Install CUDA within docker on Jetson Nano

I am trying to build a docker container on a Jetson Nano that has CUDA 10.0. Can anyone suggest how to best do this?

The only method I have found to install CUDA 10.0 on a aarch64 architecture is using the “CUDA cross-platform environment installation”. At the moment I don’t have easy access to an x86-64 system which I can build it on, so this isn’t an option for me.

On DockerHub Nvidia only provide base images with CUDA for X86-64 architecture. I assume this is because the Jetson Nano JetPack comes preinstalled with CUDA, so in most cases no-one would need to install it for an aarch64 architecture.