Skip to content

Conversation

datumbox
Copy link
Contributor

Removing performance workaround now that the pytorch/pytorch#83840 is resolved.

Copy link
Collaborator

@pmeier pmeier left a comment

Choose a reason for hiding this comment

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

Could you also remove the sample inputs that tested this specific branch:

# We have a speed hack in place for nearest interpolation and single channel images (grayscale)
for image_loader in make_image_loaders(
sizes=["random"],
color_spaces=[features.ColorSpace.GRAY],
extra_dims=VALID_EXTRA_DIMS,
):
yield ArgsKwargs(
image_loader, size=[min(image_loader.image_size) + 1], interpolation=F.InterpolationMode.NEAREST
)

Otherwise LGTM if CI is green.

@datumbox datumbox merged commit 0ab50f5 into pytorch:main Oct 10, 2022
@datumbox datumbox deleted the prototype/remove_resize_workaround branch October 10, 2022 11:32
datumbox added a commit that referenced this pull request Oct 11, 2022
@datumbox datumbox added revert(ed) For reverted PRs, and PRs that revert other PRs and removed revert(ed) For reverted PRs, and PRs that revert other PRs labels Oct 11, 2022
facebook-github-bot pushed a commit that referenced this pull request Oct 17, 2022
Summary:
* Remove performance workaround for mask resize

* Fix linter

* bug fixes

* remove unnecessary import

* Fixing linter

Reviewed By: NicolasHug

Differential Revision: D40427473

fbshipit-source-id: e7e632069c62cc4e53c31571f0c177d68ee61c3e
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants