A collection of really small language models pre-trained from scratch with open-data. Ideal for use in experimentation and evaluations.