| { | |
| "best_metric": 0.9290666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0393/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": 39.2876091003418, | |
| "learning_rate": 2.0030030030030034e-05, | |
| "loss": 0.7197, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8866666666666667, | |
| "eval_loss": 0.3398627042770386, | |
| "eval_runtime": 6.9052, | |
| "eval_samples_per_second": 543.071, | |
| "eval_steps_per_second": 8.544, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 53.6608772277832, | |
| "learning_rate": 1.0030030030030031e-05, | |
| "loss": 0.1114, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9208, | |
| "eval_loss": 0.24425150454044342, | |
| "eval_runtime": 6.7032, | |
| "eval_samples_per_second": 559.438, | |
| "eval_steps_per_second": 8.802, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.0522213838994503, | |
| "learning_rate": 3.003003003003003e-08, | |
| "loss": 0.0192, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9290666666666667, | |
| "eval_loss": 0.232973113656044, | |
| "eval_runtime": 6.7401, | |
| "eval_samples_per_second": 556.374, | |
| "eval_steps_per_second": 8.754, | |
| "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 | |
| } | |