Refactor turbomind attention by precomputing rotary embed #3129
Workflow file for this run
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: unit-test | |
on: | |
pull_request: | |
paths: | |
- ".github/workflows/unit-test.yml" | |
- "cmake/**" | |
- "src/**" | |
- "tests/**" | |
- "3rdparty/**" | |
- "lmdeploy/**" | |
- "requirements/**" | |
- "requirements.txt" | |
- "CMakeLists.txt" | |
- "setup.py" | |
push: | |
branches: | |
- main | |
paths: | |
- ".github/workflows/unit-test.yml" | |
- "cmake/**" | |
- "src/**" | |
- "tests/**" | |
- "3rdparty/**" | |
- "lmdeploy/**" | |
- "requirements/**" | |
- "requirements.txt" | |
- "CMakeLists.txt" | |
- "setup.py" | |
tags: | |
- "v*.*.*" | |
jobs: | |
unit_test: | |
runs-on: [self-hosted, linux-a100-s2] | |
timeout-minutes: 4320 # 72hours | |
container: | |
image: nvcr.io/nvidia/tritonserver:22.12-py3 | |
options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip -e CUDA_VISIBLE_DEVICES=2,3 --pull never" | |
volumes: | |
- /nvme/share_data/github-actions/pip-cache:/root/.cache/pip | |
- /nvme/share_data/github-actions/packages:/root/packages | |
- /usr/share/zoneinfo/Asia/Shanghai:/etc/localtime:ro | |
steps: | |
- name: Setup systems | |
run: | | |
rm /etc/apt/sources.list.d/cuda*.list | |
apt-get update && apt-get install -y --no-install-recommends rapidjson-dev \ | |
libgoogle-glog-dev | |
rm -rf /var/lib/apt/lists/* | |
- name: Clone repository | |
uses: actions/checkout@v2 | |
- name: Install pytorch | |
run: | | |
python3 -m pip cache dir | |
python3 -m pip install torch==2.3.1 torchvision==0.18.1 --index-url https://download.pytorch.org/whl/cu118 | |
- name: Build lmdeploy | |
run: | | |
python3 -m pip install cmake | |
python3 -m pip install -r requirements/build.txt | |
mkdir build | |
cd build | |
cp -r /root/packages/_deps _deps | |
cmake .. \ | |
-DCMAKE_BUILD_TYPE=RelWithDebInfo \ | |
-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \ | |
-DCMAKE_INSTALL_PREFIX=./install \ | |
-DBUILD_PY_FFI=ON \ | |
-DBUILD_MULTI_GPU=OFF \ | |
-DCMAKE_CUDA_FLAGS="-lineinfo" \ | |
-DUSE_NVTX=ON \ | |
-DSM=80 \ | |
-DCMAKE_CUDA_ARCHITECTURES=80 \ | |
-DBUILD_TEST=OFF | |
make -j$(nproc) && make install | |
- name: Install lmdeploy | |
run: | | |
python3 -m pip install pynvml packaging protobuf transformers_stream_generator | |
# manually install flash attn | |
python3 -m pip install /root/packages/flash_attn-2.6.3+cu118torch2.3cxx11abiFALSE-cp38-cp38-linux_x86_64.whl | |
python3 -m pip install -r requirements.txt -r requirements/test.txt | |
python3 -m pip install . | |
- name: Check env | |
run: | | |
python3 -m pip list | |
lmdeploy check_env | |
- name: Test lmdeploy csrc | |
run: | | |
#./build/bin/build/bin/unittest | |
echo "TODO" | |
- name: Test lmdeploy python UT | |
run: | | |
coverage run --branch --source lmdeploy -m pytest -rsE tests | |
coverage xml | |
coverage report -m | |
- name: Clear workfile | |
if: always() | |
run: | | |
export workdir=$(pwd) | |
cd .. | |
rm -rf $workdir | |
mkdir $workdir | |
chmod -R 777 $workdir |