IndexError: tuple index out of range despite having trained forest

[QuantumDamage]

Everything is in this gist: https://gist.github.com/QuantumDamage/4d256e9a4f02b91f332105c105b5e457

I'm just training forest and trying to see features importance with eli5.

[lopuhin]

@QuantumDamage thanks for providing the gist! So the error is

eli5.explain_weights(estimator = forest)

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-21-596faaed9271> in <module>()
----> 1 eli5.explain_weights(estimator = forest)

lib/python3.6/site-packages/singledispatch.py in wrapper(*args, **kw)
    208 
    209     def wrapper(*args, **kw):
--> 210         return dispatch(args[0].__class__)(*args, **kw)
    211 
    212     registry[object] = func

IndexError: tuple index out of range

I think that eli5.explain_weights(forest) will work correctly. The reason why is that we are using the type of the first argument to decide which implementation of explain_weights to use, but singledispatch library gets confused if the first argument is passed as a keyword argument instead. Ideally we'd like to word around this issue (or provide a better error message, but most likely it's easier to work around this).

[teabolt]

It appears that a better error message was added from Python 3.6 onward: https://bugs.python.org/issue33967 @lopuhin