[ci] limit cmake build nproc (#8100)

Author: slin1237

.github/workflows/release-docker-dev.yml

@@ -41,5 +41,5 @@ jobs:
 
       - name: Build and Push Dev Image
         run: |
-          docker buildx build --output type=image,compression=zstd . -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} -t lmsysorg/sglang:${{ matrix.variant.tag }} --no-cache
+          docker buildx build --output type=image,compression=zstd . -f docker/Dockerfile --build-arg CUDA_VERSION=${{ matrix.variant.version }} --build-arg BUILD_TYPE=${{ matrix.variant.type }} --build-arg CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) -t lmsysorg/sglang:${{ matrix.variant.tag }} --no-cache
           docker push lmsysorg/sglang:${{ matrix.variant.tag }}

docker/Dockerfile

@@ -3,6 +3,7 @@ FROM nvidia/cuda:${CUDA_VERSION}-cudnn-devel-ubuntu22.04
 
 ARG BUILD_TYPE=all
 ARG DEEPEP_COMMIT=b6ce310bb0b75079682d09bc2ebc063a074fbd58
+ARG CMAKE_BUILD_PARALLEL_LEVEL=2
 ENV DEBIAN_FRONTEND=noninteractive \
     CUDA_HOME=/usr/local/cuda \
     GDRCOPY_HOME=/usr/src/gdrdrv-2.4.4/ \
@@ -78,7 +79,7 @@ RUN wget https://developer.download.nvidia.com/compute/redist/nvshmem/3.3.9/sour
     NVSHMEM_TIMEOUT_DEVICE_POLLING=0 \
     NVSHMEM_USE_GDRCOPY=1 \
     cmake -S . -B build/ -DCMAKE_INSTALL_PREFIX=${NVSHMEM_DIR} -DCMAKE_CUDA_ARCHITECTURES=90 \
- && cmake --build build --target install -j \
+ && cmake --build build --target install -j${CMAKE_BUILD_PARALLEL_LEVEL} \
  && cd /sgl-workspace/DeepEP \
  && NVSHMEM_DIR=${NVSHMEM_DIR} pip install .