| { | |
| "best_metric": 0.030757462605834007, | |
| "best_model_checkpoint": "./vit-base-beans/checkpoint-100", | |
| "epoch": 2.3255813953488373, | |
| "global_step": 100, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0001922480620155039, | |
| "loss": 0.5188, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00018449612403100774, | |
| "loss": 0.2328, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00017674418604651164, | |
| "loss": 0.1325, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.0001689922480620155, | |
| "loss": 0.1029, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.00016124031007751937, | |
| "loss": 0.0371, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.00015348837209302327, | |
| "loss": 0.0208, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 0.00014573643410852714, | |
| "loss": 0.0248, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.000137984496124031, | |
| "loss": 0.0826, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 0.0001302325581395349, | |
| "loss": 0.0499, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.00012248062015503876, | |
| "loss": 0.0343, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "eval_accuracy": 0.9895287958115183, | |
| "eval_loss": 0.030757462605834007, | |
| "eval_runtime": 3.0912, | |
| "eval_samples_per_second": 61.788, | |
| "eval_steps_per_second": 7.764, | |
| "step": 100 | |
| } | |
| ], | |
| "max_steps": 258, | |
| "num_train_epochs": 6, | |
| "total_flos": 1.239871833833472e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |