Bidirectional LSTM in seq2seq model

Question

I am trying to build a Encoder decoder bidirectional lstm model for my text summarization task. I am facing problems with input shape. Error: An initial_state was p