Parameterization of WeibullRV missing scale?

[Armavica]

Description of your problem or feature request

I am trying to upgrade PyMC to aesara=2.8.2, and some of the tests failing are related to WeibullRV.

I think that it might have to do with commit fcd1c4c which imposes the prototype of WeibullRV.__call__(self, shape, size, **kwargs). However, the Weibull distribution is usually described by two parameters: its scale and its shape. The scale is missing from the function prototype, which might explain the errors on PyMC's side.

cc @rlouf because I think that you are the author of this commit

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

import pymc as pm # (branch Armavica:aesara-rewrite)
with pm.Model() as model:
    pm.Weibull("x", alpha=1.0, beta=1.0, shape=(3, 4))

Please provide the full traceback of any errors.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Input In [3], in <cell line: 1>()
      1 with pm.Model() as model:
----> 2     pm.Weibull("x", alpha=1.0, beta=1.0, size=(3, 4))

File ~/src/pymc/pymc/distributions/distribution.py:263, in Distribution.__new__(cls, name, rng, dims, initval, observed, total_size, transform, *args, **kwargs)
    259     raise TypeError(f"Name needs to be a string but got: {name}")
    261 # Create the RV and process dims and observed to determine
    262 # a shape by which the created RV may need to be resized.
--> 263 rv_out, dims, observed, resize_shape = _make_rv_and_resize_shape(
    264     cls=cls, dims=dims, model=model, observed=observed, args=args, **kwargs
    265 )
    267 if resize_shape:
    268     # A batch size was specified through `dims`, or implied by `observed`.
    269     rv_out = change_rv_size(rv=rv_out, new_size=resize_shape, expand=True)

File ~/src/pymc/pymc/distributions/distribution.py:165, in _make_rv_and_resize_shape(cls, dims, model, observed, args, **kwargs)
    162 """Creates the RV and processes dims or observed to determine a resize shape."""
    163 # Create the RV without dims information, because that's not something tracked at the Aesara level.
    164 # If necessary we'll later replicate to a different size implied by already known dims.
--> 165 rv_out = cls.dist(*args, **kwargs)
    166 ndim_actual = rv_out.ndim
    167 resize_shape = None

File ~/src/pymc/pymc/distributions/continuous.py:2585, in Weibull.dist(cls, alpha, beta, *args, **kwargs)
   2582 alpha = at.as_tensor_variable(floatX(alpha))
   2583 beta = at.as_tensor_variable(floatX(beta))
-> 2585 return super().dist([alpha, beta], *args, **kwargs)

File ~/src/pymc/pymc/distributions/distribution.py:351, in Distribution.dist(cls, dist_params, shape, **kwargs)
    346 create_size, ndim_expected, ndim_batch, ndim_supp = find_size(
    347     shape=shape, size=size, ndim_supp=cls.rv_op.ndim_supp
    348 )
    349 # Create the RV with a `size` right away.
    350 # This is not necessarily the final result.
--> 351 rv_out = cls.rv_op(*dist_params, size=create_size, **kwargs)
    353 # Replicate dimensions may be prepended via a shape with Ellipsis as the last element:
    354 if shape is not None and Ellipsis in shape:

TypeError: WeibullRV.__call__() got multiple values for argument 'size'

Please provide any additional information below.

Versions and main components

• Aesara version: 2.8.2
• Aesara config (python -c "import aesara; print(aesara.config)")
• Python version: 3.10.5
• Operating system: linux
• How did you install Aesara: conda

[rlouf]

Did you open an issue in PyMC?

[Armavica]

Here is my draft PR: pymc-devs/pymc#6059.

Sorry if I was unclear, but the issue is not in the current version of PyMC. It appeared when I tried to update aesara's version to 2.8.2. WeibullRV is called by PyMC with two parameters, and this worked fine until aesara<2.8.0, but now it looks like it wants only one. Since the recent release notes of aesara don't mention anything particular related to the API of WeibullRV (only docstrings), I assumed that this change was unintentional.

[rlouf]

Thanks for linking the relevant PR. The idea behind the change in the WeibullRV API was to conform to the API in NumPy. It's easy enough to rescale if necessary.

[Armavica]

Thank you, that answers my question. In this case, I will close this issue as I think that this should probably be treated on the PyMC side.

[rlouf]

It looks like the parametrization change was propagated to AePPL as well, so this should not be an issue.