This message was deleted.

Another way to do is run models are individual pods and assign those pods to the same node

Would this work if you only have a single GPU?