| { | |
| "best_metric": 0.5102489493078572, | |
| "best_model_checkpoint": "distilbert-base-uncased-finetuned-cola/run-7/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": 2.883378108768492e-05, | |
| "loss": 0.5687, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.5858158684415064e-05, | |
| "loss": 0.4948, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.4758210778236389, | |
| "eval_matthews_correlation": 0.43428613653509085, | |
| "eval_runtime": 0.7527, | |
| "eval_samples_per_second": 1385.607, | |
| "eval_steps_per_second": 87.68, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.2882536281145206e-05, | |
| "loss": 0.3622, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.9906913877875348e-05, | |
| "loss": 0.344, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.6103293895721436, | |
| "eval_matthews_correlation": 0.5102489493078572, | |
| "eval_runtime": 0.8162, | |
| "eval_samples_per_second": 1277.866, | |
| "eval_steps_per_second": 80.862, | |
| "step": 2138 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 5345, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "save_steps": 500, | |
| "total_flos": 75192065887500.0, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 3.180940349095478e-05, | |
| "num_train_epochs": 5, | |
| "per_device_train_batch_size": 8, | |
| "seed": 22 | |
| } | |
| } | |