vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 09:04:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

added skip bos for tokenize endpoint

Browse source
This commit is contained in:
Concedo 2024-06-05 10:49:11 +08:00
parent 5789417802
commit 10b148f4c2
5 changed files with 11 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

3
koboldcpp.py
View file

@ -1352,7 +1352,8 @@ Enter Prompt:<br>
try:
genparams = json.loads(body)
countprompt = genparams.get('prompt', "")
rawcountdata = handle.token_count(countprompt.encode("UTF-8"))
tcaddspecial = genparams.get('special', True)
rawcountdata = handle.token_count(countprompt.encode("UTF-8"),tcaddspecial)
countlimit = rawcountdata.count if (rawcountdata.count>=0 and rawcountdata.count<50000) else 0
# the above protects the server in case the count limit got corrupted
countdata = [rawcountdata.ids[i] for i in range(countlimit)]