| { | |
| "best_metric": 0.8968406938741306, | |
| "best_model_checkpoint": "hmBERT-CoNLL-cp2/checkpoint-500", | |
| "epoch": 2.0, | |
| "global_step": 878, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "eval_accuracy": 0.9005295743935205, | |
| "eval_f1": 0.36743180306054557, | |
| "eval_loss": 0.4116312265396118, | |
| "eval_precision": 0.36320289378493914, | |
| "eval_recall": 0.3717603500504881, | |
| "eval_runtime": 21.4683, | |
| "eval_samples_per_second": 151.386, | |
| "eval_steps_per_second": 4.751, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_accuracy": 0.9458549121918929, | |
| "eval_f1": 0.6632702571567201, | |
| "eval_loss": 0.22468078136444092, | |
| "eval_precision": 0.638387297633873, | |
| "eval_recall": 0.6901716593739482, | |
| "eval_runtime": 20.8937, | |
| "eval_samples_per_second": 155.549, | |
| "eval_steps_per_second": 4.882, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_accuracy": 0.9579845021611308, | |
| "eval_f1": 0.746131050378663, | |
| "eval_loss": 0.16239947080612183, | |
| "eval_precision": 0.7302610377054464, | |
| "eval_recall": 0.7627061595422416, | |
| "eval_runtime": 21.6024, | |
| "eval_samples_per_second": 150.446, | |
| "eval_steps_per_second": 4.722, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_accuracy": 0.9588022273275962, | |
| "eval_f1": 0.7508835374373305, | |
| "eval_loss": 0.15414278209209442, | |
| "eval_precision": 0.7338152610441767, | |
| "eval_recall": 0.7687647256815887, | |
| "eval_runtime": 21.4049, | |
| "eval_samples_per_second": 151.835, | |
| "eval_steps_per_second": 4.765, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_accuracy": 0.9642537284373661, | |
| "eval_f1": 0.7844736198320149, | |
| "eval_loss": 0.13490867614746094, | |
| "eval_precision": 0.7609555450087011, | |
| "eval_recall": 0.8094917536183104, | |
| "eval_runtime": 21.4103, | |
| "eval_samples_per_second": 151.796, | |
| "eval_steps_per_second": 4.764, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_accuracy": 0.9693547759043651, | |
| "eval_f1": 0.8115174582161179, | |
| "eval_loss": 0.12301833927631378, | |
| "eval_precision": 0.7981770833333334, | |
| "eval_recall": 0.8253113429821609, | |
| "eval_runtime": 21.3799, | |
| "eval_samples_per_second": 152.012, | |
| "eval_steps_per_second": 4.771, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_accuracy": 0.97272302480433, | |
| "eval_f1": 0.8234421364985163, | |
| "eval_loss": 0.0997077226638794, | |
| "eval_precision": 0.8069466882067852, | |
| "eval_recall": 0.8406260518343992, | |
| "eval_runtime": 21.0809, | |
| "eval_samples_per_second": 154.168, | |
| "eval_steps_per_second": 4.839, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_accuracy": 0.9731902963280246, | |
| "eval_f1": 0.8308945793149318, | |
| "eval_loss": 0.10443305224180222, | |
| "eval_precision": 0.8210647387446599, | |
| "eval_recall": 0.8409626388421407, | |
| "eval_runtime": 21.7924, | |
| "eval_samples_per_second": 149.135, | |
| "eval_steps_per_second": 4.681, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_accuracy": 0.9759939254701919, | |
| "eval_f1": 0.8507239141288069, | |
| "eval_loss": 0.08712451159954071, | |
| "eval_precision": 0.8413429888084266, | |
| "eval_recall": 0.860316391787277, | |
| "eval_runtime": 21.6266, | |
| "eval_samples_per_second": 150.278, | |
| "eval_steps_per_second": 4.716, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_accuracy": 0.9732876445621277, | |
| "eval_f1": 0.8375655648988427, | |
| "eval_loss": 0.10657607018947601, | |
| "eval_precision": 0.8288021090789257, | |
| "eval_recall": 0.8465163244698755, | |
| "eval_runtime": 21.4498, | |
| "eval_samples_per_second": 151.517, | |
| "eval_steps_per_second": 4.755, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_accuracy": 0.9765780148748102, | |
| "eval_f1": 0.8623576691225721, | |
| "eval_loss": 0.08719414472579956, | |
| "eval_precision": 0.8580473175608131, | |
| "eval_recall": 0.8667115449343655, | |
| "eval_runtime": 21.04, | |
| "eval_samples_per_second": 154.468, | |
| "eval_steps_per_second": 4.848, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_accuracy": 0.9773178614539932, | |
| "eval_f1": 0.8613053613053614, | |
| "eval_loss": 0.08342944085597992, | |
| "eval_precision": 0.8522240527182866, | |
| "eval_recall": 0.8705822955233928, | |
| "eval_runtime": 20.9406, | |
| "eval_samples_per_second": 155.201, | |
