Implementing multi-output classification

Does anyone have code they could share implementing a fastai model with multiple outputs? i.e. containing multiple softmax outputs that each on their own sum to 1.

For example if I had pictures of people and I wanted predict both their hair color and age bracket using the same model.