Deploying the trained Object2Vec and running inference