Optimize ExecuTorch fallback with CMSIS-NN on Cortex-M
Introduction
Understand and prepare the environment
Lower a model to the Cortex-M backend
Add Ethos-U delegation and CPU fallback
Inspect the generated CMSIS Pack selection
Understand the reusable CMSIS project
Build and run the mixed application
Next Steps
Optimize ExecuTorch fallback with CMSIS-NN on Cortex-M
Who is this for?
This Learning Path is for embedded machine learning developers building ExecuTorch applications for Cortex-M CPUs and Ethos-U NPUs.
What will you learn?
After completing this Learning Path, you'll be able to perform the following tasks:
- Export an ExecuTorch graph with Ethos-U, CMSIS-NN, and portable-kernel execution regions.
- Lower quantized CPU fallback operations to CMSIS-NN-backed Cortex-M operators.
- Trace the exported operators to selected components in the published ExecuTorch 1.5.1 and CMSIS-NN Packs.
- Build and run the application on a Corstone-300 Ethos-U55 Fixed Virtual Platform (FVP).
Prerequisites
Before starting, you'll need the following prerequisites:
- A Windows or Linux host machine, or an Apple silicon Mac running macOS 15 or later
- Visual Studio Code installed
- Familiarity with PyTorch
- Basic familiarity with Cortex-M processors and Ethos-U NPUs
Summary
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.
You’ll build and validate an ExecuTorch application that runs across Ethos-U and Cortex-M, with CMSIS-NN accelerating its fallback convolution. First, you’ll export the model for Cortex-M alone, then add Ethos-U delegation and identify the portable
topk operation. You’ll trace the exported operations to versioned CMSIS Pack components and follow the solution and generated AI layer into the firmware build. Finally, you’ll verify the compiled source paths and run the application on the Corstone-300 Ethos-U55 FVP.