fix flaky test for rotate_bounding_box #7362
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes https://app.circleci.com/pipelines/github/pytorch/vision/23712/workflows/71c48632-734a-42bc-b2fe-bbdf64f9dcd3/jobs/1845129/tests.
Apart from that there is also a flaky failure for the same op on
test_cuda_vs_cpu
: https://github.com/pytorch/vision/actions/runs/4292476270/jobs/7478954697#step:10:2949.That one is really weird given that we are calling the same op multiple times with the input data on same dtype and device so there shouldn't be any flakiness. I've improved the error messages to include the full input data the next time this happens, since I cannot reproduce.
The error happens for
XYWH
and internallyrotate_bounding_box
converts toXYXY
and back. The dtype istorch.int64
so there might be rounding issues. But no idea why there would be different behavior when calling the op with a batch or not.