| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 7.0, | |
| "eval_steps": 500, | |
| "global_step": 13755, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.8857142857142856e-05, | |
| "loss": 0.5664, | |
| "step": 1965 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7443843461660834, | |
| "eval_f1": 0.4215982721382289, | |
| "eval_false_negatives": 3699, | |
| "eval_false_positives": 318, | |
| "eval_loss": 0.5374670624732971, | |
| "eval_precision": 0.8215488215488216, | |
| "eval_recall": 0.28355607205113303, | |
| "eval_runtime": 750.1293, | |
| "eval_samples_per_second": 20.95, | |
| "eval_steps_per_second": 0.656, | |
| "eval_true_negatives": 10234, | |
| "eval_true_positives": 1464, | |
| "step": 1965 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.5714285714285715e-05, | |
| "loss": 0.4873, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7466751511294941, | |
| "eval_f1": 0.5101513473606497, | |
| "eval_false_negatives": 3090, | |
| "eval_false_positives": 891, | |
| "eval_loss": 0.5381917357444763, | |
| "eval_precision": 0.6993927125506073, | |
| "eval_recall": 0.40151074956420685, | |
| "eval_runtime": 749.6591, | |
| "eval_samples_per_second": 20.963, | |
| "eval_steps_per_second": 0.656, | |
| "eval_true_negatives": 9661, | |
| "eval_true_positives": 2073, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 1.257142857142857e-05, | |
| "loss": 0.3802, | |
| "step": 5895 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7215399300031817, | |
| "eval_f1": 0.5454923140839218, | |
| "eval_false_negatives": 2537, | |
| "eval_false_positives": 1839, | |
| "eval_loss": 0.6351324319839478, | |
| "eval_precision": 0.5881298992161255, | |
| "eval_recall": 0.5086190199496416, | |
| "eval_runtime": 749.7171, | |
| "eval_samples_per_second": 20.961, | |
| "eval_steps_per_second": 0.656, | |
| "eval_true_negatives": 8713, | |
| "eval_true_positives": 2626, | |
| "step": 5895 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 9.428571428571428e-06, | |
| "loss": 0.2694, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.725739739102768, | |
| "eval_f1": 0.5421712343318462, | |
| "eval_false_negatives": 2611, | |
| "eval_false_positives": 1699, | |
| "eval_loss": 0.761067271232605, | |
| "eval_precision": 0.6003293342742884, | |
| "eval_recall": 0.49428626767383305, | |
| "eval_runtime": 750.8442, | |
| "eval_samples_per_second": 20.93, | |
| "eval_steps_per_second": 0.655, | |
| "eval_true_negatives": 8853, | |
| "eval_true_positives": 2552, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 6.285714285714285e-06, | |
| "loss": 0.1848, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.7166401527203309, | |
| "eval_f1": 0.5417309869301224, | |
| "eval_false_negatives": 2531, | |
| "eval_false_positives": 1922, | |
| "eval_loss": 0.9724581241607666, | |
| "eval_precision": 0.577953447518665, | |
| "eval_recall": 0.5097811349990315, | |
| "eval_runtime": 749.9897, | |
| "eval_samples_per_second": 20.954, | |
| "eval_steps_per_second": 0.656, | |
| "eval_true_negatives": 8630, | |
| "eval_true_positives": 2632, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.1428571428571425e-06, | |
| "loss": 0.1305, | |
| "step": 11790 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.7165128857779192, | |
| "eval_f1": 0.5395348837209303, | |
| "eval_false_negatives": 2553, | |
| "eval_false_positives": 1902, | |
| "eval_loss": 1.1839183568954468, | |
| "eval_precision": 0.5784574468085106, | |
| "eval_recall": 0.505520046484602, | |
| "eval_runtime": 816.6636, | |
| "eval_samples_per_second": 19.243, | |
| "eval_steps_per_second": 0.602, | |
| "eval_true_negatives": 8650, | |
| "eval_true_positives": 2610, | |
| "step": 11790 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.0948, | |
| "step": 13755 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.7077951002227172, | |
| "eval_f1": 0.5435387673956262, | |
| "eval_false_negatives": 2429, | |
| "eval_false_positives": 2163, | |
| "eval_loss": 1.3593369722366333, | |
| "eval_precision": 0.55830100061262, | |
| "eval_recall": 0.5295370908386597, | |
| "eval_runtime": 766.2151, | |
| "eval_samples_per_second": 20.51, | |
| "eval_steps_per_second": 0.642, | |
| "eval_true_negatives": 8389, | |
| "eval_true_positives": 2734, | |
| "step": 13755 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 13755, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 7, | |
| "save_steps": 500, | |
| "total_flos": 5.828645021303194e+16, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |