How to keep torch models in memory while waiting for requests to process data

Question

I have several models for image processing. GPU memory is large enough to hold all of them at once. Currently my flask server is waiting for a request to process photos. When th