| { | |
| "best_metric": 0.6745283018867925, | |
| "best_model_checkpoint": "videomae-base-finetuned-engine-subset-20230313/checkpoint-988", | |
| "epoch": 29.00720720720721, | |
| "global_step": 1110, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.504504504504505e-06, | |
| "loss": 2.7278, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.00900900900901e-06, | |
| "loss": 2.747, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.3513513513513515e-05, | |
| "loss": 2.6212, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_accuracy": 0.37735849056603776, | |
| "eval_loss": 2.362868070602417, | |
| "eval_runtime": 69.3989, | |
| "eval_samples_per_second": 6.11, | |
| "eval_steps_per_second": 1.023, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.801801801801802e-05, | |
| "loss": 2.6599, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.2522522522522523e-05, | |
| "loss": 2.5895, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.702702702702703e-05, | |
| "loss": 2.5383, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.153153153153153e-05, | |
| "loss": 2.455, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "eval_accuracy": 0.2169811320754717, | |
| "eval_loss": 2.3674099445343018, | |
| "eval_runtime": 70.0897, | |
| "eval_samples_per_second": 6.049, | |
| "eval_steps_per_second": 1.013, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.603603603603604e-05, | |
| "loss": 2.421, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 4.0540540540540545e-05, | |
| "loss": 2.4139, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 4.5045045045045046e-05, | |
| "loss": 2.4999, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.954954954954955e-05, | |
| "loss": 2.4311, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "eval_accuracy": 0.3231132075471698, | |
| "eval_loss": 2.2190816402435303, | |
| "eval_runtime": 71.2703, | |
| "eval_samples_per_second": 5.949, | |
| "eval_steps_per_second": 0.996, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.954954954954955e-05, | |
| "loss": 2.5089, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.9049049049049054e-05, | |
| "loss": 2.4986, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.854854854854855e-05, | |
| "loss": 2.3317, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 4.804804804804805e-05, | |
| "loss": 2.2768, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "eval_accuracy": 0.3608490566037736, | |
| "eval_loss": 2.1226866245269775, | |
| "eval_runtime": 70.1636, | |
| "eval_samples_per_second": 6.043, | |
| "eval_steps_per_second": 1.012, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 4.754754754754755e-05, | |
| "loss": 2.0041, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 4.704704704704705e-05, | |
| "loss": 2.2092, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 4.654654654654655e-05, | |
| "loss": 1.9442, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 4.604604604604605e-05, | |
| "loss": 1.7528, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "eval_accuracy": 0.4363207547169811, | |
| "eval_loss": 1.729621171951294, | |
| "eval_runtime": 69.2109, | |
| "eval_samples_per_second": 6.126, | |
| "eval_steps_per_second": 1.026, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.5545545545545545e-05, | |
| "loss": 1.6692, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.5045045045045046e-05, | |
| "loss": 1.8567, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 4.4544544544544546e-05, | |
| "loss": 1.5381, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "eval_accuracy": 0.4339622641509434, | |
| "eval_loss": 1.5015883445739746, | |
| "eval_runtime": 71.2754, | |
| "eval_samples_per_second": 5.949, | |
| "eval_steps_per_second": 0.996, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 4.404404404404405e-05, | |
| "loss": 1.4846, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 4.354354354354355e-05, | |
| "loss": 1.3143, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 4.304304304304305e-05, | |
| "loss": 1.2959, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 4.254254254254255e-05, | |
| "loss": 1.407, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "eval_accuracy": 0.5448113207547169, | |
| "eval_loss": 1.2878233194351196, | |
| "eval_runtime": 70.9118, | |
| "eval_samples_per_second": 5.979, | |
| "eval_steps_per_second": 1.001, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 4.204204204204204e-05, | |
| "loss": 1.3608, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 4.1541541541541544e-05, | |
| "loss": 1.5729, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 4.1041041041041045e-05, | |
| "loss": 1.3095, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 4.0540540540540545e-05, | |
| "loss": 1.1053, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "eval_accuracy": 0.4009433962264151, | |
