| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 24.137931034482758, | |
| "global_step": 3500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 5e-05, | |
| "loss": 3.4838, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "eval_loss": 1.1867836713790894, | |
| "eval_runtime": 80.0488, | |
| "eval_samples_per_second": 20.987, | |
| "eval_steps_per_second": 2.623, | |
| "eval_wer": 0.8338501826200814, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 0.0001, | |
| "loss": 0.5623, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "eval_loss": 0.4056074321269989, | |
| "eval_runtime": 78.6534, | |
| "eval_samples_per_second": 21.36, | |
| "eval_steps_per_second": 2.67, | |
| "eval_wer": 0.3664116876852043, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 10.34, | |
| "learning_rate": 8.507462686567164e-05, | |
| "loss": 0.2018, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 10.34, | |
| "eval_loss": 0.407957524061203, | |
| "eval_runtime": 78.7154, | |
| "eval_samples_per_second": 21.343, | |
| "eval_steps_per_second": 2.668, | |
| "eval_wer": 0.3300254978981462, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 13.79, | |
| "learning_rate": 7.014925373134329e-05, | |
| "loss": 0.1217, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 13.79, | |
| "eval_loss": 0.38568004965782166, | |
| "eval_runtime": 78.9883, | |
| "eval_samples_per_second": 21.269, | |
| "eval_steps_per_second": 2.659, | |
| "eval_wer": 0.3065949968989043, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 17.24, | |
| "learning_rate": 5.5223880597014934e-05, | |
| "loss": 0.0829, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 17.24, | |
| "eval_loss": 0.42486006021499634, | |
| "eval_runtime": 79.0665, | |
| "eval_samples_per_second": 21.248, | |
| "eval_steps_per_second": 2.656, | |
| "eval_wer": 0.2981186685962373, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 20.69, | |
| "learning_rate": 4.029850746268657e-05, | |
| "loss": 0.0626, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 20.69, | |
| "eval_loss": 0.4198434352874756, | |
| "eval_runtime": 78.3866, | |
| "eval_samples_per_second": 21.432, | |
| "eval_steps_per_second": 2.679, | |
| "eval_wer": 0.2981186685962373, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 24.14, | |
| "learning_rate": 2.537313432835821e-05, | |
| "loss": 0.0488, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 24.14, | |
| "eval_loss": 0.4277154803276062, | |
| "eval_runtime": 78.4897, | |
| "eval_samples_per_second": 21.404, | |
| "eval_steps_per_second": 2.676, | |
| "eval_wer": 0.29115843153469784, | |
| "step": 3500 | |
| } | |
| ], | |
| "max_steps": 4350, | |
| "num_train_epochs": 30, | |
| "total_flos": 3.2107549243060224e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |