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

ggml backend interface wip

Browse source 
refactor ggml-cuda
This commit is contained in:
slaren 2023-07-10 17:32:06 +02:00
parent 6e7cca4047
commit 0d2b66c638
15 changed files with 4480 additions and 4782 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama.h
View file

@ -2,12 +2,7 @@
#define LLAMA_H
#include "ggml.h"
#ifdef GGML_USE_CUBLAS
#include "ggml-cuda.h"
#define LLAMA_MAX_DEVICES GGML_CUDA_MAX_DEVICES
#else
#define LLAMA_MAX_DEVICES 1
#endif // GGML_USE_CUBLAS
#include <stddef.h>
#include <stdint.h>
#include <stdbool.h>
@ -48,7 +43,7 @@
#define LLAMA_DEFAULT_SEED 0xFFFFFFFF
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL)
#if defined(GGML_USE_CUDA) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL)
// Defined when llama.cpp is compiled with support for offloading model layers to GPU.
#define LLAMA_SUPPORTS_GPU_OFFLOAD
#endif