| { | |
| "best_metric": 0.3898809254169464, | |
| "best_model_checkpoint": "/tmp/model/checkpoint-2000", | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 2000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 5e-06, | |
| "loss": 0.7412, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1e-05, | |
| "loss": 0.5394, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.5e-05, | |
| "loss": 0.4858, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2e-05, | |
| "loss": 0.4178, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.5e-05, | |
| "loss": 0.3928, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3e-05, | |
| "loss": 0.4125, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.5e-05, | |
| "loss": 0.3896, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4e-05, | |
| "loss": 0.3878, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.5e-05, | |
| "loss": 0.3794, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5e-05, | |
| "loss": 0.4038, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.9444444444444446e-05, | |
| "loss": 0.399, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.888888888888889e-05, | |
| "loss": 0.393, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 4.8333333333333334e-05, | |
| "loss": 0.4145, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.7777777777777784e-05, | |
| "loss": 0.3391, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.722222222222222e-05, | |
| "loss": 0.4212, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.666666666666667e-05, | |
| "loss": 0.3693, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 4.6111111111111115e-05, | |
| "loss": 0.3795, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.555555555555556e-05, | |
| "loss": 0.368, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 4.5e-05, | |
| "loss": 0.3376, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.3751, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.86225, | |
| "eval_auc": 0.8955589113728649, | |
| "eval_f1": 0.7137662337662338, | |
| "eval_loss": 0.3898809254169464, | |
| "eval_precision": 0.73475935828877, | |
| "eval_recall": 0.693939393939394, | |
| "eval_runtime": 640.7343, | |
| "eval_samples_per_second": 6.243, | |
| "eval_steps_per_second": 0.39, | |
| "step": 2000 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 10000, | |
| "num_train_epochs": 5, | |
| "save_steps": 500, | |
| "total_flos": 1052444221440000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |