Fixed issue with gpt2 regex custom preprocessor
This commit is contained in:
parent
753580360b
commit
36d983262e
|
|
@ -357,6 +357,7 @@ static std::vector<size_t> unicode_gpt2_regex_preprocess(const std::wstring & wt
|
||||||
token += utf_char;
|
token += utf_char;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
start += offset;
|
||||||
}
|
}
|
||||||
|
|
||||||
return bpe_offsets;
|
return bpe_offsets;
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue