I do not really understand the obviously (or actually the same?) training procedures for training a LSTM encoder-decoder.
on the one hand in the tutorial they use a f
