Signatures and behaviors of `argmax` and `argmin` are incompatible with NumPy

First, for the signature: In CuPy the signature for argmax() and argmin() is

cupy.argm*(a, axis=None, dtype=None, out=None, keepdims=False)

But in NumPy it’s

numpy.argm*(a, axis=None, out=None)

That is, dtype and keepdims should be removed.

Second, for the behavior: In CuPy the axis could be a tuple, but in Numpy it can only be an integer:

>>> import cupy as cp
>>> a = cp.arange(60).reshape(3,4,5)
>>> a.argmax(axis=(0,1))
array([11, 11, 11, 11, 11], dtype=int64)
>>> 
>>> import numpy as np
>>> a = np.arange(60).reshape(3,4,5)
>>> a.argmax(axis=(0,1))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: 'tuple' object cannot be interpreted as an integer

I think both are easy to fix. Just add a few guards prior to calling the actual workhorses.

[niboshi]

We don’t have to remove any arguments? Then, #2872 fixes the remaining issue (dtype argument of argmin/argmax).

[toslunar]

Issue numpy/numpy#8710: “Add keepdims argument to argmin and argmax”