|
- attack = dict(
- type='RFGSM',
- params=[
- dict(name='rfgsm_01', epsilon=0.1, alpha=0.5),
- dict(name='rfgsm_02', epsilon=0.15, alpha=0.5),
- dict(name='rfgsm_03', epsilon=0.2, alpha=0.5),
- dict(name='rfgsm_04', epsilon=0.25, alpha=0.5),
- dict(name='rfgsm_05', epsilon=0.3, alpha=0.5),
- dict(name='rfgsm_06', epsilon=0.35, alpha=0.5),
- dict(name='rfgsm_07', epsilon=0.4, alpha=0.5),
- ],
- IS_WHITE=True,
- IS_PYTORCH_WHITE=False,
- IS_DOCKER_BLACK=False,
- IS_TARGETED=False,
- IS_COMPARE_MODEL=False,
- )
- defense = dict(
- model='Models.UserModel.FP_resnet',
- path='Models/weights/FP_ResNet20.th'
- )
- evaluation = dict(
- type="ACAC"
- )
- model = dict(
- name="ResNet20",
- path='Models.UserModel.ResNet2',
- weights='Models/weights/resnet20_cifar.pt'
- )
- datasets = dict(
- type='cifar10',
- dict_path='test/dict_lists/cifar10_dict.txt',
- test_path=[
- "Datasets/CIFAR_cln_data/cifar10_30_origin_inputs.npy",
- "Datasets/CIFAR_cln_data/cifar10_30_origin_labels.npy",
- "Datasets/CIFAR_cln_data/cifar10_30_origin_inputs.npy",
- "Datasets/CIFAR_cln_data/cifar10_30_origin_labels.npy",
- ],
- augment=dict(
- Crop_ImageSize=(32, 32),
- Scale_ImageSize=(32, 32)
- ),
- batch_size=4,
- CAM_layer=28,
- )
- result = dict(
- IS_SAVE=False,
- save_path='test/Attack_generation/',
- save_method='.npy',
- save_visualization_base_path='test/temp/',
- black_Result_dir='..',
- )
- gpu = dict(
- nums=2,
- index='0,1'
- )
|