daily_ete_test #105
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: daily_ete_test | |
| on: | |
| workflow_dispatch: | |
| inputs: | |
| repo_org: | |
| required: false | |
| description: 'Tested repository organization name. Default is InternLM' | |
| type: string | |
| default: 'InternLM/lmdeploy' | |
| repo_ref: | |
| required: false | |
| description: 'Set branch or tag or commit id. Default is "main"' | |
| type: string | |
| default: 'main' | |
| backend: | |
| required: true | |
| description: 'Set backend testcase filter: turbomind or pytorch or turbomind, pytorch. Default is "["turbomind", "pytorch"]"' | |
| type: string | |
| default: "['turbomind', 'pytorch']" | |
| model: | |
| required: true | |
| description: 'Set testcase module filter: chat, restful, pipeline, quantization. Default contains all models' | |
| type: string | |
| default: "['quantization','convert','pipeline','restful','chat','interface-pipeline']" | |
| schedule: | |
| - cron: '00 21 * * *' | |
| env: | |
| HOST_PIP_CACHE_DIR: /nvme/github-actions/pip-cache | |
| HOST_LOCALTIME: /usr/share/zoneinfo/Asia/Shanghai | |
| jobs: | |
| test_functions: | |
| runs-on: [self-hosted, linux-a100] | |
| timeout-minutes: 240 | |
| 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/github-actions/pip-cache:/root/.cache/pip | |
| - /nvme/github-actions/packages:/root/packages | |
| - /nvme/qa_test_models:/nvme/qa_test_models | |
| - /mnt/bigdisk/qa_test_models:/mnt/bigdisk/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@v3 | |
| with: | |
| repository: ${{ github.event.inputs.repo_org || 'InternLM/lmdeploy' }} | |
| ref: ${{github.event.inputs.repo_ref || 'main'}} | |
| - name: Install pytorch | |
| run: | | |
| python3 -m pip cache dir | |
| python3 -m pip install torch==2.1.0 torchvision==0.16.0 --index-url https://download.pytorch.org/whl/cu118 | |
| - name: Build lmdeploy | |
| run: | | |
| python3 -m pip install cmake | |
| python3 -m pip install -r requirements/build.txt | |
| mkdir build | |
| cd build | |
| cmake .. \ | |
| -DCMAKE_BUILD_TYPE=RelWithDebInfo \ | |
| -DCMAKE_EXPORT_COMPILE_COMMANDS=1 \ | |
| -DCMAKE_INSTALL_PREFIX=/opt/tritonserver \ | |
| -DBUILD_PY_FFI=ON \ | |
| -DBUILD_MULTI_GPU=ON \ | |
| -DCMAKE_CUDA_FLAGS="-lineinfo" \ | |
| -DUSE_NVTX=ON \ | |
| -DSM=80 \ | |
| -DCMAKE_CUDA_ARCHITECTURES=80 \ | |
| -DBUILD_TEST=OFF | |
| make -j$(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 | |
| rm -rf allure-results | |
| - name: Test lmdeploy - quantization w4a16 | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'quantization')) | |
| run: | | |
| pytest autotest/tools/quantization/test_quantization_w4a16.py -m 'not pr_test' -n 8 --alluredir=allure-results --clean-alluredir | |
| - name: Test lmdeploy - quantization kv int8 | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'quantization')) | |
| run: | | |
| pytest autotest/tools/quantization/test_quantization_kvint8.py -n 8 --alluredir=allure-results | |
| - name: Test lmdeploy - quantization w8a8 | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'pytorch') && contains(fromJSON(github.event.inputs.model), 'quantization')) | |
| run: | | |
| pytest autotest/tools/quantization/test_quantization_w8a8.py -n 8 --alluredir=allure-results | |
| - name: Test lmdeploy - quantization kv int8 and w4a16 | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'quantization')) | |
| run: | | |
| pytest autotest/tools/quantization/test_quantization_kvint8_w4a16.py -n 8 --alluredir=allure-results | |
| - name: Test lmdeploy - convert | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'convert')) | |
| run: | | |
| pytest autotest/tools/convert -m 'not pr_test' -n 8 --alluredir=allure-results | |
| - name: Test lmdeploy - chat workspace | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'chat')) | |
| timeout-minutes: 20 | |
| run: | | |
