| { |
| "best_metric": 0.4863512385958954, |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-3/checkpoint-1069", |
| "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": 1.2538588476474406e-05, |
| "loss": 0.5581, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 8.711180516622632e-06, |
| "loss": 0.4917, |
| "step": 1000 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_loss": 0.48650237917900085, |
| "eval_matthews_correlation": 0.4863512385958954, |
| "eval_runtime": 0.7606, |
| "eval_samples_per_second": 1371.338, |
| "eval_steps_per_second": 86.777, |
| "step": 1069 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 4.883772556770861e-06, |
| "loss": 0.3582, |
| "step": 1500 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 1.0563645969190893e-06, |
| "loss": 0.3426, |
| "step": 2000 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_loss": 0.5801453590393066, |
| "eval_matthews_correlation": 0.48193970397464586, |
| "eval_runtime": 0.7674, |
| "eval_samples_per_second": 1359.149, |
| "eval_steps_per_second": 86.006, |
| "step": 2138 |
| } |
| ], |
| "logging_steps": 500, |
| "max_steps": 2138, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 2, |
| "save_steps": 500, |
| "total_flos": 75365670622848.0, |
| "train_batch_size": 8, |
| "trial_name": null, |
| "trial_params": { |
| "learning_rate": 1.6365996436326178e-05, |
| "num_train_epochs": 2, |
| "per_device_train_batch_size": 8, |
| "seed": 33 |
| } |
| } |
|
|
