Batch Data Loading and Processing

[manasRK]

Any plan to convert some of the code base (like the LSTM/NER) to allow for minibatch processing?
Currently, all of them take one instance at a time. Any pointers in that regard would be useful as well.

[rguthrie3]

Typically, minibatching adds a substantial amount of confusion in the implementation, since one has to keep track of what each axis of your tensors means. Since this is a tutorial, I intentionally did not use minibatching because I wanted to avoid making it substantially more confusing without much return on educational value. So I don't plan to do anything about minibatching in this tutorial. I don't have any specific tips for implementing minibatching. It is just about staying organized and clean with how you arrange your tensors.