Deploy ExecuTorch firmware on NXP FRDM i.MX 93 for Ethos-U65 acceleration
Introduction
Understand ExecuTorch deployment on NXP with Ethos-U
Boot the NXP FRDM i.MX 93 board
Set up the ExecuTorch build environment
Build and install ExecuTorch
Build ExecuTorch models for Ethos-U65
Build Cortex-M33 firmware for ExecuTorch
Deploy and test on FRDM-IMX93
Next Steps

  Log an issue

  Fork and edit

  Discuss on Discord

About this Learning Path

Skill level:  Introductory
Reading time:  2 hrs
Last updated:  13 Aug 2026
Skill level:
  Introductory
Reading time:
  2 hrs
Last updated:
  13 Aug 2026
Authors:Waheed Brown, Arm
Fidel Makatia Omusilibwa
Arm IP:Cortex-A Cortex-M Ethos-U
Tags: ML Linux macOS Baremetal Python PyTorch ExecuTorch Arm Compute Library GCC
Authors:
  • Waheed Brown, Arm
  • Fidel Makatia Omusilibwa
Arm IP:
Cortex-A Cortex-M Ethos-U
Tags:
ML Linux macOS Baremetal Python PyTorch ExecuTorch Arm Compute Library GCC

Who is this for?

This is an introductory topic for developers and data scientists new to TinyML who want to observe ExecuTorch performance on a physical device.

What will you learn?

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

  • Bring up a custom ExecuTorch `executor_runner` firmware on the FRDM i.MX 93 Cortex-M33 using Linux RemoteProc
  • Compile an ExecuTorch `.pte` model for Ethos-U65 and run inference with NPU acceleration
  • Understand how heterogeneous Arm systems split responsibilities across application cores, microcontrollers, and NPUs

Prerequisites

Before starting, you will need the following:

  • An NXP FRDM i.MX 93 development board
  • A USB Mini-B to USB Type-A cable, or a USB Mini-B to USB Type-C cable
  • Completion of Use Linux on an NXP FRDM i.MX 93 board (Linux setup, login access, and file transfer)
  • Basic knowledge of Machine Learning concepts
  • A host computer to compile ExecuTorch libraries

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 run an ExecuTorch model on the FRDM i.MX 93 Cortex-M33 with Ethos-U65 acceleration. First, you’ll prepare a build environment, compile executor_runner firmware and a U65-targeted .pte model. Then, you’ll transfer both artifacts to the board. You’ll start the firmware from Linux with RemoteProc and observe inference across the Cortex-M33 and NPU.

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
?
Which USB connector should I use for the serial console, and what do I need on macOS?
Use the DEBUG USB-C connector on the board. On macOS, install the Silicon Labs USB-to-UART driver and a serial terminal such as picocom (for example, install it with brew install picocom).
Why build ExecuTorch inside a Docker container on macOS?
Building inside a Docker container provides an Ubuntu build environment that matches the toolchains used in this Learning Path, and avoids gaps in macOS-native cross-compilers. The container is only for building and produces prebuilt ExecuTorch libraries and .pte files that you’ll copy to the FRDM i.MX 93.
After installing ExecuTorch, how do I confirm the package is available?
Run pip list | grep executorch and check that executorch appears in the output.
What artifacts must be present before starting the firmware with Linux RemoteProc?
Copy the U65-compiled .pte model and the executor_runner ELF to the board. The runner loads the .pte, prepares buffers, and invokes the NPU or CPU backend on the Cortex-M33.
What output should I expect when the MobileNet V2 model compiles successfully?
The compilation output reports the number of NPU operators and their coverage. For the example MobileNet V2 model, expect an output reporting 100% NPU utilization.
Next