hmm.ipynb

For me this produces a error earlier:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-9-560824d838b4> in <module>()
      9     p0 = np.ones(num_states, dtype=float) / num_states
     10     # No shape, so each state is a scalar tensor
---> 11     states = [pymc3.Categorical('s0', p = p0)]
     12     emissions = [pymc3.Categorical('z0',
     13                                    p = ifelse(eq(states[0], 0), E[0], ifelse(eq(states[0], 1), E[1], E[2])),

/homes/abie/anaconda/envs/testenv/lib/python2.7/site-packages/pymc3-3.0-py2.7.egg/pymc3/distributions/distribution.pyc in __new__(cls, name, *args, **kwargs)
     17             data = kwargs.pop('observed', None)
     18             dist = cls.dist(*args, **kwargs)
---> 19             return model.Var(name, dist, data)
     20         elif name is None:
     21             return object.__new__(cls) #for pickle

/homes/abie/anaconda/envs/testenv/lib/python2.7/site-packages/pymc3-3.0-py2.7.egg/pymc3/model.pyc in Var(self, name, dist, data)
    141         """
    142         if data is None:
--> 143             var = FreeRV(name=name, distribution=dist, model=self)
    144             self.free_RVs.append(var)
    145         else:

/homes/abie/anaconda/envs/testenv/lib/python2.7/site-packages/pymc3-3.0-py2.7.egg/pymc3/model.pyc in __init__(self, type, owner, index, name, distribution, model)
    351             self.tag.test_value = np.ones(
    352                 distribution.shape, distribution.dtype) * distribution.default()
--> 353             self.logp_elemwiset = distribution.logp(self)
    354             self.model = model
    355 

/homes/abie/anaconda/envs/testenv/lib/python2.7/site-packages/pymc3-3.0-py2.7.egg/pymc3/distributions/discrete.pyc in logp(self, value)
    287         k = self.k
    288 
--> 289         return bound(log(p[value]),
    290             value >= 0,
    291             value <= (k - 1),

IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices

I would turn your variables into one vector rather than create the lists. Unfortunately, I hit this issue when I did that here: pymc-devs/pymc#792

I have a very similar model (actually the exact topology which made this example extremely helpful). I am having the same issue as outlined above, but I am not following the suggestion of @twiecki for creating a vector instead of the list.

I changed the lists to np.array everywhere where it is possible and it is not making any difference. Any other suggestions?