onnx: an open source format for AI models that enables interoperability between different frameworks and hardware

What it solves

ONNX (Open Neural Network Exchange) addresses the problem of interoperability between different AI frameworks. It allows developers to move models between different tools, frameworks, and hardware, streamlining the path from research to production by providing a common format for AI models.

How it works

ONNX defines an open source format for AI models, covering both deep learning and traditional machine learning. It uses an extensible computation graph model, standard data types, and a set of built-in operators to represent the model's structure and logic. The project currently focuses on the capabilities needed for inferencing (scoring).

Who it’s for

AI developers who need to move their models across different frameworks, tools, and hardware to optimize performance or evolve their project's toolset.

Highlights

• Open source format for both deep learning and traditional ML models.
• Extensible computation graph model with standard data types and built-in operators.
• Widely supported across various frameworks, tools, and hardware.
• Includes programming utilities for shape and type inference, graph optimization, and opset version conversion.