[jph00]

A couple of weeks ago a new paper came out that shows how to train a high quality language model on a single GPU in one day.

https://arxiv.org/abs/2212.14034