vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

rpc : rpc_count always zero for non-RPC builds

Browse source
This commit is contained in:
Radoslav Gerganov 2024-09-03 16:58:26 +03:00
parent 7049733bcc
commit 99cd44677a
1 changed files with 8 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

8
src/llama.cpp
View file

@ -3346,7 +3346,11 @@ static size_t llama_get_device_count(const llama_model & model) {
static ggml_backend_buffer_type_t llama_default_buffer_type_offload(const llama_model & model, int gpu) { static ggml_backend_buffer_type_t llama_default_buffer_type_offload(const llama_model & model, int gpu) {
ggml_backend_buffer_type_t buft = nullptr; ggml_backend_buffer_type_t buft = nullptr;
#ifdef GGML_USE_RPC
int rpc_count = (int)model.rpc_servers.size(); int rpc_count = (int)model.rpc_servers.size();
#else
int rpc_count = 0;
#endif
int local_gpu = gpu - rpc_count; int local_gpu = gpu - rpc_count;
#if defined(GGML_USE_RPC) #if defined(GGML_USE_RPC)
if (gpu < rpc_count) { if (gpu < rpc_count) {
@ -3403,7 +3407,11 @@ static ggml_backend_buffer_type_t llama_default_buffer_type_split(const llama_mo
} }
static size_t llama_get_device_memory(const llama_model & model, int device) { static size_t llama_get_device_memory(const llama_model & model, int device) {
#ifdef GGML_USE_RPC
int rpc_count = (int)model.rpc_servers.size(); int rpc_count = (int)model.rpc_servers.size();
#else
int rpc_count = 0;
#endif
int local_device = device - rpc_count; int local_device = device - rpc_count;
#if defined(GGML_USE_RPC) #if defined(GGML_USE_RPC)
if (device < rpc_count) { if (device < rpc_count) {