bvolpato opened a new pull request, #39949: URL: https://github.com/apache/beam/pull/39949
Addresses #38782. ## Why The Faster R-CNN benchmark added in #37186 has not completed successfully in the daily workflow. Recent runs consistently reach the object-detection batch CPU step, start the Dataflow job, and then cancel it after about 1,803 seconds. This is not the GitHub Actions timeout. Each benchmark step allows 180 minutes, but `pytorch_image_object_detection.run()` has its own hard-coded 30-minute `wait_until_finish()` call. Its `finally` block cancels the still-running Dataflow job before the outer workflow timeout can help. ## What changed * Add a `--timeout_ms` option to the object-detection example, preserving the existing 30-minute default for other callers. * Set this benchmark's timeout to 60 minutes. This doubles the observed failing limit while retaining a two-hour guard before the workflow step timeout. Keeping the override in the benchmark options makes future runtime tuning possible without changing example behavior globally. ## Validation * `python -m py_compile sdks/python/apache_beam/examples/inference/pytorch_image_object_detection.py` * `yapf==0.43.0 --diff` on the changed Python file * `ruff==0.15.22 check --ignore I001,UP006` on the changed Python file. Remaining ignored findings predate this change. * `git diff --check` End-to-end validation requires the Dataflow benchmark infrastructure, so this remains draft until CI and a workflow run can validate the increased budget. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
