Skip to content

[Bugfix][ROCm] Fix gemm_a4w4 call to use updated AITER API signature #40754

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.

Sign up for GitHub

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

Merged
tjtanaa merged 2 commits into from
Apr 29, 2026
Merged

[Bugfix][ROCm] Fix gemm_a4w4 call to use updated AITER API signature #40754

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
519216e
[ROCm] Fix gemm_a4w4 to use updated AITER API signature
chelnnexy Apr 24, 2026
320a5e3
Merge branch 'main' into cheiluno/fix-aiter-gemm-a4w4-api
chelnnexy Apr 24, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 2 additions & 11 deletions vllm/model_executor/layers/quantization/quark/schemes/quark_ocp_mx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,21 +96,12 @@ def gemm_with_dynamic_quant(
x_q = x
x_s = x_scales

# 32 alignment is enough for dim0 padding of output for
# gemm_a4w4 kernel
y = torch.empty(
(M + 31) // 32 * 32,
weight.shape[0],
device=x_q.device,
dtype=out_dtype,
)

gemm_a4w4(
y = gemm_a4w4(
x_q,
weight.view(x_q.dtype),
x_s,
weight_scale.view(x_s.dtype),
y,
dtype=out_dtype,
bpreshuffle=True,
)
return y[:M]
Expand Down
Loading