Is it possible to apply saliency map or grad-cam with pycox?

Hi !
Many thanks for this great work!
I'm wondering can we have a simple example of applying saliency map or grad-cam on pycox (or torchtuples) models?
or a simple code to demonstrate how to get the loss over the correct scores, and then compute the gradients with a backward pass.

No files or tests are named. Start by reading the pycox and torchtuples model interfaces, then review how PyTorch loss selection and backward passes expose gradients. Done means a simple, documented example showing saliency-map or Grad-CAM use with these models, including how to obtain gradients for the relevant scores.

Written by the indexing model from the issue text.