| { | |
| "best_metric": 0.8629333333333333, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0568/checkpoints/checkpoint-2997", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 2997, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 88.83711242675781, | |
| "learning_rate": 9.698463103929542e-05, | |
| "loss": 2.6722, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7328, | |
| "eval_loss": 0.922845184803009, | |
| "eval_runtime": 17.2373, | |
| "eval_samples_per_second": 217.551, | |
| "eval_steps_per_second": 3.423, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 162.32318115234375, | |
| "learning_rate": 8.83022221559489e-05, | |
| "loss": 0.6886, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8056, | |
| "eval_loss": 0.5906055569648743, | |
| "eval_runtime": 17.4102, | |
| "eval_samples_per_second": 215.391, | |
| "eval_steps_per_second": 3.389, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 125.2595443725586, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 0.4502, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8370666666666666, | |
| "eval_loss": 0.4952096939086914, | |
| "eval_runtime": 17.4727, | |
| "eval_samples_per_second": 214.621, | |
| "eval_steps_per_second": 3.377, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 39.090694427490234, | |
| "learning_rate": 5.868240888334653e-05, | |
| "loss": 0.3343, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8517333333333333, | |
| "eval_loss": 0.44834887981414795, | |
| "eval_runtime": 17.3476, | |
| "eval_samples_per_second": 216.168, | |
| "eval_steps_per_second": 3.401, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 54.005367279052734, | |
| "learning_rate": 4.131759111665349e-05, | |
| "loss": 0.249, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8616, | |
| "eval_loss": 0.43427228927612305, | |
| "eval_runtime": 17.2431, | |
| "eval_samples_per_second": 217.478, | |
| "eval_steps_per_second": 3.422, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 58.777793884277344, | |
| "learning_rate": 2.500000000000001e-05, | |
| "loss": 0.2038, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.86, | |
| "eval_loss": 0.4400232136249542, | |
| "eval_runtime": 17.5758, | |
| "eval_samples_per_second": 213.361, | |
| "eval_steps_per_second": 3.357, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 77.77790069580078, | |
| "learning_rate": 1.1697777844051105e-05, | |
| "loss": 0.1693, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8597333333333333, | |
| "eval_loss": 0.4400482475757599, | |
| "eval_runtime": 17.2745, | |
| "eval_samples_per_second": 217.083, | |
| "eval_steps_per_second": 3.415, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 92.69695281982422, | |
| "learning_rate": 3.0153689607045845e-06, | |
| "loss": 0.1506, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8618666666666667, | |
| "eval_loss": 0.4376513361930847, | |
| "eval_runtime": 17.1121, | |
| "eval_samples_per_second": 219.143, | |
| "eval_steps_per_second": 3.448, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 52.924644470214844, | |
| "learning_rate": 0.0, | |
| "loss": 0.1439, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8629333333333333, | |
| "eval_loss": 0.4386611580848694, | |
| "eval_runtime": 15.1142, | |
| "eval_samples_per_second": 248.11, | |
| "eval_steps_per_second": 3.904, | |
| "step": 2997 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2997, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 7.3587863155968e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |