| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 5.0, | |
| "global_step": 5860, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.573378839590444e-05, | |
| "loss": 0.1077, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 4.1467576791808876e-05, | |
| "loss": 0.0742, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9741838326251815, | |
| "eval_f1": 0.8941176470588236, | |
| "eval_loss": 0.06532363593578339, | |
| "eval_precision": 0.9193548387096774, | |
| "eval_recall": 0.8702290076335878, | |
| "eval_runtime": 10.9317, | |
| "eval_samples_per_second": 95.228, | |
| "eval_steps_per_second": 11.983, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.720136518771331e-05, | |
| "loss": 0.0506, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 3.293515358361775e-05, | |
| "loss": 0.0396, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9747754531275211, | |
| "eval_f1": 0.8959139784946236, | |
| "eval_loss": 0.07732317596673965, | |
| "eval_precision": 0.9088132635253054, | |
| "eval_recall": 0.8833757421543681, | |
| "eval_runtime": 11.1179, | |
| "eval_samples_per_second": 93.632, | |
| "eval_steps_per_second": 11.783, | |
| "step": 2344 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 2.8668941979522183e-05, | |
| "loss": 0.0325, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 2.4402730375426623e-05, | |
| "loss": 0.0156, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 2.013651877133106e-05, | |
| "loss": 0.0153, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9738611305329963, | |
| "eval_f1": 0.8905547226386806, | |
| "eval_loss": 0.11707163602113724, | |
| "eval_precision": 0.8996105581999134, | |
| "eval_recall": 0.8816793893129771, | |
| "eval_runtime": 10.7972, | |
| "eval_samples_per_second": 96.414, | |
| "eval_steps_per_second": 12.133, | |
| "step": 3516 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.5870307167235497e-05, | |
| "loss": 0.0058, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 1.1604095563139932e-05, | |
| "loss": 0.0059, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9746678857634594, | |
| "eval_f1": 0.8941074146553598, | |
| "eval_loss": 0.13896839320659637, | |
| "eval_precision": 0.9174475680499777, | |
| "eval_recall": 0.8719253604749788, | |
| "eval_runtime": 10.8916, | |
| "eval_samples_per_second": 95.579, | |
| "eval_steps_per_second": 12.028, | |
| "step": 4688 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 7.337883959044369e-06, | |
| "loss": 0.0038, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 3.0716723549488057e-06, | |
| "loss": 0.0024, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9755284246759534, | |
| "eval_f1": 0.8987612131567707, | |
| "eval_loss": 0.15846052765846252, | |
| "eval_precision": 0.9053356282271945, | |
| "eval_recall": 0.8922815945716709, | |
| "eval_runtime": 11.3812, | |
| "eval_samples_per_second": 91.467, | |
| "eval_steps_per_second": 11.51, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "step": 5860, | |
| "total_flos": 3598808026592814.0, | |
| "train_loss": 0.030279771266536908, | |
| "train_runtime": 2342.7606, | |
| "train_samples_per_second": 20.0, | |
| "train_steps_per_second": 2.501 | |
| } | |
| ], | |
| "max_steps": 5860, | |
| "num_train_epochs": 5, | |
| "total_flos": 3598808026592814.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |