Skip to content

Latest commit

 

History

History
237 lines (176 loc) · 9.43 KB

macos.md

File metadata and controls

237 lines (176 loc) · 9.43 KB
Raw

Machine Learning Build Machine Setup for macOS

These same instructions should work for native compilation on both x86_64 and aarch64 architectures.

To ensure everything is consistent for redistributable builds we build all redistributable components from source.

NOTE: if you upgrade macOS then Xcode may need to be re-installed. Please ensure your Xcode is still valid after upgrades.

You will need the following environment variables to be defined:

  • JAVA_HOME - Should point to the JDK you want to use to run Gradle.
  • CPP_SRC_HOME - Only required if building the C++ code directly using cmake, as Gradle sets it automatically.
  • PATH - Must have /usr/local/bin before /usr/bin and /bin.

For example, you might create a .bashrc file in your home directory containing something like this:

umask 0002
export JAVA_HOME=/Library/Java/JavaVirtualMachines/jdk-17.0.1.jdk/Contents/Home
export PYTHONHOME=/Library/Frameworks/Python.framework/Versions/3.10
export PATH=$JAVA_HOME/bin:$PYTHONHOME/bin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin
# Only required if building the C++ code directly using cmake - adjust depending on the location of your Git clone
export CPP_SRC_HOME=$HOME/ml-cpp

Note, that bash doesn't read ~/.bashrc for login shells (which is what you get when you open a new Terminal window) - it reads ~/.bash_profile or ~/.profile so you should probably also create a .bash_profile file in your home directory which sources .bashrc.

General settings for building the tools

Some tools may be built via a GNU "configure" script. There are some environment variables that affect the behaviour of this. Therefore, when building ANY tool on macOS, set the following environment variables:

export SSEFLAGS=`[ $(uname -m) = x86_64 ] && echo -msse4.2`
export CPP='clang -E'
export CC=clang
export CFLAGS="-O3 $SSEFLAGS"
export CXX='clang++ -std=c++17 -stdlib=libc++'
export CXXFLAGS="-O3 $SSEFLAGS"
export CXXCPP='clang++ -std=c++17 -E'
export LDFLAGS=-Wl,-headerpad_max_install_names
unset CPATH
unset C_INCLUDE_PATH
unset CPLUS_INCLUDE_PATH
unset LIBRARY_PATH

The above environment variables only need to be set when building tools on macOS. They should NOT be set when compiling the Machine Learning source code (as this should pick up all settings from our build system).

Xcode

The first major piece of development software to install is Apple's development environment, Xcode, which can be downloaded from https://developer.apple.com/download/ . You will need to register as a developer with Apple. Alternatively, you can get the latest version of Xcode from the App Store.

  • If you are using Big Sur, you must install Xcode 13.2.x
  • If you are using Monterey, you must install Xcode 14.2.x
  • If you are using Ventura, you must install Xcode 15.2.x
  • If you are using Sonoma, you must install Xcode 15.2.x or above

Xcode is distributed as a .xip file; simply double click the .xip file to expand it, then drag Xcode.app to your /Applications directory. (Older versions of Xcode can be downloaded from here, provided you are signed in with your Apple ID.)

There are no command line tools out-of-the-box, so you'll need to install them following installation of Xcode. You can do this by running:

xcode-select --install

at the command prompt.

Boost 1.83.0

Download version 1.83.0 of Boost from https://boostorg.jfrog.io/artifactory/main/release/1.83.0/source/boost_1_83_0.tar.bz2. You must get this exact version, as the Machine Learning build system requires it.

Assuming you chose the .bz2 version, extract it to a temporary directory:

bzip2 -cd boost_1_83_0.tar.bz2 | tar xvf -

In the resulting boost_1_83_0 directory, run:

./bootstrap.sh --with-toolset=clang --without-libraries=context --without-libraries=coroutine --without-libraries=graph_parallel --without-libraries=mpi --without-libraries=python --without-icu

This should build the b2 program, which in turn is used to build Boost.

Edit boost/unordered/detail/prime_fmod.hpp and change line 134 from

    (13ul)(29ul)(53ul)(97ul)(193ul)(389ul)(769ul)(1543ul)(3079ul)(6151ul)(       \

to:

    (3ul)(13ul)(29ul)(53ul)(97ul)(193ul)(389ul)(769ul)(1543ul)(3079ul)(6151ul)(       \

To complete the build, type:

./b2 -j8 --layout=versioned --disable-icu cxxflags="-std=c++17 -stdlib=libc++ $SSEFLAGS" linkflags="-std=c++17 -stdlib=libc++ -Wl,-headerpad_max_install_names" optimization=speed inlining=full define=BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS define=BOOST_LOG_WITHOUT_DEBUG_OUTPUT define=BOOST_LOG_WITHOUT_EVENT_LOG define=BOOST_LOG_WITHOUT_SYSLOG define=BOOST_LOG_WITHOUT_IPC
sudo ./b2 install --layout=versioned --disable-icu cxxflags="-std=c++17 -stdlib=libc++ $SSEFLAGS" linkflags="-std=c++17 -stdlib=libc++ -Wl,-headerpad_max_install_names" optimization=speed inlining=full define=BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS define=BOOST_LOG_WITHOUT_DEBUG_OUTPUT define=BOOST_LOG_WITHOUT_EVENT_LOG define=BOOST_LOG_WITHOUT_SYSLOG define=BOOST_LOG_WITHOUT_IPC

to install the Boost headers and libraries.

