| { | |
| "best_metric": 0.6178978681564331, | |
| "best_model_checkpoint": "add_BERT_48_cola/checkpoint-335", | |
| "epoch": 10.0, | |
| "global_step": 670, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.9200000000000004e-05, | |
| "loss": 0.6211, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.619252622127533, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0055, | |
| "eval_samples_per_second": 520.06, | |
| "eval_steps_per_second": 4.488, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.8400000000000005e-05, | |
| "loss": 0.6175, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6525280475616455, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0086, | |
| "eval_samples_per_second": 519.264, | |
| "eval_steps_per_second": 4.481, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.76e-05, | |
| "loss": 0.6147, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6189554333686829, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0113, | |
| "eval_samples_per_second": 518.567, | |
| "eval_steps_per_second": 4.475, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.680000000000001e-05, | |
| "loss": 0.6126, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6182320713996887, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0163, | |
| "eval_samples_per_second": 517.28, | |
| "eval_steps_per_second": 4.464, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3.6e-05, | |
| "loss": 0.61, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6178978681564331, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0077, | |
| "eval_samples_per_second": 519.497, | |
| "eval_steps_per_second": 4.483, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.52e-05, | |
| "loss": 0.6104, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6183823347091675, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0141, | |
| "eval_samples_per_second": 517.837, | |
| "eval_steps_per_second": 4.468, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 3.44e-05, | |
| "loss": 0.6108, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6222911477088928, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0077, | |
| "eval_samples_per_second": 519.492, | |
| "eval_steps_per_second": 4.483, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.3600000000000004e-05, | |
| "loss": 0.6108, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6193031668663025, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0144, | |
| "eval_samples_per_second": 517.774, | |
| "eval_steps_per_second": 4.468, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 3.28e-05, | |
| "loss": 0.6093, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6289520263671875, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0156, | |
| "eval_samples_per_second": 517.475, | |
| "eval_steps_per_second": 4.465, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 3.2000000000000005e-05, | |
| "loss": 0.609, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.6912751793861389, | |
| "eval_loss": 0.6254616379737854, | |
| "eval_matthews_correlation": 0.0, | |
| "eval_runtime": 2.0168, | |
| "eval_samples_per_second": 517.154, | |
| "eval_steps_per_second": 4.463, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 670, | |
| "total_flos": 1.221287775043584e+16, | |
| "train_loss": 0.6126257312831594, | |
| "train_runtime": 604.4829, | |
| "train_samples_per_second": 707.299, | |
| "train_steps_per_second": 5.542 | |
| } | |
| ], | |
| "max_steps": 3350, | |
| "num_train_epochs": 50, | |
| "total_flos": 1.221287775043584e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |