Did you make it work with another package? Would be interested to hear how it went.
Regarding @jungpenglao 's comment that it’s not trivial to allow observed Deterministic variables, what’s the blocker?
Say we observe \tilde{x}_i = f(x, a), where x \sim N(0, b^2), a and b are stochastic variables and \tilde{x} is the observed variable. If we invert f we can write f^{-1}(\tilde{x}_i, a) \sim N(0, b^2). Given values for a and b we could calculate the likelihood easily, and also its partial derivatives with a and b (or am I missing something).
So is it possible to enter a Deterministic variable as the observed value of a stochastic variable? I might be missing something here.