Add py-torch to container so that libtorch binaries are accessible for running inference
No changes between version 4 and version 4
