| { | |
| "best_metric": 0.8430973000509424, | |
| "best_model_checkpoint": "./outputs/bert-base-uncased/fine_tuned_models/checkpoint-49088", | |
| "epoch": 5.0, | |
| "global_step": 61360, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7388593694651082e-05, | |
| "loss": 0.5823, | |
| "step": 12272 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8318899643402955, | |
| "eval_loss": 0.4404328763484955, | |
| "eval_runtime": 73.7336, | |
| "eval_samples_per_second": 133.114, | |
| "eval_steps_per_second": 8.327, | |
| "step": 12272 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.3041445270988312e-05, | |
| "loss": 0.3672, | |
| "step": 24544 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8400407539480387, | |
| "eval_loss": 0.4377327859401703, | |
| "eval_runtime": 73.6581, | |
| "eval_samples_per_second": 133.251, | |
| "eval_steps_per_second": 8.336, | |
| "step": 24544 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 8.694296847325541e-06, | |
| "loss": 0.2456, | |
| "step": 36816 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8419765664798777, | |
| "eval_loss": 0.45335260033607483, | |
| "eval_runtime": 73.7206, | |
| "eval_samples_per_second": 133.138, | |
| "eval_steps_per_second": 8.329, | |
| "step": 36816 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.3471484236627705e-06, | |
| "loss": 0.1613, | |
| "step": 49088 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8430973000509424, | |
| "eval_loss": 0.5531118512153625, | |
| "eval_runtime": 73.7124, | |
| "eval_samples_per_second": 133.153, | |
| "eval_steps_per_second": 8.33, | |
| "step": 49088 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.1522, | |
| "step": 61360 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8406520631686195, | |
| "eval_loss": 0.6579604148864746, | |
| "eval_runtime": 24.5657, | |
| "eval_samples_per_second": 399.54, | |
| "eval_steps_per_second": 49.948, | |
| "step": 61360 | |
| } | |
| ], | |
| "max_steps": 61360, | |
| "num_train_epochs": 5, | |
| "total_flos": 1.3208002736945357e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |