opencompass/.github/workflows/daily-run-test.yml

name: daily_run_test

on:
  workflow_dispatch:
    inputs:
      repo_org:
        required: false
        description: 'Tested repository organization name. Default is open-compass/opencompass'
        type: string
        default: 'open-compass/opencompass'
      repo_ref:
        required: false
        description: 'Set branch or tag or commit id. Default is "main"'
        type: string
        default: 'main'
      build_lmdeploy:
        required: false
        description: 'whether to build lmdeploy'
        type:  boolean
        default: false
      repo_org_lmdeploy:
        required: false
        description: 'Tested repository organization name. Default is internlm/lmdeploy'
        type: string
        default: 'InternLM/lmdeploy'
      repo_ref_lmdeploy:
        required: false
        description: 'Set branch or tag or commit id. Default is "main"'
        type: string
        default: 'main'
      regression_func:
        required: true
        description: 'regression functions'
        type: string
        default: "['chat_models','base_models', 'chat_obj_fullbench', 'chat_sub_fullbench', 'base_fullbench','cmd', 'api']"
      cuda_env:
        required: true
        description: "regression conda env, eg. ['dsw_cu11','dsw_cu12']"
        type: string
        default: "['dsw_cu12']"
      fullbench_eval:
        required: true
        description: 'fullbench volc functions'
        type: string
        default: "['base_long_context','base_objective','chat_long_context','chat_objective','chat_subjective']"
  schedule:
    - cron:  '15 14 * * *'

env:
  HF_DATASETS_OFFLINE: 1
  HF_EVALUATE_OFFLINE: 1
  TRANSFORMERS_OFFLINE: 1
  VLLM_USE_MODELSCOPE: false
  LMDEPLOY_USE_MODELSCOPE: false
  HF_HUB_OFFLINE: 1
  OUTPUT_FOLDER: cuda12.1_dist_${{ github.run_id }}

jobs:
  build-pypi:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v2
        with:
          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
          ref: ${{github.event.inputs.repo_ref || 'main'}}
      - name: Set up Python 3.x
        uses: actions/setup-python@v2
        with:
          python-version: 3.x
      - name: Build lagent
        run: |
          pip install wheel setuptools
          python setup.py sdist bdist_wheel
      - name: Upload Artifacts
        uses: actions/upload-artifact@v4
        with:
          if-no-files-found: error
          path: dist/*
          retention-days: 1
          name: my-artifact-${{ github.run_id }}

  build-pypi-lmdeploy:
    if: ${{!cancelled() && (github.event_name != 'schedule' && inputs.build_lmdeploy)}}
    strategy:
      matrix:
        pyver: [py310]
    runs-on: ubuntu-latest
    env:
      PYTHON_VERSION: ${{ matrix.pyver }}
      PLAT_NAME: manylinux2014_x86_64
      DOCKER_TAG: cuda12.1
    steps:
      - name: Checkout repository
        uses: actions/checkout@v3
        with:
          repository: ${{ github.event.inputs.repo_org_lmdeploy || 'InternLM/lmdeploy' }}
          ref: ${{github.event.inputs.repo_ref_lmdeploy || 'main'}}
      - name: Build
        run: |
          echo ${PYTHON_VERSION}
          echo ${PLAT_NAME}
          echo ${DOCKER_TAG}
          echo ${OUTPUT_FOLDER}
          echo ${GITHUB_RUN_ID}
          # remove -it
          sed -i 's/docker run --rm -it/docker run --rm/g' builder/manywheel/build_wheel.sh
          bash builder/manywheel/build_wheel.sh ${PYTHON_VERSION} ${PLAT_NAME} ${DOCKER_TAG} ${OUTPUT_FOLDER}
      - name: Upload Artifacts
        uses: actions/upload-artifact@v4
        with:
          if-no-files-found: error
          path: builder/manywheel/${{ env.OUTPUT_FOLDER }}
          retention-days: 1
          name: my-artifact-${{ github.run_id }}-${{ matrix.pyver }}


