There are two variations, one with Fn and one without. There's a huge performance difference between the two, so it's appealing as a target for optimizing Fn.
The only API change this has is the addition of a simple paramVector to Continuous; this is not any different for now from just making a List of param, but it provides a hook for later optimizations.
PR for awareness but this is low-risk enough I'm going to self-merge.
This adds a GLMM benchmark based on https://github.com/stan-dev/example-models/blob/master/BPA/Ch.04/GLMM_Poisson2.stan (which in turn is taken from https://www.vogelwarte.ch/de/projekte/publikationen/bpa/ ).
There are two variations, one with Fn and one without. There's a huge performance difference between the two, so it's appealing as a target for optimizing Fn.
The only API change this has is the addition of a simple
paramVector
toContinuous
; this is not any different for now from just making aList
ofparam
, but it provides a hook for later optimizations.