Infer Type issue when dtype passed is None

[anirudh2290]

Description

 data = mx.sym.var("data")
 data2 = mx.sym.var("data2")
 out = mx.sym.elemwise_add(data, data2)
 arg_types, out_types, aux_types = out.infer_type_partial(data=None, data2=np.float32)

This will fail, because of this line : https://github.com/apache/incubator-mxnet/blob/master/python/mxnet/symbol/symbol.py#L1006

Makes it inconsistent to the documentation which says :

Unknown type can be marked as None.

https://github.com/apache/incubator-mxnet/blob/master/python/mxnet/symbol/symbol.py#L896

Also, there is no way to pass unknown types to infer_type_partial because of this bug.

This is because:

>>> np.dtype(None)
dtype('float64')

https://github.com/apache/incubator-mxnet/blob/master/python/mxnet/symbol/symbol.py#L1006

The fix should be simple, but this issue has always been there assuming that np.dype(None) to be converted to float64 was always the case. Will there be users depending on this buggy behavior and are we okay with breaking these user scripts ?

@ptrendx

[samskalicky]

@lanking520 assign @anirudh2290

[anirudh2290]

un-assigning myself since i am not working on this currently. if someone is interested, please go ahead.