EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

disable sm100 for FlashMLA and fast-hadamard-transform in cuda12.6.1 (#11274)

Browse Source
This commit is contained in:
gongwei-130
2025-10-06 14:48:31 -07:00
committed by GitHub
parent 466992b2d0
commit 4aeb193fbd
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
docker/Dockerfile
View File

@@ -6,7 +6,7 @@ ARG BUILD_TYPE=all
ARG BRANCH_TYPE=remote ARG BRANCH_TYPE=remote
ARG DEEPEP_COMMIT=9af0e0d0e74f3577af1979c9b9e1ac2cad0104ee ARG DEEPEP_COMMIT=9af0e0d0e74f3577af1979c9b9e1ac2cad0104ee
ARG FLASHMLA_COMMIT=1408756a88e52a25196b759eaf8db89d2b51b5a1 ARG FLASHMLA_COMMIT=1408756a88e52a25196b759eaf8db89d2b51b5a1
ARG FAST_HADAMARD_TRANSFORM_COMMIT=f3cdeed95b0f3284b5df3da9b3311d3d0600ce2b ARG FAST_HADAMARD_TRANSFORM_COMMIT=7fd811c2b47f63b0b08d2582619f939e14dad77c
ARG CMAKE_BUILD_PARALLEL_LEVEL=2 ARG CMAKE_BUILD_PARALLEL_LEVEL=2
ARG SGL_KERNEL_VERSION=0.3.12 ARG SGL_KERNEL_VERSION=0.3.12
ENV DEBIAN_FRONTEND=noninteractive \ ENV DEBIAN_FRONTEND=noninteractive \
@@ -145,6 +145,9 @@ RUN if [ "$TARGETARCH" = "amd64" ]; then \
cd flash-mla && \ cd flash-mla && \
git checkout ${FLASHMLA_COMMIT} && \ git checkout ${FLASHMLA_COMMIT} && \
git submodule update --init --recursive && \ git submodule update --init --recursive && \
if [ "$CUDA_VERSION" = "12.6.1" ]; then \
export FLASH_MLA_DISABLE_SM100=1; \
fi && \
pip install -v . ; \ pip install -v . ; \
fi fi