Un-0

Un-0 is an image-generation model built on Kuramoto dynamics. It generates an image by integrating the phase dynamics of a population of coupled oscillators starting from random phases to a set of image latent phases which are decoded to produce an image. It comes from Unconventional, Inc., where we research dynamical systems as a computing substrate that maps onto analog and physical hardware, pointing toward dramatically lower energy for AI than today's digital accelerators.

This repository hosts the released class-conditional checkpoints for two independently trained models, one trained on CIFAR-10 (32×32), the other ImageNet (64×64). The reference implementation, training recipe, and evaluation code live at github.com/unconv-ai/Un-0.