| { | |
| "best_metric": 0.9424, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0478/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": 0.6568130254745483, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 0.6558, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9202666666666667, | |
| "eval_loss": 0.2776768207550049, | |
| "eval_runtime": 6.2212, | |
| "eval_samples_per_second": 602.778, | |
| "eval_steps_per_second": 9.484, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 1.9522514343261719, | |
| "learning_rate": 2.5136294516063797e-05, | |
| "loss": 0.1352, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9362666666666667, | |
| "eval_loss": 0.20813514292240143, | |
| "eval_runtime": 6.4152, | |
| "eval_samples_per_second": 584.546, | |
| "eval_steps_per_second": 9.197, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.38482269644737244, | |
| "learning_rate": 2.472341277037771e-10, | |
| "loss": 0.0391, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9424, | |
| "eval_loss": 0.19482260942459106, | |
| "eval_runtime": 6.1483, | |
| "eval_samples_per_second": 609.928, | |
| "eval_steps_per_second": 9.596, | |
| "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 | |
| } | |