Build ONNX Runtime for inferencing

Follow the instructions below to build ONNX Runtime to perform inference.

Contents

CPU

Basic CPU build

Prerequisites

  • Checkout the source tree:

     git clone --recursive https://github.com/Microsoft/onnxruntime.git
     cd onnxruntime
    
  • Install Python 3.x.

  • Install cmake-3.26 or higher.

    On Windows, please run

      python -m pip install cmake
      where cmake
    

    On Linux, please run

      python3 -m pip install cmake
      which cmake
    

    If the above commands failed, please manually get cmake from https://cmake.org/download/.

    After the installation, you can run

      cmake --version
    

    to verify if the installation was successful.

Build Instructions

Windows

Open Developer Command Prompt for Visual Studio version you are going to use. This will properly setup the environment including paths to your compiler, linker, utilities and header files.

.\build.bat --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync

The default Windows CMake Generator is Visual Studio 2022. For Visual Studio 2019 add --cmake_generator "Visual Studio 16 2019".

We recommend using Visual Studio 2022.

If you want to build an ARM64 binary on a Windows ARM64 machine, you can use the same command above. Just be sure that your Visual Studio, CMake and Python are all ARM64 version.

If you want to cross-compile an ARM32 or ARM64 or ARM64EC binary on a Windows x86 machine, you need to add “–arm” or “–arm64” or “–arm64ec” to the build command above.

When building on x86 Windows without “–arm” or “–arm64” or “–arm64ec” args, the built binaries will be 64-bit if your python is 64-bit, or 32-bit if your python is 32-bit,

Linux

./build.sh --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync

macOS

By default, ONNX Runtime is configured to be built for a minimum target macOS version of 10.12. The shared library in the release Nuget(s) and the Python wheel may be installed on macOS versions of 10.12+.

If you would like to use Xcode to build the onnxruntime for x86_64 macOS, please add the –user_xcode argument in the command line.

Without this flag, the cmake build generator will be Unix makefile by default.

Today, Mac computers are either Intel-Based or Apple silicon(aka. ARM) based. By default, ONNX Runtime’s build script only generate bits for the CPU ARCH that the build machine has. If you want to do cross-compiling: generate ARM binaries on a Intel-Based Mac computer, or generate x86 binaries on a Mac ARM computer, you can set the “CMAKE_OSX_ARCHITECTURES” cmake variable. For example:

Build for Intel CPUs:

./build.sh --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync --cmake_extra_defines CMAKE_OSX_ARCHITECTURES=x86_64

Build for Apple silicon CPUs:

./build.sh --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync --cmake_extra_defines CMAKE_OSX_ARCHITECTURES=arm64

Build for both:

./build.sh --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync --cmake_extra_defines CMAKE_OSX_ARCHITECTURES="x86_64;arm64"

The last command will generate a fat-binary for both CPU architectures.

Note: unit tests will be skipped due to the incompatible CPU instruction set when doing cross-compiling.

Notes

  • Please note that these instructions build the debug build, which may have performance tradeoffs. The “–config” parameter has four valid values: Debug, Release, RelWithDebInfo and MinSizeRel. Compared to “Release”, “RelWithDebInfo” not only has debug info, it also disables some inlines to make the binary easier to debug. Thus RelWithDebInfo is slower than Release.
  • To build the version from each release (which include Windows, Linux, and Mac variants), see these .yml files for reference
  • The build script runs all unit tests by default for native builds and skips tests by default for cross-compiled builds. To skip the tests, run with --build or --update --build.
  • If you need to install protobuf from source code, please note:
    • First, please open cmake/deps.txt to check which protobuf version ONNX Runtime’s offical packages use.
    • As we statically link to protobuf, on Windows protobuf’s CMake flag protobuf_BUILD_SHARED_LIBS should be turned OFF, on Linux if the option is OFF you also need to make sure PIC is enabled. After the installation, you should have the ‘protoc’ executable in your PATH. It is recommended to run ldconfig to make sure protobuf libraries are found.
    • If you installed your protobuf in a non standard location it would be helpful to set the following env var:export CMAKE_ARGS="-DONNX_CUSTOM_PROTOC_EXECUTABLE=full path to protoc" so the ONNX build can find it. Also run ldconfig <protobuf lib folder path> so the linker can find protobuf libraries.
  • If you’d like to install onnx from source code, install protobuf first and:
      export ONNX_ML=1
      python3 setup.py bdist_wheel
      pip3 install --upgrade dist/*.whl
    

