vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

increase genamt hardlimit from 0.1 to 0.2 ratio

Browse source
This commit is contained in:
Concedo 2025-06-18 19:34:12 +08:00
parent 40443a98f5
commit a8d33ebb0d
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
koboldcpp.py
View file

@ -1384,7 +1384,7 @@ def generate(genparams, stream_flag=False):
print(f"\n!!! ====== !!!\n(Warning! Request max_context_length={max_context_length} exceeds allocated context size of {maxctx}. It will be reduced to fit. Consider launching with increased --contextsize to avoid issues. This message will only show once per session.)\n!!! ====== !!!")
showmaxctxwarning = False
max_context_length = maxctx
min_remain_hardlimit = max(min(max_context_length-4, 16),int(max_context_length*0.1))
min_remain_hardlimit = max(min(max_context_length-4, 16),int(max_context_length*0.2))
min_remain_softlimit = max(min(max_context_length-4, 16),int(max_context_length*0.4))
if max_length >= (max_context_length-min_remain_softlimit):
print(f"\n!!! ====== !!!\nWarning: You are trying to generate text with max_length ({max_length}) near or exceeding max_context_length limit ({max_context_length}).\nMost of the context will be removed, and your outputs will not be very coherent.\nConsider launching with increased --contextsize to avoid issues.\n!!! ====== !!!")