| "eval_loss": 1.5210211277008057, | |
| "eval_runtime": 68.6229, | |
| "eval_samples_per_second": 6.179, | |
| "eval_steps_per_second": 1.035, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 4.0040040040040046e-05, | |
| "loss": 1.2462, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 3.953953953953955e-05, | |
| "loss": 1.114, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 3.903903903903904e-05, | |
| "loss": 1.2971, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 3.8538538538538534e-05, | |
| "loss": 1.0893, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "eval_accuracy": 0.46226415094339623, | |
| "eval_loss": 1.3901519775390625, | |
| "eval_runtime": 70.4934, | |
| "eval_samples_per_second": 6.015, | |
| "eval_steps_per_second": 1.007, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 3.8038038038038035e-05, | |
| "loss": 1.3258, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 3.7537537537537536e-05, | |
| "loss": 1.1349, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 3.7037037037037037e-05, | |
| "loss": 1.0029, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 3.653653653653654e-05, | |
| "loss": 0.8136, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "eval_accuracy": 0.4033018867924528, | |
| "eval_loss": 1.6456053256988525, | |
| "eval_runtime": 71.0727, | |
| "eval_samples_per_second": 5.966, | |
| "eval_steps_per_second": 0.999, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 3.603603603603604e-05, | |
| "loss": 1.1048, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 3.553553553553554e-05, | |
| "loss": 1.2661, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 3.503503503503503e-05, | |
| "loss": 0.9565, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "eval_accuracy": 0.5613207547169812, | |
| "eval_loss": 1.18259859085083, | |
| "eval_runtime": 70.181, | |
| "eval_samples_per_second": 6.042, | |
| "eval_steps_per_second": 1.012, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 3.453453453453453e-05, | |
| "loss": 1.0921, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 3.4034034034034034e-05, | |
| "loss": 0.7383, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 3.3533533533533535e-05, | |
| "loss": 0.926, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 3.3033033033033035e-05, | |
| "loss": 1.0147, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "eval_accuracy": 0.5117924528301887, | |
| "eval_loss": 1.2098637819290161, | |
| "eval_runtime": 70.8824, | |
| "eval_samples_per_second": 5.982, | |
| "eval_steps_per_second": 1.002, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 3.2532532532532536e-05, | |
| "loss": 0.9178, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 3.203203203203203e-05, | |
| "loss": 0.8994, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 3.153153153153153e-05, | |
| "loss": 0.8648, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 3.103103103103103e-05, | |
| "loss": 0.9125, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "eval_accuracy": 0.5495283018867925, | |
| "eval_loss": 1.1849919557571411, | |
| "eval_runtime": 70.4782, | |
| "eval_samples_per_second": 6.016, | |
| "eval_steps_per_second": 1.007, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 3.053053053053053e-05, | |
| "loss": 0.8926, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 3.0030030030030033e-05, | |
| "loss": 0.8618, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 2.952952952952953e-05, | |
| "loss": 0.8057, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 2.902902902902903e-05, | |
| "loss": 0.7091, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "eval_accuracy": 0.535377358490566, | |
| "eval_loss": 1.2324182987213135, | |
| "eval_runtime": 70.9353, | |
| "eval_samples_per_second": 5.977, | |
| "eval_steps_per_second": 1.001, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 2.852852852852853e-05, | |
| "loss": 0.74, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 2.8028028028028032e-05, | |
| "loss": 0.7484, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 2.752752752752753e-05, | |
| "loss": 0.7962, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 2.702702702702703e-05, | |
| "loss": 0.7361, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "eval_accuracy": 0.6226415094339622, | |
| "eval_loss": 1.022481083869934, | |
| "eval_runtime": 70.6845, | |
| "eval_samples_per_second": 5.998, | |
| "eval_steps_per_second": 1.004, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 2.652652652652653e-05, | |
| "loss": 0.8396, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 2.6026026026026028e-05, | |
| "loss": 0.7193, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 2.552552552552553e-05, | |
| "loss": 0.6979, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "eval_accuracy": 0.5589622641509434, | |
