Fetching the paper…

HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precision · Around