Enzyme is great and from some testing it looks like the most recent versions will work with most if not all the geometric models we currently use. Now because enzyme doesn't fully have its own rule system I can't (yet) fully switch over but I can add special rrules for geometric models to use Enzyme. This should happen soon. Especially for geometric polarized fits where we probably want to have faster gradients.
Enzyme is great and from some testing it looks like the most recent versions will work with most if not all the geometric models we currently use. Now because enzyme doesn't fully have its own rule system I can't (yet) fully switch over but I can add special rrules for geometric models to use Enzyme. This should happen soon. Especially for geometric polarized fits where we probably want to have faster gradients.