vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 09:34:37 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

alias banned_tokens with banned_strings from ST

Browse source 
increase max bans to 32 for now
This commit is contained in:
Concedo 2024-10-10 21:52:46 +08:00
parent 5c9650d68e
commit d75cbd671d
2 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
expose.h
View file

@ -2,7 +2,7 @@
#include <cstdint> #include <cstdint>
const int stop_token_max = 24; const int stop_token_max = 24;
const int ban_token_max = 24; const int ban_token_max = 32;
const int tensor_split_max = 16; const int tensor_split_max = 16;
const int logit_bias_max = 24; const int logit_bias_max = 24;
const int dry_seq_break_max = 24; const int dry_seq_break_max = 24;

5
koboldcpp.py
View file

@ -20,7 +20,7 @@ from datetime import datetime, timezone
# constants # constants
sampler_order_max = 7 sampler_order_max = 7
stop_token_max = 24 stop_token_max = 24
ban_token_max = 24 ban_token_max = 32
tensor_split_max = 16 tensor_split_max = 16
logit_bias_max = 24 logit_bias_max = 24
dry_seq_break_max = 24 dry_seq_break_max = 24
@ -909,7 +909,8 @@ def generate(genparams, is_quiet=False, stream_flag=False):
smoothing_factor = genparams.get('smoothing_factor', 0.0) smoothing_factor = genparams.get('smoothing_factor', 0.0)
logit_biases = genparams.get('logit_bias', {}) logit_biases = genparams.get('logit_bias', {})
render_special = genparams.get('render_special', False) render_special = genparams.get('render_special', False)
banned_tokens = genparams.get('banned_tokens', []) banned_strings = genparams.get('banned_strings', []) # SillyTavern uses that name
banned_tokens = genparams.get('banned_tokens', banned_strings)
bypass_eos_token = genparams.get('bypass_eos', False) bypass_eos_token = genparams.get('bypass_eos', False)
inputs = generation_inputs() inputs = generation_inputs()