Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-02-05 16:10:42 +01:00
Code Issues Packages Projects Releases Wiki Activity

added top-k sampler to improve performance

Browse Source
This commit is contained in:
VJHack 2025-01-17 01:17:40 -06:00
parent 6664d4709f
commit c6123e69b0
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
common/sampling.cpp
View File

@ -168,8 +168,9 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
if (params.mirostat == 0) { if (params.mirostat == 0) {
if(params.top_n_sigma >= 0) { if(params.top_n_sigma >= 0) {
llama_sampler_chain_add(result->chain, llama_sampler_init_temp(params.temp)); llama_sampler_chain_add(result->chain, llama_sampler_init_top_k (params.top_k));
llama_sampler_chain_add(result->chain, llama_sampler_init_top_n_sigma(params.top_n_sigma)); llama_sampler_chain_add(result->chain, llama_sampler_init_temp (params.temp));
llama_sampler_chain_add(result->chain, llama_sampler_init_top_n_sigma (params.top_n_sigma));
} else { } else {
for (const auto & cnstr : params.samplers) { for (const auto & cnstr : params.samplers) {
switch (cnstr) { switch (cnstr) {