| { | |
| "best_metric": 0.07112815976142883, | |
| "best_model_checkpoint": "../model/xlm-roberta-large_8_5e-06_512_train_french_5_512/tmp/checkpoint-6840", | |
| "epoch": 3.0, | |
| "global_step": 6840, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.5e-06, | |
| "loss": 0.1391, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7017543859649122, | |
| "eval_loss": 0.08547523617744446, | |
| "eval_runtime": 56.0096, | |
| "eval_samples_per_second": 40.707, | |
| "eval_steps_per_second": 5.088, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.0803, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7263157894736842, | |
| "eval_loss": 0.07496826350688934, | |
| "eval_runtime": 55.7373, | |
| "eval_samples_per_second": 40.906, | |
| "eval_steps_per_second": 5.113, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.5e-06, | |
| "loss": 0.0654, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7328947368421053, | |
| "eval_loss": 0.07112815976142883, | |
| "eval_runtime": 56.0517, | |
| "eval_samples_per_second": 40.677, | |
| "eval_steps_per_second": 5.085, | |
| "step": 6840 | |
| } | |
| ], | |
| "max_steps": 22800, | |
| "num_train_epochs": 10, | |
| "total_flos": 5.098754631595622e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |