Smallify: Learning Network Size while Training

Guillaume Leclerc, Manasi Vartak, Raul Castro Fernandez +2

As neural networks become widely deployed in different applications and on different hardware, it has become increasingly important to optimize inference time and model size along…