Creating networks using tf-agents for reinforcement learning


I wish to implement an encoder, decoder architecture for actor network using tensorflow-agents for td3. Looking at the tensorflow-agents actor-rnn-network code, I see that the networks is a keras subclass architecture. I am not very familiar with it, so if I need to create an encoder/decoder actor network, do I need to create two separate classes as per this link:
I am stuck with this part since if I create an encoder-decoder class separately, I will also have to change the target networks for them separately. Since I am new to this, a clarification in my misconceptions would help a lot.
Any suggestion is appreciated!
Thanks a lot!

© Copyright 2013-2019 Analytics Vidhya