Difference between these two implementations of LSTM Autoencoder?

Question

Specifically I\'m trying to understand the return_sequence argument of TensorFlow\'s version of an LSTM layer.

The docs say:

Bool