Tensorflow parallel optimization of multiple networks with combined loss

Question

I am working on a generator for neural network potentials. An example training data could look as follows: I have three different elements. Each element has its own neural n