Unpin scipy & use threadpoolctl in minimize_with_timeout #2712
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
the
CLA Signed
|
@saitcakmak has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #2712 +/- ##
=======================================
Coverage 99.98% 99.98%
=======================================
Files 202 202
Lines 18602 18604 +2
=======================================
+ Hits 18600 18602 +2
Misses 2 2 ☔ View full report in Codecov by Sentry. |
Summary: Scipy>=1.15 was leading to significant slowdowns in CI due to issues with OpenBLAS configuration (see scipy/scipy#22438). Thanks to suggestions by SciPy maintainers, we've found a solution that resolves the slowdown without modifying and global configurations for OpenBLAS. This PR wraps `minimize` calls in `minimize_with_timeout` in a `with threadpool_limits(limits=1, user_api="blas")` context, which makes OpenBLAS operate single-threaded within the context. Since PyTorch relies on OpenMP & MKL rather than OpenBLAS, this does not affect the evaluations of the acquisition functions or the MLL. Test Plan: Compare runtimes in the following tutorial workflows Baseline - with scipy 1.14.1, without `threadpool_limits`: https://github.com/pytorch/botorch/actions/runs/13044166839/job/36391816602 With scipy 1.14.1, with `threadpool_limits`: https://github.com/pytorch/botorch/actions/runs/13044167834/job/36391818438 With scipy 1.15.1, with `threadpool_limits`: https://github.com/pytorch/botorch/actions/runs/13044245918/job/36392026548 Differential Revision: D68874700 Pulled By: saitcakmak
facebook-github-bot
force-pushed
the
threadctl
branch
from
54247a6 to
8bf1230
Compare
|
This pull request was exported from Phabricator. Differential Revision: D68874700 |
|
@saitcakmak merged this pull request in acae12d. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Scipy>=1.15 was leading to significant slowdowns in CI due to issues with OpenBLAS configuration (see scipy/scipy#22438). Thanks to suggestions by SciPy maintainers, we've found a solution that resolves the slowdown without modifying and global configurations for OpenBLAS.
This PR wraps
minimizecalls inminimize_with_timeoutin awith threadpool_limits(limits=1, user_api="blas")context, which makes OpenBLAS operate single-threaded within the context. Since PyTorch relies on OpenMP & MKL rather than OpenBLAS, this does not affect the evaluations of the acquisition functions or the MLL.Test Plan:
Compare runtimes in the following tutorial workflows
Baseline - with scipy 1.14.1, without
threadpool_limits: https://github.com/pytorch/botorch/actions/runs/13044166839/job/36391816602With scipy 1.14.1, with
threadpool_limits: https://github.com/pytorch/botorch/actions/runs/13044167834/job/36391818438With scipy 1.15.1, with
threadpool_limits: https://github.com/pytorch/botorch/actions/runs/13044245918/job/36392026548