Hi!

I try to select elements from NDArray structure by indexes like that:

ix_valid = np.ix_(valid.asnumpy().astype(np.uint8) != 0)

vlabels = labels[mx.nd.array(ix_valid)]

where `valid`

is MXNET NDArray type.

But this way is slower because I spend time to convert `asnumpy`

.

Is there any way to implement this by MXNET NDArray only?

I am also interested function `cumsum`

in MXNET context:

`numpy.`

`cumsum`

( *a* , *axis=None* , *dtype=None* , *out=None* )[[source]]