Sampling from pretrained pytorch normalizing flow

Misc.
1

I have a pre-trained normalizing flow built with pytorch that I would like to sample as a prior in my inference using NUTS. Is that possible?