[Flex] Fix silent correctness w/ backpropping grads #163677
Closed
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/163677
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 2870fe4 with merge base 134dfbe ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
ydwu4
approved these changes
Sep 23, 2025
drisspg
added
release notes: nn
|
@pytorchbot merge |
Merge startedYour 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 |
dsashidh
pushed a commit
to dsashidh/pytorch
that referenced
this pull request
Sep 26, 2025
Fixes #pytorch#162228 # Summary Majority of our tests are only compiling flex-attention in isolation. This means that for fake tensor propagation the input primals and all captured buffers dont do any intermediate computation below autograd. As a result result the by happen chance match the `require_grad`ness of the eager implementation and this check will pass. However if score_mod is a the result of some other intermediate fake tensor prop then it is not guaranteed to have accurate req_gradness, which was happening here. TLDR is that this was a boot and suspenders that was actually harmful and we should just let the joint graph handle creating the correct joint graph Pull Request resolved: pytorch#163677 Approved by: https://github.com/ydwu4
jainapurva
pushed a commit
that referenced
this pull request
Sep 29, 2025
Fixes ##162228 # Summary Majority of our tests are only compiling flex-attention in isolation. This means that for fake tensor propagation the input primals and all captured buffers dont do any intermediate computation below autograd. As a result result the by happen chance match the `require_grad`ness of the eager implementation and this check will pass. However if score_mod is a the result of some other intermediate fake tensor prop then it is not guaranteed to have accurate req_gradness, which was happening here. TLDR is that this was a boot and suspenders that was actually harmful and we should just let the joint graph handle creating the correct joint graph Pull Request resolved: #163677 Approved by: https://github.com/ydwu4
|
@pytorchbot cherry-pick --onto release/2.9 --c critical |
pytorchbot
pushed a commit
that referenced
this pull request
Oct 1, 2025
Fixes ##162228 # Summary Majority of our tests are only compiling flex-attention in isolation. This means that for fake tensor propagation the input primals and all captured buffers dont do any intermediate computation below autograd. As a result result the by happen chance match the `require_grad`ness of the eager implementation and this check will pass. However if score_mod is a the result of some other intermediate fake tensor prop then it is not guaranteed to have accurate req_gradness, which was happening here. TLDR is that this was a boot and suspenders that was actually harmful and we should just let the joint graph handle creating the correct joint graph Pull Request resolved: #163677 Approved by: https://github.com/ydwu4 (cherry picked from commit e2ce79e)
Cherry picking #163677The cherry pick PR is at #164366 and it is recommended to link a critical cherry pick PR with an issue. The following tracker issues are updated: Details for Dev Infra teamRaised by workflow job |
Camyll
pushed a commit
that referenced
this pull request
Oct 1, 2025
[Flex] Fix silent correctness w/ backpropping grads (#163677) Fixes ##162228 # Summary Majority of our tests are only compiling flex-attention in isolation. This means that for fake tensor propagation the input primals and all captured buffers dont do any intermediate computation below autograd. As a result result the by happen chance match the `require_grad`ness of the eager implementation and this check will pass. However if score_mod is a the result of some other intermediate fake tensor prop then it is not guaranteed to have accurate req_gradness, which was happening here. TLDR is that this was a boot and suspenders that was actually harmful and we should just let the joint graph handle creating the correct joint graph Pull Request resolved: #163677 Approved by: https://github.com/ydwu4 (cherry picked from commit e2ce79e) Co-authored-by: drisspg <drisspguessous@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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.
Stack from ghstack (oldest at bottom):
Fixes ##162228
Summary
Majority of our tests are only compiling flex-attention in isolation. This means that for fake tensor propagation the input primals and all captured buffers dont do any intermediate computation below autograd. As a result result the by happen chance match the
require_gradness of the eager implementation and this check will pass. However if score_mod is a the result of some other intermediate fake tensor prop then it is not guaranteed to have accurate req_gradness, which was happening here.TLDR is that this was a boot and suspenders that was actually harmful and we should just let the joint graph handle creating the correct joint graph
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben @Chillee @yanboliang @BoyuanFeng