Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 01:05:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

init total weight for adaptive p

Browse Source
This commit is contained in:
Concedo
2025-12-28 15:33:06 +08:00
parent ec95655f3c
commit 21d801f6d5
3 changed files with 70 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files
embd_res/klite.embd
+61 -34
View File
File diff suppressed because one or more lines are too long
gpttype_adapter.cpp
+8 -3
View File
@@ -3441,9 +3441,6 @@ generation_outputs gpttype_generate(const generation_inputs inputs)
}
}
adaptive_p_weighted_sum = 0;
adaptive_p_total_weight = 0;
//handle custom token bans and antislop phrase banning
banned_phrases.clear();
delayed_generated_tokens_limit = 0;
@@ -3655,6 +3652,14 @@ generation_outputs gpttype_generate(const generation_inputs inputs)
kcpp_data->adaptive_target = inputs.adaptive_target;
kcpp_data->adaptive_decay = inputs.adaptive_decay;
adaptive_p_weighted_sum = 0;
adaptive_p_total_weight = 0;
if(kcpp_data->adaptive_target > 0.0f && kcpp_data->adaptive_decay<1.0f)
{
adaptive_p_weighted_sum = kcpp_data->adaptive_target / (1.0f - kcpp_data->adaptive_decay);
adaptive_p_total_weight = 1.0f / (1.0f - kcpp_data->adaptive_decay);
}
// Parse dry sequence breakers / restart sequences
kcpp_data->dry_sequence_breakers.clear();
dry_sequence_breakers.clear();
koboldcpp.py
+1
View File
@@ -1607,6 +1607,7 @@ def generate(genparams, stream_flag=False):
smoothing_curve = tryparsefloat(genparams.get('smoothing_curve', 1.0),1.0)
adaptive_target = tryparsefloat(genparams.get('adaptive_target', -1.0),-1.0)
adaptive_decay = tryparsefloat(genparams.get('adaptive_decay', 0.9),0.9)
adaptive_decay = 0.01 if adaptive_decay < 0.01 else (0.99 if adaptive_decay > 0.99 else adaptive_decay)
if adaptive_target>0 and min_p<=0 and top_p>=1.0: #adaptive p sampler requires a truncation sampler first, force a tiny min-p
min_p = 0.002
logit_biases = genparams.get('logit_bias', {})