    Then, it’s better to uninstall protobuf before you start to build ONNX Runtime, especially if you have install a different version of protobuf other than what ONNX Runtime has. —

Supported architectures and build environments

Architectures

  x86_32 x86_64 ARM32v7 ARM64 PPC64LE
Windows YES YES YES YES NO
Linux YES YES YES YES YES
macOS NO YES NO NO NO
Android NO NO YES YES NO
iOS NO NO NO YES NO

Build Environments(Host)

OS Supports CPU Supports GPU Notes
Windows 10 YES YES VS2019 through the latest VS2015 are supported
Windows 10
Subsystem for Linux
YES NO  
Ubuntu 20.x/22.x YES YES Also supported on ARM32v7 (experimental)
CentOS 7/8/9 YES YES Also supported on ARM32v7 (experimental)
macOS YES NO  

GCC 8.x and below are not supported.
If you want to build a binary for a 32-bit architecture, you might have to do cross-compiling since a 32-bit compiler might not have enough memory to run the build.
Building the code on Android/iOS is not supported. You need to use a Windows, Linux or macOS device to do so.

OS/Compiler Supports VC Supports GCC Supports Clang
Windows 10 YES Not tested Not tested
Linux NO YES(gcc>=8) Not tested
macOS NO Not tested YES (Minimum version required not ascertained)

Target Environments

You can build the code for

  • Windows
  • Linux
  • MacOS
  • Android
  • iOS
  • WebAssembly

At runtime:

  • The minimum supported Windows version is Windows 10.
  • The minimum supported CentOS version is 7.
  • The minimum supported Ubuntu version is 16.04.

Common Build Instructions

Description Command Additional details
Basic build build.bat (Windows)
./build.sh (Linux)
 
Release build --config Release Release build. Other valid config values are RelWithDebInfo and Debug.
Build using parallel processing --parallel This is strongly recommended to speed up the build.
Build Shared Library --build_shared_lib  
Enable Training support --enable_training  

APIs and Language Bindings

API Command Additional details
Python --build_wheel  
C# and C Nuget packages --build_nuget Builds C# bindings and creates nuget package. Implies --build_shared_lib
Detailed instructions can be found below.
WindowsML --use_winml
--use_dml
--build_shared_lib
WindowsML depends on DirectML and the OnnxRuntime shared library
Java --build_java Creates an onnxruntime4j.jar in the build directory, implies --build_shared_lib
Compiling the Java API requires gradle v6.1+ to be installed in addition to the usual requirements.
Node.js --build_nodejs Build Node.js binding. Implies --build_shared_lib

Build Nuget packages

Currently only supported on Windows and Linux.

Prerequisites
  • dotnet is required for building csharp bindings and creating managed nuget package. Follow the instructions here to download dotnet. Tested with versions 2.1 and 3.1.
  • nuget.exe. Follow the instructions here to download nuget
    • On Windows, downloading nuget is straightforward and simply following the instructions above should work.
    • On Linux, nuget relies on Mono runtime and therefore this needs to be setup too. Above link has all the information to setup Mono and nuget. The instructions can directly be found here. In some cases it is required to run sudo apt-get install mono-complete after installing mono.
Build Instructions
Windows
.\build.bat --build_nuget
Linux
./build.sh --build_nuget

Nuget packages are created under \nuget-artifacts


Other build options

Reduced Operator Kernel Build

Reduced Operator Kernel builds allow you to customize the kernels in the build to provide smaller binary sizes.

DebugNodeInputsOutputs

OnnxRuntime supports build options for enabling debugging of intermediate tensor shapes and data.

Build Instructions

Set onnxruntime_DEBUG_NODE_INPUTS_OUTPUT to build with this enabled.

Linux
./build.sh --cmake_extra_defines onnxruntime_DEBUG_NODE_INPUTS_OUTPUTS=1
Windows
.\build.bat --cmake_extra_defines onnxruntime_DEBUG_NODE_INPUTS_OUTPUTS=1

Configuration

The debug dump behavior can be controlled with several environment variables. See onnxruntime/core/framework/debug_node_inputs_outputs_utils.h for details.

Examples

To specify that node output data should be dumped (to stdout by default), set this environment variable:

ORT_DEBUG_NODE_IO_DUMP_OUTPUT_DATA=1

To specify that node output data should be dumped to files for nodes with name “Foo” or “Bar”, set these environment variables:

ORT_DEBUG_NODE_IO_DUMP_OUTPUT_DATA=1
ORT_DEBUG_NODE_IO_NAME_FILTER="Foo;Bar"
ORT_DEBUG_NODE_IO_DUMP_DATA_TO_FILES=1

ARM

There are a few options for building for ARM.

Cross compiling for ARM with simulation (Linux/Windows)

EASY, SLOW, RECOMMENDED

This method relies on qemu user mode emulation. It allows you to compile using a desktop or cloud VM through instruction level simulation. You’ll run the build on x86 CPU and translate every ARM instruction to x86. This is much faster than compiling natively on a low-end ARM device and avoids out-of-memory issues that may be encountered. The resulting ONNX Runtime Python wheel (.whl) file is then deployed to an ARM device where it can be invoked in Python 3 scripts.

Here is an example for Raspberrypi3 and Raspbian. Note: this does not work for Raspberrypi 1 or Zero, and if your operating system is different from what the dockerfile uses, it also may not work.

The build process can take hours.

Cross compiling on Linux

Difficult, fast

This option is very fast and allows the package to be built in minutes, but is challenging to setup. If you have a large code base (e.g. you are adding a new execution provider to onnxruntime), this may be the only feasible method.

  1. Get the corresponding toolchain.

    TLDR; Go to https://www.linaro.org/downloads/, get “64-bit Armv8 Cortex-A, little-endian” and “Linux Targeted”, not “Bare-Metal Targeted”. Extract it to your build machine and add the bin folder to your $PATH env. Then skip this part.

    You can use GCC or Clang. Both work, but instructions here are based on GCC.

    In GCC terms:

    • “build” describes the type of system on which GCC is being configured and compiled.
    • “host” describes the type of system on which GCC runs.
    • “target” to describe the type of system for which GCC produce code.

    When not cross compiling, usually “build” = “host” = “target”. When you do cross compile, usually “build” = “host” != “target”. For example, you may build GCC on x86_64, then run GCC on x86_64, then generate binaries that target aarch64. In this case,”build” = “host” = x86_64 Linux, target is aarch64 Linux.

    You can either build GCC from source code by yourself, or get a prebuilt one from a vendor like Ubuntu, linaro. Choosing the same compiler version as your target operating system is best. If ths is not possible, choose the latest stable one and statically link to the GCC libs.

    When you get the compiler, run aarch64-linux-gnu-gcc -v This should produce an output like below:

     Using built-in specs.
     COLLECT_GCC=/usr/bin/aarch64-linux-gnu-gcc
     COLLECT_LTO_WRAPPER=/usr/libexec/gcc/aarch64-linux-gnu/9/lto-wrapper
     Target: aarch64-linux-gnu
     Configured with: ../gcc-9.2.1-20190827/configure --bindir=/usr/bin --build=x86_64-redhat-linux-gnu --datadir=/usr/share --disable-decimal-float --disable-dependency-tracking --disable-gold --disable-libgcj --disable-libgomp --disable-libmpx --disable-libquadmath --disable-libssp --disable-libunwind-exceptions --disable-shared --disable-silent-rules --disable-sjlj-exceptions --disable-threads --with-ld=/usr/bin/aarch64-linux-gnu-ld --enable-__cxa_atexit --enable-checking=release --enable-gnu-unique-object --enable-initfini-array --enable-languages=c,c++ --enable-linker-build-id --enable-lto --enable-nls --enable-obsolete --enable-plugin --enable-targets=all --exec-prefix=/usr --host=x86_64-redhat-linux-gnu --includedir=/usr/include --infodir=/usr/share/info --libexecdir=/usr/libexec --localstatedir=/var --mandir=/usr/share/man --prefix=/usr --program-prefix=aarch64-linux-gnu- --sbindir=/usr/sbin --sharedstatedir=/var/lib --sysconfdir=/etc --target=aarch64-linux-gnu --with-bugurl=http://bugzilla.redhat.com/bugzilla/ --with-gcc-major-version-only --with-isl --with-newlib --with-plugin-ld=/usr/bin/aarch64-linux-gnu-ld --with-sysroot=/usr/aarch64-linux-gnu/sys-root --with-system-libunwind --with-system-zlib --without-headers --enable-gnu-indirect-function --with-linker-hash-style=gnu
     Thread model: single
     gcc version 9.2.1 20190827 (Red Hat Cross 9.2.1-3) (GCC)
    

    Check the value of --build, --host, --target, and if it has special args like --with-arch=armv8-a, --with-arch=armv6, --with-tune=arm1176jz-s, --with-fpu=vfp, --with-float=hard.

    You must also know what kind of flags your target hardware need, which can differ greatly. For example, if you just get the normal ARMv7 compiler and use it for Raspberry Pi V1 directly, it won’t work because Raspberry Pi only has ARMv6. Generally every hardware vendor will provide a toolchain; check how that one was built.

    A target env is identifed by:

    • Arch: x86_32, x86_64, armv6,armv7,arvm7l,aarch64,…
    • OS: bare-metal or linux.
    • Libc: gnu libc/ulibc/musl/…
    • ABI: ARM has mutilple ABIs like eabi, eabihf…

    You can get all these information from the previous output, please be sure they are all correct.

  2. Get a pre-compiled protoc:

    Get this from https://github.com/protocolbuffers/protobuf/releases/download/v3.18.1/protoc-3.18.1-linux-x86_64.zip and unzip after downloading. The version must match the one onnxruntime is using in cmake/external/protobuf. Currently we are using 3.18.1.

  3. (Optional) Setup sysroot to enable python extension. Skip if not using Python.

    Dump the root file system of the target operating system to your build machine. We’ll call that folder “sysroot” and use it for build onnxruntime python extension. Before doing that, you should install python3 dev package(which contains the C header files) and numpy python package on the target machine first.

    Below are some examples.

    If the target OS is raspbian-buster, please download the RAW image from their website then run:

     $ fdisk -l 2020-02-13-raspbian-buster.img
    

    Disk 2020-02-13-raspbian-buster.img: 3.54 GiB, 3787456512 bytes, 7397376 sectors Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disklabel type: dos Disk identifier: 0xea7d04d6

    Device Boot Start End Sectors Size Id Type
    2020-02-13-raspbian-buster.img1   8192 532479 524288 256M c W95 FAT32 (LBA)
    2020-02-13-raspbian-buster.img2   532480 7397375 6864896 3.3G 83 Linux

    You’ll find the the root partition starts at the 532480 sector, which is 532480 * 512=272629760 bytes from the beginning.

    Then run:

     $ mkdir /mnt/pi
     $ mount -r -o loop,offset=272629760 2020-02-13-raspbian-buster.img /mnt/pi
    

    You’ll see all raspbian files at /mnt/pi. However you can’t use it yet. Because some of the symlinks are broken, you must fix them first.

    In /mnt/pi, run

     $ find . -type l -exec realpath  {} \; |grep 'No such file'
    

    It will show which are broken. Then you can fix them by running:

     $ mkdir /mnt/pi2
     $ cd /mnt/pi2
     $ sudo tar -C /mnt/pi -cf - . | sudo tar --transform 'flags=s;s,^/,/mnt/pi2/,' -xf -
    

    Then /mnt/pi2 is the sysroot folder you’ll use in the next step.

    If the target OS is Ubuntu, you can get an image from https://cloud-images.ubuntu.com/. But that image is in qcow2 format. Please convert it before run fdisk and mount.

     qemu-img convert -p -O raw ubuntu-18.04-server-cloudimg-arm64.img ubuntu.raw
    

    The remaining part is similar to raspbian.

    If the target OS is manylinux2014, you can get it by: Install qemu-user-static from apt or dnf. Then run the docker

    Ubuntu:

     docker run -v /usr/bin/qemu-aarch64-static:/usr/bin/qemu-aarch64-static -it --rm quay.io/pypa/manylinux2014_aarch64 /bin/bash
    

    The “-v /usr/bin/qemu-aarch64-static:/usr/bin/qemu-aarch64-static” arg is not needed on Fedora.

    Then, inside the docker, run

     cd /opt/python
     ./cp35-cp35m/bin/python -m pip install numpy==1.16.6
     ./cp36-cp36m/bin/python -m pip install numpy==1.16.6
     ./cp37-cp37m/bin/python -m pip install numpy==1.16.6
     ./cp38-cp38/bin/python -m pip install numpy==1.16.6
    

    These commands will take a few hours because numpy doesn’t have a prebuilt package yet. When completed, open a second window and run

     docker ps
    

    From the output:

     CONTAINER ID        IMAGE                                COMMAND             CREATED             STATUS              PORTS               NAMES
     5a796e98db05        quay.io/pypa/manylinux2014_aarch64   "/bin/bash"         3 minutes ago       Up 3 minutes                            affectionate_cannon
    

    You’ll see the docker instance id is: 5a796e98db05. Use the following command to export the root filesystem as the sysroot for future use.

     docker export 5a796e98db05 -o manylinux2014_aarch64.tar
    
  4. Generate CMake toolchain file Save the following content as tool.cmake

     SET(CMAKE_SYSTEM_NAME Linux)
     SET(CMAKE_SYSTEM_VERSION 1)
     SET(CMAKE_C_COMPILER aarch64-linux-gnu-gcc)
     SET(CMAKE_CXX_COMPILER aarch64-linux-gnu-g++)
     SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
     SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
     SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
     SET(CMAKE_FIND_ROOT_PATH_MODE_PACKAGE ONLY)
     SET(CMAKE_FIND_ROOT_PATH /mnt/pi)
    

    If you don’t have a sysroot, you can delete the last line.

