vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 23:24:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add mmap prefetch and unloading

Browse source
This commit is contained in:
Lizonghang 2024-10-25 16:33:56 +04:00
parent ba5117581e
commit c97ea10617
7 changed files with 161 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

1
common/common.h
View file

@ -147,6 +147,7 @@ struct gpt_params {
int32_t n_layer_window = 32; // number of layers to process in each compute
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
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)