| { | |
| "best_metric": 0.055517952889204025, | |
| "best_model_checkpoint": "XLM_RU_EN_WMT2021_WO_LF_100/checkpoint-500", | |
| "epoch": 2.2831050228310503, | |
| "global_step": 500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 4.94337899543379e-06, | |
| "loss": 0.0826, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "eval_loss": 0.055517952889204025, | |
| "eval_mse": 0.05551794916391373, | |
| "eval_pearson_correlation": 0.4512787180760665, | |
| "eval_rmse": 0.23562246561050415, | |
| "eval_runtime": 9.8163, | |
| "eval_samples_per_second": 101.871, | |
| "eval_spearman_corr": 0.3608326915950213, | |
| "eval_steps_per_second": 3.26, | |
| "learning_rate": 4.94337899543379e-06, | |
| "step": 500 | |
| } | |
| ], | |
| "max_steps": 43800, | |
| "num_train_epochs": 200, | |
| "total_flos": 2767652060607936.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |