LLM Whisper Models Evaluation #154
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: LLM Whisper Models Evaluation | |
| # Cancel previous runs in the PR when you push new commits | |
| concurrency: | |
| group: ${{ github.workflow }}-llm-nightly-test-${{ github.event.pull_request.number || github.run_id }} | |
| cancel-in-progress: true | |
| permissions: | |
| contents: read | |
| # Controls when the action will run. | |
| on: | |
| schedule: | |
| - cron: "00 13 * * *" # GMT time, 13:00 GMT == 21:00 China | |
| pull_request: | |
| branches: [main] | |
| paths: | |
| - ".github/workflows/llm-whisper-evaluation.yml" | |
| # Allows you to run this workflow manually from the Actions tab | |
| workflow_dispatch: | |
| inputs: | |
| model_name: | |
| description: 'Model names, separated by comma and must be quoted.' | |
| required: true | |
| type: string | |
| precision: | |
| description: 'Precisions, separated by comma and must be quoted.' | |
| required: true | |
| type: string | |
| task: | |
| description: 'Tasks, separated by comma and must be quoted.' | |
| required: true | |
| type: string | |
| runs-on: | |
| description: 'Labels to filter the runners, separated by comma and must be quoted.' | |
| default: "accuracy" | |
| required: false | |
| type: string | |
| # A workflow run is made up of one or more jobs that can run sequentially or in parallel | |
| jobs: | |
| llm-cpp-build: # please uncomment it for PR tests | |
| uses: ./.github/workflows/llm-binary-build.yml | |
| # Set the testing matrix based on the event (schedule, PR, or manual dispatch) | |
| set-matrix: | |
| runs-on: ubuntu-latest | |
| outputs: | |
| model_name: ${{ steps.set-matrix.outputs.model_name }} | |
| precision: ${{ steps.set-matrix.outputs.precision }} | |
| task: ${{ steps.set-matrix.outputs.task }} | |
| runner: ${{ steps.set-matrix.outputs.runner }} | |
| steps: | |
| - name: set-env | |
| env: | |
| MATRIX_MODEL_NAME: '["whisper-tiny", "whisper-small", "whisper-medium", "whisper-base"]' | |
| MATRIX_TASK: '["librispeech"]' | |
| MATRIX_PRECISION: '["sym_int4", "fp8_e5m2"]' | |
| LABELS: '["self-hosted", "llm", "perf"]' | |
| run: | | |
| echo "model_name=$MATRIX_MODEL_NAME" >> $GITHUB_ENV | |
| echo "task=$MATRIX_TASK" >> $GITHUB_ENV | |
| echo "precision=$MATRIX_PRECISION" >> $GITHUB_ENV | |
| echo "runner=$LABELS" >> $GITHUB_ENV | |
| - name: set-matrix | |
| id: set-matrix | |
| run: | | |
| echo "model_name=$model_name" >> $GITHUB_OUTPUT | |
| echo "task=$task" >> $GITHUB_OUTPUT | |
| echo "precision=$precision" >> $GITHUB_OUTPUT | |
| echo "runner=$runner" >> $GITHUB_OUTPUT | |
| llm-whisper-evaluation: | |
| # if: ${{ github.event.schedule || github.event.inputs.artifact == 'llm-whisper-evaluation' || github.event.inputs.artifact == 'all' }} # please comment it for PR tests | |
| needs: [llm-cpp-build, set-matrix] # please uncomment it for PR tests | |
| # needs: [set-matrix] # please comment it for PR tests | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| python-version: ["3.11"] | |
| model_name: ${{ fromJson(needs.set-matrix.outputs.model_name) }} | |
| task: ${{ fromJson(needs.set-matrix.outputs.task) }} | |
| precision: ${{ fromJson(needs.set-matrix.outputs.precision) }} | |
| device: [xpu] | |
| runs-on: ${{ fromJson(needs.set-matrix.outputs.runner) }} | |
| env: | |
| ANALYTICS_ZOO_ROOT: ${{ github.workspace }} | |
| ORIGIN_DIR: /mnt/disk1/models | |
| steps: | |
| - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3 | |
| - name: Set up Python ${{ matrix.python-version }} | |
| uses: actions/setup-python@v4 | |
| with: | |
| python-version: ${{ matrix.python-version }} | |
| - name: Install dependencies | |
| shell: bash | |
| run: | | |
| python -m pip install --upgrade pip | |
| python -m pip install --upgrade wheel | |
| python -m pip install --upgrade pandas | |
| python -m pip install --upgrade datasets | |
