any way to do reconstructions?

[jcpeterson]

Do you already have code to do reconstructions? I think it would be enough to have code to get image encodings for images given a trained model, since there's already enough code for generating images given a random sample using numpy...etc.

[jcpeterson]

Checking in on this