| pytest autotest/tools/chat/test_command_chat_workspace.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/chat/test_command_chat_workspace.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - chat hf turbomind | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'chat')) | |
| timeout-minutes: 20 | |
| run: | | |
| pytest autotest/tools/chat/test_command_chat_hf_turbomind.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/chat/test_command_chat_hf_turbomind.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - chat hf torch | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'pytorch') && contains(fromJSON(github.event.inputs.model), 'chat')) | |
| timeout-minutes: 20 | |
| run: | | |
| pytest autotest/tools/chat/test_command_chat_hf_pytorch.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/chat/test_command_chat_hf_pytorch.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - pipeline turbomind | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'pipeline')) | |
| timeout-minutes: 25 | |
| run: | | |
| pytest autotest/tools/pipeline/test_pipeline_chat_turbomind.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/pipeline/test_pipeline_chat_turbomind.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - restful turbomind | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'restful')) | |
| timeout-minutes: 30 | |
| run: | | |
| pytest autotest/tools/restful/test_restful_chat_turbomind.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/restful/test_restful_chat_turbomind.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - interface pipeline turbomind case | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'turbomind') && contains(fromJSON(github.event.inputs.model), 'interface-pipeline')) | |
| timeout-minutes: 20 | |
| run: | | |
| pytest autotest/interface/pipeline/test_pipeline_turbomind_func.py -m 'not pr_test' --alluredir=allure-results | |
| - name: Test lmdeploy - pipeline torch | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'pytorch') && contains(fromJSON(github.event.inputs.model), 'pipeline')) | |
| timeout-minutes: 25 | |
| run: | | |
| pytest autotest/tools/pipeline/test_pipeline_chat_pytorch.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/pipeline/test_pipeline_chat_pytorch.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - restful torch | |
| continue-on-error: true | |
| if: github.event_name == 'schedule' || (contains(fromJSON(github.event.inputs.backend), 'pytorch') && contains(fromJSON(github.event.inputs.model), 'restful')) | |
| timeout-minutes: 40 | |
| run: | | |
| pytest autotest/tools/restful/test_restful_chat_pytorch.py -m 'gpu_num_1 and not pr_test' -n 8 --alluredir=allure-results | |
| pytest autotest/tools/restful/test_restful_chat_pytorch.py -m 'gpu_num_2 and not pr_test' -n 4 --alluredir=allure-results | |
| - name: Test lmdeploy - rerun all fail cases | |
| timeout-minutes: 30 | |
| run: | | |
| pytest autotest --lf --alluredir=allure-results | |
| - 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 $workdir | |
| test_triton: | |
| runs-on: [self-hosted, linux-a100] | |
| timeout-minutes: 30 | |
| env: | |
| HF_MODEL: /nvme/qa_test_models/internlm-chat-20b | |
| WORKDIR: /nvme/qa_test_models/triton_workspace | |
| TB_MODEL: internlm-chat-20b-fp16-tp2 | |
| GRPC_PORT: 33337 | |
| steps: | |
| - name: Clone repository | |
| uses: actions/checkout@v2 | |
| - name: Create test container | |
| run: | | |
| export CONTAINER_ID=$(docker create \ | |
| --rm \ | |
| --gpus='"device=4,5"' \ | |
| --shm-size 16g \ | |
| --cap-add=SYS_PTRACE \ | |
| --cap-add=SYS_ADMIN \ | |
| --security-opt seccomp=unconfined \ | |
| --name "lmdeploy-ci-triton-$GITHUB_RUN_ID" \ | |
| --workdir /__w/lmdeploy/lmdeploy \ | |
| --env NCCL_LAUNCH_MODE=GROUP \ | |
| -v $(pwd)/../../:/__w \ | |
| -v ${HF_MODEL}:/root/workspace/hf_model \ | |
| -v ${WORKDIR}:/root/workspace/workdir \ | |
| -v ${HOST_PIP_CACHE_DIR}:/root/.cache/pip \ | |
