| { | |
| "best_metric": 0.9106666666666666, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0641/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": 7.908380508422852, | |
| "learning_rate": 6.00900900900901e-05, | |
| "loss": 0.8029, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.844, | |
| "eval_loss": 0.5099542737007141, | |
| "eval_runtime": 12.2827, | |
| "eval_samples_per_second": 305.307, | |
| "eval_steps_per_second": 4.803, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 9.180096626281738, | |
| "learning_rate": 3.0090090090090093e-05, | |
| "loss": 0.2025, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8957333333333334, | |
| "eval_loss": 0.3544307351112366, | |
| "eval_runtime": 12.2955, | |
| "eval_samples_per_second": 304.99, | |
| "eval_steps_per_second": 4.799, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.07338438928127289, | |
| "learning_rate": 9.00900900900901e-08, | |
| "loss": 0.0414, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9106666666666666, | |
| "eval_loss": 0.3123619258403778, | |
| "eval_runtime": 12.437, | |
| "eval_samples_per_second": 301.519, | |
| "eval_steps_per_second": 4.744, | |
| "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 | |
| } | |