[CI/Build] Limit wheel size check to CUDA 13 (#50357)
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com> Co-authored-by: OpenAI Codex <codex@openai.com>
Tyler Michael Smith committed
Jul 29, 2026 at 23:10 UTC
1ad5182ba95a6f1de23b537d57b860082912b28e
3 files changed
+4
-6
.buildkite/scripts/hardware_ci/run-gh200-test.sh
-1
@@ -15,7 +15,6 @@ DOCKER_BUILDKIT=1 docker build . \
15
-t gh200-test \
16
--build-arg max_jobs=66 \
17
--build-arg nvcc_threads=2 \
18
- --build-arg RUN_WHEEL_CHECK=false \
18
--build-arg torch_cuda_arch_list="9.0+PTX"
19
20
# Setup cleanup
docker/Dockerfile
+3
-2
@@ -585,13 +585,14 @@ COPY --from=extensions-build /tmp/ep_kernels_workspace/dist /tmp/ep_kernels_work
585
RUN sha256sum /tmp/ep_kernels_workspace/dist/*.whl \
586
> /tmp/ep_kernels_workspace/dist/wheels.sha256
587
588
-# Check the size of the wheel if RUN_WHEEL_CHECK is true
588
+# Check the size of the CUDA 13 wheel uploaded to PyPI
589
COPY .buildkite/check-wheel-size.py check-wheel-size.py
590
# sync the default value with .buildkite/check-wheel-size.py
591
ARG VLLM_MAX_SIZE_MB=500
592
ENV VLLM_MAX_SIZE_MB=$VLLM_MAX_SIZE_MB
593
ARG RUN_WHEEL_CHECK=true
594
-RUN if [ "$RUN_WHEEL_CHECK" = "true" ]; then \
594
+RUN if [ "$RUN_WHEEL_CHECK" = "true" ] && \
595
+ [ "${CUDA_VERSION%%.*}" = "13" ]; then \
596
python3 check-wheel-size.py dist; \
597
else \
598
echo "Skipping wheel size check."; \
docs/getting_started/installation/gpu.cuda.inc.md
+1
-3
@@ -384,8 +384,7 @@ A docker container can be built for aarch64 systems such as the Nvidia Grace-Hop
384
-t vllm/vllm-gh200-openai:latest \
385
--build-arg max_jobs=66 \
386
--build-arg nvcc_threads=2 \
387
- --build-arg torch_cuda_arch_list="9.0 10.0+PTX" \
388
- --build-arg RUN_WHEEL_CHECK=false
387
+ --build-arg torch_cuda_arch_list="9.0 10.0+PTX"
388
```
389
390
For (G)B300, we recommend using CUDA 13, as shown in the following command.
@@ -398,7 +397,6 @@ For (G)B300, we recommend using CUDA 13, as shown in the following command.
397
--build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 \
398
--build-arg max_jobs=256 \
399
--build-arg nvcc_threads=2 \
401
- --build-arg RUN_WHEEL_CHECK=false \
400
--build-arg torch_cuda_arch_list='9.0 10.0+PTX' \
401
--platform "linux/arm64" \
402
--tag vllm/vllm-gb300-openai:latest \