Separate encoder from autoencoder train model

[ptrblck]

If you separated your encoder and decoder like in this thread, you could just call model.encoder(data) to get the encoder features.