| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 5.0, | |
| "global_step": 1375, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3e-05, | |
| "loss": 1.8697, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_exact_match": 74.49385052034059, | |
| "eval_f1": 83.73659826874903, | |
| "eval_runtime": 5.3211, | |
| "eval_samples_per_second": 1996.969, | |
| "eval_steps_per_second": 6.39, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3e-05, | |
| "loss": 1.0286, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_exact_match": 78.61873226111636, | |
| "eval_f1": 86.81427416943762, | |
| "eval_runtime": 3.8646, | |
| "eval_samples_per_second": 2749.59, | |
| "eval_steps_per_second": 8.798, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.8267, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_exact_match": 79.34720908230842, | |
| "eval_f1": 87.37493005554087, | |
| "eval_runtime": 3.8851, | |
| "eval_samples_per_second": 2735.061, | |
| "eval_steps_per_second": 8.751, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.6826, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_exact_match": 79.02554399243141, | |
| "eval_f1": 87.1065115863637, | |
| "eval_runtime": 3.8434, | |
| "eval_samples_per_second": 2764.747, | |
| "eval_steps_per_second": 8.846, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.5675, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_exact_match": 79.20529801324503, | |
| "eval_f1": 87.42053692550513, | |
| "eval_runtime": 3.8716, | |
| "eval_samples_per_second": 2744.592, | |
| "eval_steps_per_second": 8.782, | |
| "step": 1375 | |
| } | |
| ], | |
| "max_steps": 5500, | |
| "num_train_epochs": 20, | |
| "total_flos": 165447852883968.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |