How to train two half cheetha opengym environments using same DDPG ( neural networks) simultaneously in reinforcement learning?

Question

How to train two half cheetha opengym environments with single neural network of DDPG in reinforcement learning? Objective is to implement single DDPG (one main actor, one main critic, one target actor, one target critic neural networks) network training multiple half cheetha environments, half cheetha has 6 continous action space and 17 continuous state space.

How the input/output layer Neurons will look while training two half cheetha environment simultaneously, for two half cheetahs it would be 12 out neurons(6 each) in that case.For fully connected network the inputs of one half cheetah env will influence other half cheetha and vice versa.