Skip to main content

Model Compression

Model compression techniques are mainly classified into quantization (Zafrir et al.,2019), pruning (Hoefler et al., 2021; Vadera and Ameen, 2022), Neural Architecture Search(NAS) (Sun et al., 2019), and knowledge distillation 

Model Quantization

Quantization Granularity Quantization is a magic spell to reduce the memory footprint of a model...

Neural Arch Search

Knowledge Distillation

Pruning

reference

量化技术背景 从CNN量化说起 在传统CNN网络中，为了加速网络的推理速度，一种非常有效的方法是INT8量化 ，即将权重与激活(feature map) 的浮点数值量化成8-bit整型表示...
Back to top