| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 1.8226342885780673, |
| "global_step": 5000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00015, |
| "loss": 3.0516, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.18, |
| "eval_loss": 0.7620939016342163, |
| "eval_phone_accuracy": 0.8364582892453549, |
| "eval_runtime": 14.2416, |
| "eval_samples_per_second": 14.043, |
| "eval_steps_per_second": 3.511, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.0003, |
| "loss": 0.734, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.36, |
| "eval_loss": 0.29917111992836, |
| "eval_phone_accuracy": 0.9144199432852246, |
| "eval_runtime": 14.9273, |
| "eval_samples_per_second": 13.398, |
| "eval_steps_per_second": 3.35, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00026656263932233615, |
| "loss": 0.5259, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.55, |
| "eval_loss": 0.23695439100265503, |
| "eval_phone_accuracy": 0.9256994116900157, |
| "eval_runtime": 13.1882, |
| "eval_samples_per_second": 15.165, |
| "eval_steps_per_second": 3.791, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.0002331252786446723, |
| "loss": 0.4523, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.73, |
| "eval_loss": 0.2072543352842331, |
| "eval_phone_accuracy": 0.931984593896813, |
| "eval_runtime": 14.2481, |
| "eval_samples_per_second": 14.037, |
| "eval_steps_per_second": 3.509, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00019968791796700847, |
| "loss": 0.4134, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.91, |
| "eval_loss": 0.19657564163208008, |
| "eval_phone_accuracy": 0.935307063952258, |
| "eval_runtime": 12.2098, |
| "eval_samples_per_second": 16.38, |
| "eval_steps_per_second": 4.095, |
| "step": 2500 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 0.0001662505572893446, |
| "loss": 0.3782, |
| "step": 3000 |
| }, |
| { |
| "epoch": 1.09, |
| "eval_loss": 0.18357256054878235, |
| "eval_phone_accuracy": 0.9371481779320269, |
| "eval_runtime": 11.6127, |
| "eval_samples_per_second": 17.222, |
| "eval_steps_per_second": 4.306, |
| "step": 3000 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 0.00013281319661168078, |
| "loss": 0.3544, |
| "step": 3500 |
| }, |
| { |
| "epoch": 1.28, |
| "eval_loss": 0.17402812838554382, |
| "eval_phone_accuracy": 0.9397511321792864, |
| "eval_runtime": 11.6142, |
| "eval_samples_per_second": 17.22, |
| "eval_steps_per_second": 4.305, |
| "step": 3500 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 9.937583593401693e-05, |
| "loss": 0.333, |
| "step": 4000 |
| }, |
| { |
| "epoch": 1.46, |
| "eval_loss": 0.16247966885566711, |
| "eval_phone_accuracy": 0.941782706225928, |
| "eval_runtime": 11.5423, |
| "eval_samples_per_second": 17.328, |
| "eval_steps_per_second": 4.332, |
| "step": 4000 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 6.593847525635309e-05, |
| "loss": 0.3177, |
| "step": 4500 |
| }, |
| { |
| "epoch": 1.64, |
| "eval_loss": 0.15813709795475006, |
| "eval_phone_accuracy": 0.9438354425022221, |
| "eval_runtime": 11.9528, |
| "eval_samples_per_second": 16.732, |
| "eval_steps_per_second": 4.183, |
| "step": 4500 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 3.2501114578689255e-05, |
| "loss": 0.3088, |
| "step": 5000 |
| }, |
| { |
| "epoch": 1.82, |
| "eval_loss": 0.1539994180202484, |
| "eval_phone_accuracy": 0.9445761205400601, |
| "eval_runtime": 11.5818, |
| "eval_samples_per_second": 17.269, |
| "eval_steps_per_second": 4.317, |
| "step": 5000 |
| } |
| ], |
| "max_steps": 5486, |
| "num_train_epochs": 2, |
| "total_flos": 2.4334288540465897e+19, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
