Can not correctly sample high dimensional RVs from customized likelihood

Questions
2

Hello,

I encountered a similar error, did you figure out what might happen in your code? My code is something like below:

        with pm.Model() as m:
            sigma = pm.Uniform('sigma',lower=0,upper=1)
            nc = pm.HalfNormal('nc',sigma=sigma,observed=y)
            nc_hat = pm.Deterministic('nc_hat',pm.math.sum(nc)/len(y))
            psi = pm.Beta('alpha',alpha=2,beta=nc_hat*10)
            mu = pm.Gamma('beta',alpha=nc_hat*10,beta=1)
            c = pm.ZeroInflatedPoisson('c',psi=psi,mu=mu,observed=x)
        with m:
            step = pm.NUTS()
            trace = pm.sample(100,step=step,return_inferencedata=False,cores=1)

Error:

  File "/data/salomonis2/software/snaf/gtex.py", line 369, in tumor_specificity
    c = pm.ZeroInflatedPoisson('c',psi=psi,mu=mu,observed=x)
  File "/data/salomonis2/LabFiles/Frank-Li/refactor/neo_env/lib/python3.7/site-packages/pymc3/distributions/distribution.py", line 121, in __new__
    dist = cls.dist(*args, **kwargs)
  File "/data/salomonis2/LabFiles/Frank-Li/refactor/neo_env/lib/python3.7/site-packages/pymc3/distributions/distribution.py", line 130, in dist
    dist.__init__(*args, **kwargs)
TypeError: __init__() missing 1 required positional argument: 'theta'