vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 22:24:55 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

support different window sizes

Browse source
This commit is contained in:
Lizonghang 2024-10-26 12:34:14 +04:00
parent 5685cb87ed
commit 76a7fc7527
6 changed files with 200 additions and 127 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.h
View file

@ -144,7 +144,7 @@ struct gpt_sampler_params {
struct gpt_params {
int32_t n_world = 1; // number of devices to use
int32_t rank = 0; // my rank for distributed inference
int32_t n_layer_window = 32; // number of layers to process in each compute
uint32_t n_layer_window[32] = {32}; // layer window size on each node
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