A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 14:45 UTC
[ROCm][CI/Build] Fix the pytest hook to properly print out the summary (#38585)
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Gregory Shtrasberg committed
Apr 3, 2026 at 04:24 UTC
a7d79fa1331912d533cc0adbf31552ea761104a5
1 file changed
+14
-1
docker/Dockerfile.rocm
+14
-1
@@ -390,7 +390,20 @@ ENV MIOPEN_DEBUG_CONV_GEMM=0
390
RUN mkdir src && mv vllm src/vllm
391
392
# This is a workaround to ensure pytest exits with the correct status code in CI tests.
393
-RUN echo "import os\n\ndef pytest_sessionfinish(session, exitstatus):\n os._exit(int(exitstatus))" > /vllm-workspace/conftest.py
393
+RUN cat << 'EOF' > /vllm-workspace/conftest.py
394
+import os
395
+
396
+_exit_code = 1
397
+
398
+def pytest_sessionfinish(session, exitstatus):
399
+ global _exit_code
400
+ _exit_code = int(exitstatus)
401
+
402
+def pytest_unconfigure(config):
403
+ sys.stdout.flush()
404
+ sys.stderr.flush()
405
+ os._exit(_exit_code)
406
+EOF
407
408
# -----------------------
409
# Final vLLM image