Improving Neural Network Quantization without Retraining using Outlier Channel Splitting
Ritchie Zhao · Yuwei Hu · Jordan Dotzel · Christopher De Sa · Zhiru Zhang

Tue Jun 11th 05:05 -- 05:10 PM @ Grand Ballroom

Quantization can improve the execution latency and energy efficiency of neural networks on both commodity GPUs and specialized accelerators. The majority of existing literature focuses on training quantized DNNs, while this work examines the less-studied topic of quantizing a floating-point model without (re)training. DNN weights and activations follow a bell-shaped distribution post-training, while practical hardware uses a linear quantization grid. This leads to challenges in dealing with outliers in the distribution. Prior work has addressed this by clipping the outliers or using specialized hardware. In this work, we propose outlier channel splitting (OCS), which duplicates channels containing outliers, then halves the channel values. The network remains functionally identical, but affected outliers are moved toward the center of the distribution. OCS requires no additional training and works on commodity hardware. Experimental evaluation on ImageNet classification and language modeling shows that OCS can outperform state-of-the-art clipping techniques with only minor overhead.

Author Information

Ritchie Zhao (Cornell University)
Yuwei Hu (Cornell University)
Jordan Dotzel (Cornell University)
Christopher De Sa (Cornell)
Zhiru Zhang (Cornell Univeristy)

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

More from the Same Authors