Create a health check for nvidia containers that will check nvidia-smi every ~10 minutes and restart the container if the connection is broken.