mirror of
https://github.com/ollama/ollama.git
synced 2026-03-27 02:58:43 +07:00
Add experimental MLX backend and engine with imagegen support (#13648)
* WIP - MLX backend with gemma3 * MLX: add cmake and go tag build toggles To build the new MLX backend code: cmake --preset MLX cmake --build --preset MLX --parallel cmake --install build --component MLX go build -tags mlx . Note: the main.go entrypoint for the MLX engine will change in a follow up commit. * add experimental image generation runtime * add experimental image generation runtime * MLX: wire up cuda build for linux * MLX: get dependencies correct and dedup This is still too large for a unified github artifact, but is now "correct" for the mlx_cuda_v13 directory. * fix relative link bug in dedup * Add darwin build and readme * add go build tag for mlx dependent code and wire up build_darwin.sh * lint cleanup * macos: build mlx for x86 This will be CPU only. * cuda build instructions and fix drift from mlx bump * stale comment * Delete agent helper doc * Clean up readme.md * Revise README for tokenizer clarity and details Updated README to clarify tokenizer functionality and removed correctness section. --------- Co-authored-by: jmorganca <jmorganca@gmail.com>
This commit is contained in:
@@ -2,6 +2,22 @@ cmake_minimum_required(VERSION 3.21)
|
||||
|
||||
project(Ollama C CXX)
|
||||
|
||||
# Handle cross-compilation on macOS: when CMAKE_OSX_ARCHITECTURES is set to a
|
||||
# single architecture different from the host, override CMAKE_SYSTEM_PROCESSOR
|
||||
# to match. This is necessary because CMAKE_SYSTEM_PROCESSOR defaults to the
|
||||
# host architecture, but downstream projects (like MLX) use it to detect the
|
||||
# target architecture.
|
||||
if(CMAKE_OSX_ARCHITECTURES AND NOT CMAKE_OSX_ARCHITECTURES MATCHES ";")
|
||||
# Single architecture specified
|
||||
if(CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" AND NOT CMAKE_SYSTEM_PROCESSOR STREQUAL "x86_64")
|
||||
message(STATUS "Cross-compiling for x86_64: overriding CMAKE_SYSTEM_PROCESSOR from ${CMAKE_SYSTEM_PROCESSOR} to x86_64")
|
||||
set(CMAKE_SYSTEM_PROCESSOR "x86_64")
|
||||
elseif(CMAKE_OSX_ARCHITECTURES STREQUAL "arm64" AND NOT CMAKE_SYSTEM_PROCESSOR STREQUAL "arm64")
|
||||
message(STATUS "Cross-compiling for arm64: overriding CMAKE_SYSTEM_PROCESSOR from ${CMAKE_SYSTEM_PROCESSOR} to arm64")
|
||||
set(CMAKE_SYSTEM_PROCESSOR "arm64")
|
||||
endif()
|
||||
endif()
|
||||
|
||||
include(CheckLanguage)
|
||||
include(GNUInstallDirs)
|
||||
|
||||
@@ -12,7 +28,7 @@ set(BUILD_SHARED_LIBS ON)
|
||||
|
||||
set(CMAKE_CXX_STANDARD 17)
|
||||
set(CMAKE_CXX_STANDARD_REQUIRED ON)
|
||||
set(CMAKE_CXX_EXTENSIONS OFF)
|
||||
set(CMAKE_CXX_EXTENSIONS ON) # Recent versions of MLX Requires gnu++17 extensions to compile properly
|
||||
|
||||
set(GGML_BUILD ON)
|
||||
set(GGML_SHARED ON)
|
||||
@@ -147,14 +163,48 @@ if(CMAKE_HIP_COMPILER)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
find_package(Vulkan)
|
||||
if(Vulkan_FOUND)
|
||||
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/ml/backend/ggml/ggml/src/ggml-vulkan)
|
||||
install(TARGETS ggml-vulkan
|
||||
RUNTIME_DEPENDENCIES
|
||||
PRE_INCLUDE_REGEXES vulkan
|
||||
PRE_EXCLUDE_REGEXES ".*"
|
||||
RUNTIME DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT Vulkan
|
||||
LIBRARY DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT Vulkan
|
||||
)
|
||||
if(NOT APPLE)
|
||||
find_package(Vulkan)
|
||||
if(Vulkan_FOUND)
|
||||
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/ml/backend/ggml/ggml/src/ggml-vulkan)
|
||||
install(TARGETS ggml-vulkan
|
||||
RUNTIME_DEPENDENCIES
|
||||
PRE_INCLUDE_REGEXES vulkan
|
||||
PRE_EXCLUDE_REGEXES ".*"
|
||||
RUNTIME DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT Vulkan
|
||||
LIBRARY DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT Vulkan
|
||||
)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
option(MLX_ENGINE "Enable MLX backend" OFF)
|
||||
|
||||
if(MLX_ENGINE)
|
||||
message(STATUS "Setting up MLX (this takes a while...)")
|
||||
add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/x/ml/backend/mlx)
|
||||
|
||||
# Find CUDA toolkit if MLX is built with CUDA support
|
||||
find_package(CUDAToolkit)
|
||||
|
||||
install(TARGETS mlx mlxc
|
||||
RUNTIME_DEPENDENCIES
|
||||
DIRECTORIES ${CUDAToolkit_BIN_DIR} ${CUDAToolkit_BIN_DIR}/x64 ${CUDAToolkit_LIBRARY_DIR}
|
||||
PRE_INCLUDE_REGEXES cublas cublasLt cudart nvrtc cudnn nccl
|
||||
PRE_EXCLUDE_REGEXES ".*"
|
||||
RUNTIME DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT MLX
|
||||
LIBRARY DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT MLX
|
||||
FRAMEWORK DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT MLX
|
||||
)
|
||||
|
||||
# Manually install cudart and cublas since they might not be picked up as direct dependencies
|
||||
if(CUDAToolkit_FOUND)
|
||||
file(GLOB CUDART_LIBS
|
||||
"${CUDAToolkit_LIBRARY_DIR}/libcudart.so*"
|
||||
"${CUDAToolkit_LIBRARY_DIR}/libcublas.so*")
|
||||
if(CUDART_LIBS)
|
||||
install(FILES ${CUDART_LIBS}
|
||||
DESTINATION ${OLLAMA_INSTALL_DIR}
|
||||
COMPONENT MLX)
|
||||
endif()
|
||||
endif()
|
||||
endif()
|
||||
Reference in New Issue
Block a user