| "eval_loss": 1.0737839937210083, | |
| "eval_runtime": 69.8991, | |
| "eval_samples_per_second": 6.066, | |
| "eval_steps_per_second": 1.016, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 2.502502502502503e-05, | |
| "loss": 0.6869, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 2.4524524524524527e-05, | |
| "loss": 0.6704, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 2.4024024024024024e-05, | |
| "loss": 0.8011, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 2.3523523523523525e-05, | |
| "loss": 0.5265, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "eval_accuracy": 0.5872641509433962, | |
| "eval_loss": 1.1062343120574951, | |
| "eval_runtime": 70.6095, | |
| "eval_samples_per_second": 6.005, | |
| "eval_steps_per_second": 1.006, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 2.3023023023023026e-05, | |
| "loss": 0.6415, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 2.2522522522522523e-05, | |
| "loss": 0.7402, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 2.2022022022022024e-05, | |
| "loss": 0.6275, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 2.1521521521521524e-05, | |
| "loss": 0.5651, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "eval_accuracy": 0.5801886792452831, | |
| "eval_loss": 1.1402446031570435, | |
| "eval_runtime": 70.0342, | |
| "eval_samples_per_second": 6.054, | |
| "eval_steps_per_second": 1.014, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.102102102102102e-05, | |
| "loss": 0.5542, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.0520520520520522e-05, | |
| "loss": 0.6903, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.0020020020020023e-05, | |
| "loss": 0.7, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 1.951951951951952e-05, | |
| "loss": 0.7182, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "eval_accuracy": 0.5801886792452831, | |
| "eval_loss": 1.097399115562439, | |
| "eval_runtime": 70.689, | |
| "eval_samples_per_second": 5.998, | |
| "eval_steps_per_second": 1.004, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 1.9019019019019018e-05, | |
| "loss": 0.5921, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 1.8518518518518518e-05, | |
| "loss": 0.5428, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 1.801801801801802e-05, | |
| "loss": 0.7289, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 1.7517517517517516e-05, | |
| "loss": 0.6582, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "eval_accuracy": 0.6179245283018868, | |
| "eval_loss": 1.0529001951217651, | |
| "eval_runtime": 70.8949, | |
| "eval_samples_per_second": 5.981, | |
| "eval_steps_per_second": 1.001, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 1.7017017017017017e-05, | |
| "loss": 0.5001, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.6516516516516518e-05, | |
| "loss": 0.5387, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.6016016016016015e-05, | |
| "loss": 0.5709, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "eval_accuracy": 0.6344339622641509, | |
| "eval_loss": 0.965450644493103, | |
| "eval_runtime": 69.2205, | |
| "eval_samples_per_second": 6.125, | |
| "eval_steps_per_second": 1.026, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 1.5515515515515516e-05, | |
| "loss": 0.7102, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.5015015015015016e-05, | |
| "loss": 0.4746, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.4514514514514515e-05, | |
| "loss": 0.611, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.4014014014014016e-05, | |
| "loss": 0.4808, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "eval_accuracy": 0.6226415094339622, | |
| "eval_loss": 1.0440747737884521, | |
| "eval_runtime": 70.6133, | |
| "eval_samples_per_second": 6.005, | |
| "eval_steps_per_second": 1.005, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 1.3513513513513515e-05, | |
| "loss": 0.5081, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.3013013013013014e-05, | |
| "loss": 0.6717, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.2512512512512515e-05, | |
| "loss": 0.343, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 1.2012012012012012e-05, | |
| "loss": 0.5816, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "eval_accuracy": 0.6438679245283019, | |
| "eval_loss": 0.9445228576660156, | |
| "eval_runtime": 70.5922, | |
| "eval_samples_per_second": 6.006, | |
| "eval_steps_per_second": 1.006, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 1.1511511511511513e-05, | |
| "loss": 0.4179, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 1.1011011011011012e-05, | |
| "loss": 0.4346, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 1.051051051051051e-05, | |
| "loss": 0.4038, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 1.0010010010010011e-05, | |
