Bug Fix: Always index into squence-like outputs #201
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Outdated suggestions cannot be applied.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.
Some operators are intended to produce 1 or more outputs and will always return a
Sequence[torch.Tensor]
. Therefore, whenlen(output_values) == 1
, the code breaks since it treated it as a single output tensor. I fix this issue by modifying the condition to index into output.Include op types are:
['If', 'Loop', 'Scan', 'SequenceMap', 'Split']