From 9cc43ca4e1fa4abc2628e66f654d33fd0779eeca Mon Sep 17 00:00:00 2001 From: Even Zhou Date: Wed, 2 Sep 2026 08:57:59 +0800 Subject: [PATCH] [NPU] Update sgl-kernel-npu version to 2026.9.0 and move memfabric deps into pyproject (#37399) --- .github/workflows/release-docker-npu-nightly.yml | 2 +- .github/workflows/release-docker-npu.yml | 2 +- docker/npu.Dockerfile | 6 ------ python/pyproject_npu.toml | 2 ++ scripts/ci/npu/npu_ci_install_dependency.sh | 8 +------- 5 files changed, 5 insertions(+), 15 deletions(-) diff --git a/.github/workflows/release-docker-npu-nightly.yml b/.github/workflows/release-docker-npu-nightly.yml index 5a0dfe6ab..5855fd729 100644 --- a/.github/workflows/release-docker-npu-nightly.yml +++ b/.github/workflows/release-docker-npu-nightly.yml @@ -80,6 +80,6 @@ jobs: push: ${{ github.repository == 'sgl-project/sglang' && github.event_name != 'pull_request' }} provenance: false build-args: | - SGLANG_KERNEL_NPU_TAG=20260826 + SGLANG_KERNEL_NPU_TAG=2026.9.0 CANN_VERSION=${{ matrix.cann_version }} DEVICE_TYPE=${{ matrix.device_type }} diff --git a/.github/workflows/release-docker-npu.yml b/.github/workflows/release-docker-npu.yml index f0b44409e..eaa13da93 100644 --- a/.github/workflows/release-docker-npu.yml +++ b/.github/workflows/release-docker-npu.yml @@ -87,7 +87,7 @@ jobs: push: ${{ github.repository == 'sgl-project/sglang' && github.event_name != 'pull_request' }} provenance: false build-args: | - SGLANG_KERNEL_NPU_TAG=20260826 + SGLANG_KERNEL_NPU_TAG=2026.9.0 CANN_VERSION=${{ matrix.cann_version }} DEVICE_TYPE=${{ matrix.device_type }} SGLANG_TAG=${{ steps.version.outputs.version }} diff --git a/docker/npu.Dockerfile b/docker/npu.Dockerfile index 39ebaf042..de8f48ac9 100644 --- a/docker/npu.Dockerfile +++ b/docker/npu.Dockerfile @@ -72,12 +72,6 @@ ENV LANGUAGE=en_US:en ENV LC_ALL=en_US.UTF-8 -### Install MemFabric -RUN ${PIP_INSTALL} memfabric-hybrid==1.0.8 - -### Install zbal -RUN ${PIP_INSTALL} memfabric-zbal==1.2.0 - ### Install SGLang Model Gateway RUN ${PIP_INSTALL} sglang-router diff --git a/python/pyproject_npu.toml b/python/pyproject_npu.toml index 5c56fe986..5832cd90a 100644 --- a/python/pyproject_npu.toml +++ b/python/pyproject_npu.toml @@ -66,6 +66,8 @@ dependencies = [ "uvloop", "xxhash", "xgrammar==0.2.1", + "memfabric-hybrid==1.1.4", + "memfabric-zbal==1.2.0", ] [project.optional-dependencies] diff --git a/scripts/ci/npu/npu_ci_install_dependency.sh b/scripts/ci/npu/npu_ci_install_dependency.sh index 75cb418ca..89578cc77 100755 --- a/scripts/ci/npu/npu_ci_install_dependency.sh +++ b/scripts/ci/npu/npu_ci_install_dependency.sh @@ -41,10 +41,6 @@ export PATH="${CARGO_HOME:-$HOME/.cargo}/bin:${PATH}" ${UV_PIP_INSTALL} pybind11 pyyaml decorator scipy attrs psutil -### Install MemFabric -${UV_PIP_INSTALL} memfabric-hybrid==1.0.8 - - ### Install PyTorch and PTA PYTORCH_VERSION="2.10.0" TORCHVISION_VERSION="0.25.0" @@ -54,14 +50,12 @@ PTA_URL="https://gitcode.com/Ascend/pytorch/releases/download/v26.0.0-pytorch2.1 # GitCode does not allow UV downloads. ${PIP_INSTALL} ${PTA_URL} -### Install zbal -${UV_PIP_INSTALL} memfabric-zbal==1.2.0 ### Install Triton-Ascend ${PIP_INSTALL} triton-ascend==3.2.1.dev20260530 --extra-index-url=https://mirrors.huaweicloud.com/ascend/repos/pypi/nightly --trusted-host triton-ascend.osinfra.cn ### Install sgl-kernel-npu -SGLANG_KERNEL_NPU_TAG="20260826" +SGLANG_KERNEL_NPU_TAG="2026.9.0" mkdir sgl-kernel-npu (cd sgl-kernel-npu && wget "${GITHUB_PROXY_URL:=""}https://github.com/sgl-project/sgl-kernel-npu/releases/download/${SGLANG_KERNEL_NPU_TAG}/sgl-kernel-npu-${SGLANG_KERNEL_NPU_TAG}-torch${PYTORCH_VERSION}-py311-cann9.0.0-${DEVICE_TYPE}-$(arch).zip" \ && unzip ./sgl-kernel-npu-${SGLANG_KERNEL_NPU_TAG}-torch${PYTORCH_VERSION}-py311-cann9.0.0-${DEVICE_TYPE}-$(arch).zip \