AIExplainer
Machine Learning Intermediate 1 min read

What is a data parallelism?

A way of scaling training or inference that replicates an entire model onto multiple devices and then passes a subset of the input data to each device.