Fix custom prefixes in BertWordPieceTokenizer - #2413
Open
HaokaiDing wants to merge 1 commit into
Open
HaokaiDing wants to merge 1 commit into
HaokaiDing wants to merge 1 commit into
Conversation
Author
|
The conflicts come from #2391 removing I prepared the same fix against Since the 0.23.2 release notes call it the last v0 release, would you accept this as a backport, and which base should this PR target? I'll keep the current base unchanged until the intended target is clear. |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #2412.
BertWordPieceTokenizerpasseswordpieces_prefixto its decoder but leaves the WordPiece model using##. Loading a vocabulary trained with a custom prefix such as@@consequently turns known words into[UNK].Forward the prefix to both WordPiece constructor branches. Add tests for default/custom prefixes that check tokens, IDs, offsets and decoding, plus a training/save/reload round trip.
Validation:
tests/implementations/test_bert_wordpiece.py.tokenizers 0.23.2native extension; the local runner verifies the wrapper's source path and native module identity. Current Rust was not rebuilt and network-dependent tests were not run.