Add example for handling large datasets

[nginyc]

Currently, the example models load all training examples into memory before conducting training - this is not possible with large datasets.

We want to add a model example that trains on a relatively large dataset where it partially loads the dataset into memory.