pr_run_test #452
This file contains hidden or 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_run_test | |
| on: | |
| pull_request: | |
| branches: | |
| - "main" | |
| paths-ignore: | |
| - "docs/**" | |
| - "**.md" | |
| workflow_dispatch: | |
| schedule: | |
| - cron: '56 01 * * *' | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }} | |
| cancel-in-progress: true | |
| env: | |
| BASE_SCORE: '{"MMBench_V11_MINI":{"Qwen2-VL-7B-Instruct":0.8727272727272727,"InternVL2_5-8B":0.89090909,"llava_onevision_qwen2_7b_si":0.8363636363636363},"MMStar_MINI":{"Qwen2-VL-7B-Instruct":0.6266666666666667,"InternVL2_5-8B":0.6333333333333333,"llava_onevision_qwen2_7b_si":0.49333333333333335},"AI2D_MINI":{"Qwen2-VL-7B-Instruct":0.7975708502024291,"InternVL2_5-8B":0.854251012145749,"llava_onevision_qwen2_7b_si":0.8178137651821862},"OCRBench_MINI":{"Qwen2-VL-7B-Instruct":16.6,"InternVL2_5-8B":16.7,"llava_onevision_qwen2_7b_si":13.0}}' | |
| HF_HUB_CACHE: /fs-computility/llm/shared/llmeval/models/opencompass_hf_hub | |
| HF_HUB_OFFLINE: 1 | |
| CONDA_PATH: /fs-computility/llm/qa-llm-cicd/miniconda3 | |
| CONDA_ENV: vlm_pr_test | |
| jobs: | |
| vlm_test: | |
| if: ${{!cancelled()}} | |
| runs-on: [volc_cu12_mllm] | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| model: [Qwen/Qwen2-VL-7B-Instruct,OpenGVLab/InternVL2_5-8B,lmms-lab/llava-onevision-qwen2-7b-si] | |
| dataset: ["MMBench_V11_MINI MMStar_MINI AI2D_MINI","OCRBench_MINI"] | |
| steps: | |
| - name: clone_repo | |
| uses: actions/checkout@v3 | |
| - name: evaluation_model | |
| uses: nick-fields/retry@v3 | |
| with: | |
| max_attempts: 3 | |
| timeout_minutes: 30 | |
| command: | | |
| . ${{env.CONDA_PATH}}/bin/activate | |
| conda activate ${{env.CONDA_ENV}} | |
| pip uninstall vlmeval -y | |
| pip install -e . | |
| pre_model=$(echo ${{matrix.model}} | awk -F'/' '{print $1}') | |
| if [ "${{matrix.model}}" = "lmms-lab/llava-onevision-qwen2-7b-si" ];then | |
| model_name="llava_onevision_qwen2_7b_si" | |
| else | |
| model_name=$(echo ${{matrix.model}} | awk -F'/' '{print $2}') | |
| fi | |
| pip list | |
| nvidia-smi | |
| LOG=$(python run.py --data ${{matrix.dataset}} --model $model_name 2>&1) | |
| echo "$LOG" | |
| if echo "$LOG" | grep -q "CUDA out of memory"; then | |
| sleep 300 | |
| exit 1 # retry becuase of oom | |
| fi | |
| - name: assert_result | |
| run: | | |
| . ${{env.CONDA_PATH}}/bin/activate | |
| conda activate ${{env.CONDA_ENV}} | |
| if [ "${{matrix.model}}" = "lmms-lab/llava-onevision-qwen2-7b-si" ];then | |
| model_name="llava_onevision_qwen2_7b_si" | |
| else | |
| model_name=$(echo ${{matrix.model}} | awk -F'/' '{print $2}') | |
| fi | |
| python .github/scripts/assert_score.py --dataset "${{matrix.dataset}}" --base_score $BASE_SCORE --model-name $model_name |