How to use Tesla M10 with CUDA on a windows server 2012?

Hi, I know the Tesla M10 is not the perfect device for this, but our client installed this device. We need to run a python program that uses Pytorch with CUDA on this GPU. I read that the M10 can use either Passthrough or Grid but I have no experience with this kind of GPU.
Can I use this GPU as a normal desktop GPU to run inference? It’s fine if it can only utilize 1 of the 4 units of the GPU.