| { | |
| "best_metric": 0.11092716455459595, | |
| "best_model_checkpoint": "BERT-HARD-balanced-2/checkpoint-500", | |
| "epoch": 2.5210084033613445, | |
| "global_step": 1500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 7.198879551820729e-06, | |
| "loss": 0.1421, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_accuracy": 0.9602649006622517, | |
| "eval_f1": 0.9607751575998131, | |
| "eval_loss": 0.11092716455459595, | |
| "eval_precision": 0.9485938220378054, | |
| "eval_recall": 0.9732734153263954, | |
| "eval_runtime": 6.7747, | |
| "eval_samples_per_second": 1248.164, | |
| "eval_steps_per_second": 9.89, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 4.397759103641457e-06, | |
| "loss": 0.0981, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "eval_accuracy": 0.962038789025544, | |
| "eval_f1": 0.9620881067674502, | |
| "eval_loss": 0.11493375152349472, | |
| "eval_precision": 0.9608398207124321, | |
| "eval_recall": 0.9633396404919584, | |
| "eval_runtime": 6.7211, | |
| "eval_samples_per_second": 1258.119, | |
| "eval_steps_per_second": 9.969, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 1.5966386554621848e-06, | |
| "loss": 0.0794, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "eval_accuracy": 0.962038789025544, | |
| "eval_f1": 0.9624077760861928, | |
| "eval_loss": 0.12313080579042435, | |
| "eval_precision": 0.9531431222454188, | |
| "eval_recall": 0.9718543046357616, | |
| "eval_runtime": 6.7353, | |
| "eval_samples_per_second": 1255.468, | |
| "eval_steps_per_second": 9.948, | |
| "step": 1500 | |
| } | |
| ], | |
| "max_steps": 1785, | |
| "num_train_epochs": 3, | |
| "total_flos": 2.2352839715103744e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |