| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.9411362080766599, |
| "global_step": 5500, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.09, |
| "learning_rate": 3.7713894592744696e-06, |
| "loss": 0.0, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 3.4291581108829565e-06, |
| "loss": 0.0, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 3.086926762491444e-06, |
| "loss": 0.0, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 2.744695414099931e-06, |
| "loss": 0.0, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 2.402464065708419e-06, |
| "loss": 0.0, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 2.060232717316906e-06, |
| "loss": 0.0, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 1.7180013689253934e-06, |
| "loss": 0.0, |
| "step": 3500 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 1.375770020533881e-06, |
| "loss": 0.0, |
| "step": 4000 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 1.0335386721423683e-06, |
| "loss": 0.0, |
| "step": 4500 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 6.913073237508556e-07, |
| "loss": 0.0, |
| "step": 5000 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 3.490759753593429e-07, |
| "loss": 0.0, |
| "step": 5500 |
| } |
| ], |
| "max_steps": 5844, |
| "num_train_epochs": 1, |
| "total_flos": 9.373616201532211e+16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
