| { | |
| "best_metric": 0.9064, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0739/checkpoints/checkpoint-999", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 999, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 22.325693130493164, | |
| "learning_rate": 6.762244255374248e-05, | |
| "loss": 1.651, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8301333333333333, | |
| "eval_loss": 0.6351649761199951, | |
| "eval_runtime": 6.0149, | |
| "eval_samples_per_second": 623.455, | |
| "eval_steps_per_second": 9.809, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.767637312412262, | |
| "learning_rate": 2.262266506445742e-05, | |
| "loss": 0.4095, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8925333333333333, | |
| "eval_loss": 0.386909157037735, | |
| "eval_runtime": 6.4806, | |
| "eval_samples_per_second": 578.648, | |
| "eval_steps_per_second": 9.104, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.8459718227386475, | |
| "learning_rate": 2.2251071493339937e-10, | |
| "loss": 0.1707, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9064, | |
| "eval_loss": 0.33924761414527893, | |
| "eval_runtime": 5.8876, | |
| "eval_samples_per_second": 636.934, | |
| "eval_steps_per_second": 10.021, | |
| "step": 999 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 999, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "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": 4.94223961867776e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |