root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 20:04:35 +00:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

llama : fix non-CUDA build

Browse Source
This commit is contained in:
Georgi Gerganov 2023-10-29 11:12:03 +02:00
parent b4ad03b3a7
commit 25cfbf6776
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 14 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
llama.cpp
View File

@ -5553,6 +5553,13 @@ static struct ggml_cgraph * llama_build_graph(
{ OFFLOAD_FUNC_NR, "GPU (CUDA) NR" }, { OFFLOAD_FUNC_NR, "GPU (CUDA) NR" },
{ OFFLOAD_FUNC_EMB, "GPU (CUDA) EMB" }, { OFFLOAD_FUNC_EMB, "GPU (CUDA) EMB" },
{ OFFLOAD_FUNC_OUT, "GPU (CUDA) OUT" }, { OFFLOAD_FUNC_OUT, "GPU (CUDA) OUT" },
#else
{ OFFLOAD_FUNC, "CPU" },
{ OFFLOAD_FUNC_KQ, "CPU" },
{ OFFLOAD_FUNC_V, "CPU" },
{ OFFLOAD_FUNC_NR, "CPU" },
{ OFFLOAD_FUNC_EMB, "CPU" },
{ OFFLOAD_FUNC_OUT, "CPU" },
#endif // GGML_USE_CUBLAS #endif // GGML_USE_CUBLAS
}; };
@ -5707,6 +5714,12 @@ static struct ggml_cgraph * llama_build_graph(
offload_func_t func = ggml_offload_nop; offload_func_t func = ggml_offload_nop;
#ifdef GGML_USE_CUBLAS
static offload_func_t ggml_offload_gpu = ggml_cuda_assign_buffers_no_alloc;
#else
static offload_func_t ggml_offload_gpu = ggml_offload_nop;
#endif
switch (func_e) { switch (func_e) {
case OFFLOAD_FUNC_NOP: case OFFLOAD_FUNC_NOP:
case OFFLOAD_FUNC_OUT: func = ggml_offload_nop; break; case OFFLOAD_FUNC_OUT: func = ggml_offload_nop; break;
@ -5714,7 +5727,7 @@ static struct ggml_cgraph * llama_build_graph(
case OFFLOAD_FUNC_KQ: case OFFLOAD_FUNC_KQ:
case OFFLOAD_FUNC_V: case OFFLOAD_FUNC_V:
case OFFLOAD_FUNC_NR: case OFFLOAD_FUNC_NR:
case OFFLOAD_FUNC_EMB: func = ggml_cuda_assign_buffers_no_alloc; break; case OFFLOAD_FUNC_EMB: func = ggml_offload_gpu; break;
default: GGML_ASSERT(false); default: GGML_ASSERT(false);
} }