| { | |
| "best_metric": 0.9373333333333334, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0697/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": 12.061416625976562, | |
| "learning_rate": 6.666666666666667e-05, | |
| "loss": 0.6361, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9221333333333334, | |
| "eval_loss": 0.2635498642921448, | |
| "eval_runtime": 7.5503, | |
| "eval_samples_per_second": 496.67, | |
| "eval_steps_per_second": 7.814, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.21975940465927124, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.1266, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9301333333333334, | |
| "eval_loss": 0.2408057451248169, | |
| "eval_runtime": 7.7983, | |
| "eval_samples_per_second": 480.873, | |
| "eval_steps_per_second": 7.566, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.18101432919502258, | |
| "learning_rate": 0.0, | |
| "loss": 0.0445, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9373333333333334, | |
| "eval_loss": 0.22564542293548584, | |
| "eval_runtime": 7.6396, | |
| "eval_samples_per_second": 490.862, | |
| "eval_steps_per_second": 7.723, | |
| "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 | |
| } | |