Ë
    ÛÓ^j;  ã                   ó$   — d Z ddlmZ eZeZddgZy)al  
Compatibility shims for BART tokenizers in v5.

In v5 we consolidate on the tokenizers-library backend and remove separate
"slow" vs "fast" implementations. BART uses the same byte-level BPE
tokenizer as RoBERTa, so we expose `BartTokenizer` and `BartTokenizerFast`
as aliases to `RobertaTokenizer` to preserve the public API expected by
existing code and tests.
é   )ÚRobertaTokenizerÚBartTokenizerÚBartTokenizerFastN)Ú__doc__Úroberta.tokenization_robertar   Ú_RobertaTokenizerr   r   Ú__all__© ó    úu/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/bart/tokenization_bart.pyú<module>r      s(   ðñõ Qð "€Ø%Ð àÐ/Ð
0�r   