SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Bug fix: token classification pipeline while passing offset_mapping (#22034)

Browse Source 
fix slow tokenizers with passing offset_mapping
This commit is contained in:
Ceyda Cinarel
2023-03-09 06:21:46 +09:00
committed by GitHub
parent 1cbac6867b
commit 3ec8171bed
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/transformers/pipelines/token_classification.py
View File

@@ -304,7 +304,9 @@ class TokenClassificationPipeline(Pipeline):
start_ind = start_ind.item() start_ind = start_ind.item()
end_ind = end_ind.item() end_ind = end_ind.item()
word_ref = sentence[start_ind:end_ind] word_ref = sentence[start_ind:end_ind]
if getattr(self.tokenizer._tokenizer.model, "continuing_subword_prefix", None): if getattr(self.tokenizer, "_tokenizer", None) and getattr(
self.tokenizer._tokenizer.model, "continuing_subword_prefix", None
):
# This is a BPE, word aware tokenizer, there is a correct way # This is a BPE, word aware tokenizer, there is a correct way
# to fuse tokens # to fuse tokens
is_subword = len(word) != len(word_ref) is_subword = len(word) != len(word_ref)