| { |
| "best_metric": 0.666749451550147, |
| "best_model_checkpoint": "final_models/glue_models/structroberta_s1_final//finetune/main_verb_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": 0.6683009266853333, |
| "eval_f1": 0.666749451550147, |
| "eval_loss": 3.100526809692383, |
| "eval_mcc": 0.3365875778956931, |
| "eval_runtime": 47.6628, |
| "eval_samples_per_second": 509.245, |
| "eval_steps_per_second": 63.656, |
| "step": 400 |
| }, |
| { |
| "epoch": 7.25, |
| "learning_rate": 1.3768115942028985e-05, |
| "loss": 0.0094, |
| "step": 500 |
| }, |
| { |
| "epoch": 10.0, |
| "step": 690, |
| "total_flos": 6823634144670720.0, |
| "train_loss": 0.006852422207863866, |
| "train_runtime": 428.4842, |
| "train_samples_per_second": 190.835, |
| "train_steps_per_second": 1.61 |
| } |
| ], |
| "max_steps": 690, |
| "num_train_epochs": 10, |
| "total_flos": 6823634144670720.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
