This appears to be a diff that was exported from phabricator, but the PR author does not have sufficient permissions to run CI. @alexsibir, please do step 2 of internal wiki to get write access so you do not need to get CI approvals in the future. If you think this is a mistake, please contact the Pytorch Dev Infra team.

The committers listed above are authorized under a signed CLA.

• ✅ login: alexsibir / name: Alex Sibiryakov (32afe38)

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/165446

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 32afe38 with merge base e6f766c ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@alexsibir has exported this pull request. If you are a Meta employee, you can view the originating Diff in D84612194.

Would it be possible to add a test for this?

@pytorchbot merge

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Note: by looks of it the PR shouldn't have changed the behavior, nDims is used before that batch loop starts, that's fine. Then inside batch loop it is potentially changed befind if isInOutAligned conditional, and also used behind stride_size > 1 conditional and behind if (memory_format != c10::MemoryFormat::Contiguous) conditional where changed value could mean trouble. But isInOutAligned can be true only if stride_size=1 and memory_format is c10::MemoryFormat::Contiguous, here's how it's set in the beginning

bool isInOutAligned = isContig && at::native::memory::get_alignment(data) >= alignment && memory_format == c10::MemoryFormat::Contiguous
(isContig is set to stride_size == 1 above), so both those uses should be unaffected. But the author claims it fixes their usecase. Test would really be useful.

@pytorchbot cherry-pick --onto release/2.9 --fixes "fixes silent correctness bug" -c regression

Cherry picking #165446

Command git -C /home/runner/work/pytorch/pytorch cherry-pick -x 4f400ab520f0151c8f01d7c305637276e4a222ca returned non-zero exit code 1

Auto-merging aten/src/ATen/native/cuda/Shape.cu
CONFLICT (content): Merge conflict in aten/src/ATen/native/cuda/Shape.cu
error: could not apply 4f400ab520f... Fix: nDims is mutated inside the loop in Shape.cu (#165446)
hint: After resolving the conflicts, mark them with
hint: "git add/rm <pathspec>", then run
hint: "git cherry-pick --continue".
hint: You can instead skip this commit with "git cherry-pick --skip".
hint: To abort and get back to the state before "git cherry-pick",
hint: run "git cherry-pick --abort".
hint: Disable this message with "git config set advice.mergeConflict false"