Using the WebNN Execution Provider

This document explains how to use the WebNN execution provider in ONNX Runtime.

Contents

• Basics
  • What is WebNN? Should I use it?
  • How to use WebNN EP in ONNX Runtime Web
• Keep tensor data on WebNN MLTensor (IO binding)
  • Create input tensor from a MLTensor
  • Use pre-allocated MLTensor tensors