| { |
| "best_metric": 0.6659030476580078, |
| "best_model_checkpoint": "final_models/structroberta_sx2_final/finetune/control_raising_relative_token_position/checkpoint-400", |
| "epoch": 10.0, |
| "global_step": 690, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 5.8, |
| "eval_accuracy": 0.6670743227005005, |
| "eval_f1": 0.6659030476580078, |
| "eval_loss": 2.979311466217041, |
| "eval_mcc": 0.3342410383312042, |
| "eval_runtime": 51.3619, |
| "eval_samples_per_second": 477.611, |
| "eval_steps_per_second": 59.714, |
| "step": 400 |
| }, |
| { |
| "epoch": 7.25, |
| "learning_rate": 1.3768115942028985e-05, |
| "loss": 0.0108, |
| "step": 500 |
| }, |
| { |
| "epoch": 10.0, |
| "step": 690, |
| "total_flos": 7480378396293120.0, |
| "train_loss": 0.00783546825034031, |
| "train_runtime": 423.7038, |
| "train_samples_per_second": 192.729, |
| "train_steps_per_second": 1.628 |
| } |
| ], |
| "max_steps": 690, |
| "num_train_epochs": 10, |
| "total_flos": 7480378396293120.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
