onnx/optimizer

ONNX Optimizer

What it solves

It provides a way to perform arbitrary optimizations on ONNX models to improve their efficiency. By offering a shared library of optimization passes, it prevents different ONNX backend implementations from having to reimplement the same optimizations repeatedly.

How it works

The project is a C++ library that allows users to apply a list of prepackaged optimization passes to an ONNX graph. It can be used via a Python API or a command-line interface to transform an input model into an optimized output model.

Who it’s for

It is designed for developers and engineers working with ONNX models who want to optimize their models for better performance or those who wish to implement and integrate new optimization passes.

Highlights

  • C++ library with Python bindings.
  • Includes a growing list of prepackaged optimization passes.
  • Supports both a Python API and a command-line interface.
  • Allows for the definition and implementation of custom optimization passes.

Related

  • Project
    Artelnics/opennn

    A high-performance C++ library for implementing neural networks used in advanced analytics for business, healthcare, and engineering.

  • Project
    microsoft/onnxruntime-extensions

    A C/C++ library that extends ONNX Runtime with custom operators to handle pre- and post-processing for vision, text, and NLP models within the ONNX graph.

  • Project
    onnx/onnx-mlir

    A compiler that transforms ONNX graphs into optimized, low-runtime code using LLVM/MLIR technology for efficient model deployment.

  • Project
    inisis/OnnxSlim

    A tool for simplifying and slimming ONNX models to reduce operator count and improve inference speed while maintaining accuracy.

  • Project
    onnx/models

    A curated collection of pre-trained, state-of-the-art machine learning models in the ONNX format, enabling cross-framework compatibility and deployment.