20231088/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CI/BUILD] Support non-AVX512 vLLM building and testing (#5574)

Browse Source
This commit is contained in:
Jie Fu (傅杰) 2024-06-18 02:36:10 +08:00 committed by GitHub
parent 728c4c8a06
commit ab66536dbf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 20 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
.buildkite/run-cpu-test.sh
View File

@ -4,17 +4,20 @@ set -ex
# Try building the docker image # Try building the docker image
docker build -t cpu-test -f Dockerfile.cpu . docker build -t cpu-test -f Dockerfile.cpu .
docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" -t cpu-test-avx2 -f Dockerfile.cpu .
# Setup cleanup # Setup cleanup
remove_docker_container() { docker rm -f cpu-test || true; } remove_docker_container() { docker rm -f cpu-test cpu-test-avx2 || true; }
trap remove_docker_container EXIT trap remove_docker_container EXIT
remove_docker_container remove_docker_container
# Run the image # Run the image
docker run -itd -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus=48-95 --cpuset-mems=1 --network host -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --name cpu-test cpu-test docker run -itd -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus=48-95 --cpuset-mems=1 --network host -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --name cpu-test cpu-test
docker run -itd -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus=48-95 --cpuset-mems=1 --network host -e HF_TOKEN --env VLLM_CPU_KVCACHE_SPACE=4 --name cpu-test-avx2 cpu-test-avx2
# offline inference # offline inference
docker exec cpu-test bash -c "python3 examples/offline_inference.py" docker exec cpu-test bash -c "python3 examples/offline_inference.py"
docker exec cpu-test-avx2 bash -c "python3 examples/offline_inference.py"
# Run basic model test # Run basic model test
docker exec cpu-test bash -c "cd tests; docker exec cpu-test bash -c "cd tests;

4
Dockerfile.cpu
View File

@ -21,6 +21,10 @@ WORKDIR /workspace/vllm
RUN pip install -v -r requirements-cpu.txt --extra-index-url https://download.pytorch.org/whl/cpu RUN pip install -v -r requirements-cpu.txt --extra-index-url https://download.pytorch.org/whl/cpu
# Support for building with non-AVX512 vLLM: docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" ...
ARG VLLM_CPU_DISABLE_AVX512
ENV VLLM_CPU_DISABLE_AVX512=${VLLM_CPU_DISABLE_AVX512}
RUN VLLM_TARGET_DEVICE=cpu python3 setup.py install RUN VLLM_TARGET_DEVICE=cpu python3 setup.py install
WORKDIR /workspace/ WORKDIR /workspace/

13
cmake/cpu_extension.cmake
View File

@ -33,10 +33,21 @@ function (find_isa CPUINFO TARGET OUT)
endif() endif()
endfunction() endfunction()
function (is_avx512_disabled OUT)
set(DISABLE_AVX512 $ENV{VLLM_CPU_DISABLE_AVX512})
if(DISABLE_AVX512 AND DISABLE_AVX512 STREQUAL "true")
set(${OUT} ON PARENT_SCOPE)
else()
set(${OUT} OFF PARENT_SCOPE)
endif()
endfunction()
is_avx512_disabled(AVX512_DISABLED)
find_isa(${CPUINFO} "avx2" AVX2_FOUND) find_isa(${CPUINFO} "avx2" AVX2_FOUND)
find_isa(${CPUINFO} "avx512f" AVX512_FOUND) find_isa(${CPUINFO} "avx512f" AVX512_FOUND)
if (AVX512_FOUND) if (AVX512_FOUND AND NOT AVX512_DISABLED)
list(APPEND CXX_COMPILE_FLAGS list(APPEND CXX_COMPILE_FLAGS
"-mavx512f" "-mavx512f"
"-mavx512vl" "-mavx512vl"