[BUG] Enabling cudf.pandas leads to exception when using a Numpy array about cudf HOT 3 OPEN

Leaving some notes after a first pass investigation:

ExponentialSmoothing.__init__ is quietly a pandas function since it uses the pandas.util._decorators.deprecate_kwarg decorator. Seems like self as an arg to __init__ isn't being proxied correctly when falling back to a "slow" path.

Removing the @deprecate_kwarg allows ExponentialSmoothing to construct correctly

from cudf.

Simpler repo:

In [1]: import cudf.pandas
   ...: cudf.pandas.install()
   ...: import pandas as pd

In [2]: class Foo:
   ...:     @pd.util._decorators.deprecate_kwarg("not", "important")
   ...:     def __init__(self, val, important=None):
   ...:         self.val = val
   ...: 

In [3]: Foo(1)
TypeError: Foo.__init__() missing 1 required positional argument: 'val'

from cudf.

(Apologies for the crappy issue title (none at all), tried to improve it)

from cudf.