Fitting a distribution with custom functions

Ah OK. It looks like I need the partial derivatives with respect to every parameter and only one of them is easy to compute. I’ll try switching to a non-gradient sampling method.