Keras regression | Get a single prediction from a model fitted with multiple y parameters

Question

I have the following code to create a custom loss function using an additional feature from the dataset (which is not an input feature) as an argument. The code works fine w