Your server should be more than enough, though if you are sharing the server with others who are also working it hard, you may need to coordinate with them. I don’t think pytorch does multi-gpu on a single model, but another user might be using tensorflow and hogging both gpus.
If you are on your own server rather than paperspace:
git clone https://github.com/fastai/fastai.git
conda env update
source activate fastai
use the url given in the terminal if jupyter doesn’t automatically pop up in your browser
–edit - I see up top that Jeremy says the paperspace script should work for any ubuntu computer, so I guess try that first