February 28, 2025

Setting up a TinyML dev environment on Apple Silicon

TinyMLSetupTutorial

Apple Silicon Macs are excellent TinyML development machines — but the toolchain has enough arm64 / x86_64 gotchas to waste a morning if you don’t know where to look.

This is the setup I use in 2025. Everything below has been tested on an M3 MacBook Pro running macOS Sequoia 15.

Prerequisites

# Homebrew (if not already installed)
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Python 3.11 via pyenv (avoid system Python)
brew install pyenv
pyenv install 3.11.9
pyenv global 3.11.9

TensorFlow Lite for Microcontrollers

The tflite-micro library ships as a pure C++ source tree — no pip package. Clone it and use the provided Makefile to generate a project for your target.

git clone https://github.com/tensorflow/tflite-micro.git
cd tflite-micro
make -f tensorflow/lite/micro/tools/make/Makefile \
  TARGET=arduino \
  generate_arduino_zip_library

The output gen/arduino_zip_lib/tensorflow_lite.zip can be imported directly into Arduino IDE via Sketch → Include Library → Add .ZIP Library.

Edge Impulse CLI

npm install -g edge-impulse-cli
edge-impulse-daemon   # follow prompts to connect your device

Apple Silicon note: the edge-impulse-cli binary includes a bundled native addon. If you see an Unsupported architecture error, run npm install -g edge-impulse-cli --target_arch=arm64 explicitly.

Arduino IDE 2.x

Download the .dmg directly from arduino.cc — the Homebrew cask sometimes lags behind. Under Preferences → Additional Boards Manager URLs, add the Nicla Vision and ESP32 board definitions:

https://downloads.arduino.cc/packages/package_index.json
https://raw.githubusercontent.com/espressif/arduino-esp32/gh-pages/package_esp32_index.json

Verify the full chain

Connect an Arduino Nicla Vision, open the ei_image_classification example from the Edge Impulse library, and flash. If you see inference output over Serial at 115200 baud, the chain is working end-to-end.

← Notes