| { |
| "best_metric": 0.47078712112764887, |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-0/checkpoint-2140", |
| "epoch": 4.0, |
| "eval_steps": 500, |
| "global_step": 2140, |
| "is_hyper_param_search": true, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.93, |
| "learning_rate": 6.355229248248623e-05, |
| "loss": 0.5639, |
| "step": 500 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_loss": 0.5066079497337341, |
| "eval_matthews_correlation": 0.3915828968181581, |
| "eval_runtime": 0.697, |
| "eval_samples_per_second": 1496.499, |
| "eval_steps_per_second": 94.697, |
| "step": 535 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 4.8942570072719275e-05, |
| "loss": 0.3737, |
| "step": 1000 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_loss": 0.5427648425102234, |
| "eval_matthews_correlation": 0.4465618323468758, |
| "eval_runtime": 0.7057, |
| "eval_samples_per_second": 1478.025, |
| "eval_steps_per_second": 93.528, |
| "step": 1070 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 3.433284766295233e-05, |
| "loss": 0.2331, |
| "step": 1500 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_loss": 0.6696511507034302, |
| "eval_matthews_correlation": 0.44079021569318316, |
| "eval_runtime": 0.736, |
| "eval_samples_per_second": 1417.068, |
| "eval_steps_per_second": 89.671, |
| "step": 1605 |
| }, |
| { |
| "epoch": 3.74, |
| "learning_rate": 1.9723125253185378e-05, |
| "loss": 0.1434, |
| "step": 2000 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_loss": 0.8634324669837952, |
| "eval_matthews_correlation": 0.47078712112764887, |
| "eval_runtime": 0.7355, |
| "eval_samples_per_second": 1417.988, |
| "eval_steps_per_second": 89.729, |
| "step": 2140 |
| } |
| ], |
| "logging_steps": 500, |
| "max_steps": 2675, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "save_steps": 500, |
| "total_flos": 170832234090192.0, |
| "train_batch_size": 16, |
| "trial_name": null, |
| "trial_params": { |
| "learning_rate": 7.816201489225318e-05, |
| "num_train_epochs": 5, |
| "per_device_train_batch_size": 16, |
| "seed": 37 |
| } |
| } |
|
|
