ONNX export

Converting a trained model into the framework-independent ONNX format for portable inference deployment.

ONNX (Open Neural Network Exchange) is a framework-independent format for a trained model's architecture and weights. ONNX export converts a model trained in one framework (e.g. PyTorch) into this shared format so it can be run for inference by any ONNX-compatible runtime, decoupling deployment from the training framework.

Examples

A model trained in PyTorch is exported to ONNX so it can be served in a production system built on an entirely different runtime.