Support "reflect" padding mode #326
Merged
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
For efficiency and simplicity, reflection padding is only supported for the last 2 dimensions of the input. Any additional dimensions are treated as batch dimensions which must have the same size in the input and output. PyTorch has a constraint that the amount of padding must be less than the size of a dimension, which is useful since it avoids the need for padding to wrap around by taking an expensive modulus of the dimension size. NumPy however allows wrap-around, and the example in the ONNX spec relies on this. In this implementation, wrap-around is supported but it might be necessary to split the cases in future for efficiency, to avoid a modulus in the common case where the padding size is less than the dimension size.
robertknight
force-pushed
the
pad-reflect
branch
from
246c4b0
to
72f78fb
Compare
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Outdated suggestions cannot be applied.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.
For efficiency and simplicity, reflection padding is only supported for the last
2 dimensions of the input. Any additional dimensions are treated as batch
dimensions which must have the same size in the input and output.
PyTorch has a constraint that the amount of padding must be less than the size
of a dimension, which is useful since it avoids the need for padding to wrap
around by taking an expensive modulus of the dimension size. NumPy however
allows wrap-around, and the example in the ONNX spec relies on this. In this
implementation, wrap-around is supported but it might be necessary to split the
cases in future for efficiency, to avoid a modulus in the common case where the
padding size is less than the dimension size.