per_second": 4.77,
"step": 700
},
{
"epoch": 1.65,
"eval_accuracy": 0.9852420077099802,
"eval_f1": 0.9115851108322878,
"eval_loss": 0.05955210700631142,
"eval_precision": 0.9062032263429237,
"eval_recall": 0.9170313025917199,
"eval_runtime": 21.1164,
"eval_samples_per_second": 153.909,
"eval_steps_per_second": 4.83,
"step": 725
},
{
"epoch": 1.71,
"eval_accuracy": 0.9848331451267474,
"eval_f1": 0.9068602904356534,
"eval_loss": 0.0579909048974514,
"eval_precision": 0.8995033112582781,
"eval_recall": 0.9143386065297879,
"eval_runtime": 21.1108,
"eval_samples_per_second": 153.95,
"eval_steps_per_second": 4.832,
"step": 750
},
{
"epoch": 1.77,
"eval_accuracy": 0.9857676881741365,
"eval_f1": 0.9126685087498954,
"eval_loss": 0.05822019651532173,
"eval_precision": 0.9081819696717214,
"eval_recall": 0.9171995960955908,
"eval_runtime": 22.2525,
"eval_samples_per_second": 146.051,
"eval_steps_per_second": 4.584,
"step": 775
},
{
"epoch": 1.82,
"eval_accuracy": 0.9852420077099802,
"eval_f1": 0.9100617386951443,
"eval_loss": 0.0588374100625515,
"eval_precision": 0.9023825281270682,
"eval_recall": 0.9178727701110737,
"eval_runtime": 21.7438,
"eval_samples_per_second": 149.468,
"eval_steps_per_second": 4.691,
"step": 800
},
{
"epoch": 1.88,
"eval_accuracy": 0.9855535220591098,
"eval_f1": 0.911860174781523,
"eval_loss": 0.0592406764626503,
"eval_precision": 0.9020253581425984,
"eval_recall": 0.9219118142039717,
"eval_runtime": 21.1485,
"eval_samples_per_second": 153.675,
"eval_steps_per_second": 4.823,
"step": 825
},
{
"epoch": 1.94,
"eval_accuracy": 0.9852225380631595,
"eval_f1": 0.911764705882353,
"eval_loss": 0.06002892181277275,
"eval_precision": 0.9054098904746101,
"eval_recall": 0.9182093571188152,
"eval_runtime": 22.0668,
"eval_samples_per_second": 147.28,
"eval_steps_per_second": 4.622,
"step": 850
},
{
"epoch": 1.99,
"eval_accuracy": 0.9860792025232662,
"eval_f1": 0.913464751085867,
"eval_loss": 0.05682874843478203,
"eval_precision": 0.9067993366500829,
"eval_recall": 0.9202288791652642,
"eval_runtime": 20.9827,
"eval_samples_per_second": 154.89,
"eval_steps_per_second": 4.861,
"step": 875
},
{
"epoch": 2.05,
"eval_accuracy": 0.9860792025232662,
"eval_f1": 0.9171483622350675,
"eval_loss": 0.0570731945335865,
"eval_precision": 0.9130942452043369,
"eval_recall": 0.9212386401884888,
"eval_runtime": 21.3364,
"eval_samples_per_second": 152.322,
"eval_steps_per_second": 4.781,
"step": 900
},
{
"epoch": 2.11,
"eval_accuracy": 0.9857676881741365,
"eval_f1": 0.9156969443281708,
"eval_loss": 0.0577477402985096,
"eval_precision": 0.9110444777611194,
"eval_recall": 0.920397172669135,
"eval_runtime": 21.9855,
"eval_samples_per_second": 147.825,
"eval_steps_per_second": 4.639,
"step": 925
},
{
"epoch": 2.16,
"eval_accuracy": 0.986040263229625,
"eval_f1": 0.9184714441006773,
"eval_loss": 0.06048833578824997,
"eval_precision": 0.912747216220708,
"eval_recall": 0.9242679232581622,
"eval_runtime": 21.8831,
"eval_samples_per_second": 148.516,
"eval_steps_per_second": 4.661,
"step": 950
},
{
"epoch": 2.22,
"eval_accuracy": 0.986682761574705,
"eval_f1": 0.9166318254034618,
"eval_loss": 0.05745207890868187,
"eval_precision": 0.9109190626558086,
"eval_recall": 0.922416694715584,
"eval_runtime": 21.8724,
"eval_samples_per_second": 148.589,
"eval_steps_per_second": 4.663,
"step": 975
},
{
"epoch": 2.28,
"learning_rate": 1.2034927866362947e-05,
"loss": 0.0392,
"step": 1000
},
{
"epoch": 2.28,
"eval_accuracy": 0.9862154900510105,
"eval_f1": 0.9181643400484828,
"eval_loss": 0.0572221502661705,
"eval_precision": 0.9121408403919614,
"eval_recall": 0.9242679232581622,
"eval_runtime": 21.5878,
"eval_samples_per_second": 150.548,
"eval_steps_per_second": 4.725,
"step": 1000
},
{
"epoch": 2.33,
"eval_accuracy": 0.9869748062770142,
"eval_f1": 0.9211694730669348,
"eval_loss": 0.056650131940841675,
