Skip to content

Commit 0ab78c7

Browse files
authored
[tritonbench] Fix the B200 benchmark workflow (#113)
1 parent d370d70 commit 0ab78c7

File tree

1 file changed

+31
-13
lines changed

1 file changed

+31
-13
lines changed

.github/workflows/tritonbench.yml

Lines changed: 31 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -63,6 +63,10 @@ jobs:
6363
matrix: ${{ fromJson(needs.set-parameters.outputs.benchmark_matrix) }}
6464
fail-fast: false
6565
runs-on: ${{ matrix.runner }}
66+
env:
67+
TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN: ${{ secrets.TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN }}
68+
JOB_NAME: tritonbench-${{ matrix.runner }}-benchmark-periodic
69+
RUNNER_TYPE: ${{ matrix.runner }}
6670
environment: pytorch-x-vllm
6771
permissions:
6872
id-token: write
@@ -125,6 +129,14 @@ jobs:
125129
fi
126130
echo "DEVICE_TYPE=$DEVICE_TYPE" >> $GITHUB_ENV
127131
132+
- name: Install dependencies
133+
shell: bash
134+
working-directory: triton-benchmarks/tritonbench
135+
run: |
136+
set -eux
137+
138+
pip install -r .ci/upload/requirements.txt
139+
128140
- name: Setup CUDA GPU_FLAG for docker run
129141
if: env.DEVICE_NAME == 'cuda'
130142
run: |
@@ -144,13 +156,15 @@ jobs:
144156
exit 1
145157
fi
146158
147-
DOCKER_IMAGE="meta-pytorch/tritonbench:${IMAGE_SUFFIX}"
159+
CONDA_ENV="triton-main"
160+
DOCKER_IMAGE="ghcr.io/meta-pytorch/tritonbench:${IMAGE_SUFFIX}"
148161
echo "DOCKER_IMAGE=$DOCKER_IMAGE" >> "$GITHUB_ENV"
149-
echo "CONDA_ENV=triton-main" >> "$GITHUB_ENV"
162+
echo "CONDA_ENV=$CONDA_ENV" >> "$GITHUB_ENV"
150163
echo "Using docker image: $DOCKER_IMAGE "
151164
echo "Using conda env: $CONDA_ENV "
152165
153166
- name: Run TritonBench benchmark
167+
working-directory: triton-benchmarks/tritonbench
154168
run: |
155169
set -eux
156170
@@ -170,12 +184,15 @@ jobs:
170184
)
171185
172186
docker exec -t -w /tmp/workspace "${container_name}" bash -c " \
173-
set -eux && cd /workspace/tritonbench &&
174-
bash .ci/tritonbench/run-benchmark.sh ${{ matrix.BENCHMARKS }} --conda-env ${{ env.CONDA_ENV }} "
175-
176-
docker exec -t -w /tmp/workspace "${container_name}" bash -c " \
177-
set -eux && cd /workspace/tritonbench && mv .benchmarks /tmp/workspace/triton-benchmarks/tritonbench/results
178-
"
187+
set -eux && cd /workspace/tritonbench && \
188+
bash .ci/tritonbench/run-benchmark.sh ${{ matrix.BENCHMARKS }} --conda-env ${{ env.CONDA_ENV }} && \
189+
sudo mv /workspace/tritonbench/.benchmarks /tmp/workspace/triton-benchmarks/tritonbench/results-${{ env.CONDA_ENV }} && \
190+
sudo chmod -R 777 /tmp/workspace/triton-benchmarks/tritonbench/results-${{ env.CONDA_ENV }} "
191+
192+
# post-process result.json
193+
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
194+
python3 ./.ci/test_infra/oss_ci_benchmark_v3.py --json ${latest_result_json} \
195+
--add-github-env --output ${latest_result_json}
179196
180197
- name: Authenticate with AWS
181198
# AWS CUDA runners already have access to the bucket via its runner IAM role
@@ -191,22 +208,23 @@ jobs:
191208
- uses: actions/upload-artifact@v4
192209
if: always()
193210
with:
194-
name: tritonbench-results
195-
path: triton-benchmarks/tritonbench/results
211+
name: ${{ env.JOB_NAME }}
212+
path: triton-benchmarks/tritonbench/results-${{ env.CONDA_ENV }}
196213
retention-days: 30
197214

198215
- name: Upload result to Scribe
216+
if: ${{ env.TRITONBENCH_SCRIBE_GRAPHQL_ACCESS_TOKEN != '' }}
199217
working-directory: triton-benchmarks/tritonbench
200218
run: |
201-
latest_result_json=$(find ./results/${TRITONBENCH_SIDE_A_ENV} -name "result.json" | sort -r | head -n 1)
219+
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
202220
python3 ./.ci/upload/scribe.py --json ${latest_result_json}
203221
204222
- name: Rewrite Tritonbench result json to ClickHouse style
205223
working-directory: triton-benchmarks/tritonbench
206224
run: |
207-
latest_result_json=$(find ./results/${TRITONBENCH_SIDE_A_ENV} -name "result.json" | sort -r | head -n 1)
225+
latest_result_json=$(find ./results-${CONDA_ENV} -name "result.json" | sort -r | head -n 1)
208226
python3 ./.ci/test_infra/oss_ci_benchmark_v3.py --json ${latest_result_json} \
209-
--output clickhouse-results/result-${TRITONBENCH_SIDE_A_ENV}.json
227+
--output clickhouse-results/result-${CONDA_ENV}.json
210228
211229
- name: Upload result to ClickHouse
212230
uses: pytorch/test-infra/.github/actions/upload-benchmark-results@main

0 commit comments

Comments
 (0)