ONNX Runtime is a cross-platform, high performance scoring engine for ML models. You can get Python bindings for Linux, Windows, Mac on x64 and arm64 platform from pypi. The project does not provide pre-built packages for Raspbery Pi Linux 32bit (ARM32v7) as of today. Instead it has an instruction on how to build ONNX Runtime from the source code.

Building package takes a lot of time. This repository provides pre-built Python ONNX Runtime Python wheel packages for Raspberry pi for convenience.

There is NO WARRANTY of any kind at all.

This directory has built packages.

• wheels/bullseye/
  • Built python wheel packages for Raspberry Pi OS based on Debian Bullseye (Relase date: 2021-10-30~).
• wheels/buster/
  • Built python wheel packages for Raspbian Buster and Raspberry Pi OS based on Debian Buster.
• wheels/stretch/
  • Built python wheel packages for Raspbian Stretch.

See your raspberry pi's /etc/debian_version file to find your OS version.

Wheels are built on Raspberry Pi with this procedure since v1.6.0.
Building ONNX Runtime wheel for Raspberry Pi 32bit on Docker with Qemu stopped working since 2020 summer.

A script to build ONNX Runtime wheels for armv7-linux on Docker with Qemu.
It builds wheels for Python 3.8, 3.9, 3.10 and 3.11.
You need docker to run it.

Usage:

$ sudo ./build.sh

A Dockerfile template used in build.sh.