ML-Compiler-Bridge is a compiler agnostic library to aid in ML-Enabled Compiler Optimizations. ML-Compiler-Bridge supports both training and inference scenarios. Library exposes Python and C/C++ APIs to interface with the Python-based ML models and a C/C++ compiler. This design allows ML model development within a traditional Python framework while making end-to-end integration with an optimizing compiler possible and efficient.

This repo contains the source code and relevant information described in our paper, "The Next 700 ML-Enabled Compiler Optimizations" (arxiv). Please see here for documentation and other details.

‍The Next 700 ML-Enabled Compiler Optimizations, S. VenkataKeerthy, Siddharth Jain, Umesh Kalvakuntla, Pranav Sai Gorantla, Rajiv Shailesh Chitale, Eugene Brevdo, Albert Cohen, Mircea Trofin and Ramakrishna Upadrasta. CC 2024.

Features

Unified Framework: Comes with a suite of two inter-process and two in-process model runners and three serialization-deserialization mechanisms to support interleaved and non-interleaved communication between models and compiler.
Multi-language Support: Exposes C++ and C APIs to interface model runners and serializers with the compilers and Python APIs to interface inter-process model runners with ML models.
Compiler and ML-Framework Independence: Provides compiler and ML-Framework independent APIs, and supports easier integration with compilers like LLVM, MLIR, and Pluto and ML Frameworks like TensorFlow, PyTorch, JAX, etc.
Deeper Integration: Enables deeper integration of ML models within the compiler in a framework-independent manner to support easier inference in case of ML driven compiler optimizations.

Requirements

cmake (>= 3.10)
GNU Make (4.2.1)
LLVM (10.X) - src, release
Python (3.10), C++17
gRPC v1.58 and protobuf v23.4 - for gRPC Model Runner
- Building GRPC from Source: Please follow Build GRPC with cmake v1.58 (protobuf v23.4) to build GRPC from source.
- In the above tutorial setting DCMAKE_INSTALL_PREFIX may not be necessary and the default install prefix can be used.
- The following dependencies will be required for Python: pip install grpcio-tools.
ONNXRuntime v1.13.1
TensorFlow - for TF Model Runner (AOT flow)
- Tested with TensorFlow 2.13.0
Other python requirements are available in mlbridge.yml
- Conda/Anaconda based virtual environment is assumed

(Experiments are done on an Ubuntu 20.04 machine)

Setup

ML-Compiler-Bridge can be built as a stand-alone library to generate .a files that can in turn be linked with any compiler.

mkdir build && cd build
cmake [-DCMAKE_BUILD_TYPE=Release|Debug] [-DCMAKE_INSTALL_PREFIX=<Install_path>] [-DMLBRIDGE_ENABLE_TEST=ON|OFF] -DONNXRUNTIME_ROOTDIR=<Path to ONNX install dir> -DPROTOS_DIRECTORY=<Path to protobuf files> -DTENSORFLOW_AOT_PATH=<Path to TensorFlow pip install dir> ../
make -j [&& make install]
pip install compilerinterface

This process would generate libMLCompilerBridge.a and libMLCompilerBridgeC.a libraries under build/lib directory, required headers under build/include directory. libMLCompilerBridgeC.a exposes C APIs for using with C-based compilers like Pluto, where as libMLCompilerBridge.a exposes C++ APIs that can be used with any compiler written in C++.

Python end points are available under CompilerInterface. They can be downloaded as a package from pypi.

To ensure the correctness, run make verify-all. This would need enabling tests in cmake (-DMLBRIDGE_ENABLE_TEST=ON) and PROTOS_DIRECTORY should point to test/protos.

Using ML-Compiler-Bridge with LLVM and MLIR

ML-Compiler-Bridge can be integrated and built along with the LLVM project. This can be done by adding this repository as a new project and setting LLVM_MLBRIDGE option to ON.

You can check the CMakeLists.txt of the ml-llvm-project repository which demonstrates such an integration.

The passes that need to make use of this library can then just link with LLVMMLBridge.

Example CMakeLists.txt of an LLVM pass that would use the library is shown below.

 .txt
add_llvm_component_library(LLVMMLPass
pass.cpp
ml.cpp
 
ADDITIONAL_HEADER_DIRS
  ${CMAKE_CURRENT_SOURCE_DIR}/includes
 
DEPENDS
  LLVMMLBridge
  intrinsics_gen
)
target_link_libraries(LLVMMLPass PRIVATE LLVMMLBridge)

To use TensorFlow AOT Model Runner, you need to make use of tf_find_and_compile method exposed in cmake/modules/TensorFlowCompile.cmake in the CMakeLists.txt of your pass with appropriate arguments. An example of integrating TF AOT Model with inlining pass is shown here.

Artifacts

Libraries are autogenerated for every relevant check-in with GitHub actions. Such generated artifacts are tagged along with the successful runs of [Publish]() action.

Citation

@inproceedings{venkatakeerthy-2024-MLCompilerBridge,
author = {VenkataKeerthy, S. and Jain, Siddharth and Kalvakuntla, Umesh and Gorantla, Pranav Sai and Chitale, Rajiv Shailesh and Brevdo, Eugene and Cohen, Albert and Trofin, Mircea and Upadrasta, Ramakrishna},
title = {The Next 700 ML-Enabled Compiler Optimizations},
year = {2024},
isbn = {9798400705076},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3640537.3641580},
doi = {10.1145/3640537.3641580},
booktitle = {Proceedings of the 33rd ACM SIGPLAN International Conference on Compiler Construction},
pages = {238–249},
numpages = {12},
keywords = {Machine Learning for Compiler Optimizations, ONNX, Pipes, TensorFlow AOT, gRPC},
location = {<conf-loc>, <city>Edinburgh</city>, <country>United Kingdom</country>, </conf-loc>},
series = {CC 2024}
}

Contributions

Please feel free to raise issues to file a bug, pose a question, or initiate any related discussions. Pull requests are welcome :)

License

ML-Compiler-Bridge is released under Apache 2.0 license with LLVM Exceptions. See LICENSE file for more details.