Good evening! What is the better way of storing trained model for object classification mobile app? Directly in device memory or on remote google server? The dataset is large and we need the lowest processing time so i think google's server is good way, but there are other things like internet speed etc. What are your thoughts?