Abstract: Mixed-precision quantization is a popular approach for compressing deep neural networks (DNNs). However, it is challenging to scale the performance efficiently with mixed-precision DNNs ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results