Text this: A snapshot of parallelism in distributed deep learning training