vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-11 08:34:33 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

keep the output layer weights in shared memory by default

Browse source
This commit is contained in:
Lizonghang 2025-01-25 23:31:43 +04:00
parent f3dd5776eb
commit 1ca9a43bd1
4 changed files with 16 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.h
View file

@ -148,7 +148,7 @@ struct gpt_params {
std::string master_ip = "localhost"; // ip address of the master node
std::string next_node_ip = "localhost"; // ip address of my next node
bool unload = false; // unload layer weights after use or not
bool keep_out_in_metal = false; // whether to keep output weights in metal memory, not by default
bool keep_out_in_metal = true; // whether to keep output weights in metal memory, true by default
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