Error after trying to evaluate my RaResNet50 on pgd

[headscott]

I trained a RaResNet50, by using make experiments/RaResNet50/.done_train

Afterwards, I wanted to evaluate it with make experiments/RaResNet50/.done_test_pgd, but it gave me this error:

Error executing job with overrides: ['train_test=test', '++name=RaResNet50', '++visualization.entity=', '++attack.train.eps=5', '++attack.test.eps=4', '++attack.test.eps=2', '++attack.test.gamma=1', '++attack.test.step=10', 'model=model', '++model.kwargs.depths=[5,8,13,1]', '++model.kwargs.stage_widths=[288,576,1120,2160]', '++model.kwargs.group_widths=[36,72,140,270]', '++model.kwargs.stem_downsample_factor=2', '++model.kwargs.strides.0=2', '++model.kwargs.stem_width=96', '++model.kwargs.se_ratio=0.25', '++model.kwargs.se_activation._target_=hydra.utils.get_class', '++model.kwargs.se_activation.path=torch.nn.ReLU', '++model.kwargs.activation_layer.0.path=torch.nn.SiLU', '++model.kwargs.activation_layer.1.path=torch.nn.SiLU', '++model.kwargs.activation_layer.2.path=torch.nn.SiLU', '++model.kwargs.norm_layer.0.path=torch.nn.Identity']
Traceback (most recent call last):
  File "/home/theiner/scripts/ModelleRobustbench/robust-principles/robustarch/main.py", line 107, in main
    train_loader, test_loader = get_datasets(cfg)
  File "/home/theiner/scripts/ModelleRobustbench/robust-principles/robustarch/utils.py", line 283, in get_datasets
    return get_imagenet(
  File "/home/theiner/scripts/ModelleRobustbench/robust-principles/robustarch/utils.py", line 260, in get_imagenet
    train_dataset = datasets.ImageFolder(root=train_path, transform=train_transform)
  File "/home/theiner/anaconda3/envs/ra-principles/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 328, in __init__
    super().__init__(
  File "/home/theiner/anaconda3/envs/ra-principles/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 149, in __init__
    classes, class_to_idx = self.find_classes(self.root)
  File "/home/theiner/anaconda3/envs/ra-principles/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 234, in find_classes
    return find_classes(directory)
  File "/home/theiner/anaconda3/envs/ra-principles/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 41, in find_classes
    classes = sorted(entry.name for entry in os.scandir(directory) if entry.is_dir())
FileNotFoundError: [Errno 2] No such file or directory: '../imagenet/train'

how can I fix it? Training just worked fine, so the path of ImageNet should be correct

[headscott]

Okay I think I could fix it: I had to add ++dataset.data_dir="$(BASE)/imagenet" to TEST in the Makefile:

so from

TEST = cd $(SRC) && $(PYTHON) -m main train_test=test ++name=$(NAME) ++visualization.entity=$(WANDB_ACCOUNT)
to

TEST = cd $(SRC) && $(PYTHON) -m main train_test=test ++name=$(NAME) ++visualization.entity=$(WANDB_ACCOUNT) ++dataset.data_dir="$(BASE)/imagenet"