wkcn commented on a change in pull request #16923: [WIP] Fallback mechanism for
mx.np operators
URL: https://github.com/apache/incubator-mxnet/pull/16923#discussion_r351558270
##########
File path: python/mxnet/numpy/multiarray.py
##########
@@ -202,13 +220,13 @@ def __array_ufunc__(self, ufunc, method, *inputs,
**kwargs): # pylint: disable=
name = ufunc.__name__
mx_ufunc = _NUMPY_ARRAY_UFUNC_DICT.get(name, None)
if mx_ufunc is None:
- raise ValueError('mxnet.numpy operator `{}` has not been
registered in '
- 'the _NUMPY_ARRAY_UFUNC_LIST. Please make
sure you are '
- 'using NumPy >= 1.15.0 and the operator
implementation '
- 'is compatible with NumPy. Then add the
operator name '
- 'to the list.'
- .format(name))
- return mx_ufunc(*inputs, **kwargs)
+ # try to fallback to official NumPy op
+ onp_op = _get_np_op(name)
+ new_inputs = [arg.asnumpy() if isinstance(arg, ndarray) else
arg for arg in inputs]
+ out = onp_op(*new_inputs, **kwargs)
Review comment:
It will break the computational graph, and could not compute the gradient.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
