µ-Serve: Serving Deep Learning Models Without Breaking the Bank (or the Planet!)
µ-Serve: Serving Deep Learning Models Without Breaking the Bank (or the Planet!) Hey friend, ever think about the energy costs of running all those fancy deep learning models? It’s a HUGE deal, especially as these models get bigger and more complex. Researchers at the University of Illinois and IBM have been tackling this problem, and their solution is pretty cool. The issue is simple: serving deep learning models (making them available for use) uses a lot of power, mostly from…