[Feature] support alpacaeval #66
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_run_test | |
| on: | |
| pull_request: | |
| paths-ignore: | |
| - 'README.md' | |
| - 'README_zh-CN.md' | |
| - 'docs/**' | |
| - 'configs/**' | |
| - 'tools/**' | |
| workflow_dispatch: | |
| schedule: | |
| - cron: '56 20 * * *' | |
| concurrency: | |
| group: ${{ github.workflow }}-${{ github.ref }} | |
| cancel-in-progress: true | |
| jobs: | |
| pr_run_test: | |
| runs-on: self-hosted | |
| timeout-minutes: 20 | |
| steps: | |
| - name: Clone repository | |
| uses: actions/checkout@v2 | |
| - name: Prepare - create conda env and install code | |
| run: | | |
| eval "$(conda shell.bash hook)" | |
| conda create --name opencompass_regression --clone opencompass_base | |
| conda activate opencompass_regression | |
| python3 -m pip install -e . | |
| conda info --envs | |
| - name: Prepare - prepare data and hf model | |
| run: | | |
| cp -r /cpfs01/user/qa-llm-cicd/data . | |
| rm -rf ~/.cache/huggingface/hub -f && mkdir ~/.cache -p && mkdir ~/.cache/huggingface -p | |
| ln -s /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub ~/.cache/huggingface/hub | |
| export HF_DATASETS_OFFLINE=1; export TRANSFORMERS_OFFLINE=1; | |
| - name: Run test | |
| run: | | |
| eval "$(conda shell.bash hook)" | |
| conda activate opencompass_regression | |
| conda info --envs | |
| rm -rf regression_result | |
| python3 run.py --models hf_internlm_chat_7b --datasets siqa_gen --work-dir regression_result --debug | |
| - name: Get result | |
| if: always() | |
| run: | | |
| score=$(sed -n '$p' regression_result/*/summary/*.csv | awk -F ',' '{print $NF}') | |
| if (( ${score%.*} >= 70 && ${score%.*} <= 80 )); then | |
| echo "score is $score between 70 and 80" | |
| else | |
| echo "score is $score not between 70 and 80" | |
| exit 1 | |
| fi | |
| rm -rf regression_result | |
| - name: Remove Conda Env | |
| if: always() | |
| run: | | |
| eval "$(conda shell.bash hook)" | |
| conda env remove --name opencompass_regression | |
| conda info --envs |