| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "global_step": 59, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0001347830758006758, | |
| "loss": 2.4476, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0001277026099836151, | |
| "loss": 2.2935, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00011645752234985997, | |
| "loss": 2.0005, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0001018401897048198, | |
| "loss": 2.1957, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 8.488061161115849e-05, | |
| "loss": 2.028, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 6.677383222936865e-05, | |
| "loss": 1.9452, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.8795732412502015e-05, | |
| "loss": 1.9326, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.221312570298567e-05, | |
| "loss": 2.1213, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.81944932507936e-05, | |
| "loss": 2.0756, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 7.7276476511192e-06, | |
| "loss": 1.9295, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.550127458494346e-06, | |
| "loss": 1.7807, | |
| "step": 55 | |
| } | |
| ], | |
| "max_steps": 59, | |
| "num_train_epochs": 1, | |
| "total_flos": 61403627520000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |