We’re working on ways to keep memory requirements trim, even when processing the entire network, by only handling subsets at a time. 🤞🏼