decoder_initial_state

Is this some left-over from previous versions of the code ?
https://github.com/JayParks/tf-seq2seq/blob/master/seq2seq_model.py#L359

Gets overwritten at #L393

If attention was not used, decoder_initial_state would simply be (a tiled) encoder_last_state ?