| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 11.450381679389313, | |
| "global_step": 1500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9964473845399976, | |
| "eval_f1": 0.7887323943661972, | |
| "eval_loss": 0.02695615217089653, | |
| "eval_precision": 0.8484848484848485, | |
| "eval_recall": 0.7368421052631579, | |
| "eval_runtime": 1.6552, | |
| "eval_samples_per_second": 59.207, | |
| "eval_steps_per_second": 19.937, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9969374004655152, | |
| "eval_f1": 0.8533333333333334, | |
| "eval_loss": 0.022969316691160202, | |
| "eval_precision": 0.8648648648648649, | |
| "eval_recall": 0.8421052631578947, | |
| "eval_runtime": 1.6508, | |
| "eval_samples_per_second": 59.365, | |
| "eval_steps_per_second": 19.99, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9968761484748254, | |
| "eval_f1": 0.8211920529801323, | |
| "eval_loss": 0.023406116291880608, | |
| "eval_precision": 0.8266666666666667, | |
| "eval_recall": 0.8157894736842105, | |
| "eval_runtime": 1.6791, | |
| "eval_samples_per_second": 58.364, | |
| "eval_steps_per_second": 19.653, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 3.727735368956743e-05, | |
| "loss": 0.0151, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9977949283351709, | |
| "eval_f1": 0.832214765100671, | |
| "eval_loss": 0.01629047282040119, | |
| "eval_precision": 0.8493150684931506, | |
| "eval_recall": 0.8157894736842105, | |
| "eval_runtime": 1.6475, | |
| "eval_samples_per_second": 59.485, | |
| "eval_steps_per_second": 20.031, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9976724243537914, | |
| "eval_f1": 0.8609271523178809, | |
| "eval_loss": 0.016509264707565308, | |
| "eval_precision": 0.8666666666666667, | |
| "eval_recall": 0.8552631578947368, | |
| "eval_runtime": 1.6672, | |
| "eval_samples_per_second": 58.78, | |
| "eval_steps_per_second": 19.793, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9976111723631018, | |
| "eval_f1": 0.8535031847133758, | |
| "eval_loss": 0.01784520410001278, | |
| "eval_precision": 0.8271604938271605, | |
| "eval_recall": 0.881578947368421, | |
| "eval_runtime": 1.6462, | |
| "eval_samples_per_second": 59.532, | |
| "eval_steps_per_second": 20.046, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.99797868430724, | |
| "eval_f1": 0.9103448275862069, | |
| "eval_loss": 0.021767688915133476, | |
| "eval_precision": 0.9565217391304348, | |
| "eval_recall": 0.868421052631579, | |
| "eval_runtime": 1.6645, | |
| "eval_samples_per_second": 58.876, | |
| "eval_steps_per_second": 19.825, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 2.455470737913486e-05, | |
| "loss": 0.0004, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9981624402793091, | |
| "eval_f1": 0.9115646258503401, | |
| "eval_loss": 0.01889556646347046, | |
| "eval_precision": 0.9436619718309859, | |
| "eval_recall": 0.881578947368421, | |
| "eval_runtime": 1.6651, | |
| "eval_samples_per_second": 58.855, | |
| "eval_steps_per_second": 19.819, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9981011882886194, | |
| "eval_f1": 0.8918918918918918, | |
| "eval_loss": 0.020160716027021408, | |
| "eval_precision": 0.9166666666666666, | |
| "eval_recall": 0.868421052631579, | |
| "eval_runtime": 1.6829, | |
| "eval_samples_per_second": 58.232, | |
| "eval_steps_per_second": 19.609, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.99797868430724, | |
| "eval_f1": 0.8918918918918918, | |
| "eval_loss": 0.021420367062091827, | |
| "eval_precision": 0.9166666666666666, | |
| "eval_recall": 0.868421052631579, | |
| "eval_runtime": 1.656, | |
| "eval_samples_per_second": 59.18, | |
| "eval_steps_per_second": 19.928, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.9982236922699987, | |
| "eval_f1": 0.9251700680272109, | |
| "eval_loss": 0.01935696415603161, | |
| "eval_precision": 0.9577464788732394, | |
| "eval_recall": 0.8947368421052632, | |
| "eval_runtime": 1.6602, | |
| "eval_samples_per_second": 59.028, | |
| "eval_steps_per_second": 19.877, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 11.45, | |
| "learning_rate": 1.1832061068702292e-05, | |
| "loss": 0.0001, | |
| "step": 1500 | |
| } | |
| ], | |
| "max_steps": 1965, | |
| "num_train_epochs": 15, | |
| "total_flos": 693341757784872.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |