greengrass ml component

Hello I am studying Greengrass machine learning components. I have a question. I'm going to distribute Greengrass custom machine learning components to Raspberry Pi now. The artifacts associated with the recipe in this component contain inference codes uploaded to S3. And this inference code has a function of determining who is by photographing the user's face in real time. I'm curious. If the components are distributed well in Raspberry Pi, will the inference code continue to run? Or, do I have to deploy components every time I try to run an inference? Thank you.