MLPerf inference SDXL #13
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: MLPerf inference SDXL | |
| #off now as we have SCC24 test doing the same | |
| on: | |
| schedule: | |
| - cron: "1 2 * * *" | |
| jobs: | |
| build_reference: | |
| if: github.repository_owner == 'gateoverflow_off' | |
| runs-on: [ self-hosted, linux, x64 ] | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| python-version: [ "3.12" ] | |
| backend: [ "pytorch" ] | |
| precision: [ "float16" ] | |
| steps: | |
| - name: Test MLPerf Inference SDXL Reference | |
| run: | | |
| source gh_action/bin/deactivate || python3 -m venv gh_action | |
| source gh_action/bin/activate | |
| export CM_REPOS=$HOME/GH_CM | |
| python3 -m pip install cm4mlops | |
| cm pull repo | |
| cm run script --tags=run-mlperf,inference,_submission,_short --submitter="MLCommons" --docker --model=sdxl --backend=${{ matrix.backend }} --device=cuda --scenario=Offline --test_query_count=1 --precision=${{ matrix.precision }} --target_qps=1 --quiet --docker_it=no --docker_cm_repo=gateoverflow@cm4mlops --adr.compiler.tags=gcc --hw_name=gh_action --docker_dt=yes --results_dir=$HOME/gh_action_results --submission_dir=$HOME/gh_action_submissions --clean | |
| build_nvidia: | |
| if: github.repository_owner == 'gateoverflow_off' | |
| runs-on: [ self-hosted, linux, x64 ] | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| python-version: [ "3.12" ] | |
| backend: [ "tensorrt" ] | |
| precision: [ "float16" ] | |
| implementation: [ "nvidia" ] | |
| steps: | |
| - name: Test MLPerf Inference SDXL Nvidia | |
| run: | | |
| source gh_action/bin/deactivate || python3 -m venv gh_action | |
| source gh_action/bin/activate | |
| export CM_REPOS=$HOME/GH_CM | |
| cm pull repo | |
| cm run script --tags=run-mlperf,inference,_submission,_short --submitter="MLCommons" --docker --model=sdxl --implementation=${{ matrix.implementation }} --backend=${{ matrix.backend }} --device=cuda --scenario=Offline --test_query_count=1 --precision=${{ matrix.precision }} --target_qps=1 --quiet --docker_it=no --docker_cm_repo=gateoverflow@cm4mlops --adr.compiler.tags=gcc --hw_name=gh_action --docker_dt=yes --results_dir=$HOME/gh_action_results --submission_dir=$HOME/gh_action_submissions --clean |