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Outdated suggestions cannot be applied.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.
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.
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
Adding new rocm triton flash attention kernel #4
Adding new rocm triton flash attention kernel #4
Changes from 1 commit
e569133
0e63661
c89c0e3
d4cb905
1fff99a
a3bc6c8
8cd9653
52af554
File filter
Filter by extension
Conversations
Jump to
There are no files selected for viewing
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have some questions: can flash-attn-func and triton fa co-exist? or should we allow user to install them at the same time in docker file?
Also, what is the steps to validate this PR ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Some background
flash_attn_func is Tri Dao's "reference" implementation of Flash Attention. It calls the cutlass version in the backend.
If
is_hip()
, flash_attn_func calls the CK version in the backend.Cutlass version I believe is a pip package and one can just
pip install flash-attention
. I think for CK we need to install from our fork but I'm a little less familiar with that.Triton is completely separate and does not intersect with either the front end flash_attn_func or cutlass / CK. It requires a Triton install. And the kernel, which is available in this PR.
So, yes both can co-exist but only one is needed.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Add a flag to skip