Welcome to the ExecuTorch Documentation

ExecuTorch is PyTorch’s solution to training and inference on the Edge.

Key Value Propositions

• Portability: Compatibility with a wide variety of computing platforms, from high-end mobile phones to highly constrained embedded systems and microcontrollers.

• Productivity: Enabling developers to use the same toolchains and Developer Tools from PyTorch model authoring and conversion, to debugging and deployment to a wide variety of platforms.

• Performance: Providing end users with a seamless and high-performance experience due to a lightweight runtime and utilizing full hardware capabilities such as CPUs, NPUs, and DSPs.

ExecuTorch provides support for:

• Strong Model Support LLMs (Large Language Models), CV (Computer Vision), ASR (Automatic Speech Recognition), TTS (Text To Speech)

• All Major Platforms Android, Mac, Linux, Windows

• Rich Acceleration Support Apple, Arm, Cadence, MediaTek, Qualcomm, Vulkan, XNNPACK

Documentation Navigation

Introduction

• Overview

• How it Works

• Getting Started with Architecture

• Concepts

Usage

• Getting Started

• Using Executorch Export

• Using Executorch on Android

• Using Executorch on iOS

• Using Executorch with C++

• Runtime Integration

• Troubleshooting

• Building from Source

• FAQs

Examples

• Android Demo Apps

• iOS Demo Apps

Backends

• Overview

• XNNPACK

• Core ML

• MPS

• Vulkan

• ARM Ethos-U

• Qualcomm

• MediaTek

• Cadence

Developer Tools

• Overview

• Bundled IO

• ETRecord

• ETDump

• Runtime Profiling

• Model Debugging

• Model Inspector

• Memory Planning Inspection

• Delegate Debugging

• Tutorial

Runtime

• Overview

• Extension Module

• Extension Tensor

• Running a Model (C++ Tutorial)

• Backend Delegate Implementation and Linking

• Platform Abstraction Layer

Portable C++ Programming

• PTE File Format

API Reference

• Export to Executorch API Reference

• Executorch Runtime API Reference

• Runtime Python API Reference

• API Life Cycle

• Javadoc

Quantization

• Overview

Kernel Library

• Overview

• Custom ATen Kernel

• Selective Build

Working with LLMs

• Llama

• Llama on Android

• Llama on iOS

• Llama on Android via Qualcomm backend

• Intro to LLMs in Executorch

Backend Development

• Delegates Integration

• XNNPACK Reference

• Dependencies

• Compiler Delegate and Partitioner

• Debug Backend Delegate

IR Specification

• EXIR

• Ops Set Definition

Compiler Entry Points

• Backend Dialect

• Custom Compiler Passes

• Memory Planning

Contributing

• Contributing