LightOnML documentation

LightOnML is a high level machine learning-oriented library that allows to control LightOn’s Optical Processing Units (OPUs). Our current product performs random projections using light.

LightOn’s OPUs are available through LightOn’s Cloud service, or physically with a LightOn Appliance .

The library is developed for Python 3, with interfaces for scikit-learn and pytorch to offer to ML researchers and practitioners a familiar and user-friendly environment.

It is available on Github and PyPI, so it can be installed with pip install lightonml.


  • Run large-scale non-linear and linear random projections using LightOn’s Aurora OPUs

  • Simulate these projections on any machine without access to an OPU

  • Encode input data in a binary form using various encoders, for OPU input

Below shows a minimal usage example of lightonml’s OPU class:

OPU library showcase

Version 1.4.2