vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 09:34:37 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fixed compile error: GGML_BACKEND_TYPE_GPU (+1 squashed commits)

Browse source 
Squashed commits:

[00ca282a] fixed compile error: LLAMA_SPLIT_MODE_ROW
This commit is contained in:
Concedo 2024-02-26 10:49:02 +08:00
parent 3ccaf8e09a
commit d47e13c892
2 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

4
gpttype_adapter.cpp
View file

@ -970,9 +970,9 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
model_params.main_gpu = cu_parseinfo_maindevice; model_params.main_gpu = cu_parseinfo_maindevice;
#if defined(GGML_USE_CUBLAS) #if defined(GGML_USE_CUBLAS)
model_params.split_mode = (inputs.use_rowsplit?llama_split_mode::LLAMA_SPLIT_ROW:llama_split_mode::LLAMA_SPLIT_LAYER); model_params.split_mode = (inputs.use_rowsplit?llama_split_mode::LLAMA_SPLIT_MODE_ROW:llama_split_mode::LLAMA_SPLIT_MODE_LAYER);
#else #else
model_params.split_mode = llama_split_mode::LLAMA_SPLIT_LAYER; model_params.split_mode = llama_split_mode::LLAMA_SPLIT_MODE_LAYER;
#endif #endif
llama_ctx_params.n_batch = kcpp_params->n_batch; llama_ctx_params.n_batch = kcpp_params->n_batch;

2
llama.cpp
View file

@ -2863,7 +2863,7 @@ struct llama_model_loader {
bool shouldoffload = (layernum>=0 && clblast_offload_fallback_layers>layernum); bool shouldoffload = (layernum>=0 && clblast_offload_fallback_layers>layernum);
if(shouldoffload) if(shouldoffload)
{ {
cur->backend = GGML_BACKEND_GPU; cur->backend = GGML_BACKEND_TYPE_GPU;
ggml_cl_transform_tensor(cur->data, cur); ggml_cl_transform_tensor(cur->data, cur);
} }
} }