"eval_precision": 0.9170975813177648,
"eval_recall": 0.9252776842813868,
"eval_runtime": 21.5295,
"eval_samples_per_second": 150.956,
"eval_steps_per_second": 4.738,
"step": 1025
},
{
"epoch": 2.39,
"eval_accuracy": 0.9870916241579378,
"eval_f1": 0.9243514644351465,
"eval_loss": 0.0570232979953289,
"eval_precision": 0.9192743009320905,
"eval_recall": 0.9294850218781555,
"eval_runtime": 21.5481,
"eval_samples_per_second": 150.825,
"eval_steps_per_second": 4.734,
"step": 1050
},
{
"epoch": 2.45,
"eval_accuracy": 0.986682761574705,
"eval_f1": 0.9215079829474212,
"eval_loss": 0.05837239325046539,
"eval_precision": 0.9154625477495433,
"eval_recall": 0.9276337933355773,
"eval_runtime": 21.2288,
"eval_samples_per_second": 153.094,
"eval_steps_per_second": 4.805,
"step": 1075
},
{
"epoch": 2.51,
"eval_accuracy": 0.9867022312215257,
"eval_f1": 0.9226653289858708,
"eval_loss": 0.059072330594062805,
"eval_precision": 0.9167635819903639,
"eval_recall": 0.9286435543588017,
"eval_runtime": 22.3224,
"eval_samples_per_second": 145.594,
"eval_steps_per_second": 4.569,
"step": 1100
},
{
"epoch": 2.56,
"eval_accuracy": 0.9874031385070675,
"eval_f1": 0.9246323529411764,
"eval_loss": 0.057747144252061844,
"eval_precision": 0.9181878526385662,
"eval_recall": 0.931167956916863,
"eval_runtime": 21.9739,
"eval_samples_per_second": 147.903,
"eval_steps_per_second": 4.642,
"step": 1125
},
{
"epoch": 2.62,
"eval_accuracy": 0.987033215217476,
"eval_f1": 0.9233344492802142,
"eval_loss": 0.057037804275751114,
"eval_precision": 0.9184149184149184,
"eval_recall": 0.9283069673510602,
"eval_runtime": 21.3838,
"eval_samples_per_second": 151.984,
"eval_steps_per_second": 4.77,
"step": 1150
},
{
"epoch": 2.68,
"eval_accuracy": 0.987227911685682,
"eval_f1": 0.9244541119384256,
"eval_loss": 0.05629764869809151,
"eval_precision": 0.9191482282482116,
"eval_recall": 0.929821608885897,
"eval_runtime": 21.8946,
"eval_samples_per_second": 148.438,
"eval_steps_per_second": 4.659,
"step": 1175
},
{
"epoch": 2.73,
"eval_accuracy": 0.9871500330983995,
"eval_f1": 0.9246449456975774,
"eval_loss": 0.05648628994822502,
"eval_precision": 0.918049104180491,
"eval_recall": 0.9313362504207338,
"eval_runtime": 21.2748,
"eval_samples_per_second": 152.763,
"eval_steps_per_second": 4.794,
"step": 1200
},
{
"epoch": 2.79,
"eval_accuracy": 0.9873447295666057,
"eval_f1": 0.9243767776476494,
"eval_loss": 0.055928729474544525,
"eval_precision": 0.9189953426480373,
"eval_recall": 0.929821608885897,
"eval_runtime": 21.5883,
"eval_samples_per_second": 150.545,
"eval_steps_per_second": 4.725,
"step": 1225
},
{
"epoch": 2.85,
"eval_accuracy": 0.9873057902729645,
"eval_f1": 0.9238748536054878,
"eval_loss": 0.05620228126645088,
"eval_precision": 0.9184963406520292,
"eval_recall": 0.9293167283742848,
"eval_runtime": 21.3558,
"eval_samples_per_second": 152.184,
"eval_steps_per_second": 4.776,
"step": 1250
},
{
"epoch": 2.9,
"eval_accuracy": 0.9871500330983995,
"eval_f1": 0.922961104140527,
"eval_loss": 0.05639491230249405,
"eval_precision": 0.9175120572093797,
"eval_recall": 0.928475260854931,
"eval_runtime": 21.3856,
"eval_samples_per_second": 151.972,
"eval_steps_per_second": 4.77,
"step": 1275
},
{
"epoch": 2.96,
"eval_accuracy": 0.9873057902729645,
"eval_f1": 0.9237330657300552,
"eval_loss": 0.05629625543951988,
"eval_precision": 0.9180518617021277,
"eval_recall": 0.9294850218781555,
"eval_runtime": 21.6582,
"eval_samples_per_second": 150.059,
"eval_steps_per_second": 4.71,
"step": 1300
},
{
"epoch": 3.0,
"step": 1317,
"total_flos": 1368370122693444.0,
"train_loss": 0.07309725869308391,
"train_runtime": 1743.2547,
"train_samples_per_second": 24.163,
"train_steps_per_second": 0.755
}
],
"max_steps": 1317,
"num_train_epochs": 3,
"total_flos": 1368370122693444.0,
"trial_name": null,
"trial_params": null
}