| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 99.61089494163424, | |
| "global_step": 3200, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 12.45, | |
| "learning_rate": 0.00023999999999999998, | |
| "loss": 3.6403, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 12.45, | |
| "eval_loss": 0.3551729917526245, | |
| "eval_runtime": 131.9123, | |
| "eval_samples_per_second": 9.082, | |
| "eval_steps_per_second": 1.137, | |
| "eval_wer": 0.5929875585716594, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 24.9, | |
| "learning_rate": 0.0002666666666666666, | |
| "loss": 0.292, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 24.9, | |
| "eval_loss": 0.28114205598831177, | |
| "eval_runtime": 130.8031, | |
| "eval_samples_per_second": 9.159, | |
| "eval_steps_per_second": 1.147, | |
| "eval_wer": 0.4598481176280498, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 37.35, | |
| "learning_rate": 0.00022222222222222218, | |
| "loss": 0.1561, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 37.35, | |
| "eval_loss": 0.32178691029548645, | |
| "eval_runtime": 131.3931, | |
| "eval_samples_per_second": 9.118, | |
| "eval_steps_per_second": 1.142, | |
| "eval_wer": 0.438762320245597, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 49.81, | |
| "learning_rate": 0.00017777777777777776, | |
| "loss": 0.1001, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 49.81, | |
| "eval_loss": 0.3630038797855377, | |
| "eval_runtime": 131.8456, | |
| "eval_samples_per_second": 9.086, | |
| "eval_steps_per_second": 1.138, | |
| "eval_wer": 0.4234932945548554, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 62.26, | |
| "learning_rate": 0.0001333333333333333, | |
| "loss": 0.0715, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 62.26, | |
| "eval_loss": 0.4096166789531708, | |
| "eval_runtime": 131.7338, | |
| "eval_samples_per_second": 9.094, | |
| "eval_steps_per_second": 1.139, | |
| "eval_wer": 0.4266440458878656, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 74.71, | |
| "learning_rate": 8.888888888888888e-05, | |
| "loss": 0.0549, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 74.71, | |
| "eval_loss": 0.42000046372413635, | |
| "eval_runtime": 131.9589, | |
| "eval_samples_per_second": 9.079, | |
| "eval_steps_per_second": 1.137, | |
| "eval_wer": 0.4240588140248829, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 87.16, | |
| "learning_rate": 4.444444444444444e-05, | |
| "loss": 0.0446, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 87.16, | |
| "eval_loss": 0.43966689705848694, | |
| "eval_runtime": 132.8166, | |
| "eval_samples_per_second": 9.02, | |
| "eval_steps_per_second": 1.129, | |
| "eval_wer": 0.42317014057198254, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 99.61, | |
| "learning_rate": 0.0, | |
| "loss": 0.0408, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 99.61, | |
| "eval_loss": 0.435197114944458, | |
| "eval_runtime": 132.4475, | |
| "eval_samples_per_second": 9.045, | |
| "eval_steps_per_second": 1.133, | |
| "eval_wer": 0.4142834060429795, | |
| "step": 3200 | |
| } | |
| ], | |
| "max_steps": 3200, | |
| "num_train_epochs": 100, | |
| "total_flos": 1.9009113406470115e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |