arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.14645cs.CV
收藏

REx:无数据残差量化误差扩展

REx: Data-Free Residual Quantization Error Expansion

Sorbonne Université(索邦大学);CNRS(法国国家科学研究中心);ISIR(智能系统与机器人研究所);Datakalab

Edouard Yvinec, Arnaud Dapgony, Matthieu Cord, Kevin Bailly

AI总结:

本文提出无数据量化方法 REx,通过残差误差扩展、组稀疏性和集成近似实现灵活适应不同位宽与设备,并在多种视觉与 NLP 任务、架构及位宽上取得优越性能。

摘要:

深度神经网络(DNNs)在计算机视觉和自然语言处理中无处不在,但面临较高的推理成本。该问题可以通过量化来解决,量化即将浮点运算转换为较低位宽格式。随着对隐私权的关注日益增加,我们将研究重点放在无数据方法上。然而,这类技术存在对目标设备适应性不足的问题,因为硬件通常仅支持特定的位宽。因此,为了适应各种设备,量化方法应具备足够的灵活性,以便针对每个位宽和目标设备找到良好的精度与速度权衡。为了实现这一目标,我们提出了 REx,一种利用残差误差扩展、组稀疏性以及集成近似以实现更好并行化的量化方法。REx 具有强有力的理论保证作为支撑,并在每个基准测试应用(从视觉到 NLP 任务)、架构(ConvNets、transformers)和位宽(从 int8 到三值量化)上均取得了优越的性能。

英文摘要:

Deep neural networks (DNNs) are ubiquitous in computer vision and natural language processing, but suffer from high inference cost. This problem can be addressed by quantization, which consists in converting floating point operations into a lower bit-width format. With the growing concerns on privacy rights, we focus our efforts on data-free methods. However, such techniques suffer from their lack of adaptability to the target devices, as a hardware typically only support specific bit widths. Thus, to adapt to a variety of devices, a quantization method shall be flexible enough to find good accuracy v.s. speed trade-offs for every bit width and target device. To achieve this, we propose REx, a quantization method that leverages residual error expansion, along with group sparsity and an ensemble approximation for better parallelization. REx is backed off by strong theoretical guarantees and achieves superior performance on every benchmarked application (from vision to NLP tasks), architecture (ConvNets, transformers) and bit-width (from int8 to ternary quantization).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