| { | |
| "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": 1.8887, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_exact_match": 75.80889309366131, | |
| "eval_f1": 84.41435788887438, | |
| "eval_runtime": 4.2189, | |
| "eval_samples_per_second": 2519.399, | |
| "eval_steps_per_second": 8.059, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3e-05, | |
| "loss": 1.0166, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_exact_match": 79.08230842005676, | |
| "eval_f1": 86.94262157084606, | |
| "eval_runtime": 3.7419, | |
| "eval_samples_per_second": 2840.535, | |
| "eval_steps_per_second": 9.086, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.8084, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_exact_match": 80.03784295175024, | |
| "eval_f1": 87.72233558421931, | |
| "eval_runtime": 3.7476, | |
| "eval_samples_per_second": 2836.222, | |
| "eval_steps_per_second": 9.072, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.661, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_exact_match": 80.1608325449385, | |
| "eval_f1": 87.6426540801448, | |
| "eval_runtime": 3.707, | |
| "eval_samples_per_second": 2867.288, | |
| "eval_steps_per_second": 9.172, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.5426, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_exact_match": 80.2081362346263, | |
| "eval_f1": 87.74581825580132, | |
| "eval_runtime": 3.6892, | |
| "eval_samples_per_second": 2881.145, | |
| "eval_steps_per_second": 9.216, | |
| "step": 1375 | |
| } | |
| ], | |
| "max_steps": 5500, | |
| "num_train_epochs": 20, | |
| "total_flos": 165523350355968.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |