Caesar AI Atlas

Data Parallelism

Caesar AI Atlas Definition

Data parallelism is a scaling technique in which the same model is replicated across multiple devices while different subsets of input data are processed in parallel. It can accelerate training or inference and support larger batch sizes, provided the model can fit on each device.

Other Definitions

Data Parallelism Source

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. Data parallelism can enable training and inference on very large batch sizes; however, data parallelism requires that the model be small enough to fit on all devices. Data parallelism typically speeds training and inference. See also model parallelism.

Data Parallelism Source

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. Data parallelism can enable training and inference on very large batch sizes; however, data parallelism requires that the model be small enough to fit on all devices. Data parallelism typically speeds training and inference.

Related Terms