You use ray tune to define the search space, and it will automatically use the idle resources of the entire cluster to start 500 parallel training experiments to help you find the best set of parameters. Model online: The trained model is deployed into an online server using ray serve.
Ray Davies
