| { | |
| "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": 2.8412, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_loss": 1.5813477039337158, | |
| "eval_phone_accuracy": 0.5792233379941528, | |
| "eval_runtime": 11.4324, | |
| "eval_samples_per_second": 17.494, | |
| "eval_steps_per_second": 4.374, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003, | |
| "loss": 1.482, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_loss": 0.8674991726875305, | |
| "eval_phone_accuracy": 0.7458370408033558, | |
| "eval_runtime": 10.836, | |
| "eval_samples_per_second": 18.457, | |
| "eval_steps_per_second": 4.614, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00026656263932233615, | |
| "loss": 1.0056, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_loss": 0.5736156702041626, | |
| "eval_phone_accuracy": 0.8264268463200711, | |
| "eval_runtime": 10.9515, | |
| "eval_samples_per_second": 18.262, | |
| "eval_steps_per_second": 4.566, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0002331252786446723, | |
| "loss": 0.7876, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_loss": 0.43626296520233154, | |
| "eval_phone_accuracy": 0.8651752044404898, | |
| "eval_runtime": 11.1035, | |
| "eval_samples_per_second": 18.012, | |
| "eval_steps_per_second": 4.503, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.00019968791796700847, | |
| "loss": 0.6745, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_loss": 0.3738095760345459, | |
| "eval_phone_accuracy": 0.8828545400618618, | |
| "eval_runtime": 10.7535, | |
| "eval_samples_per_second": 18.599, | |
| "eval_steps_per_second": 4.65, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.0001662505572893446, | |
| "loss": 0.5965, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "eval_loss": 0.3428588807582855, | |
| "eval_phone_accuracy": 0.8905342993940935, | |
| "eval_runtime": 10.807, | |
| "eval_samples_per_second": 18.507, | |
| "eval_steps_per_second": 4.627, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.00013281319661168078, | |
| "loss": 0.5429, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "eval_loss": 0.29903942346572876, | |
| "eval_phone_accuracy": 0.9031926613279099, | |
| "eval_runtime": 11.2035, | |
| "eval_samples_per_second": 17.852, | |
| "eval_steps_per_second": 4.463, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 9.937583593401693e-05, | |
| "loss": 0.5022, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "eval_loss": 0.2947227954864502, | |
| "eval_phone_accuracy": 0.9031185119274607, | |
| "eval_runtime": 10.8322, | |
| "eval_samples_per_second": 18.464, | |
| "eval_steps_per_second": 4.616, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 6.593847525635309e-05, | |
| "loss": 0.4767, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "eval_loss": 0.2698778808116913, | |
| "eval_phone_accuracy": 0.9127261556713698, | |
| "eval_runtime": 11.228, | |
| "eval_samples_per_second": 17.813, | |
| "eval_steps_per_second": 4.453, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 3.2501114578689255e-05, | |
| "loss": 0.4565, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "eval_loss": 0.2591952979564667, | |
| "eval_phone_accuracy": 0.9146540400830473, | |
| "eval_runtime": 11.459, | |
| "eval_samples_per_second": 17.454, | |
| "eval_steps_per_second": 4.363, | |
| "step": 5000 | |
| } | |
| ], | |
| "max_steps": 5486, | |
| "num_train_epochs": 2, | |
| "total_flos": 1.8035219197147087e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |