ONNX MLIR

The Open Neural Network Exchange implementation in MLIR (http://onnx.ai/onnx-mlir/).

System	Build Status
x86-Linux
s390-Linux
x86-Windows
x86-macOS

Prebuilt Container

An easy way to get started with ONNX-MLIR is to use a prebuilt docker image. These images are created as a result of a successful merge build on the trunk. This means that the latest image represents the tip of the trunk. Currently there are images for amd64, ppc64le and IBM System Z respectively saved in Docker Hub as onnxmlirczar/onnx-mlir-build:amd64, onnxmlirczar/onnx-mlir-build:ppc64le and onnxmlirczar/onnx-mlir-build:s390x. To use one of these images either pull it directly from Docker Hub, launch a container and run an interactive bash shell in it, or use it as the base image in a dockerfile. The container contains the full build tree including the prerequisites and a clone of the source code. The source can be modified and onnx-mlir rebuilt from within the container, so it is possible to use it as a development environment. It is also possible to attach vscode to the running container. An example Dockerfile and vscode configuration files can be seen in the docs folder. The Dockerfile is shown here.

FROM onnxmlirczar/onnx-mlir-build:amd64

WORKDIR /build
ENV HOME=/build
ENV PYENV_ROOT=$HOME/.pyenv
ENV PATH=$PYENV_ROOT/shims:$PYENV_ROOT/bin:$PATH
RUN pyenv global 3.7.0
RUN pyenv rehash

ENV PATH=$PATH:/build/bin
RUN apt-get update
RUN apt-get install -y python-numpy
RUN apt-get install -y python3-pip
RUN apt-get install -y gdb
RUN apt-get install -y lldb
RUN apt-get install -y emacs
WORKDIR /build/.vscode
ADD .vscode /build/.vscode
WORKDIR /build

Prerequisites

gcc >= 6.4
libprotoc >= 3.11.0
cmake >= 3.15.4

At any point in time, ONNX MLIR depends on a specific commit of the LLVM project that has been shown to work with the project. Periodically the maintainers need to move to a more recent LLVM level. Among other things, this requires that the commit string in utils/clone-mlir.sh be updated. A consequence of making this change is that the TravisCI build will fail until the Docker images that contain the prereqs are rebuilt. There is a GitHub workflow that rebuilds this image for the amd64 architecture, but currently the ppc64le and s390x images must be rebuilt manually. The Dockerfiles to accomplish that are in the repo.

Installation on UNIX

MLIR

Firstly, install MLIR (as a part of LLVM-Project):

git clone https://github.com/llvm/llvm-project.git
# Check out a specific branch that is known to work with ONNX MLIR.
cd llvm-project && git checkout d28af7c654d8db0b68c175db5ce212d74fb5e9bc && cd ..

mkdir llvm-project/build
cd llvm-project/build
cmake -G Ninja ../llvm \
   -DLLVM_ENABLE_PROJECTS=mlir \
   -DLLVM_BUILD_EXAMPLES=ON \
   -DLLVM_TARGETS_TO_BUILD="host" \
   -DCMAKE_BUILD_TYPE=Release \
   -DLLVM_ENABLE_ASSERTIONS=ON \
   -DLLVM_ENABLE_RTTI=ON

cmake --build . --target -- ${MAKEFLAGS}
cmake --build . --target check-mlir

ONNX-MLIR (this project)

The following environment variables can be set before building onnx-mlir (or alternatively, they need to be passed as CMake variables):

• LLVM_PROJ_SRC should point to the llvm-project src directory (e.g., llvm-project/).
• LLVM_PROJ_BUILD should point to the llvm-project build directory (e.g., llvm-project/build).

To build ONNX-MLIR, use the following command:

git clone --recursive https://github.com/onnx/onnx-mlir.git

# Export environment variables pointing to LLVM-Projects.
export LLVM_PROJ_SRC=$(pwd)/llvm-project/
export LLVM_PROJ_BUILD=$(pwd)/llvm-project/build

mkdir onnx-mlir/build && cd onnx-mlir/build
cmake ..
cmake --build .

# Run FileCheck tests:
export LIT_OPTS=-v
cmake --build . --target check-onnx-lit

If you are running on OSX Big Sur, you need to add -DCMAKE_CXX_COMPILER=/usr/bin/c++ to the cmake .. command due to changes in the compilers. After the above commands succeed, an onnx-mlir executable should appear in the bin directory.

Installation on Windows

Building onnx-mlir on Windows requires building some additional prerequisites that are not available by default.

Note that the instructions in this file assume you are using Visual Studio 2019 Community Edition. It is recommended that you have the Desktop development with C++ and Linux development with C++ workloads installed. This ensures you have all toolchains and libraries needed to compile this project and its dependencies on Windows.

Run all the commands from a shell started from "Developer Command Prompt for VS 2019".

Protobuf

Build protobuf as a static library.

git clone --recurse-submodules https://github.com/protocolbuffers/protobuf.git
REM Check out a specific branch that is known to work with ONNX MLIR.
REM This corresponds to the v3.11.4 tag
cd protobuf && git checkout d0bfd5221182da1a7cc280f3337b5e41a89539cf && cd ..

set root_dir=%cd%
md protobuf_build
cd protobuf_build
call cmake %root_dir%\protobuf\cmake -G "Visual Studio 16 2019" -A x64 -T host=x64 ^
   -DCMAKE_INSTALL_PREFIX="%root_dir%\protobuf_install" ^
   -DCMAKE_BUILD_TYPE=Release ^
   -Dprotobuf_BUILD_EXAMPLES=OFF ^
   -Dprotobuf_BUILD_SHARED_LIBS=OFF ^
   -Dprotobuf_BUILD_TESTS=OFF ^
   -Dprotobuf_MSVC_STATIC_RUNTIME=OFF ^
   -Dprotobuf_WITH_ZLIB=OFF

call cmake --build . --config Release -- /m
call cmake --build . --config Release --target install -- /m

Before running CMake for onnx-mlir, ensure that the bin directory to this protobuf is before any others in your PATH:

set PATH=%root_dir%\protobuf_install\bin;%PATH%

MLIR

Install MLIR (as a part of LLVM-Project):

git clone https://github.com/llvm/llvm-project.git
# Check out a specific branch that is known to work with ONNX MLIR.
cd llvm-project && git checkout d28af7c654d8db0b68c175db5ce212d74fb5e9bc && cd ..

set root_dir=%cd%
md llvm-project\build
cd llvm-project\build
call cmake %root_dir%\llvm-project\llvm -G "Visual Studio 16 2019" -A x64 -T host=x64 ^
   -DCMAKE_INSTALL_PREFIX="%root_dir%\llvm-project\build\install" ^
   -DLLVM_ENABLE_PROJECTS=mlir ^
   -DLLVM_BUILD_EXAMPLES=ON ^
   -DLLVM_TARGETS_TO_BUILD="host" ^
   -DCMAKE_BUILD_TYPE=Release ^
   -DLLVM_ENABLE_ASSERTIONS=ON ^
   -DLLVM_ENABLE_RTTI=ON ^
   -DLLVM_ENABLE_ZLIB=OFF

call cmake --build . --config Release -- /m
call cmake --build . --config Release --target install -- /m
call cmake --build . --config Release --target check-mlir -- /m

ONNX-MLIR (this project)

The following environment variables can be set before building onnx-mlir (or alternatively, they need to be passed as CMake variables):

• LLVM_PROJ_BUILD: Path to the build directory for LLVM (e.g. c:/repos/llvm-project/build)
• LLVM_PROJ_SRC: Path to the source directory for LLVM (e.g. c:/repos/llvm-project)

This project uses lit (LLVM's Integrated Tester) for unit tests. When running CMake, we can also specify the path to the lit tool from LLVM using the LLVM_EXTERNAL_LIT define but it is not required as long as LLVM_PROJ_BUILD is set correctly.

