A stream of payload parts. Each part contains a portion of the response for a streaming inference request.