damccorm commented on code in PR #40816:
URL: https://github.com/apache/beam/pull/40816#discussion_r4538999911
##########
sdks/python/apache_beam/ml/inference/base_test.py:
##########
@@ -1201,13 +1223,13 @@ def
test_run_inference_timeout_does_garbage_collection(self):
with TestPipeline() as pipeline:
# Start with bad example which gets timed out.
# Then provide plenty of time for GC to happen.
- examples = [20] + [1] * 15
- expected_good = [1] * 15
+ examples = [20] + [1] * 20
+ expected_good = [1] * 20
expected_bad = [20]
pcoll = pipeline | 'start' >> beam.Create(examples)
main, other = pcoll | base.RunInference(
FakeSlowModelHandler(
- 0, True, tmp_path)).with_exception_handling(timeout=5)
+ 0, True, tmp_path)).with_exception_handling(timeout=9)
Review Comment:
For context: bumped this because model loading does take more time now due
to spawning processes.
This should be acceptable since it is relatively small and non-recurring
(around 4-5 seconds of fixed overhead per model process the first time it is
loaded.)
--
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]
