We should run GMM for smaller batch sizes, e.g. 100.
We should run GMM for smaller batch sizes, e.g. 100.