| python -m pip install --upgrade evaluate | |
| python -m pip install --upgrade soundfile | |
| python -m pip install --upgrade librosa | |
| python -m pip install --upgrade jiwer | |
| # please uncomment it and comment the "Install IPEX-LLM from Pypi" part for PR tests | |
| - name: Download llm binary | |
| uses: ./.github/actions/llm/download-llm-binary | |
| - name: Run LLM install (all) test | |
| uses: ./.github/actions/llm/setup-llm-env | |
| with: | |
| extra-dependency: "xpu_2.1" | |
| # - name: Install IPEX-LLM from Pypi | |
| # shell: bash | |
| # run: | | |
| # pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/ | |
| # - name: Test installed xpu version | |
| # shell: bash | |
| # run: | | |
| # source /opt/intel/oneapi/setvars.sh | |
| # bash python/llm/test/run-llm-install-tests.sh | |
| - name: Run whisper evaluation | |
| shell: bash | |
| run: | | |
| source /opt/intel/oneapi/setvars.sh | |
| export USE_XETLA=OFF | |
| export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1 | |
| echo "MODEL_PATH=${ORIGIN_DIR}/${{ matrix.model_name }}/" >> "$GITHUB_ENV" | |
| MODEL_PATH=${ORIGIN_DIR}/${{ matrix.model_name }}/ | |
| export LIBRISPEECH_DATASET_PATH=/mnt/disk1/datasets/librispeech | |
| cd python/llm/dev/benchmark/whisper | |
| python run_whisper.py --model_path ${MODEL_PATH} --data_type other --device xpu --load_in_low_bit ${{ matrix.precision }} --save_result | |
| - uses: actions/upload-artifact@v3 | |
| with: | |
| name: whisper_results | |
| path: | |
| ${{ github.workspace }}/python/llm/dev/benchmark/whisper/results/** | |
| llm-whisper-summary: | |
| if: ${{github.event_name == 'schedule' || github.event_name == 'pull_request'}} | |
| needs: [set-matrix, llm-whisper-evaluation] | |
| runs-on: ["self-hosted", "llm", "perf"] | |
| steps: | |
| - uses: actions/checkout@f43a0e5ff2bd294095638e18286ca9a3d1956744 # actions/checkout@v3 | |
| - name: Set up Python 3.11 | |
| uses: actions/setup-python@v4 | |
| with: | |
| python-version: 3.11 | |
| - name: Set output path | |
| shell: bash | |
| run: | | |
| DATE=$(date +%Y-%m-%d) | |
| OUTPUT_PATH="results_$DATE" | |
| echo "OUTPUT_PATH=$OUTPUT_PATH" >> $GITHUB_ENV | |
| NIGHTLY_FOLDER="/mnt/disk1/whisper_nightly_gpu" | |
| echo "NIGHTLY_FOLDER=$NIGHTLY_FOLDER" >> $GITHUB_ENV | |
| PR_FOLDER="/mnt/disk1/whisper_pr_gpu" | |
| echo "PR_FOLDER=$PR_FOLDER" >> $GITHUB_ENV | |
| - name: Download all results for nightly run | |
| if: github.event_name == 'schedule' | |
| uses: actions/download-artifact@v3 | |
| with: | |
| name: whisper_results | |
| path: ${{ env.NIGHTLY_FOLDER}}/${{ env.OUTPUT_PATH }} | |
| - name: Download all results for pr run | |
| if: github.event_name == 'pull_request' | |
| uses: actions/download-artifact@v3 | |
| with: | |
| name: whisper_results | |
| path: ${{ env.PR_FOLDER}}/${{ env.OUTPUT_PATH }} | |
| - name: Summarize the results for nightly run | |
| if: github.event_name == 'schedule' | |
| shell: bash | |
| run: | | |
| cp -r /mnt/disk1/datasets/whisper_fp16_results/* /mnt/disk1/whisper_nightly_gpu/${{ env.OUTPUT_PATH }} | |
| pip install pandas==1.5.3 | |
| python ${{ github.workspace }}/python/llm/dev/benchmark/whisper/whisper_concat_csv.py -i ${{ env.NIGHTLY_FOLDER}}/${{ env.OUTPUT_PATH }} -o ${{ env.NIGHTLY_FOLDER}} | |
| python ${{ github.workspace }}/python/llm/dev/benchmark/whisper/whisper_csv_to_html.py -f ${{ env.NIGHTLY_FOLDER}} | |
| - name: Summarize the results for pull request | |
| if: github.event_name == 'pull_request' | |
| shell: bash | |
| run: | | |
| cp -r /mnt/disk1/datasets/whisper_fp16_results/* /mnt/disk1/whisper_pr_gpu/${{ env.OUTPUT_PATH }} | |
| pip install pandas==1.5.3 | |
| python ${{ github.workspace }}/python/llm/dev/benchmark/whisper/whisper_concat_csv.py -i ${{ env.PR_FOLDER}}/${{ env.OUTPUT_PATH }} -o ${{ env.PR_FOLDER}} | |
| python ${{ github.workspace }}/python/llm/dev/benchmark/whisper/whisper_csv_to_html.py -f ${{ env.PR_FOLDER}} |