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

allow splitmode to be set independently, enable tensor parallelism

Browse Source
This commit is contained in:
Concedo
2026-05-02 16:41:28 +08:00
parent 7e98e06075
commit 8b62e7b667
3 changed files with 45 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gpttype_adapter.cpp
+1 -6
View File
@@ -2453,12 +2453,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
#endif
model_params.main_gpu = kcpp_parseinfo_maindevice;
#if defined(GGML_USE_CUDA)
model_params.split_mode = (inputs.use_rowsplit?llama_split_mode::LLAMA_SPLIT_MODE_ROW:llama_split_mode::LLAMA_SPLIT_MODE_LAYER);
#else
model_params.split_mode = llama_split_mode::LLAMA_SPLIT_MODE_LAYER;
#endif
model_params.split_mode = (inputs.splitmode>0?((llama_split_mode)(inputs.splitmode)):llama_split_mode::LLAMA_SPLIT_MODE_LAYER);
llama_ctx_params.n_batch = kcpp_data->n_batch;
llama_ctx_params.n_ubatch = kcpp_data->n_ubatch;