| { | |
| "best_metric": 0.9173333333333333, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0193/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": 19.830650329589844, | |
| "learning_rate": 0.000375, | |
| "loss": 0.783, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8173333333333334, | |
| "eval_loss": 0.6350002884864807, | |
| "eval_runtime": 7.45, | |
| "eval_samples_per_second": 503.355, | |
| "eval_steps_per_second": 7.919, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.017280852422118187, | |
| "learning_rate": 0.00012500000000000006, | |
| "loss": 0.2816, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8944, | |
| "eval_loss": 0.3488812744617462, | |
| "eval_runtime": 7.8767, | |
| "eval_samples_per_second": 476.089, | |
| "eval_steps_per_second": 7.49, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 1.7541874647140503, | |
| "learning_rate": 0.0, | |
| "loss": 0.0605, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9173333333333333, | |
| "eval_loss": 0.2743903398513794, | |
| "eval_runtime": 7.1839, | |
| "eval_samples_per_second": 521.999, | |
| "eval_steps_per_second": 8.213, | |
| "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 | |
| } | |