Skip to content

mha fp8 support descale#1459

Merged
valarLip merged 15 commits intomainfrom
rocking/fp8-dquant-fmha-fwd
Nov 28, 2025
Merged

mha fp8 support descale#1459
valarLip merged 15 commits intomainfrom
rocking/fp8-dquant-fmha-fwd

Conversation

@rocking5566
Copy link
Contributor

Motivation

support descale for flash_attn_fp8_pertensor_func and flash_attn_varlen_fp8_pertensor_func

Submission Checklist

AITER_LOG_MORE=1 pytest op_tests/test_mha_fp8.py
AITER_LOG_MORE=1 pytest op_tests/test_mha_varlen_fp8.py

@rocking5566 rocking5566 changed the title [draft] mha fp8 support descale mha fp8 support descale Nov 24, 2025
valarLip
valarLip previously approved these changes Nov 24, 2025
@gyohuangxin
Copy link
Member

@gyohuangxin
Copy link
Member

gyohuangxin commented Nov 26, 2025

Please also resolve the conflicts, thanks!

@rocking5566
Copy link
Contributor Author

@rocking5566 Prebuild failed in https://github.com/ROCm/aiter/actions/runs/19674977504/job/56353662339?pr=1459#step:7:99, can you check?

The prebuilt fail is due to this commit of CK
de6a9590abe907283e189abba1b487f8e5562d1b
I am communicating with the author now

args.d_qk,
args.d_v,
args.d_qkv,
args.d_qkv,
Copy link
Collaborator

Choose a reason for hiding this comment

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

??

@valarLip valarLip merged commit 9ec5c74 into main Nov 28, 2025
22 checks passed
@valarLip valarLip deleted the rocking/fp8-dquant-fmha-fwd branch November 28, 2025 08:36
farlukas pushed a commit that referenced this pull request Dec 4, 2025
* update ck

* Support descale for mha_fwd

* Support descale for mha_varlen_fwd

* update ck

* update ck

* update ck

* update CK

* Fix compile error for batch prefill

---------

Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>
Co-authored-by: Xin Huang <Xin.Huang@amd.com>
nsusanto pushed a commit that referenced this pull request Dec 4, 2025
* update ck

* Support descale for mha_fwd

* Support descale for mha_varlen_fwd

* update ck

* update ck

* update ck

* update CK

* Fix compile error for batch prefill

---------

Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>
Co-authored-by: Xin Huang <Xin.Huang@amd.com>
zhuyuhua-v pushed a commit that referenced this pull request Dec 17, 2025
* update ck

* Support descale for mha_fwd

* Support descale for mha_varlen_fwd

* update ck

* update ck

* update ck

* update CK

* Fix compile error for batch prefill

---------

Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>
Co-authored-by: Xin Huang <Xin.Huang@amd.com>
valarLip added a commit that referenced this pull request Mar 18, 2026
* update ck

* Support descale for mha_fwd

* Support descale for mha_varlen_fwd

* update ck

* update ck

* update ck

* update CK

* Fix compile error for batch prefill

---------

Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>
Co-authored-by: Xin Huang <Xin.Huang@amd.com>
valarLip added a commit that referenced this pull request Mar 18, 2026
* update ck

* Support descale for mha_fwd

* Support descale for mha_varlen_fwd

* update ck

* update ck

* update ck

* update CK

* Fix compile error for batch prefill

---------

Co-authored-by: Lingpeng Jin <103567126+valarLip@users.noreply.github.com>
Co-authored-by: Xin Huang <Xin.Huang@amd.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants