Try training models as SHARCNET

[lintool]

We have access to this GPU cluster: https://docs.computecanada.ca/wiki/Graham

Let's try to see how difficult it would be to take advantage of these resources...

Assigning to @dishant-mittal