@hameerabbasi I want to try with NumPy. Do you have an implementation of NEP 18?

We don't have any implementation of NEP 18 yet, beyond the proof of concept implementation referenced in the NEP (which doesn't actually implement the NumPy API).

OK, I am waiting it.

Please read numpy/numpy#12028.

The development branch of numpy now has support for __array_function__. It would be great if you could give this a try and report back any issues!

Thanks, I will try it.

This PR is working with NumPy master branch. @shoyer

import numpy
import cupy

a = cupy.arange(10).reshape(2,5)
print(numpy.argmax(a))         #=> array(9)
print(numpy.argmax(a,axis=0))  #=> array([1, 1, 1, 1, 1])

LGTM! I suppose that it is reasonable to merge this feature soon as __array_function__ will be experimentally supported since comming NumPy 1.16.0.