llama.cpp_build

Here you can find scripts and instructions how to build llama.cpp with diferent backends: CPU, VULKAN, SYCL, OpenVINO.

Usage instructions:

git clone https://github.com/avbelova/llama.cpp_build.git
cd llama.cpp_build

CPU

Windows

build_cpu.bat

Linux

build_cpu.sh

GPU-VULKAN

Windows

Install VULKAN SDK. The following script downloads VULKAN SDK installer and runs it. Please follow the installer steps manually.

install_vulkan.bat

Build

build_vulkan.bat

Linux

install_vulkan.sh
build_vulkan.sh

GPU-SYCL

Windows

Install Intel® oneAPI Base Toolkit manually or using the script:

install_oneapi.bat

Set up needed environment

"C:\Program Files (x86)\Intel\oneAPI\setvars.bat"

Use build script from cloned llama.cpp

cd llama.cpp
.\examples\sycl\win-build-sycl.bat

OpenVINO

Windows

Prerequisites:

• MSVC (script is verified with MSVC 2022).
• OpenVINO (for C++)

Execute the following commands in MSVC Developer PowerShell:

Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope Process -Force
build_openvino.ps1