I am quite new to gRPC. I created my model serving Docker image. The model has a unique pre-processing pipeline (hence not using the TensorFlow serving).