vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-10 23:54:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add args -k and --force

Browse source
This commit is contained in:
Zonghang Li 2025-03-11 20:44:36 +04:00
parent 9cbdf01645
commit bcfdace59b
5 changed files with 27 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.h
View file

@ -149,7 +149,9 @@ struct gpt_params {
std::string next_node_ip = "localhost"; // ip address of my next node
bool prefetch = false; // prefetch layer weights
bool keep_out_in_metal = true; // whether to keep output weights in metal memory, true by default
bool force = false; // force to start prefetching after computation
int32_t gpu_mem = 999.0; // gpu memory to use, in GiB
int32_t n_cycles = 0; // number of cycles to output one token
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)