| { |
| "best_metric": 0.5268164418763311, |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-6/checkpoint-1605", |
| "epoch": 3.0, |
| "eval_steps": 500, |
| "global_step": 1605, |
| "is_hyper_param_search": true, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.93, |
| "learning_rate": 1.3543559458705673e-05, |
| "loss": 0.5213, |
| "step": 500 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_loss": 0.48579296469688416, |
| "eval_matthews_correlation": 0.4326078706030828, |
| "eval_runtime": 0.7405, |
| "eval_samples_per_second": 1408.497, |
| "eval_steps_per_second": 89.128, |
| "step": 535 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 1.0430097514175634e-05, |
| "loss": 0.3536, |
| "step": 1000 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_loss": 0.4809291660785675, |
| "eval_matthews_correlation": 0.5083396333380269, |
| "eval_runtime": 0.8098, |
| "eval_samples_per_second": 1287.978, |
| "eval_steps_per_second": 81.502, |
| "step": 1070 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 7.316635569645593e-06, |
| "loss": 0.2508, |
| "step": 1500 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_loss": 0.5551302433013916, |
| "eval_matthews_correlation": 0.5268164418763311, |
| "eval_runtime": 1.2059, |
| "eval_samples_per_second": 864.931, |
| "eval_steps_per_second": 54.732, |
| "step": 1605 |
| } |
| ], |
| "logging_steps": 500, |
| "max_steps": 2675, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "save_steps": 500, |
| "total_flos": 129007464042276.0, |
| "train_batch_size": 16, |
| "trial_name": null, |
| "trial_params": { |
| "learning_rate": 1.6657021403235714e-05, |
| "num_train_epochs": 5, |
| "per_device_train_batch_size": 16, |
| "seed": 10 |
| } |
| } |
|
|
