Skip to content
Snippets Groups Projects
Normal view Permalink
CIFAR10-AIR-__FGSM____PGD__.log 3.27 KiB
Newer Older
Mina Moshfegh's avatar
Upload New File
Mina Moshfegh committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 
[2025-02-20 21:24:28] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] === Configuration ===
[2025-02-20 21:24:28] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Config(seed=42, device='cuda', epochs=1, batch_size=128, learning_rate=0.01, momentum=0.9, weight_decay=0.0005, epsilon=0.03137254901960784, alpha=0.00784313725490196, num_steps=10, random_init=True, defense_method='AIR', lambda_SD=1.0, lambda_IR=1.0, lambda_AR=1.0, lambda_Reg=1.0, alpha_range=(0.3, 0.7), use_rdrop=True, iso_noise_std=0.01, iso_clamp_min=-3, iso_clamp_max=3, iso_p_flip=0.5, iso_flip_dim=3, iso_p_rotation=0.5, iso_max_rotation=10, iso_p_crop=0.5, iso_p_erase=0.5, dataset='CIFAR10', data_root='./data', num_workers=2, lambda_lfl=1.0, feature_lambda=1.0, freeze_classifier=True, joint_lambda=0.5, adv_lambda=0.8, feat_lambda=1.0, lambda_ewc=100.0, attack_sequence=('FGSM', 'PGD'))
[2025-02-20 21:24:31] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] --- Task 1 / 2: Attack = FGSM ---
[2025-02-20 21:24:31] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Using AIR Defense...
[2025-02-20 21:24:32] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 0, Loss: 2.3806
[2025-02-20 21:25:45] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 100, Loss: 2.1954
[2025-02-20 21:27:00] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 200, Loss: 2.0442
[2025-02-20 21:28:14] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 300, Loss: 1.7253
[2025-02-20 21:29:21] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Epoch [1/1] - Train Loss: 2.0699
[2025-02-20 21:29:24] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Clean Test Accuracy: 38.37%
[2025-02-20 21:29:24] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Model saved at: ./models/CIFAR10_AIR_FGSM_task0.pt
[2025-02-20 21:29:27] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Clean Accuracy after Task 1, Attack=FGSM: 38.37%
[2025-02-20 21:29:38] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Robust Accuracy on FGSM after Task 1: 17.81%
[2025-02-20 21:30:58] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Robust Accuracy on PGD after Task 1: 16.97%
[2025-02-20 21:30:58] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] --- Task 2 / 2: Attack = PGD ---
[2025-02-20 21:30:58] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Using AIR Defense...
[2025-02-20 21:31:00] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 0, Loss: 1.7935
[2025-02-20 21:33:49] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 100, Loss: 3.0950
[2025-02-20 21:36:37] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 200, Loss: 2.5866
[2025-02-20 21:39:25] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Batch 300, Loss: 2.1758
[2025-02-20 21:41:56] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Epoch [1/1] - Train Loss: 2.7837
[2025-02-20 21:41:58] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Clean Test Accuracy: 10.77%
[2025-02-20 21:41:59] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Model saved at: ./models/CIFAR10_AIR_PGD_task1.pt
[2025-02-20 21:42:01] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Clean Accuracy after Task 2, Attack=PGD: 10.77%
[2025-02-20 21:42:12] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Robust Accuracy on FGSM after Task 2: 34.21%
[2025-02-20 21:43:33] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Robust Accuracy on PGD after Task 2: 29.21%
[2025-02-20 21:43:33] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] === All tasks completed! ===
[2025-02-20 21:43:33] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] === Final Results ===
[2025-02-20 21:43:33] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] []
[2025-02-20 21:43:33] [CIFAR10-AIR-('FGSM', 'PGD')] [INFO] Done!