Run Cuda HPC applications from prepared docker containers.

https://ngc.nvidia.com/registry

For example LAMMPS.