Fix a regex issue in tokenize_dialogue.
The existing regex would fail if using character names that start with numbers, for example: 9S or 2B.
This commit is contained in:
@@ -532,7 +532,7 @@ if args.chat or args.cai_chat:
|
||||
dialogue = re.sub('<START>', '', dialogue)
|
||||
dialogue = re.sub('<start>', '', dialogue)
|
||||
dialogue = re.sub('(\n|^)[Aa]non:', '\\1You:', dialogue)
|
||||
dialogue = re.sub('(\n|^)\[CHARACTER\]:', f'\\1{name2}:', dialogue)
|
||||
dialogue = re.sub('(\n|^)\[CHARACTER\]:', f'\\g<1>{name2}:', dialogue)
|
||||
idx = [m.start() for m in re.finditer(f"(^|\n)({name1}|{name2}):", dialogue)]
|
||||
if len(idx) == 0:
|
||||
return _history
|
||||
|
||||
Reference in New Issue
Block a user