  prepare_env:
    if: ${{!cancelled()}}
    needs: ['build-pypi', 'build-pypi-lmdeploy']
    strategy:
      fail-fast: false
      matrix:
        cuda_env: ${{ fromJSON(inputs.cuda_env || '["dsw_cu12"]')}}
    runs-on: ${{ matrix.cuda_env }}
    env:
      CONDA_ENV: opencompass_regression
      PIP_CACHE_PATH: /cpfs01/user/qa-llm-cicd/.cache/pip
    environment: 'prod'
    timeout-minutes: 240 #4hours
    steps:
      - name: Clone repository
        uses: actions/checkout@v2
        with:
          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
          ref: ${{github.event.inputs.repo_ref || 'main'}}
      - name: Download Artifacts
        uses: actions/download-artifact@v4
        with:
          name: my-artifact-${{ github.run_id }}
      - name:  Remove Conda Env
        if: always()
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda env remove -y --name ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
      - name: Prepare - create conda env and install torch - cu11
        if: ${{matrix.cuda_env == 'dsw_cu11'}}
        uses: nick-fields/retry@v3
        id: retry1
        with:
          max_attempts: 3
          timeout_minutes: 40
          command: |
            . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
            conda create -y --name ${{env.CONDA_ENV}}_${{ matrix.cuda_env }} python=3.10
            conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
            pip install -r /cpfs01/shared/public/qa-llm-cicd/requirements-cu11.txt --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install opencompass*.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install /cpfs01/user/qa-llm-cicd/packages/lmdeploy-0.6.1+cu118-cp310-cp310-manylinux2014_x86_64.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install /cpfs01/user/qa-llm-cicd/packages/vllm-0.6.1.post1+cu118-cp310-cp310-manylinux1_x86_64.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            pip uninstall torch torchvision torchaudio -y
            pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --cache-dir ${{env.PIP_CACHE_PATH}} --index-url https://download.pytorch.org/whl/cu118
            FLASH_ATTENTION_FORCE_BUILD=TRUE pip install /cpfs01/user/qa-llm-cicd/packages/flash_attn-2.7.0.post2+cu11torch2.5cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
            pip install /cpfs01/user/qa-llm-cicd/packages/xformers-0.0.28.post3-cp310-cp310-manylinux_2_28_x86_64.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            conda info --envs
            pip list
      - name: Prepare - create conda env and install torch - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12'}}
        uses: nick-fields/retry@v3
        id: retry2
        with:
          max_attempts: 3
          timeout_minutes: 40
          command: |
            . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
            conda create -y --name ${{env.CONDA_ENV}}_${{ matrix.cuda_env }} python=3.10
            conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
            pip install -r /cpfs01/shared/public/qa-llm-cicd/requirements-cu12.txt --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install opencompass*.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install opencompass[lmdeploy] --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install opencompass[vllm] --cache-dir ${{env.PIP_CACHE_PATH}}
            pip install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --cache-dir ${{env.PIP_CACHE_PATH}}
            FLASH_ATTENTION_FORCE_BUILD=TRUE pip install /cpfs01/user/qa-llm-cicd/packages/flash_attn-2.7.0.post2+cu12torch2.5cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
            pip install /cpfs01/user/qa-llm-cicd/packages/xformers-0.0.28.post3-cp310-cp310-manylinux_2_28_x86_64.whl --cache-dir ${{env.PIP_CACHE_PATH}}
            conda info --envs
            pip list
      - name: Prepare - reinstall lmdeploy - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        uses: actions/download-artifact@v4
        with:
          name: my-artifact-${{ github.run_id }}-py310
      - name: Prepare - reinstall lmdeploy - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          pip install lmdeploy-*.whl --no-deps

  daily_run_test:
    if: ${{!cancelled()}}
    needs: prepare_env
    strategy:
      fail-fast: false
      matrix:
        cuda_env: ${{ fromJSON(inputs.cuda_env || '["dsw_cu12"]')}}
        regression_func: ${{fromJSON(github.event.inputs.regression_func || '["chat_models","base_models","chat_obj_fullbench","chat_sub_fullbench","base_fullbench","cmd","api"]')}}
    runs-on: ${{ matrix.cuda_env }}
    env:
      CONDA_ENV: opencompass_regression
      PIP_CACHE_PATH: /cpfs01/user/qa-llm-cicd/.cache/pip
      HF_CACHE_PATH: /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub
      HUGGINGFACE_HUB_CACHE: /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub
      HF_HUB_CACHE: /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub
      COMPASS_DATA_CACHE: /cpfs01/shared/public/llmeval/compass_data_cache
      REPORT_ROOT: /cpfs01/shared/public/qa-llm-cicd/report
    environment: 'prod'
    timeout-minutes: 240 #4hours
    steps:
      - name: Clone repository
        uses: actions/checkout@v2
        with:
          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
          ref: ${{github.event.inputs.repo_ref || 'main'}}
      - name: Prepare - prepare data and hf model
        run: |
          rm -rf ~/.cache/huggingface/hub -f && mkdir ~/.cache -p && mkdir ~/.cache/huggingface -p
          ln -s ${{env.HF_CACHE_PATH}} ~/.cache/huggingface/hub
      - name:  Run command testcase
        if: matrix.regression_func == 'cmd'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          export from_tf=TRUE
          python tools/list_configs.py internlm2_5 mmlu
          opencompass --models hf_internlm2_5_7b hf_internlm2_1_8b --datasets race_ppl demo_gsm8k_chat_gen --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd1_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd1_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m case1 -s -v --color=yes .github/scripts/oc_score_assert.py
          opencompass --models hf_internlm2_5_7b_chat hf_internlm2_chat_1_8b --datasets race_gen demo_gsm8k_chat_gen -a lmdeploy --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd2_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd2_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m case2 -s -v --color=yes .github/scripts/oc_score_assert.py
          opencompass --datasets race_ppl demo_gsm8k_chat_gen --hf-type base --hf-path internlm/internlm2_5-7b --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd3_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd3_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m case3 -s -v --color=yes .github/scripts/oc_score_assert.py
          opencompass --datasets race_gen demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm2_5-7b-chat --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd4_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd4_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m case4 -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run chat model test
        if: matrix.regression_func == 'chat_models'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          opencompass .github/scripts/eval_regression_chat.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m chat -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run base model test
        if: matrix.regression_func == 'base_models'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          opencompass .github/scripts/eval_regression_base.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/base_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/base_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m base -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run chat model test - fullbench
        if: matrix.regression_func == 'chat_obj_fullbench'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          opencompass .github/scripts/eval_regression_chat_objective_fullbench.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_obj_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_obj_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m chat_obj_fullbench -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run chat model test - fullbench
        if: matrix.regression_func == 'chat_sub_fullbench'
        env:
          COMPASS_DATA_CACHE: /cpfs01/shared/public/llmeval/compass_data_cache_subset
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          opencompass .github/scripts/eval_regression_chat_subjective_fullbench.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_sub_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/chat_sub_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m chat_sub_fullbench -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run base model test - fullbench
        if: matrix.regression_func == 'base_fullbench'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          opencompass .github/scripts/eval_regression_base_fullbench.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/base_full_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/base_full_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m base_fullbench -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run model test - api
        if: matrix.regression_func == 'api'
        run: |
          . /cpfs01/shared/public/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.CONDA_ENV}}_${{ matrix.cuda_env }}
          conda info --envs
          lmdeploy serve api_server internlm/internlm2_5-7b-chat --max-batch-size 256 --model-name internlm2 > ${{env.REPORT_ROOT}}/${{ github.run_id }}/restful.log  2>&1  &
          echo "restful_pid=$!" >> "$GITHUB_ENV"
          sleep 120s
          opencompass .github/scripts/eval_regression_api.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/api_${{ matrix.cuda_env }} --reuse --max-num-workers 2 --dump-eval-details
          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/api_${{ matrix.cuda_env }}/*/summary regression_result_daily
          python -m pytest -m api -s -v --color=yes .github/scripts/oc_score_assert.py
      - name:  Run model test - api kill
        if: always() && matrix.regression_func == 'api'
        run: |
          kill -15 "$restful_pid"

