Running docker on system with Nvidia GPU - nvidia driver installation


I am trying to deploy my docker (Not an nvidia-docker) container on a system with Nvidia Tesla P40 GPU. I wanted to confirm whether the Nvidia driver version needs to be EXACTLY the same inside my container as that on the target host. (for the GPU to be used / detected from inside the container)

Also, what are the minimum requirements I would need to have any application running inside the container to pick up the GPU during runtime ?


Yes, it needs to be EXACTLY the same.