Timezone: »

 
Poster
Differentiable Dynamic Quantization with Mixed Precision and Adaptive Resolution
zhaoyang zhang · Wenqi Shao · Jinwei Gu · Xiaogang Wang · Ping Luo

Thu Jul 22 09:00 AM -- 11:00 AM (PDT) @ None #None

Model quantization is challenging due to many tedious hyper-parameters such as precision (bitwidth), dynamic range (minimum and maximum discrete values) and stepsize (interval between discrete values). Unlike prior arts that carefully tune these values, we present a fully differentiable approach to learn all of them, named Differentiable Dynamic Quantization (DDQ), which has several benefits. (1) DDQ is able to quantize challenging lightweight architectures like MobileNets, where different layers prefer different quantization parameters. (2) DDQ is hardware-friendly and can be easily implemented using low-precision matrix-vector multiplication, making it capable in many hardware such as ARM. (3) Extensive experiments show that DDQ outperforms prior arts on many networks and benchmarks, especially when models are already efficient and compact. e.g., DDQ is the first approach that achieves lossless 4-bit quantization for MobileNetV2 on ImageNet.

Author Information

zhaoyang zhang (The Chinese University of Hong Kong)
Wenqi Shao (The Chinese University of HongKong)
Jinwei Gu (Sensebrain)
Xiaogang Wang (Chinese University of Hong Kong, Hong Kong)
Ping Luo (The University of Hong Kong)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors