[ICLR 2020] Once for All: Train One Network and Specialize it for Efficient Deployment
Official Pytorch implementation of Once-for-All v0.1