Fix AllocatorConfig parse roundup division bug

[guangyey]

Stack from ghstack (oldest at bottom):

• -> Fix AllocatorConfig parse roundup division bug #165304
• Refine CUDA BackendStaticInitializer for allocator select #165298
• Introduce a generic API torch._C._accelerator_setAllocatorSettings #165291
• Deprecate overlapped functions in CUDAAllocatorConfig #165289
• Refine Allocator Config error message friendly #165288

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/165304

• 📄 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.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• [ROCm][CI] Machines under the label linux.rocm.gpu.2 are undergoing maintenance.

✅ No Failures

As of commit c0bf0d1 with merge base 556fc09 ():
💚 Looks good so far! There are no failures yet. 💚

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

[albanD]

@albanD has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[joshuuuasu]

Hi @guangyey, thanks very much for taking extra efforts to split the original PR into the stack! I have completed the validation for the whole stack internally, and the performance regression that we previously observed is resolved on this new stack. Therefore, the new stack is in good shape to merge from our side, except the bug I found on the second last PR #165298, as shown in the attached screenshot (our internal tool is broken right now so I cannot import my fix to github).

In retrospection, do you have any insight on what the changes between this diff and the original PR might be the root cause for this issue? Thanks again!

[albanD]

@pytorchbot merge

[pytorchmergebot]

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

Advanced Debugging

Check the merge workflow status
here

[guangyey]

@joshuuuasu Thanks for your verification and fix. I update the bug fix on #165298.
Actually, I’m not sure why the regression occurred or why it went away. I just tried to preserve the original behavior as much as possible in this new stack.