Behrens' Bayesian Learner Model : How to share parameters between steps?

I am very interested to give Turing a try actually, so I just might.