RoIHeads: don't remove empty boxes in postprocess #5
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
pytorch#1019 introduced a change that removes empty boxes in postprocessing on the basis that, previously, empty boxes would actually reach this postprocessing step, and that the model could therefore output empty boxes (even though NMS would've most likely filtered them out). It's essentially a safety check that'd be seldom needed. However, that filtering causes dynamicity on the TPU (because the number of empty boxes, if any, would be unknown). And in any case, we recently introduced a change in PR pytorch#4 that purposefully pads the boxes tensor with empty boxes, to avoid dynamicity. Therefore there's no point trying to remove boxes that we use as padding, we'll just filter those out of the output on the CPU.
jysohn23
approved these changes
Dec 11, 2019
taylanbil
approved these changes
Dec 14, 2019
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
pytorch#1019 introduced a change that removes empty boxes in
postprocessing on the basis that, previously, empty boxes would actually
reach this postprocessing step, and that the model could therefore
output empty boxes (even though NMS would've most likely filtered them
out). It's essentially a safety check that'd be seldom needed.
However, that filtering causes dynamicity on the TPU (because the number
of empty boxes, if any, would be unknown). And in any case, we recently
introduced a change in PR #4 that purposefully pads the boxes tensor
with empty boxes, to avoid dynamicity. Therefore there's no point trying
to remove boxes that we use as padding, we'll just filter those out of
the output on the CPU.