| { | |
| "best_metric": 0.9570666666666666, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0161/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": 20.831274032592773, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.8453, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9474666666666667, | |
| "eval_loss": 0.2085845172405243, | |
| "eval_runtime": 5.9177, | |
| "eval_samples_per_second": 633.689, | |
| "eval_steps_per_second": 9.97, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.12023206800222397, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 0.1026, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9522666666666667, | |
| "eval_loss": 0.17338864505290985, | |
| "eval_runtime": 6.2104, | |
| "eval_samples_per_second": 603.824, | |
| "eval_steps_per_second": 9.5, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.437683641910553, | |
| "learning_rate": 0.0, | |
| "loss": 0.0301, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9570666666666666, | |
| "eval_loss": 0.15266525745391846, | |
| "eval_runtime": 5.71, | |
| "eval_samples_per_second": 656.74, | |
| "eval_steps_per_second": 10.333, | |
| "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 | |
| } | |