| "eval_steps_per_second": 4.871, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_accuracy": 0.9782718741482029, | |
| "eval_f1": 0.868680182043856, | |
| "eval_loss": 0.08315661549568176, | |
| "eval_precision": 0.8544685007325411, | |
| "eval_recall": 0.8833726018175698, | |
| "eval_runtime": 21.3907, | |
| "eval_samples_per_second": 151.935, | |
| "eval_steps_per_second": 4.768, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_accuracy": 0.9787002063782563, | |
| "eval_f1": 0.8685364441135103, | |
| "eval_loss": 0.07756651937961578, | |
| "eval_precision": 0.8541903986981285, | |
| "eval_recall": 0.8833726018175698, | |
| "eval_runtime": 21.9914, | |
| "eval_samples_per_second": 147.785, | |
| "eval_steps_per_second": 4.638, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_accuracy": 0.98010202094934, | |
| "eval_f1": 0.8760357971494862, | |
| "eval_loss": 0.07601791620254517, | |
| "eval_precision": 0.8628795298726738, | |
| "eval_recall": 0.8895994614607876, | |
| "eval_runtime": 21.6439, | |
| "eval_samples_per_second": 150.158, | |
| "eval_steps_per_second": 4.713, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_accuracy": 0.9823994392741716, | |
| "eval_f1": 0.8887781377190795, | |
| "eval_loss": 0.06729913502931595, | |
| "eval_precision": 0.8774807282269969, | |
| "eval_recall": 0.9003702457085156, | |
| "eval_runtime": 20.8932, | |
| "eval_samples_per_second": 155.553, | |
| "eval_steps_per_second": 4.882, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_accuracy": 0.9816595926949885, | |
| "eval_f1": 0.8882013546283134, | |
| "eval_loss": 0.06811415404081345, | |
| "eval_precision": 0.8826657802891806, | |
| "eval_recall": 0.8938067990575563, | |
| "eval_runtime": 21.516, | |
| "eval_samples_per_second": 151.051, | |
| "eval_steps_per_second": 4.741, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "eval_accuracy": 0.9824189089209922, | |
| "eval_f1": 0.8896695943120033, | |
| "eval_loss": 0.06585536897182465, | |
| "eval_precision": 0.8844170962913687, | |
| "eval_recall": 0.8949848535846516, | |
| "eval_runtime": 21.1661, | |
| "eval_samples_per_second": 153.547, | |
| "eval_steps_per_second": 4.819, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "eval_accuracy": 0.9831587555001752, | |
| "eval_f1": 0.8923128175231116, | |
| "eval_loss": 0.06903570890426636, | |
| "eval_precision": 0.8832646331409728, | |
| "eval_recall": 0.9015483002356109, | |
| "eval_runtime": 21.4553, | |
| "eval_samples_per_second": 151.478, | |
| "eval_steps_per_second": 4.754, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.152619589977221e-05, | |
| "loss": 0.1399, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "eval_accuracy": 0.983217164440637, | |
| "eval_f1": 0.8968406938741306, | |
| "eval_loss": 0.0666249617934227, | |
| "eval_precision": 0.8931730929727926, | |
| "eval_recall": 0.9005385392123864, | |
| "eval_runtime": 21.9318, | |
| "eval_samples_per_second": 148.187, | |
| "eval_steps_per_second": 4.651, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "eval_accuracy": 0.982477317861454, | |
| "eval_f1": 0.8943538268506901, | |
| "eval_loss": 0.06665532290935516, | |
| "eval_precision": 0.8890736737069682, | |
| "eval_recall": 0.8996970716930327, | |
| "eval_runtime": 21.7087, | |
| "eval_samples_per_second": 149.71, | |
| "eval_steps_per_second": 4.699, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "eval_accuracy": 0.9820100463377595, | |
| "eval_f1": 0.8851177106729889, | |
| "eval_loss": 0.0698634684085846, | |
| "eval_precision": 0.8751439381477216, | |
| "eval_recall": 0.8953214405923932, | |
| "eval_runtime": 21.8836, | |
| "eval_samples_per_second": 148.513, | |
| "eval_steps_per_second": 4.661, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "eval_accuracy": 0.984015419960282, | |
| "eval_f1": 0.900702106318957, | |
| "eval_loss": 0.06173388287425041, | |
| "eval_precision": 0.8947193623380937, | |
| "eval_recall": 0.9067653988556041, | |
| "eval_runtime": 21.5086, | |
| "eval_samples_per_second": 151.102, | |
| "eval_steps_per_second": 4.742, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "eval_accuracy": 0.9840543592539231, | |
| "eval_f1": 0.9028686462003019, | |
| "eval_loss": 0.06334252655506134, | |
| "eval_precision": 0.9, | |
| "eval_recall": 0.9057556378323797, | |
| "eval_runtime": 21.1745, | |
| "eval_samples_per_second": 153.487, | |
