| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 5.0, | |
| "global_step": 610, | |
| "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": 1.1931, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_exact_match": 69.10384655047953, | |
| "eval_f1": 72.2870181759377, | |
| "eval_runtime": 3.6823, | |
| "eval_samples_per_second": 2633.395, | |
| "eval_steps_per_second": 8.419, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.5454, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_exact_match": 69.03165927606476, | |
| "eval_f1": 72.42603653430285, | |
| "eval_runtime": 3.3791, | |
| "eval_samples_per_second": 2869.736, | |
| "eval_steps_per_second": 9.174, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.4824, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_exact_match": 69.98040631123028, | |
| "eval_f1": 73.0604212623475, | |
| "eval_runtime": 3.3845, | |
| "eval_samples_per_second": 2865.162, | |
| "eval_steps_per_second": 9.16, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.4426, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_exact_match": 70.16603073115397, | |
| "eval_f1": 73.34933825934681, | |
| "eval_runtime": 3.3883, | |
| "eval_samples_per_second": 2861.942, | |
| "eval_steps_per_second": 9.149, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3e-05, | |
| "loss": 0.4111, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_exact_match": 70.56821697432196, | |
| "eval_f1": 73.46129953714774, | |
| "eval_runtime": 3.4026, | |
| "eval_samples_per_second": 2849.873, | |
| "eval_steps_per_second": 9.111, | |
| "step": 610 | |
| } | |
| ], | |
| "max_steps": 1220, | |
| "num_train_epochs": 10, | |
| "total_flos": 73112944115712.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |