Add unigram sampling (alpha, nbest_size)#1994
Merged
ArthurZucker merged 8 commits intohuggingface:mainfrom Apr 8, 2026
Merged
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Collaborator
ArthurZucker
left a comment
ArthurZucker
left a comment
There was a problem hiding this comment.
Thanks! Do you mind adding a test on python side bindings/python/tests/bindings/test_models.py!
Also update the doc of PyUnigram to mention these new args and the new behavior!
I am not super super familiar with unigram / nbest etc but happy to have parity!
Contributor
Author
@ArthurZucker Is updating the docstring sufficient? I wasn't sure if the documentation is autogenerated. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
I noticed that
models.Unigramdoesn't support sampling, which enables subword regularization (arguably one of the main reasons to choose the Unigram model). I checked GitHub and there are multiple closed issues on this topic (#730, #849). In one of these issues, it was mentioned that the sampling code has already been implemented inlattice.rsand simply needs to be exposed through Python. I filled in the missing details and added asample_nbestfunction for parity with Google's implementation. I also copied the interface for BPE dropout as closely as possible, including getters and setters for the new sampling parameters.