Skip to content

No shuffle in eval #5

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
hadarshavit wants to merge 2 commits into
base: dev
Choose a base branch
from
Open

No shuffle in eval #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3406ec1
No shuffle in eval
hadarshavit Mar 14, 2025
23b843f
Dont shuffle val loader
hadarshavit Mar 15, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions CTRAIN/data_loaders/data_loaders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,7 @@ def load_mnist(batch_size=64, normalise=True, train_transforms=[], val_split=Tru
test_dataset = datasets.MNIST(root=data_root, train=False, transform=test_transform)
if val_split:
train_dataset, val_dataset = random_split(train_dataset, [0.8, 0.2])
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=True, num_workers=2)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=2)


train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=2)
Expand Down Expand Up @@ -137,7 +137,7 @@ def load_cifar10(batch_size=64, normalise=True, train_transforms=[transforms.Ran
test_dataset = datasets.CIFAR10(root=data_root, train=False, transform=test_transform, download=True)
if val_split:
train_dataset, val_dataset = random_split(train_dataset, [0.8, 0.2])
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=True, num_workers=2)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=2)
val_loader.mean, val_loader.std = mean, std


Expand Down Expand Up @@ -229,7 +229,7 @@ def load_gtsrb(batch_size=64, normalise=True, train_transforms=[transforms.Rando
train_dataset = Subset(train_dataset_ori, train_ids)
val_dataset = Subset(train_dataset_ori, val_ids)

val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False)
val_loader.mean, val_loader.std = mean, std

else:
Expand Down Expand Up @@ -371,7 +371,7 @@ def load_tinyimagenet(batch_size=64, normalise=True, train_transforms=[transform
test_dataset = datasets.ImageFolder(root=data_root + '/tiny-imagenet-200/val/images', transform=test_transform)
if val_split:
train_dataset, val_dataset = random_split(train_dataset, [0.8, 0.2])
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=True, num_workers=2)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=2)
val_loader.mean, val_loader.std = mean, std


Expand Down
11 changes: 7 additions & 4 deletions CTRAIN/eval/eval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -320,15 +320,18 @@ def eval_adaptive(model, eps, data_loader, n_classes=10, test_samples=np.inf, de
certified = torch.tensor([], device=device)
total_images = 0

ibp_data_loader = DataLoader(data_loader.dataset, batch_size=data_loader.batch_size, shuffle=False)
ibp_data_loader.max, ibp_data_loader.min, ibp_data_loader.std = data_loader.max, data_loader.min, data_loader.std

crown_data_loader = DataLoader(data_loader.dataset, batch_size=1, shuffle=False)
crown_data_loader.max, crown_data_loader.min, crown_data_loader.std = data_loader.max, data_loader.min, data_loader.std

for batch_idx, (data, targets) in tqdm(enumerate(data_loader)):
for batch_idx, (data, targets) in tqdm(enumerate(ibp_data_loader)):
certified_idx = torch.zeros(len(data), device=device, dtype=torch.bool)

ptb = PerturbationLpNorm(eps=eps, norm=np.inf, x_L=torch.clamp(data - eps, data_loader.min, data_loader.max).to(device), x_U=torch.clamp(data + eps, data_loader.min, data_loader.max).to(device))
ptb = PerturbationLpNorm(eps=eps, norm=np.inf, x_L=torch.clamp(data - eps, ibp_data_loader.min, ibp_data_loader.max).to(device), x_U=torch.clamp(data + eps, ibp_data_loader.min, ibp_data_loader.max).to(device))
data, targets = data.to(device), targets.to(device)
if batch_idx * data_loader.batch_size >= test_samples:
if batch_idx * ibp_data_loader.batch_size >= test_samples:
continue

total_images += len(targets)
Expand All @@ -346,7 +349,7 @@ def eval_adaptive(model, eps, data_loader, n_classes=10, test_samples=np.inf, de

data = data.to('cpu')
certified_idx = certified_idx.to("cpu")
ptb = PerturbationLpNorm(eps=eps, norm=np.inf, x_L=torch.clamp(data[~certified_idx] - eps, data_loader.min, data_loader.max).to(device), x_U=torch.clamp(data[~certified_idx] + eps, data_loader.min, data_loader.max).to(device))
ptb = PerturbationLpNorm(eps=eps, norm=np.inf, x_L=torch.clamp(data[~certified_idx] - eps, ibp_data_loader.min, ibp_data_loader.max).to(device), x_U=torch.clamp(data[~certified_idx] + eps, ibp_data_loader.min, ibp_data_loader.max).to(device))
data = data.to(device)
certified_idx = certified_idx.to(device)

Expand Down