| "loss": 0.5057, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "eval_accuracy": 0.6320754716981132, | |
| "eval_loss": 1.0248113870620728, | |
| "eval_runtime": 71.1172, | |
| "eval_samples_per_second": 5.962, | |
| "eval_steps_per_second": 0.998, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 9.509509509509509e-06, | |
| "loss": 0.393, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 9.00900900900901e-06, | |
| "loss": 0.4297, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 8.508508508508508e-06, | |
| "loss": 0.5053, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 8.008008008008007e-06, | |
| "loss": 0.6253, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "eval_accuracy": 0.660377358490566, | |
| "eval_loss": 0.9517852067947388, | |
| "eval_runtime": 70.664, | |
| "eval_samples_per_second": 6.0, | |
| "eval_steps_per_second": 1.005, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 25.01, | |
| "learning_rate": 7.507507507507508e-06, | |
| "loss": 0.4143, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 25.02, | |
| "learning_rate": 7.007007007007008e-06, | |
| "loss": 0.6003, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 25.03, | |
| "learning_rate": 6.506506506506507e-06, | |
| "loss": 0.6841, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 25.03, | |
| "eval_accuracy": 0.6745283018867925, | |
| "eval_loss": 0.8912516236305237, | |
| "eval_runtime": 70.8493, | |
| "eval_samples_per_second": 5.985, | |
| "eval_steps_per_second": 1.002, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 6.006006006006006e-06, | |
| "loss": 0.2104, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 5.505505505505506e-06, | |
| "loss": 0.6054, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 26.02, | |
| "learning_rate": 5.005005005005006e-06, | |
| "loss": 0.6045, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 26.03, | |
| "learning_rate": 4.504504504504505e-06, | |
| "loss": 0.5933, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 26.03, | |
| "eval_accuracy": 0.6438679245283019, | |
| "eval_loss": 0.9012843370437622, | |
| "eval_runtime": 71.2482, | |
| "eval_samples_per_second": 5.951, | |
| "eval_steps_per_second": 0.997, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 4.004004004004004e-06, | |
| "loss": 0.4052, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 3.503503503503504e-06, | |
| "loss": 0.4422, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 27.02, | |
| "learning_rate": 3.003003003003003e-06, | |
| "loss": 0.4861, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 27.03, | |
| "learning_rate": 2.502502502502503e-06, | |
| "loss": 0.389, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 27.03, | |
| "eval_accuracy": 0.6627358490566038, | |
| "eval_loss": 0.9089723229408264, | |
| "eval_runtime": 71.3879, | |
| "eval_samples_per_second": 5.939, | |
| "eval_steps_per_second": 0.995, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 2.002002002002002e-06, | |
| "loss": 0.4572, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 1.5015015015015015e-06, | |
| "loss": 0.213, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 28.02, | |
| "learning_rate": 1.001001001001001e-06, | |
| "loss": 0.6041, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 28.03, | |
| "learning_rate": 5.005005005005005e-07, | |
| "loss": 0.3705, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 28.03, | |
| "eval_accuracy": 0.6721698113207547, | |
| "eval_loss": 0.8935866355895996, | |
| "eval_runtime": 70.2956, | |
| "eval_samples_per_second": 6.032, | |
| "eval_steps_per_second": 1.01, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 0.0, | |
| "loss": 0.6043, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "eval_accuracy": 0.6721698113207547, | |
| "eval_loss": 0.8942137956619263, | |
| "eval_runtime": 69.8197, | |
| "eval_samples_per_second": 6.073, | |
| "eval_steps_per_second": 1.017, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "step": 1110, | |
| "total_flos": 8.191332105366897e+18, | |
| "train_loss": 1.059841639716346, | |
| "train_runtime": 5977.6934, | |
| "train_samples_per_second": 1.114, | |
| "train_steps_per_second": 0.186 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "eval_accuracy": 0.6745283018867925, | |
| "eval_loss": 0.8912516236305237, | |
| "eval_runtime": 68.8583, | |
| "eval_samples_per_second": 6.158, | |
| "eval_steps_per_second": 1.031, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "eval_accuracy": 0.6745283018867925, | |
| "eval_loss": 0.8912516236305237, | |
| "eval_runtime": 68.7355, | |
| "eval_samples_per_second": 6.169, | |
| "eval_steps_per_second": 1.033, | |
| "step": 1110 | |
| } | |
| ], | |
| "max_steps": 1110, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 8.191332105366897e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |