8000 [release/2.3] [ROCM] Properly disable Flash Attention/Efficient Attention with environment variables by xinyazhang · Pull Request #1571 · ROCm/pytorch · GitHub
[go: up one dir, main page]
Skip to content

[release/2.3] [ROCM] Properly disable Flash Attention/Efficient Attention with environment variables #1571

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 6 commits into from
Sep 11, 2024
Merged

[release/2.3] [ROCM] Properly disable Flash Attention/Efficient Attention with environment variables #1571

merged 6 commits into from
Sep 11, 2024

Conversation

xinyazhang
Copy link

Now USE_FLASH_ATTENTION=0 USE_MEM_EFF_ATTENTION=0 python setup.py can compile correctly.

This is cherry-picked version of pytorch#133866

@jithunnair-amd jithunnair-amd changed the title [ROCM] Properly disable Flash Attention/Efficient Attention with environment variables [release/2.3] [ROCM] Properly disable Flash Attention/Efficient Attention with environment variables Sep 11, 2024
jithunnair-amd
jithunnair-amd reviewed Sep 11, 2024
View reviewed changes
aten/src/ATen/native/transformers/cuda/sdp_utils.cpp Outdated
"Mem Efficient attention was not compiled for current AMD GPU architecture. Attempting to run on architecture ", dprops->gcnArchName);
}
return false;
}
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@pruthvistony Isn't this where we would need to add a return true to address the "Control reaching end of non-void function" error?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No its NOT here.

@pruthvistony
Copy link
Collaborator

For 2.3 mem_efficient should always be false.
@xinyazhang , please update the PR.

@xinyazhang xinyazhang force-pushed the xinyazhang/internal-2.3-nofa branch from 56f999f to 2c4693b Compare September 11, 2024 15:56
@pruthvistony pruthvistony merged commit 1b935e2 into release/2.3 Sep 11, 2024
@pruthvistony pruthvistony deleted the xinyazhang/internal-2.3-nofa branch September 11, 2024 18:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@jithunnair-amd jithunnair-amd jithunnair-amd left review comments

@pruthvistony pruthvistony pruthvistony approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@xinyazhang @pruthvistony @jithunnair-amd
0