Providing initial values in init_params for compositional sampling

Hi there,

when I’m passing initial values init_vals = [A, init_beta, init_eta, init_sigma, init_z] to the init_params argumenet in sampling() everything works as expected:
chain = sample(model, HMC(), MCMCThreads, ... , init_params = Iterators.repeated(init_vals) )

When I extend the model to include a Bernoulli on the ss parameter I change the sampling statement to:
chain = sample(model, Gibbs(HMC(), MH(:ss)), ... )

How do I include the initial value for ss?
Adding it into the init_vals vector does not seem to work.
It returns
nested task error: Provided initial value doesn't match the dimension of the model
This is not surprising, as the init_ss can’t be assigned to just the MH sampler.

Any insights?

See this unanswered related question