| { | |
| "best_metric": 0.9413333333333334, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0579/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": 1.0803372859954834, | |
| "learning_rate": 9e-05, | |
| "loss": 0.5774, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9341333333333334, | |
| "eval_loss": 0.24395310878753662, | |
| "eval_runtime": 11.2444, | |
| "eval_samples_per_second": 333.5, | |
| "eval_steps_per_second": 5.247, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 2.42409610748291, | |
| "learning_rate": 9e-05, | |
| "loss": 0.1103, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9354666666666667, | |
| "eval_loss": 0.2358597368001938, | |
| "eval_runtime": 11.2436, | |
| "eval_samples_per_second": 333.523, | |
| "eval_steps_per_second": 5.247, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.022000601515173912, | |
| "learning_rate": 9e-05, | |
| "loss": 0.0571, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9413333333333334, | |
| "eval_loss": 0.22172445058822632, | |
| "eval_runtime": 11.8561, | |
| "eval_samples_per_second": 316.294, | |
| "eval_steps_per_second": 4.976, | |
| "step": 999 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1332, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 4.94223961867776e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |