Implementing Cross-Validation in Edward Mixture Models

Hi everyone,

I am trying to implement Cross-validation with some kind of mini-batch sampling in the Edward Mixture Model. I have a very large dataset and I am trying to achieve some kind of evaluation measure for how many clusters to go with.
Does it make sense to you, to evaluate the optimal number of clusters through CV and mini-batches when having a very large dataset (I have concluded it to be the best option), and does anyone have any success implementing this?

I would be very grateful!