| { | |
| "best_metric": 0.9106666666666666, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0417/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": 10.762717247009277, | |
| "learning_rate": 8.895562228895563e-05, | |
| "loss": 0.9144, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8104, | |
| "eval_loss": 0.6309072971343994, | |
| "eval_runtime": 7.2846, | |
| "eval_samples_per_second": 514.784, | |
| "eval_steps_per_second": 8.099, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 34.901371002197266, | |
| "learning_rate": 7.784451117784451e-05, | |
| "loss": 0.3333, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8229333333333333, | |
| "eval_loss": 0.6075742244720459, | |
| "eval_runtime": 7.3093, | |
| "eval_samples_per_second": 513.042, | |
| "eval_steps_per_second": 8.072, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 86.8698501586914, | |
| "learning_rate": 6.67334000667334e-05, | |
| "loss": 0.1852, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8576, | |
| "eval_loss": 0.5035261511802673, | |
| "eval_runtime": 7.2875, | |
| "eval_samples_per_second": 514.58, | |
| "eval_steps_per_second": 8.096, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.5270398855209351, | |
| "learning_rate": 5.562228895562229e-05, | |
| "loss": 0.1096, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8672, | |
| "eval_loss": 0.5222706198692322, | |
| "eval_runtime": 7.0125, | |
| "eval_samples_per_second": 534.757, | |
| "eval_steps_per_second": 8.414, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.0034986543469130993, | |
| "learning_rate": 4.451117784451118e-05, | |
| "loss": 0.0673, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8789333333333333, | |
| "eval_loss": 0.4766990840435028, | |
| "eval_runtime": 6.994, | |
| "eval_samples_per_second": 536.171, | |
| "eval_steps_per_second": 8.436, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.31739485263824463, | |
| "learning_rate": 3.340006673340007e-05, | |
| "loss": 0.0345, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8904, | |
| "eval_loss": 0.44725093245506287, | |
| "eval_runtime": 7.1293, | |
| "eval_samples_per_second": 526.0, | |
| "eval_steps_per_second": 8.276, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.008587081916630268, | |
| "learning_rate": 2.2288955622288957e-05, | |
| "loss": 0.0148, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9008, | |
| "eval_loss": 0.46140801906585693, | |
| "eval_runtime": 7.4573, | |
| "eval_samples_per_second": 502.861, | |
| "eval_steps_per_second": 7.912, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.4167099595069885, | |
| "learning_rate": 1.1177844511177845e-05, | |
| "loss": 0.0046, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9096, | |
| "eval_loss": 0.425393283367157, | |
| "eval_runtime": 7.0743, | |
| "eval_samples_per_second": 530.088, | |
| "eval_steps_per_second": 8.34, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.011372236534953117, | |
| "learning_rate": 6.673340006673341e-08, | |
| "loss": 0.0009, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9106666666666666, | |
| "eval_loss": 0.4087434709072113, | |
| "eval_runtime": 7.1306, | |
| "eval_samples_per_second": 525.901, | |
| "eval_steps_per_second": 8.274, | |
| "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 | |
| } | |