Standard · Established · Advanced

ONNX

Also known as: Open Neural Network Exchange

An open format for representing trained models so they can be moved between frameworks and runtimes.

What ONNX is

ONNX decouples where a model is trained from where it runs, which matters when training happens in Python but serving happens elsewhere.

How it works

A trained model is exported to the ONNX graph format and executed by a runtime that applies hardware-specific optimisations across CPUs, GPUs and accelerators.

Why it matters

It is the practical path to deploying models on hardware and in languages the training framework does not support.

Common uses

→Cross-platform deployment
→Edge and mobile inference
→Runtime optimisation

Strengths

✓Portable
✓Optimised runtimes for many targets

Watch for

✓Unsupported operators break export
✓Version mismatches are common

Continue exploring

Model Serving Edge AI Inference PyTorch