Numpy Operator Inconsistency

[stu1130]

argmax

The dtype return be argmax is float32 instead of int

>>> d = mx.np.array([0, 1])
>>> mx.np.argmax(d)
array(1.)
>>> mx.np.argmax(d).dtype
<class 'numpy.float32'>

repeat

import mxnet as mx
from mxnet import util


@util.use_np
def func():
    x = mx.np.array([[1,2],[3,4]])
    print(mx.np.repeat(x, [1, 2], axis=0))

func()

mxnet.base.MXNetError: Invalid Parameter format for repeats expect int but value='[1, 2]', in operator repeat(name="", axis="0", repeats="[1, 2]")

softmax

import mxnet as mx
from mxnet import util

@util.use_np
def func():
    x = mx.np.array([], ctx=mx.gpu(0)).reshape(2, 0, 0)
    print(mx.npx.softmax(x))

func()

Floating point exception (core dumped)

mean

import mxnet as mx
from mxnet import util


@util.use_np
def func():
    x = mx.np.array([], ctx=mx.gpu(0)).reshape(2, 0, 0)
    print(mx.np.mean(x))
func()

0.0 (expect NaN)

prod

import mxnet as mx
from mxnet import util


@util.use_np
def func():
    x = mx.np.array([], ctx=mx.gpu(0)).reshape(2, 0, 0)
    print(mx.np.prod(x))
func()

0.0 (expect 1.0)

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended label(s): Bug

[stu1130]

@mxnet-label-bot add [Numpy]

[sxjscience]

duplicate: #16048

[samskalicky]

@zachgk assign [@reminisce ]

[reminisce]

transpose fixed in #16414.

[haojin2]

Issues with np.transpose and np.split are fixed, removing those parts from the issue.

[yzhliu]

@Yiyan66 is looking into it.

[yzhliu]

closed by #18286 #18371