  fullbench_run_test:
    if: ${{!cancelled()}}
    needs: ['build-pypi', 'build-pypi-lmdeploy']
    env:
      FULLBENCH_CONDA_ENV: regression_test
      FULLBENCH_REPORT_ROOT: /fs-computility/llm/qa-llm-cicd/eval_report/regression
      COMPASS_DATA_CACHE: /fs-computility/llm/shared/llmeval/datasets/compass_data_cache
    strategy:
      fail-fast: false
      matrix:
        function_type: ${{fromJSON(github.event.inputs.fullbench_eval || '["base_long_context","base_objective","chat_long_context","chat_objective","chat_subjective"]')}}
    runs-on: volc_cu12
    environment: 'prod'
    timeout-minutes: 360 #6hours
    steps:
      - name: Clone repository
        uses: actions/checkout@v2
        with:
          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
          ref: ${{github.event.inputs.repo_ref || 'main'}}
      - name: Download Artifacts
        uses: actions/download-artifact@v4
        with:
          name: my-artifact-${{ github.run_id }}
      - name: Prepare - reinstall opencompass - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        run: |
          . /fs-computility/llm/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.FULLBENCH_CONDA_ENV}}
          pip install opencompass*.whl --no-deps
      - name: Prepare - reinstall lmdeploy - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        uses: actions/download-artifact@v4
        with:
          name: my-artifact-${{ github.run_id }}-py310
      - name: Prepare - reinstall lmdeploy - cu12
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        run: |
          . /fs-computility/llm/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.FULLBENCH_CONDA_ENV}}
          pip install lmdeploy-*.whl --no-deps
      - name: Conda env
        if: ${{matrix.cuda_env == 'dsw_cu12' && inputs.build_lmdeploy}}
        run: |
          . /fs-computility/llm/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.FULLBENCH_CONDA_ENV}}
          conda info --envs
          pip list
      - name:  Run command testcase
        run: |
          . /fs-computility/llm/qa-llm-cicd/miniconda3/bin/activate
          conda activate ${{env.FULLBENCH_CONDA_ENV}}
          conda info --envs
          export from_tf=TRUE
          opencompass /fs-computility/llm/qa-llm-cicd/ocplayground/template/regression/eval_${{ matrix.function_type }}.py --work-dir ${{env.FULLBENCH_REPORT_ROOT}}/${{ github.run_id }}/${{ matrix.function_type }} --reuse
          rm regression_result_daily -f && ln -s ${{env.FULLBENCH_REPORT_ROOT}}/${{ github.run_id }}/${{ matrix.function_type }}/*/summary regression_result_daily
          python -m pytest -m ${{ matrix.function_type }} -s -v --color=yes .github/scripts/oc_score_assert.py


  notify_to_feishu:
    if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
    needs: [daily_run_test, fullbench_run_test]
    environment: 'prod'
    timeout-minutes: 5
    runs-on: self-hosted
    steps:
      - name: notify
        run: |
          curl -X POST -H "Content-Type: application/json" -d '{"msg_type":"post","content":{"post":{"zh_cn":{"title":"Opencompass- 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.USER_ID }}'"}]]}}}}'  ${{ secrets.WEBHOOK_URL }}