Difference between validation and training loss in Tensorflow with custom loss function

Question

My Tensorflow model consists of two seperate models which each result in a Dense(1) layer and get summed up in the loss function to be compared to only one valu