# enables you to change the behavior of ``torch.compile`` across different # calls to your model without having to reapply ``torch.compile`` to your model. # This recipe provides some examples on how ...