| -v ${HOST_LOCALTIME}:/etc/localtime:ro \ | |
| openmmlab/lmdeploy:latest tail -f /dev/null \ | |
| ) | |
| docker start $CONTAINER_ID | |
| echo "CONTAINER_ID=$CONTAINER_ID" | |
| echo "CONTAINER_ID=$CONTAINER_ID" >> $GITHUB_ENV | |
| - name: Build lmdeploy from source | |
| run: | | |
| docker exec $CONTAINER_ID mkdir build | |
| docker exec --workdir /__w/lmdeploy/lmdeploy/build \ | |
| --env http_proxy=${{secrets.PROXY}} \ | |
| --env https_proxy=${{secrets.PROXY}} \ | |
| --env HTTP_PROXY=${{secrets.PROXY}} \ | |
| --env HTTPS_PROXY=${{secrets.PROXY}} \ | |
| --env no_proxy="localhost,127.0.0.1" \ | |
| --env NO_PROXY="localhost,127.0.0.1" \ | |
| $CONTAINER_ID 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 | |
| docker exec --workdir /__w/lmdeploy/lmdeploy/build $CONTAINER_ID make -j$(nproc) | |
| docker exec --workdir /__w/lmdeploy/lmdeploy/build $CONTAINER_ID make install \ | |
| --env http_proxy=${{secrets.PROXY}} \ | |
| --env https_proxy=${{secrets.PROXY}} \ | |
| --env HTTP_PROXY=${{secrets.PROXY}} \ | |
| --env HTTPS_PROXY=${{secrets.PROXY}} | |
| - name: Install lmdeploy | |
| run: | | |
| docker exec \ | |
| --env http_proxy=${{secrets.PROXY}} \ | |
| --env https_proxy=${{secrets.PROXY}} \ | |
| $CONTAINER_ID python3 -m pip install tritonclient[grpc] protobuf | |
| docker exec \ | |
| --env http_proxy=${{secrets.PROXY}} \ | |
| --env https_proxy=${{secrets.PROXY}} \ | |
| $CONTAINER_ID python3 -m pip install -r requirements/test.txt | |
| docker exec \ | |
| --env http_proxy=${{secrets.PROXY}} \ | |
| --env https_proxy=${{secrets.PROXY}} \ | |
| $CONTAINER_ID python3 -m pip install . | |
| docker exec $CONTAINER_ID lmdeploy check_env | |
| - name: Convert to turbomind model | |
| run: | | |
| docker exec $CONTAINER_ID \ | |
| lmdeploy convert \ | |
| internlm-chat-20b \ | |
| /root/workspace/hf_model \ | |
| --tp 2 \ | |
| --dst-path /root/workspace/workdir/${TB_MODEL} | |
| - name: Start triton server service | |
| run: | | |
| docker exec --detach $CONTAINER_ID bash -c \ | |
| "tritonserver \ | |
| --model-repository=/root/workspace/workdir/${TB_MODEL}/model_repository \ | |
| --allow-http=0 \ | |
| --allow-grpc=1 \ | |
| --grpc-port=${GRPC_PORT} \ | |
| --log-verbose=0 \ | |
| --allow-metrics=1 > run.log 2>&1 ; touch finish.txt" | |
| # wait for triton server to fully start up | |
| sleep 180s | |
| # print triton server log file | |
| cat run.log | |
| python3 -c 'import os; assert not os.path.exists("finish.txt"), "Failed to start tritonserver"' | |
| - name: Test triton server | |
| run: | | |
| docker exec \ | |
| --env no_proxy="localhost,127.0.0.1" \ | |
| --env NO_PROXY="localhost,127.0.0.1" \ | |
| $CONTAINER_ID python3 .github/scripts/test_triton_server.py --port ${GRPC_PORT} | |
| # print triton server log file | |
| cat run.log | |
| - name: Clear workfile | |
| if: always() | |
| run: | | |
| export workdir=$(pwd) | |
| docker exec --workdir /__w/lmdeploy $CONTAINER_ID rm -rf lmdeploy | |
| mkdir $workdir | |
| chmod -R 777 $workdir | |
| docker exec --workdir /__w/lmdeploy $CONTAINER_ID rm -rf /root/workspace/workdir/${TB_MODEL} | |
| docker stop $CONTAINER_ID | |
| notify_to_feishu: | |
| if: always() && !cancelled() && (github.ref_name == 'develop' || github.ref_name == 'main') | |
| needs: [test_functions, test_triton] | |
| timeout-minutes: 5 | |
| runs-on: [self-hosted, linux-a100] | |
| steps: | |
| - name: fail notify | |
| if: contains(needs.*.result, 'failure') | |
| run: | | |
| curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"Lmdeploy- daily test failed","content":[[{"tag":"text","text":"branch: ${{github.ref_name}}, run action: ${{github.workflow}} failed. "},{"tag":"a","text":"Please click here for details ","href":"https://github.com/'${{ github.repository }}'/actions/runs/'${GITHUB_RUN_ID}'"},{"tag":"at","user_id":"'${{ secrets.FEISHU_USER_ID }}'"}]]}}}}' ${{ secrets.FEISHU_WEBHOOK_URL }} |