[WIP]Enable assymetric heads for D_qk and D_v for micro kernel sdpa#5301
Open
h-sadia wants to merge 5 commits into
Open
[WIP]Enable assymetric heads for D_qk and D_v for micro kernel sdpa#5301h-sadia wants to merge 5 commits into
h-sadia wants to merge 5 commits into
Conversation
github-actions
Bot
added
platform:gpu-intel
h-sadia
force-pushed
the
hsadia/assym_heads_sdpa
branch
from
4a844b2 to
0fa2486
Compare
h-sadia
force-pushed
the
hsadia/assym_heads_sdpa
branch
from
f3a67be to
943a4cc
Compare
h-sadia
changed the title
h-sadia
force-pushed
the
hsadia/assym_heads_sdpa
branch
from
943a4cc to
f3a67be
Compare
h-sadia
changed the title
TaoLv
reviewed
Jun 12, 2026
| # f16 inputs + f32 intermediates + f16 outputs | ||
| --reset --op-kind=1:Multiply,1:Divide --case=complex_fusion/mha/sdpa-plain-simplified-f16-f32.json | ||
| # Asymmetric heads: Q/K head_size=64, V head_size=128 | ||
| --reset --op-kind=1:Multiply,1:Divide --case=complex_fusion/mha/sdpa-plain-asymm-heads-f16-f32.json |
Contributor
There was a problem hiding this comment.
We already have test cases to cover d_qk != d_v.
oneDNN/tests/benchdnn/inputs/graph/complex_fusion/harness_mha_all
Lines 211 to 215 in 8cde797
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Fused micro kernel doesn't allow different head sizes of QK and V tensors. This PR focuses on enabling that as per this request here: https://jira.devtools.intel.com/browse/MFDNN-14385
N.B: Will continue testing it with the tests present in Graph API and extend testing starting ww25.2
Fixes # (github issue)
Checklist
General
make testandmake test_benchdnn_*) pass locally for each commit?