You use Azure Machine Learning designer to create and run a training pipeline. You then create a real-time inference pipeline that must be deployed as a web service.

What must you do before you deploy the real-time inference pipeline?