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

Add GGML_USE_BLAS flag to llama.cpp and update BLAS documentation

Browse source
This commit is contained in:
simonteozw 2024-08-18 23:47:24 +08:00
parent e11bd856d5
commit 4a4a3420de
2 changed files with 9 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

8
docs/build.md
View file

@ -125,6 +125,14 @@ This provides BLAS acceleration using only the CPU. Make sure to have OpenBLAS i
make GGML_OPENBLAS=1 make GGML_OPENBLAS=1
``` ```
- On Mac (with Intel GPU):
```bash
brew install openblas
EXPORT PKG_CONFIG_PATH=<openblas_path>
make GGML_OPENBLAS=1 GGML_NO_METAL=1
```
- Using `CMake` on Linux: - Using `CMake` on Linux:
```bash ```bash

2
src/llama.cpp
View file

@ -17424,7 +17424,7 @@ bool llama_supports_mlock(void) {
bool llama_supports_gpu_offload(void) { bool llama_supports_gpu_offload(void) {
#if defined(GGML_USE_CUDA) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \ #if defined(GGML_USE_CUDA) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \
defined(GGML_USE_SYCL) || defined(GGML_USE_KOMPUTE) || defined(GGML_USE_RPC) defined(GGML_USE_SYCL) || defined(GGML_USE_KOMPUTE) || defined(GGML_USE_RPC) || defined(GGML_USE_BLAS)
// Defined when llama.cpp is compiled with support for offloading model layers to GPU. // Defined when llama.cpp is compiled with support for offloading model layers to GPU.
return true; return true;
#else #else