Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 17:25:07 +02:00
Code Issues Packages Projects Releases Wiki Activity

remove clblast, part 2

Browse Source
This commit is contained in:
Concedo
2026-01-23 14:09:46 +08:00
parent 7f485e5287
commit 5c6cc02985
64 changed files with 4 additions and 30034 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gpttype_adapter.cpp
+2 -17
View File
@@ -158,7 +158,7 @@ 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)
#if defined(GGML_USE_BLAS) || defined(GGML_USE_CUDA) || defined(GGML_USE_VULKAN) || defined(GGML_USE_SYCL)
return 1;
#else
return 0;
@@ -2389,21 +2389,6 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
}
}
#if defined(GGML_USE_CLBLAST)
if(file_format==FileFormat::GGUF_GENERIC && model_params.n_gpu_layers>0)
{
if(file_format_meta.model_architecture == GGUFArch::ARCH_FALCON)
{
printf("\nOpenCL does not support GPU Layer offloading for this model architecture! GPU Offload has been disabled.\n");
model_params.n_gpu_layers = 0;
}
else if(file_format_meta.n_expert_count>1)
{
printf("\nOpenCL cannot use regular GPU offloading for this model architecture. A fallback GPU offloader will be used with degraded performance.\n");
}
}
#endif
#if defined(GGML_USE_CUDA)
if(kcpp_parseinfo_maindevice>0)
{
@@ -3278,7 +3263,7 @@ int GetThreadsToUse(bool blasmode)
{
if (blasmode)
{
#if defined(GGML_USE_CLBLAST) || defined(GGML_USE_CUDA) || defined(GGML_USE_VULKAN)
#if defined(GGML_USE_CUDA) || defined(GGML_USE_VULKAN)
return kcpp_data->n_blasthreads;
#else
return std::min(kcpp_data->n_blasthreads, 4);