Simplifies running deep learning experiments on bare metal and cloud GPU clusters of any size.