To accelerate the inference of deep neural networks (DNNs), quantization with low-bitwidth numbers is actively researched. A prominent challenge is to quantize the DNN models into low-bitwidth numbers without significant accuracy degradation, especially at very low bitwidths (< 8 bits). This work targets an adaptive data representation with variable-length encoding called DyBit. DyBit can dynamically adjust the precision and range of separate bit-field to be adapted to the DNN weights/activations distribution. We also propose a hardware-aware quantization framework with a mixed-precision accelerator to trade-off the inference accuracy and speedup. Experimental results demonstrate that the inference accuracy via DyBit is 1.997% higher than the state-of-the-art at 4-bit quantization, and the proposed framework can achieve up to 8.1x speedup compared with the original model.
@article{arxiv.2302.12510,
title = {DyBit: Dynamic Bit-Precision Numbers for Efficient Quantized Neural Network Inference},
author = {Jiajun Zhou and Jiajun Wu and Yizhao Gao and Yuhao Ding and Chaofan Tao and Boyu Li and Fengbin Tu and Kwang-Ting Cheng and Hayden Kwok-Hay So and Ngai Wong},
journal= {arXiv preprint arXiv:2302.12510},
year = {2024}
}