vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

PR #4766

Browse source
This commit is contained in:
Jared Van Bortel 2024-01-10 11:29:04 -05:00
parent 3773e1afe7
commit 1eb8804c18
18 changed files with 2183 additions and 2081 deletions
Download patch file Download diff file Expand all files Collapse all files

1
common/common.h
View file

@ -59,6 +59,7 @@ struct gpt_params {
float p_split = 0.1f; // speculative decoding split probability
int32_t n_gpu_layers = -1; // number of layers to store in VRAM (-1 - use default)
int32_t n_gpu_layers_draft = -1; // number of layers to store in VRAM for the draft model (-1 - use default)
llama_split_mode split_mode = LLAMA_SPLIT_LAYER; // how to split the model across GPUs
int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors
float tensor_split[LLAMA_MAX_DEVICES] = {0}; // how split tensors should be distributed across GPUs
int32_t n_beams = 0; // if non-zero then use beam search of given width.