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Outdated suggestions cannot be applied.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.
Following discussion from #79, #75, and #78 I've revised the attribution map to track both input and output token indices. As this made the attributions get a bit complicated, I added two dataclasses
Attribution
andAttributionMap
to help make it clear what the meaning is of the various elements of the attribution. One (maybe) disadvantage is that it can be hard to know what theindex
means because it comes from the tokenizer within the library. One possible change would be to make index be strictly ascending or switch to string index.Here are examples:
Encoder
Decoder