NotImplementedError: Separable Conv1D is not supported in current MXNet

[GundamBen]

Description

I open an issue in mxnet for the crash that my code works in other keras backend but crashes in mxnet. I wonder whether the support lacks in mxnet or in keras-mxnet.

Code and reproduce

More details are in the link above.
Since some advises me update keras-mxnet to check whether it is fixed. I update it from 2.2.4.2 to 2.2.4.3 (latest version) but it does not work. The crash still exists.

Environment

keras-mxnet==2.2.4.2 (updated to 2.2.4.3)
mxnet version 1.8.0
cuda version 10.1

Thanks in advance!