| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "global_step": 62, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0001350100903542216, | |
| "loss": 3.0284, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00012858017786751834, | |
| "loss": 2.887, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00011832078520391764, | |
| "loss": 2.857, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00010488693110842961, | |
| "loss": 2.6828, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 8.913631023597238e-05, | |
| "loss": 2.8563, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 7.207453298233569e-05, | |
| "loss": 2.7969, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 5.4790921521917906e-05, | |
| "loss": 2.5781, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.838896120314628e-05, | |
| "loss": 2.6729, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 2.3915847685255567e-05, | |
| "loss": 2.5355, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.2295627933180862e-05, | |
| "loss": 2.6718, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 4.270203734710291e-06, | |
| "loss": 2.5471, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.519644153159924e-07, | |
| "loss": 2.6351, | |
| "step": 60 | |
| } | |
| ], | |
| "max_steps": 62, | |
| "num_train_epochs": 1, | |
| "total_flos": 64669777920000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |