Segmentation fault: 11 on CUDA docker image

Hi,

Running:

import mxnet as mx
a = mx.nd.zeros(10, ctx=mx.gpu())

Yields:

Segmentation fault: 11
Segmentation fault (core dumped)

I have tested the following images:
nvidia/cuda:10.1-cudnn7-devel-ubuntu16.04
nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04
nvidia/cuda:10.0-cudnn7-devel-ubuntu18.04

I installed the right version of mxnet in a separate conda environment for each image.