Get prediction result into density function to compute log likelihood on test data

Question

I am working on count models and I am doing various regressions (Poisson, Negative Binomial, ZeroInflated, etc.)
I have a training set (on which I fit the parameters)