How to add a classifier in a Unet style model in fastai framework

HI I am working on a kaggle competition, where we are trying to predict which pixel contains salt. It is a segmentation problem. I am using Unet model similar to what is taught in the lesson 14. I am trying to add a classifier from the encoder module and add that to the loss function. The output of the model is a tuple where the first element contains the segmented output and the second contains the classifier output. Inside the Stepper class, the framework checks if the output is a tuple and if it is tuple then passes only the 1st value to the loss function. Looking for ideas on how to include the classifier output to the loss function.