BF16I4 Preshuffled Grouped Gemm #3917
Closed
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
|
This pull request was exported from Phabricator. Differential Revision: D72337760 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Summary: X-link: facebookresearch/FBGEMM#1003 This diff adds a preshuffled BF16I4 mixed dtype kernel using cutlass. Performance is quite compelling and shows substantial speedups for some shapes compared to bf16 x bf16 gemm backed by cublas. Notably, this preshuffle approach is 1.5-2X faster than the standard bf16i4 gemm for most shapes. Compared to other mixed dtype kernels like marlin and machete, we see that this new kernel is probably the best average performer. {F1976677491} Reviewed By: jianyuh, jiawenliu64 Differential Revision: D72270467
Summary: X-link: facebookresearch/FBGEMM#1006 This diff adds a preshuffled variant of BF16I4 Grouped Gemm. Notably, cutlass does not currently support zero points for grouped gemm, so this kernel must be used without them. That said, the accuracy of the kernel appears reasonable and the performance is very compelling. {F1976716898} Reviewed By: jiawenliu64 Differential Revision: D72337760
jwfromm
force-pushed
the
export-D72337760
branch
from
61b5a99 to
f0dfb6f
Compare
|
This pull request was exported from Phabricator. Differential Revision: D72337760 |
|
This pull request has been merged in c407f65. |
q10
pushed a commit
to q10/FBGEMM
that referenced
this pull request
Apr 10, 2025
Summary: Pull Request resolved: facebookresearch/FBGEMM#1006 X-link: pytorch#3917 This diff adds a preshuffled variant of BF16I4 Grouped Gemm. Notably, cutlass does not currently support zero points for grouped gemm, so this kernel must be used without them. That said, the accuracy of the kernel appears reasonable and the performance is very compelling. {F1976716898} Reviewed By: jiawenliu64 Differential Revision: D72337760 fbshipit-source-id: a2cf9e913d095da42f1cf88a5c08dbbe1f2794c9
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.
Summary:
This diff adds a preshuffled variant of BF16I4 Grouped Gemm. Notably, cutlass does not currently support zero points for grouped gemm, so this kernel must be used without them. That said, the accuracy of the kernel appears reasonable and the performance is very compelling.
{F1976716898}
Differential Revision: D72337760