| { | |
| "best_metric": 0.8968, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0685/checkpoints/checkpoint-1332", | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 1332, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 1.5681734085083008, | |
| "learning_rate": 5.974873734152915e-05, | |
| "loss": 1.6127, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8216, | |
| "eval_loss": 0.6881412267684937, | |
| "eval_runtime": 6.9118, | |
| "eval_samples_per_second": 542.55, | |
| "eval_steps_per_second": 8.536, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 3.1477534770965576, | |
| "learning_rate": 3.508254927997938e-05, | |
| "loss": 0.4496, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8621333333333333, | |
| "eval_loss": 0.49533724784851074, | |
| "eval_runtime": 6.583, | |
| "eval_samples_per_second": 569.652, | |
| "eval_steps_per_second": 8.963, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.8939096927642822, | |
| "learning_rate": 1.0309702649891636e-05, | |
| "loss": 0.2132, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8845333333333333, | |
| "eval_loss": 0.39009028673171997, | |
| "eval_runtime": 7.054, | |
| "eval_samples_per_second": 531.61, | |
| "eval_steps_per_second": 8.364, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 6.415047645568848, | |
| "learning_rate": 9.734847288167535e-11, | |
| "loss": 0.1068, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8968, | |
| "eval_loss": 0.36407026648521423, | |
| "eval_runtime": 6.6001, | |
| "eval_samples_per_second": 568.173, | |
| "eval_steps_per_second": 8.939, | |
| "step": 1332 | |
| } | |
| ], | |
| "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": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 6.58965282490368e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |