reduce torchengine prefill mem usage #47
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: pr_ete_test | |
| on: | |
| pull_request: | |
| paths: | |
| - ".github/workflows/pr_ete_test.yml" | |
| - "cmake/**" | |
| - "src/**" | |
| - "autotest/**" | |
| - "3rdparty/**" | |
| - "lmdeploy/**" | |
| - "requirements/**" | |
| - "requirements.txt" | |
| - "CMakeLists.txt" | |
| - "setup.py" | |
| workflow_dispatch: | |
| env: | |
| HOST_PIP_CACHE_DIR: /nvme/github-actions/pip-cache | |
| HOST_LOCALTIME: /usr/share/zoneinfo/Asia/Shanghai | |
| jobs: | |
| pr_functions_test: | |
| runs-on: [self-hosted, linux-a100-pr] | |
| timeout-minutes: 120 | |
| env: | |
| REPORT_DIR: /nvme/qa_test_models/test-reports | |
| container: | |
| image: nvcr.io/nvidia/tritonserver:22.12-py3 | |
| options: "--gpus=all --ipc=host --user root -e PIP_CACHE_DIR=/root/.cache/pip" | |
| volumes: | |
| - /nvme/share_data/github-actions/pip-cache:/root/.cache/pip | |
| - /nvme/share_data/github-actions/packages:/root/packages | |
| - /nvme/qa_test_models:/nvme/qa_test_models | |
| - /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 libgl1 openjdk-8-jre-headless | |
| dpkg -i /root/packages/allure_2.24.1-1_all.deb | |
| 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==2.1.0 torchvision==0.16.0 --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 | |
| cmake .. \ | |
| -DCMAKE_BUILD_TYPE=RelWithDebInfo \ | |
| -DCMAKE_EXPORT_COMPILE_COMMANDS=1 \ | |
| -DCMAKE_INSTALL_PREFIX=/opt/tritonserver \ | |
| -DBUILD_PY_FFI=ON \ | |
| -DBUILD_MULTI_GPU=ON \ | |
| -DCMAKE_CUDA_FLAGS="-lineinfo" \ | |
| -DUSE_NVTX=ON \ | |
| -DSM=80 \ | |
| -DCMAKE_CUDA_ARCHITECTURES=80 \ | |
| -DBUILD_TEST=OFF | |
| make -j$(nproc) && make install | |
| - name: Install lmdeploy | |
| run: | | |
| python3 -m pip install packaging protobuf transformers_stream_generator transformers datasets | |
| # manually install flash attn | |
| # the install packeage from. https://github.com/Dao-AILab/flash-attention/releases/download/v2.3.6/flash_attn-2.3.6+cu118torch2.0cxx11abiFALSE-cp38-cp38-linux_x86_64.whl | |
| python3 -m pip install /root/packages/flash_attn-2.3.6+cu118torch2.1cxx11abiFALSE-cp38-cp38-linux_x86_64.whl | |
| python3 -m pip install -r requirements.txt -r requirements/test.txt | |
| python3 -m pip install . | |
| - name: Check env | |
| run: | | |
| python3 -m pip list | |
| lmdeploy check_env | |
| - name: Test lmdeploy | |
| timeout-minutes: 120 | |
| run: CUDA_VISIBLE_DEVICES=5,6 pytest autotest -m pr_test --alluredir=allure-results --clean-alluredir | |
| - name: Generate reports | |
| if: always() | |
| run: | | |
| export date_today="$(date +'%Y%m%d-%H%M%S')" | |
| export report_dir="$REPORT_DIR/$date_today" | |
| echo "Save report to $ALLURE_DIR" | |
| allure generate -c -o $report_dir | |
| - name: Clear workfile | |
| if: always() | |
| run: | | |
| export workdir=$(pwd) | |
| cd .. | |
| rm -rf $workdir | |
| mkdir $workdir | |
| chmod -R 777 $workdir |