CMake

CMake version 3.19.2 is the minimum required to build ml-cpp. Download the graphical installer for version 3.23.2 from https://github.com/Kitware/CMake/releases/download/v3.23.2/cmake-3.23.2-macos-universal.dmg (or get a more recent version).

Open the .dmg and install the application it by dragging it to the Applications folder.

Then make the cmake program accessible to programs that look in /usr/local/bin:

sudo mkdir -p /usr/local/bin
sudo ln -s /Applications/CMake.app/Contents/bin/cmake /usr/local/bin/cmake

Python 3.10

PyTorch currently requires Python 3.7 or higher; we use version 3.10.

Download the graphical installer for Python 3.10.9 from https://www.python.org/ftp/python/3.10.9/python-3.10.9-macos11.pkg.

Install using all the default options. When the installer completes a Finder window pops up. Double click the Install Certificates.command file in this folder to install the SSL certificates Python needs.

PyTorch 2.3.1

PyTorch requires that certain Python modules are installed. To install them:

sudo /Library/Frameworks/Python.framework/Versions/3.10/bin/pip3.10 install numpy ninja pyyaml setuptools cffi typing_extensions future six requests dataclasses

Then obtain the PyTorch code:

git clone --depth=1 --branch=v2.3.1 https://github.com/pytorch/pytorch.git
cd pytorch
git submodule sync
git submodule update --init --recursive

Edit torch/csrc/jit/codegen/fuser/cpu/fused_kernel.cpp and replace all occurrences of system( with strlen(. This file is used to compile fused CPU kernels, which we do not expect to be doing and never want to do for security reasons. Replacing the calls to system() ensures that a heuristic virus scanner looking for potentially dangerous function calls in our shipped product will not encounter these functions that run external processes.

Edit tools/setup_helpers/cmake.py and add "DNNL_TARGET_ARCH" to the list of environment variables that get passed through to CMake (around line 215).

Edit CMakeLists.txt and change set(CMAKE_LINK_WHAT_YOU_USE TRUE) to set(CMAKE_LINK_WHAT_YOU_USE FALSE)

For compilation on Apple silicon Macs edit third_party/ideep/include/ideep/utils.hpp and add:

inline void to_bytes(bytestring& bytes, const unsigned long arg) {
  auto as_cstring = reinterpret_cast<const char*>(&arg);
  bytes.append(as_cstring, sizeof(unsigned long));
}

at around line 189. This is necessary to resolve a template specialization issue.

For compilation on macOS Mojave (10.14)

  • Edit torch/csrc/jit/api/module.h and change 
    Module(Module&&) noexcept = default;`
Module& operator=(Module&&) noexcept = default;
    to 
    Module(Module&&) = default;
Module& operator=(Module&&) noexcept = default;
  • Edit third_party/onnx/CMakeLists.txt and
    • add find_package(Boost 1.83.0 EXACT REQUIRED COMPONENTS filesystem) before target_include_directories(onnx PUBLIC (around line 531)
    • add $<BUILD_INTERFACE:${Boost_INCLUDE_DIRS}> after $<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}> (around line 536)
  • Edit third_party/onnx/onnx/common/file_utils.h and change the lines 
    #include <filesystem>
#include <fstream>
    to 
    #include <boost/filesystem/operations.hpp>
#include <boost/filesystem/fstream.hpp>
    and change 
    std::filesystem::path proto_u8_path = std::filesystem::u8path(proto_path);
std::fstream proto_stream(proto_u8_path, std::ios::in | std::ios::binary);
    to 
    boost::filesystem::path proto_u8_path = boost::filesystem::u8path(proto_path);
boost::filesystem::fstream proto_stream(proto_u8_path, std::ios::in | std::ios::binary);

Build as follows:

export BLAS=vecLib
export BUILD_TEST=OFF
export BUILD_CAFFE2=OFF
export USE_NUMPY=OFF
export USE_DISTRIBUTED=OFF
[ $(uname -m) != x86_64 ] && export DNNL_TARGET_ARCH=AARCH64
export USE_MKLDNN=ON
export USE_QNNPACK=OFF
export USE_PYTORCH_QNNPACK=OFF
[ $(uname -m) = x86_64 ] && export USE_XNNPACK=OFF
export PYTORCH_BUILD_VERSION=2.3.1
export PYTORCH_BUILD_NUMBER=1
/Library/Frameworks/Python.framework/Versions/3.10/bin/python3.10 setup.py install

Once built copy headers and libraries to system directories:

sudo mkdir -p /usr/local/include/pytorch
sudo cp -r torch/include/* /usr/local/include/pytorch/
sudo cp torch/lib/libtorch_cpu.dylib /usr/local/lib/
sudo cp torch/lib/libc10.dylib /usr/local/lib/