  5. Run CMake and make

    Append -DONNX_CUSTOM_PROTOC_EXECUTABLE=/path/to/protoc -DCMAKE_TOOLCHAIN_FILE=path/to/tool.cmake to your cmake args, run cmake and make to build it. If you want to build Python package as well, you can use cmake args like:

    -Donnxruntime_GCC_STATIC_CPP_RUNTIME=ON -DCMAKE_BUILD_TYPE=Release -Dprotobuf_WITH_ZLIB=OFF -DCMAKE_TOOLCHAIN_FILE=path/to/tool.cmake -Donnxruntime_ENABLE_PYTHON=ON -DPYTHON_EXECUTABLE=/mnt/pi/usr/bin/python3 -Donnxruntime_BUILD_SHARED_LIB=OFF -Donnxruntime_DEV_MODE=OFF -DONNX_CUSTOM_PROTOC_EXECUTABLE=/path/to/protoc "-DPYTHON_INCLUDE_DIR=/mnt/pi/usr/include;/mnt/pi/usr/include/python3.7m" -DNUMPY_INCLUDE_DIR=/mnt/pi/folder/to/numpy/headers
    

    After running cmake, run

    $ make
    
  6. (Optional) Build Python package

    Copy the setup.py file from the source folder to the build folder and run

    python3 setup.py bdist_wheel -p linux_aarch64
    

    If targeting manylinux, unfortunately their tools do not work in the cross-compiling scenario. Run it in a docker like:

    docker run  -v /usr/bin/qemu-aarch64-static:/usr/bin/qemu-aarch64-static -v `pwd`:/tmp/a -w /tmp/a --rm quay.io/pypa/manylinux2014_aarch64 /opt/python/cp37-cp37m/bin/python3 setup.py bdist_wheel
    

    This is not needed if you only want to target a specfic Linux distribution (i.e. Ubuntu).

Native compiling on Linux ARM device

Easy, slower

Docker build runs on a Raspberry Pi 3B with Raspbian Stretch Lite OS (Desktop version will run out memory when linking the .so file) will take 8-9 hours in total.

sudo apt-get update
sudo apt-get install -y \
    sudo \
    build-essential \
    curl \
    libcurl4-openssl-dev \
    libssl-dev \
    wget \
    python3 \
    python3-pip \
    python3-dev \
    git \
    tar

pip3 install --upgrade pip
pip3 install --upgrade setuptools
pip3 install --upgrade wheel
pip3 install numpy

# Build the latest cmake
mkdir /code
cd /code
wget https://github.com/Kitware/CMake/releases/download/v3.26.4/cmake-3.26.4.tar.gz
tar zxf cmake-3.26.4.tar.gz

cd /code/cmake-3.26.4
./configure --system-curl
make
sudo make install

# Prepare onnxruntime Repo
cd /code
git clone --recursive https://github.com/Microsoft/onnxruntime.git

# Start the basic build
cd /code/onnxruntime
./build.sh --config MinSizeRel --update --build

# Build Shared Library
./build.sh --config MinSizeRel --build_shared_lib

# Build Python Bindings and Wheel
./build.sh --config MinSizeRel --enable_pybind --build_wheel

# Build Output
ls -l /code/onnxruntime/build/Linux/MinSizeRel/*.so
ls -l /code/onnxruntime/build/Linux/MinSizeRel/dist/*.whl

Cross compiling on Windows

Using Visual C++ compilers

  1. Download and install Visual C++ compilers and libraries for ARM(64). If you have Visual Studio installed, please use the Visual Studio Installer (look under the section Individual components after choosing to modify Visual Studio) to download and install the corresponding ARM(64) compilers and libraries.

  2. Use .\build.bat and specify --arm or --arm64 as the build option to start building. Preferably use Developer Command Prompt for VS or make sure all the installed cross-compilers are findable from the command prompt being used to build using the PATH environmant variable.


Mobile

Please see Build for Android and Build for iOS

Web

Please see Build for Web