How to use the bidirectional_dynamic_rnn with pytorch?

1

I want to use the initial_state.

            outputs, states = tf.nn.bidirectional_dynamic_rnn(cell_fw, cell_bw,encoder_conv_output,sequence_length=input_lengths, initial_state_fw=initial_state_fw,initial_state_bw=initial_state_bw,dtype=tf.float32)