| { | |
| "best_metric": 0.446244823436024, | |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-4/checkpoint-2138", | |
| "epoch": 2.0, | |
| "eval_steps": 500, | |
| "global_step": 2138, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 3.970511622906794e-05, | |
| "loss": 0.5963, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.75851173801461e-05, | |
| "loss": 0.5523, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.5371980667114258, | |
| "eval_matthews_correlation": 0.33521556405671804, | |
| "eval_runtime": 0.7629, | |
| "eval_samples_per_second": 1367.128, | |
| "eval_steps_per_second": 86.51, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5465118531224263e-05, | |
| "loss": 0.3918, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.3451196823024276e-06, | |
| "loss": 0.3683, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.6115844249725342, | |
| "eval_matthews_correlation": 0.446244823436024, | |
| "eval_runtime": 0.8682, | |
| "eval_samples_per_second": 1201.294, | |
| "eval_steps_per_second": 76.017, | |
| "step": 2138 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2138, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 75396458944020.0, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 5.182511507798978e-05, | |
| "num_train_epochs": 2, | |
| "per_device_train_batch_size": 8, | |
| "seed": 28 | |
| } | |
| } | |