Skip to content

[WIP][tritonbench] Fix workflow #113

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xuzhao9 wants to merge 7 commits into
base: main
Choose a base branch
from
Open

[WIP][tritonbench] Fix workflow #113

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
eebfe02
fix conda env
xuzhao9 Dec 4, 2025
2a994cd
fix test
xuzhao9 Dec 4, 2025
2ef16fd
move result as root
xuzhao9 Dec 6, 2025
2f7dded
move directory with sudo
xuzhao9 Dec 6, 2025
d3df6b9
install requirements to upload
xuzhao9 Dec 6, 2025
8c7f8b8
fix benchmark
xuzhao9 Dec 6, 2025
0399a49
add github env
xuzhao9 Dec 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 28 additions & 13 deletions .github/workflows/tritonbench.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,6 +63,8 @@ jobs:
matrix: ${{ fromJson(needs.set-parameters.outputs.benchmark_matrix) }}
fail-fast: false
runs-on: ${{ matrix.runner }}
env:
TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN: ${{ secrets.TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN }}
environment: pytorch-x-vllm
permissions:
id-token: write
Expand Down Expand Up @@ -125,6 +127,14 @@ jobs:
fi
echo "DEVICE_TYPE=$DEVICE_TYPE" >> $GITHUB_ENV

- name: Install dependencies
shell: bash
working-directory: triton-benchmarks/tritonbench
run: |
set -eux

pip install -r .ci/upload/requirements.txt

- name: Setup CUDA GPU_FLAG for docker run
if: env.DEVICE_NAME == 'cuda'
run: |
Expand All @@ -144,13 +154,15 @@ jobs:
exit 1
fi

DOCKER_IMAGE="meta-pytorch/tritonbench:${IMAGE_SUFFIX}"
CONDA_ENV="triton-main"
DOCKER_IMAGE="ghcr.io/meta-pytorch/tritonbench:${IMAGE_SUFFIX}"
echo "DOCKER_IMAGE=$DOCKER_IMAGE" >> "$GITHUB_ENV"
echo "CONDA_ENV=triton-main" >> "$GITHUB_ENV"
echo "CONDA_ENV=$CONDA_ENV" >> "$GITHUB_ENV"
echo "Using docker image: $DOCKER_IMAGE "
echo "Using conda env: $CONDA_ENV "

- name: Run TritonBench benchmark
working-directory: triton-benchmarks/tritonbench
run: |
set -eux

Expand All @@ -170,12 +182,14 @@ jobs:
)

docker exec -t -w /tmp/workspace "${container_name}" bash -c " \
set -eux && cd /workspace/tritonbench &&
bash .ci/tritonbench/run-benchmark.sh ${{ matrix.BENCHMARKS }} --conda-env ${{ env.CONDA_ENV }} "

docker exec -t -w /tmp/workspace "${container_name}" bash -c " \
set -eux && cd /workspace/tritonbench && mv .benchmarks /tmp/workspace/triton-benchmarks/tritonbench/results
"
set -eux && cd /workspace/tritonbench && \
bash .ci/tritonbench/run-benchmark.sh ${{ matrix.BENCHMARKS }} --conda-env ${{ env.CONDA_ENV }} && \
sudo mv /workspace/tritonbench/.benchmarks /tmp/workspace/triton-benchmarks/tritonbench/results-${{ env.CONDA_ENV }} "

# post-process result.json
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
python3 ./.ci/test_infra/oss_ci_benchmark_v3.py --json ${latest_result_json} \
--add-github-env --output ${latest_result_json}

- name: Authenticate with AWS
# AWS CUDA runners already have access to the bucket via its runner IAM role
Expand All @@ -191,22 +205,23 @@ jobs:
- uses: actions/upload-artifact@v4
if: always()
with:
name: tritonbench-results
path: triton-benchmarks/tritonbench/results
name: tritonbench-${{ matrix.runner }}-benchmark-single
path: triton-benchmarks/tritonbench/results-${{ env.CONDA_ENV }}
retention-days: 30

- name: Upload result to Scribe
if: ${{ env.TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN != '' }}
working-directory: triton-benchmarks/tritonbench
run: |
latest_result_json=$(find ./results/${TRITONBENCH_SIDE_A_ENV} -name "result.json" | sort -r | head -n 1)
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
python3 ./.ci/upload/scribe.py --json ${latest_result_json}

- name: Rewrite Tritonbench result json to ClickHouse style
working-directory: triton-benchmarks/tritonbench
run: |
latest_result_json=$(find ./results/${TRITONBENCH_SIDE_A_ENV} -name "result.json" | sort -r | head -n 1)
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
python3 ./.ci/test_infra/oss_ci_benchmark_v3.py --json ${latest_result_json} \
--output clickhouse-results/result-${TRITONBENCH_SIDE_A_ENV}.json
--output clickhouse-results/result-${CONDA_ENV}.json

- name: Upload result to ClickHouse
uses: pytorch/test-infra/.github/actions/upload-benchmark-results@main
Expand Down