[Misc] Set default backend to SDPA for get_vit_attn_backend #12235
Conversation
This file contains 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
|
👋 Hi! Thank you for contributing to the vLLM project. Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can do one of these:
🚀 |
ywang96
reviewed
Jan 21, 2025
wangxiyuan
force-pushed
the
get_vit_fix
branch
2 times, most recently
from
6fd86ac to
5ef64af
Compare
wangxiyuan
changed the title
Signed-off-by: wangxiyuan <[email protected]>
wangxiyuan
force-pushed
the
get_vit_fix
branch
from
5ef64af to
eb77619
Compare
ywang96
approved these changes
Jan 21, 2025
|
CI failure seems doesn't relate to this PR |
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Outdated suggestions cannot be applied.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.
Xformers works only on GPU, while torch SDPA is more generic. This PR set the default attention backend to SDPA for qwen-vl, so that most non-GPU platform(especially out-of-tree ) can work by default with SDPA.