Thomas Christie
Thomas Christie
Great, thanks, that worked! And yeah, I think having `nnx.jvp` would probably be a bit more intuitive - I actually searched to see if this existed whilst debugging.
@thomaspinder have just added some skeleton code for adding priors to kernel parameters. As suggested, I was going to deal with priors by adding a `prior` field to the `param_field`...
Will we be wanting to enable users to specify whether their prior is over the *constrained* or *unconstrained* parameters, or just force users to place the prior over the *constrained*...