Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-18 16:55:14 +02:00
Code Issues Packages Projects Releases Wiki Activity

smartcache allow custom number of slots

Browse Source
This commit is contained in:
Concedo
2026-01-02 17:19:40 +08:00
parent 7e1ae49e7d
commit d8942cde14
3 changed files with 29 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gpttype_adapter.cpp
+9 -2
View File
@@ -154,8 +154,8 @@ static int delayed_generated_tokens_limit = 0;
std::deque<std::string> delayed_generated_tokens; //for use with antislop sampling
static std::map<int,std::vector<int>> antislop_banned_token_ids; //first is the npast position, second is the array of banned ids at that index
const int savestate_limit = 5;
static savestate_data savestates[savestate_limit];
static int savestate_limit = 0;
static std::vector<savestate_data> savestates;
inline int kcpp_cpu_has_blas(void) {
#if defined(GGML_USE_BLAS) || defined(GGML_USE_CUDA) || defined(GGML_USE_VULKAN) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_SYCL)
@@ -2143,6 +2143,13 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
kcpp_data->use_contextshift = inputs.use_contextshift;
kcpp_data->use_fastforward = inputs.use_fastforward;
kcpp_data->smartcache = inputs.smartcache;
//prepare savestate slots
savestate_limit = inputs.smartcacheslots;
savestates.resize(savestate_limit);
if(kcpp_data->smartcache)
{
printf("SmartCache: Prepared %d KV slots\n",savestate_limit);
}
kcpp_pipeline_parallelism = inputs.pipelineparallel;
if(!kcpp_data->use_fastforward && kcpp_data->smartcache)
{