Edge AI on Arm: PyTorch and ExecuTorch rock-paper-scissors
Introduction
Set up your environment
Train and Test the rock-paper-scissors Model
Run the model on Corstone-320 FVP
Next Steps

  Log an issue

  Fork and edit

  Discuss on Discord

About this Learning Path

Skill level:  Introductory
Reading time:  1 hr
Last updated:  13 Aug 2026
Skill level:
  Introductory
Reading time:
  1 hr
Last updated:
  13 Aug 2026
Author:Dominica Abena O. Amanfo
Arm IP:Cortex-M Ethos-U
Tags: ML Linux tinyML Computer Vision Edge AI CNN PyTorch ExecuTorch
Author:
  • Dominica Abena O. Amanfo
Arm IP:
Cortex-M Ethos-U
Tags:
ML Linux tinyML Computer Vision Edge AI CNN PyTorch ExecuTorch

Who is this for?

This is an introductory topic for machine learning developers who want to deploy TinyML models on Arm-based edge devices using PyTorch and ExecuTorch.

What will you learn?

Upon completion of this Learning Path, you will be able to:

  • Train a small Convolutional Neural Network (CNN) for image classification using PyTorch
  • Use synthetic data generation for training a model when real data is limited
  • Convert and optimize a PyTorch model to an ExecuTorch program (`.pte`) for Arm-based devices
  • Run the trained model locally as an interactive mini-game to demonstrate inference

Prerequisites

Before starting, you will need the following:

Summary

AI-assisted

This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
You’ll build a rock-paper-scissors prototype with PyTorch and ExecuTorch on Arm. First, you’ll create a lightweight image-classification CNN, train it, export a .pte program, and run the command-line mini-game. Then, you’ll compile the model with the Ahead-of-Time Arm compiler and run it on the Corstone-320 Fixed Virtual Platform with Ethos-U delegation.

Frequently asked questions

AI-assisted

These FAQs were drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
Where should I put the rock–paper–scissors script before training?
Place the script in the ExecuTorch repository under the Arm examples directory at $HOME/executorch/examples/arm.
How do I launch the mini‑game after training?
Run the script with the play option. The mini‑game uses the best weights found on disk.
How do I know the export to ExecuTorch worked?
A successful export produces a .pte file. Check that the file is created in the expected location after running the export step.
Do I need a dataset or camera for training?
You don’t need an external dataset or camera because you’ll use synthetic data generation for training. The model learns to classify images of the letters R, P, and S.
Which log messages confirm that the model starts on the Corstone-320 FVP?
Check the FVP output for the model loading and Running method forward messages. An EthosUBackend.cpp initialization message confirms that the Ethos-U backend starts during inference.
Next