Error using pm.Data to supply mu parameter to Normal

[rpgoldman]

Description of your problem

When I create a pm.Data with a vector of values for the mean of a Normal, I get a type error. When I supply the same vector of values without packaging into pm.Data, this works fine.

Please provide a minimal, self-contained, and reproducible example.

import numpy as np
import pymc3 as pm

# the following gives: TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''
with pm.Model():
    x = pm.Data("x", [1.0, 2.0, 3.0])
    y = pm.Normal("y", mu=x, shape=3)

# same type error
with pm.Model():
    x = pm.Data("x", np.array([1.0, 2.0, 3.0]))
    y = pm.Normal("y", mu=x, shape=3)

# works fine
with pm.Model():
    x = np.array([1.0, 2.0, 3.0])
    y = pm.Normal("y", mu=x, shape=3)

Please provide the full traceback.

>>> Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/tmp/data_mu_bug.py", line 11, in <module>
  File "/Users/rpg/src/pymc3/pymc3/distributions/distribution.py", line 66, in __new__
    return model.Var(name, dist, data, total_size)
  File "/Users/rpg/src/pymc3/pymc3/model.py", line 998, in Var
    total_size=total_size, model=self)
  File "/Users/rpg/src/pymc3/pymc3/model.py", line 1450, in __init__
    distribution.shape, distribution.dtype) * distribution.default()
  File "/Users/rpg/src/pymc3/pymc3/distributions/distribution.py", line 91, in default
    return np.asarray(self.get_test_val(self.testval, self.defaults), self.dtype)
  File "/Users/rpg/src/pymc3/pymc3/distributions/distribution.py", line 96, in get_test_val
    if hasattr(self, v) and np.all(np.isfinite(self.getattr_value(v))):
TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''

Please provide any additional information below.

Versions and main components

• PyMC3 Version: master
• Theano Version: 1.0.4
• Python Version: 3.6.8
• Operating system: MacOS
• How did you install PyMC3: pip, editable

[AlexAndorra]

@rpgoldman, do you have some ideas of how to fix this? Maybe I can include it in the pm.Data integer PR (#3816) that I'm gonna take over.

[AlexAndorra]

I think I found a fix 🎉 I'll include it in my upcoming PR that enables pm.Data to be used with integer variables.