One Network to Fit All Hardware: New MIT AutoML Method Trains 14X Faster Than SOTA NAS

Open in new window