How to get the weight matrices of intermediate layers in Bidirectional RNNs during training in keras?

Question

I understand, that with popular deep learning API Keras, one can have access to intermediate layer features using Functional API, but how to get the weights of these interme