How do I train a pretrained JIT model?

I am trying to perform transfer learning using one of the silerio STT models located in the torch hub. The model loads as a JIT model which I am entirely unfamiliar with.

Does anyone know how I can work with this model as I would a traditional pytorch module? I want to be able to freeze / unfreeze layers, remove layers, etc, as well as use it in a training loop.

Thank you for your help!