vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 23:44:33 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add metal mem limit

Browse source
This commit is contained in:
Lizonghang 2025-01-23 16:08:52 +04:00
parent 33429ec4e1
commit 78a544d716
5 changed files with 102 additions and 67 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.h
View file

@ -149,7 +149,7 @@ struct gpt_params {
std::string next_node_ip = "localhost"; // ip address of my next node
bool unload = false; // unload layer weights after use or not
bool keep_inp_out_in_metal = false; // whether to keep input/output weight in metal, not by default
int32_t cuda_mem = 999.0; // cuda memory to use, in GiB
int32_t gpu_mem = 999.0; // gpu memory to use, in GiB
int32_t n_predict = -1; // new tokens to predict
int32_t n_ctx = 0; // context size
int32_t n_batch = 2048; // logical batch size for prompt processing (must be >=32 to use BLAS)