Improving performance of item response model in Turing.jl

Have you tried using MvNormal priors, specifically using FillArrays and I from LinearAlgebra as in this post? Regularized horseshoe prior