How does TensorFlow's MultiRnnCell work?

Question

Could someone help to explain the inner mechanism of TensorFlow's tf.contrib.rnn.MultiRnnCell?

For example, if I wanted to stack up two basic RNN cells into a MultiRnnCell, what would be the input and output of each basic RNN cell?

I would like to know the details of how it works.

Answer 1

Study this blog post as well as the provided implementation. It describes in detail how use MultiRNNCell to stack multiple RNN cells.