| { | |
| "best_metric": 0.3506666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0699/checkpoints/checkpoint-2997", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 2997, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 20.064695358276367, | |
| "learning_rate": 0.0004451117784451118, | |
| "loss": 3.7634, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.1344, | |
| "eval_loss": 3.3126235008239746, | |
| "eval_runtime": 11.8494, | |
| "eval_samples_per_second": 316.473, | |
| "eval_steps_per_second": 4.979, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 13.832483291625977, | |
| "learning_rate": 0.0003895562228895562, | |
| "loss": 3.1858, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.1856, | |
| "eval_loss": 3.1087238788604736, | |
| "eval_runtime": 11.3158, | |
| "eval_samples_per_second": 331.395, | |
| "eval_steps_per_second": 5.214, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 15.541133880615234, | |
| "learning_rate": 0.0003340006673340007, | |
| "loss": 2.982, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.2208, | |
| "eval_loss": 2.9015910625457764, | |
| "eval_runtime": 11.8961, | |
| "eval_samples_per_second": 315.229, | |
| "eval_steps_per_second": 4.96, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 18.539718627929688, | |
| "learning_rate": 0.0002784451117784451, | |
| "loss": 2.8144, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.26826666666666665, | |
| "eval_loss": 2.7212135791778564, | |
| "eval_runtime": 11.9993, | |
| "eval_samples_per_second": 312.518, | |
| "eval_steps_per_second": 4.917, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 27.653806686401367, | |
| "learning_rate": 0.00022288955622288957, | |
| "loss": 2.6496, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.2976, | |
| "eval_loss": 2.600856304168701, | |
| "eval_runtime": 11.7047, | |
| "eval_samples_per_second": 320.384, | |
| "eval_steps_per_second": 5.041, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 22.705095291137695, | |
| "learning_rate": 0.00016733400066733402, | |
| "loss": 2.476, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.31146666666666667, | |
| "eval_loss": 2.511509656906128, | |
| "eval_runtime": 11.7973, | |
| "eval_samples_per_second": 317.869, | |
| "eval_steps_per_second": 5.001, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 26.835336685180664, | |
| "learning_rate": 0.00011177844511177845, | |
| "loss": 2.2885, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.33386666666666664, | |
| "eval_loss": 2.4169158935546875, | |
| "eval_runtime": 11.8247, | |
| "eval_samples_per_second": 317.132, | |
| "eval_steps_per_second": 4.99, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 16.47260284423828, | |
| "learning_rate": 5.622288955622289e-05, | |
| "loss": 2.0424, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.3466666666666667, | |
| "eval_loss": 2.38747501373291, | |
| "eval_runtime": 11.8728, | |
| "eval_samples_per_second": 315.848, | |
| "eval_steps_per_second": 4.969, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 31.294137954711914, | |
| "learning_rate": 6.67334000667334e-07, | |
| "loss": 1.7559, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.3506666666666667, | |
| "eval_loss": 2.4410762786865234, | |
| "eval_runtime": 12.1188, | |
| "eval_samples_per_second": 309.438, | |
| "eval_steps_per_second": 4.868, | |
| "step": 2997 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2997, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9, | |
| "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": 1.482671885603328e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |