[inductor] fix MA on poor gpu #145133
Closed
[inductor] fix MA on poor gpu #145133
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/145133
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (3 Unrelated Failures)As of commit 69006be with merge base 0f051ea ( FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
jansel
approved these changes
Jan 18, 2025
|
@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 |
AnantGulati
pushed a commit
to AnantGulati/pytorch
that referenced
this pull request
Jan 22, 2025
Found this bug when debugging a MA issue in CI that can not be repro-ed on devgpu. On GPU with less than 68 SMs (like NVidia L4 used in CI), running torch compile in max-autotune mode may result in the following confusing error https://gist.github.com/shunting314/370f42f547e3367a3773237942725a86 complaining about layout: ``` torch._inductor.exc.InductorError: LoweringException: AssertionError: convert FlexibleLayout to FixedLayout first ``` The reason is, even if we don't pick Triton template, Inductor still returns a MultiTemplateBuffer for tuned addmm. MultiTemplateBuffer.get_reads called from Reduction.num_splits may indexing a FlexibleLayout which results in the error aforementioned. The issue does not appear on devgpu because we freeze the layout of addmm inputs when rendering triton templates. Pull Request resolved: pytorch#145133 Approved by: https://github.com/jansel
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.
Stack from ghstack (oldest at bottom):
Found this bug when debugging a MA issue in CI that can not be repro-ed on devgpu.
On GPU with less than 68 SMs (like NVidia L4 used in CI), running torch compile in max-autotune mode may result in the following confusing error https://gist.github.com/shunting314/370f42f547e3367a3773237942725a86 complaining about layout:
The reason is, even if we don't pick Triton template, Inductor still returns a MultiTemplateBuffer for tuned addmm. MultiTemplateBuffer.get_reads called from Reduction.num_splits may indexing a FlexibleLayout which results in the error aforementioned.
The issue does not appear on devgpu because we freeze the layout of addmm inputs when rendering triton templates.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov