vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-18 23:49:46 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

ace qwen rep pen for codes

Browse source
This commit is contained in:
Concedo 2026-03-02 21:18:06 +08:00
parent de9840afac
commit ae67caa2f7
2 changed files with 8 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

2
gpttype_adapter.cpp
View file

@ -551,7 +551,7 @@ void ContextRewind(std::vector<int> &embd, std::vector<int> &current_context_tok
}
if(file_format == FileFormat::RWKV_1 || file_format==FileFormat::RWKV_2 || is_recurrent)
{
if(!showed_rnn_warning)
if(!showed_rnn_warning && debugmode==1 && !is_quiet)
{
showed_rnn_warning = true;
printf("\nWARNING: RNN models do not support context rewind!\n");

9
otherarch/acestep/ace-qwen3.cpp
View file

@ -1086,7 +1086,7 @@ static std::vector<std::string> run_phase2_batch(
for (int v = 0; v < AUDIO_CODE_BASE; v++)
if (v != TOKEN_IM_END) lg[v] = -1e9f;
int tok = kcpp_quick_sample(lg.data(),V,std::vector<int32_t>(),1.02f,top_p,30,temperature,acestep_lm_rng);
int tok = kcpp_quick_sample(lg.data(),V,std::vector<int32_t>(),1.03f,top_p,30,temperature,acestep_lm_rng);
seqs[i].last_token = tok;
if (tok == TOKEN_IM_END) {
@ -1124,6 +1124,7 @@ static std::vector<std::string> run_phase2_batch(
for (int i = 0; i < N; i++)
if (seqs[i].done) n_active--;
std::vector<int32_t> quicklastntoks;
for (int step = 0; step < max_tokens && n_active > 0; step++) {
// Collect tokens (done sequences feed their last token, result ignored)
for (int i = 0; i < N; i++)
@ -1157,7 +1158,11 @@ static std::vector<std::string> run_phase2_batch(
for (int v = 0; v < AUDIO_CODE_BASE; v++)
if (v != TOKEN_IM_END) lc[v] = -1e9f;
int tok = kcpp_quick_sample(lc,V,std::vector<int32_t>(),1.02f,top_p,30,temperature,acestep_lm_rng);
int tok = kcpp_quick_sample(lc,V,quicklastntoks,1.03f,top_p,30,temperature,acestep_lm_rng);
quicklastntoks.push_back(tok);
if (quicklastntoks.size()>32) {
quicklastntoks.erase(quicklastntoks.begin());
}
seqs[i].last_token = tok;
if (tok == TOKEN_IM_END) {