it seems that when a model contains a HyperGeometric distribution, model_to_graphiz() fails due to AttributeError.
Could it be that some of the parameters are stored as Numpy NdArray instead of Theano Tensor (i.e. missing eval() method) ?
I am a bit clueless here, any insight?
with pm.Model() as model: x = pm.HyperGeometric('test', 5, 10, 100) pm.model_to_graphviz(model)