Skip to content

Fix: don't call round() on float images for ResizeV2 #7669

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

Merged
merged 1 commit into from
Jun 13, 2023
Merged

Fix: don't call round() on float images for ResizeV2 #7669

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 10 additions & 0 deletions test/test_transforms_v2_functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1395,3 +1395,13 @@ def test_memory_format_consistency_resize_image_tensor(test_id, info, args_kwarg
assert expected_stride == output_stride, error_msg_fn("")
else:
assert False, error_msg_fn("")


def test_resize_float16_no_rounding():
# Make sure Resize() doesn't round float16 images
# Non-regression test for https://github.com/pytorch/vision/issues/7667

img = torch.randint(0, 256, size=(1, 3, 100, 100), dtype=torch.float16)
out = F.resize(img, size=(10, 10))
assert out.dtype == torch.float16
assert (out.round() - out).sum() > 0
Copy link
Member Author

@NicolasHug NicolasHug Jun 13, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This fails on main

4 changes: 3 additions & 1 deletion torchvision/transforms/v2/functional/_geometry.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -228,7 +228,9 @@ def resize_image_tensor(
if need_cast:
if interpolation == InterpolationMode.BICUBIC and dtype == torch.uint8:
image = image.clamp_(min=0, max=255)
image = image.round_().to(dtype=dtype)
if dtype in (torch.uint8, torch.int8, torch.int16, torch.int32, torch.int64):
image = image.round_()
image = image.to(dtype=dtype)

return image.reshape(shape[:-3] + (num_channels, new_height, new_width))

Expand Down