Setting Number of Bias Inputs for Two Fully Connected RNNs

Question

Below is the architecture of a model I’m implementing. Two generator networks (Generator 1 and Generator 2) are fully connected and only Generator 2 connects to an output unit. Sometimes, input goes through Generator 1 (Input A), and then the output from Generator 1 acts as the input to Generator 2 (Input A’). Other times, input goes through Generator 2 directly (Input B).

Here is my question: I know that when setting up a single neural network, 1 layer of bias should be added as input. Therefore, there will be 1 bias layer added to Input A. However, for Input A’, the input neurons already included the original bias layer from Input A, do I still need to add one more bias layer in addition to Input A’? I’d also like to know the reasoning behind whether to add the additional bias layer or not.