| "eval_steps_per_second": 4.817, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "eval_accuracy": 0.9843269343094116, | |
| "eval_f1": 0.9040387182910546, | |
| "eval_loss": 0.06390959024429321, | |
| "eval_precision": 0.8965574313141343, | |
| "eval_recall": 0.9116459104678559, | |
| "eval_runtime": 21.7996, | |
| "eval_samples_per_second": 149.086, | |
| "eval_steps_per_second": 4.679, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "eval_accuracy": 0.9845021611307971, | |
| "eval_f1": 0.9040501043841336, | |
| "eval_loss": 0.06244850531220436, | |
| "eval_precision": 0.8972318912647108, | |
| "eval_recall": 0.910972736452373, | |
| "eval_runtime": 21.6575, | |
| "eval_samples_per_second": 150.064, | |
| "eval_steps_per_second": 4.71, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "eval_accuracy": 0.9841517074880262, | |
| "eval_f1": 0.90302066772655, | |
| "eval_loss": 0.06191794201731682, | |
| "eval_precision": 0.8979863538026294, | |
| "eval_recall": 0.9081117468865701, | |
| "eval_runtime": 21.8896, | |
| "eval_samples_per_second": 148.473, | |
| "eval_steps_per_second": 4.66, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "eval_accuracy": 0.9842685253689498, | |
| "eval_f1": 0.904538603249037, | |
| "eval_loss": 0.06151856482028961, | |
| "eval_precision": 0.9001666666666667, | |
| "eval_recall": 0.9089532144059239, | |
| "eval_runtime": 21.9582, | |
| "eval_samples_per_second": 148.008, | |
| "eval_steps_per_second": 4.645, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "eval_accuracy": 0.9849889023013123, | |
| "eval_f1": 0.9082384460817148, | |
| "eval_loss": 0.060051579028367996, | |
| "eval_precision": 0.9036987670776407, | |
| "eval_recall": 0.9128239649949512, | |
| "eval_runtime": 21.5715, | |
| "eval_samples_per_second": 150.661, | |
| "eval_steps_per_second": 4.728, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_accuracy": 0.984949963007671, | |
| "eval_f1": 0.9085891109810152, | |
| "eval_loss": 0.058469709008932114, | |
| "eval_precision": 0.9030756442227764, | |
| "eval_recall": 0.9141703130259172, | |
| "eval_runtime": 21.259, | |
| "eval_samples_per_second": 152.877, | |
| "eval_steps_per_second": 4.798, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "eval_accuracy": 0.9850862505354153, | |
| "eval_f1": 0.9089084065244667, | |
| "eval_loss": 0.05824408307671547, | |
| "eval_precision": 0.903542324962581, | |
| "eval_recall": 0.9143386065297879, | |
| "eval_runtime": 21.7846, | |
| "eval_samples_per_second": 149.188, | |
| "eval_steps_per_second": 4.682, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "eval_accuracy": 0.9852809470036213, | |
| "eval_f1": 0.9100183977253721, | |
| "eval_loss": 0.0580282025039196, | |
| "eval_precision": 0.9044215425531915, | |
| "eval_recall": 0.9156849545607539, | |
| "eval_runtime": 21.4108, | |
| "eval_samples_per_second": 151.792, | |
| "eval_steps_per_second": 4.764, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "eval_accuracy": 0.9851446594758771, | |
| "eval_f1": 0.9096682543661736, | |
| "eval_loss": 0.058311279863119125, | |
| "eval_precision": 0.903402489626556, | |
| "eval_recall": 0.9160215415684955, | |
| "eval_runtime": 21.1305, | |
| "eval_samples_per_second": 153.806, | |
| "eval_steps_per_second": 4.827, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "eval_accuracy": 0.9853782952377244, | |
| "eval_f1": 0.9113564141160729, | |
| "eval_loss": 0.05776725709438324, | |
| "eval_precision": 0.905751329787234, | |
| "eval_recall": 0.9170313025917199, | |
| "eval_runtime": 21.1284, | |
| "eval_samples_per_second": 153.821, | |
| "eval_steps_per_second": 4.828, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_accuracy": 0.9852420077099802, | |
| "eval_f1": 0.9112356730527901, | |
| "eval_loss": 0.05755457654595375, | |
| "eval_precision": 0.9060056562967892, | |
| "eval_recall": 0.9165264220801077, | |
| "eval_runtime": 21.4936, | |
| "eval_samples_per_second": 151.208, | |
| "eval_steps_per_second": 4.746, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "step": 878, | |
| "total_flos": 913386421477854.0, | |
| "train_loss": 0.09878404265384196, | |
| "train_runtime": 1144.6421, | |
| "train_samples_per_second": 24.533, | |
| "train_steps_per_second": 0.767 | |
| } | |
| ], | |
| "max_steps": 878, | |
| "num_train_epochs": 2, | |
| "total_flos": 913386421477854.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |