Summary
Multi-GPU machines are becoming much more common. Training deep learning models across multiple-GPUs is something that is often discussed.
But what about preprocessing data by leveraging the resources of more than a single GPU?
NVTabular + Dask makes this super easy!