To build ONNX MLIR, use the following commands:

git clone --recursive https://github.com/onnx/onnx-mlir.git

set root_dir=%cd%

md onnx-mlir\build
cd onnx-mlir\build
call cmake %root_dir%\onnx-mlir -G "Visual Studio 16 2019" -A x64 -T host=x64 ^
   -DCMAKE_BUILD_TYPE=Release ^
   -DCMAKE_PREFIX_PATH=%root_dir%\protobuf_install ^
   -DLLVM_LIT_ARGS=-v ^
   -DLLVM_PROJ_BUILD=%root_dir%\llvm-project\build ^
   -DLLVM_PROJ_SRC=%root_dir%\llvm-project

call cmake --build . --config Release --target onnx-mlir -- /m

To test ONNX MLIR, use the following command:

call cmake --build . --config Release --target check-onnx-lit -- /m

After the above commands succeed, an onnx-mlir executable should appear in the bin directory.

Using ONNX-MLIR

The usage of onnx-mlir is as such:

OVERVIEW: ONNX MLIR modular optimizer driver

USAGE: onnx-mlir [options] <input file>

OPTIONS:

Generic Options:

  --help        - Display available options (--help-hidden for more)
  --help-list   - Display list of available options (--help-list-hidden for more)
  --version     - Display the version of this program

ONNX MLIR Options:
These are frontend options.

  Choose target to emit:
      --EmitONNXBasic - Ingest ONNX and emit the basic ONNX operations without inferred shapes.
      --EmitONNXIR    - Ingest ONNX and emit corresponding ONNX dialect.
      --EmitMLIR      - Lower model to MLIR built-in transformation dialect.
      --EmitLLVMIR    - Lower model to LLVM IR (LLVM dialect).
      --EmitLib       - Lower model to LLVM IR, emit (to file) LLVM bitcode for model, compile and link it to a shared library.

Example

For example, to lower an ONNX model (e.g., add.onnx) to ONNX dialect, use the following command:

./onnx-mlir --EmitONNXIR add.onnx

The output should look like:

module {
  func @main_graph(%arg0: tensor<10x10x10xf32>, %arg1: tensor<10x10x10xf32>) -> tensor<10x10x10xf32> {
    %0 = "onnx.Add"(%arg0, %arg1) : (tensor<10x10x10xf32>, tensor<10x10x10xf32>) -> tensor<10x10x10xf32>
    return %0 : tensor<10x10x10xf32>
  }
}

Troubleshooting

If the latest LLVM project fails to work due to the latest changes to the MLIR subproject please consider using a slightly older version of LLVM. One such version, which we use, can be found here.

Installing third_party ONNX for Backend Tests or Rebuilding ONNX Operations

Backend tests are triggered by make check-onnx-backend in the build directory and require a few preliminary steps to run successfully. Similarily, rebuilding the ONNX operations in ONNX-MLIR from their ONNX descriptions is triggered by make OMONNXOpsIncTranslation.

You will need to install python 3.x if its not default in your environment, and possibly set the cmake PYTHON_EXECUTABLE varialbe in your top cmake file.

You will also need pybind11 which may need to be installed (mac: brew install pybind for example) and you may need to indicate where to find the software (Mac, POWER, possibly other platforms: export pybind11_DIR=<your path to pybind>). Then install the third_party/onnx software (Mac: pip install -e third_party/onnx) typed in the top directory.

On Macs/POWER and possibly other platforms, there is currently an issue that arises when installing ONNX. If you get an error during the build, try a fix where you edit the top CMakefile as reported in this PR: https://github.com/onnx/onnx/pull/2482/files.

Slack channel

We have a slack channel established under the Linux Foundation AI and Data Workspace, named #onnx-mlir-discussion. THis channel can be used for asking quick questions related to this project. A direct link is here.