Skip to content

Add FP8 Support For CK Tile Group GEMM#475

Open
aris134 wants to merge 15 commits intodevfrom
amartin/ck-grouped-gemm-fp8
Open

Add FP8 Support For CK Tile Group GEMM#475
aris134 wants to merge 15 commits intodevfrom
amartin/ck-grouped-gemm-fp8

Conversation

@aris134
Copy link
Copy Markdown

@aris134 aris134 commented Mar 6, 2026 •
edited
Loading

Description

Please include a brief summary of the changes, relevant motivation and context.

Fixes https://github.com/ROCm/frameworks-internal/issues/15787

TODO:

  • Add support for gfx942/gfx950
  • Performance analysis and tuning

Type of change

  • Documentation change (change only to the documentation, either a fix or a new content)
  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • Infra/Build change
  • Code refactoring

Changes

Please list the changes introduced in this PR:

  • Enables mixed precision (fp8/bf8) support for CK tile grouped GEMM with tensor quantization on gfx942/gfx950

Checklist:

  • I have read and followed the contributing guidelines
  • The functionality is complete
  • I have commented my code, particularly in hard-to-understand areas
  • I have made corresponding changes to the documentation
  • My changes generate no new warnings
  • I have added tests that prove my fix is effective or that my feature works
  • New and existing unit tests pass locally with my changes

@aris134 aris134 self-assigned this Mar 6, 2026
@aris134 aris134 force-pushed the amartin/ck-grouped-gemm-fp8 branch 2 times, most recently from c834302 to 32f2ac3 Compare March 11, 2026 15:51
@aris134 aris134 marked this pull request as ready for review March 24, 2026 16:07
ipanfilo
ipanfilo requested changes Mar 24, 2026
View reviewed changes
@aris134 aris134 requested a review from ipanfilo March 27, 2026 15:59
ipanfilo
ipanfilo reviewed Mar 30, 2026
View reviewed changes
@aris134 aris134 requested a review from ipanfilo March 30, 2026 20:26
@matthiasdiener matthiasdiener added the ci-level 1 CI test level 1 label Mar 31, 2026
@matthiasdiener matthiasdiener self-requested a review March 31, 2026 18:51
@aris134 aris134 requested a review from matthiasdiener March 31, 2026 22:08
matthiasdiener
matthiasdiener reviewed Mar 31, 2026
View reviewed changes
@aris134 aris134 requested a review from matthiasdiener April 1, 2026 14:37
matthiasdiener
matthiasdiener reviewed Apr 1, 2026
View reviewed changes
@aris134 aris134 requested a review from matthiasdiener April 1, 2026 18:28
@aris134 aris134 requested a review from matthiasdiener April 1, 2026 19:37
matthiasdiener
matthiasdiener approved these changes Apr 1, 2026
View reviewed changes
Copy link
Copy Markdown
Contributor

@matthiasdiener matthiasdiener left a comment

Choose a reason for hiding this comment

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

LGTM!

@aris134 aris134 force-pushed the amartin/ck-grouped-gemm-fp8 branch from bd76659 to 5632303 Compare April 2, 2026 15:08
ipanfilo
ipanfilo reviewed Apr 2, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@ipanfilo ipanfilo left a comment

Choose a reason for hiding this comment

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

Looks good in general. Few small details

wangye805
wangye805 requested changes Apr 2, 2026
View reviewed changes
ipanfilo
ipanfilo approved these changes Apr 6, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@ipanfilo ipanfilo left a comment

Choose a reason for hiding this comment

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

With some nits

transformer_engine/common/gemm/ck_grouped_gemm/ck_grouped_gemm.cpp
Copy link
Copy Markdown
Collaborator

@ipanfilo ipanfilo Apr 6, 2026

Choose a reason for hiding this comment

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

Please use git to move the files so the changes are properly tracked

Copy link
Copy Markdown
Author

@aris134 aris134 Apr 6, 2026

Choose a reason for hiding this comment

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

git mv was used, but because the file diverged significantly, git did not preserve rename detection.

@aris134 aris134 force-pushed the amartin/ck-grouped-gemm-fp8 branch 2 times, most recently from 056ae5d to 391f22a Compare April 6, 2026 19:50
@aris134 aris134 added ci-level 3 CI test level 3 and removed ci-level 1 CI test level 1 labels Apr 6, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@matthiasdiener matthiasdiener matthiasdiener approved these changes

@wangye805 wangye805 wangye805 approved these changes

@ipanfilo ipanfilo ipanfilo approved these changes

@wenchenvincent wenchenvincent Awaiting requested review from wenchenvincent wenchenvincent is a code owner

Assignees

@aris134 aris134

Labels

ci-level 3 CI test level 3

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@aris134 @matthiasdiener @wangye805 @ipanfilo