Skip to content

[AMD] change fused rms quant interface for aiter upgrade #14497

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

Draft
yctseng0211 wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

[AMD] change fused rms quant interface for aiter upgrade #14497

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
16dd160
fix for interface changed from aiter
yctseng0211 Dec 5, 2025
d927909
upgrade aiter version
yctseng0211 Dec 5, 2025
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
2 changes: 1 addition & 1 deletion docker/rocm.Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,7 @@ ENV BUILD_TRITON="0"
ENV BUILD_LLVM="0"
ENV BUILD_AITER_ALL="0"
ENV BUILD_MOONCAKE="1"
ENV AITER_COMMIT="v0.1.7.post2"
ENV AITER_COMMIT="v0.1.7.post4"
ENV NO_DEPS_FLAG=""
# ===============================
# Chosen arch and args
Expand Down
2 changes: 1 addition & 1 deletion python/sglang/srt/models/deepseek_v2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1836,7 +1836,7 @@ def forward_absorb_prepare(
current_stream.wait_stream(self.alt_stream)
else:
if _use_aiter_gfx95 and self.q_b_proj.weight.dtype == torch.uint8:
q, k_nope, *_ = fused_rms_mxfp4_quant(
q, _, k_nope, *_ = fused_rms_mxfp4_quant(
q,
self.q_a_layernorm.weight,
self.q_a_layernorm.variance_epsilon,
Expand Down
Loading