I’m developing inference engine, so I actually do understand how it works. As well as other types of parallelism and how exactly they do different trade offs