[论文解读] HAQ: Hardware-Aware Automated Quantization

Kuan Wang, Zhijian Liu+3 more|arXiv (Cornell University)|Nov 21, 2018

Advanced Neural Network Applications参考文献 22被引用 58

一句话总结

HAQ 提出了一种基于强化学习的框架，通过硬件模拟器引入实时硬件反馈（延迟和能效），自动确定深度神经网络（DNNs）的最优混合精度量化策略。与 8 位量化相比，该方法在极小精度损失下实现了 1.4–1.95 倍的延迟降低和 1.9 倍的能效提升，同时揭示了边缘与云环境硬件架构在最优策略上的显著差异。

ABSTRACT

Model quantization is a widely used technique to compress and accelerate deep neural network (DNN) inference. Emergent DNN hardware accelerators begin to support mixed precision (1-8 bits) to further improve the computation efficiency, which raises a great challenge to find the optimal bitwidth for each layer: it requires domain experts to explore the vast design space trading off among accuracy, latency, energy, and model size, which is both time-consuming and sub-optimal. Conventional quantization algorithm ignores the different hardware architectures and quantizes all the layers in a uniform way. In this paper, we introduce the Hardware-Aware Automated Quantization (HAQ) framework which leverages the reinforcement learning to automatically determine the quantization policy, and we take the hardware accelerator's feedback in the design loop. Rather than relying on proxy signals such as FLOPs and model size, we employ a hardware simulator to generate direct feedback signals (latency and energy) to the RL agent. Compared with conventional methods, our framework is fully automated and can specialize the quantization policy for different neural network architectures and hardware architectures. Our framework effectively reduced the latency by 1.4-1.95x and the energy consumption by 1.9x with negligible loss of accuracy compared with the fixed bitwidth (8 bits) quantization. Our framework reveals that the optimal policies on different hardware architectures (i.e., edge and cloud architectures) under different resource constraints (i.e., latency, energy and model size) are drastically different. We interpreted the implication of different quantization policies, which offer insights for both neural network architecture design and hardware architecture design.

研究动机与目标

• 解决在多样化硬件加速器上手动探索混合精度量化庞大设计空间的挑战。
• 克服传统量化方法采用统一位宽且忽略硬件特定性能特征的局限性。
• 实现自动化、硬件定制化的量化策略，同时优化延迟、能效、模型大小和精度。
• 揭示在不同资源约束下，边缘与云环境硬件架构对量化策略的根本性差异。

提出的方法

• 使用强化学习智能体在 DNN 各层中搜索最优每层位宽（1–8 位）。
• 将硬件模拟器集成到强化学习循环中，直接提供延迟和能效消耗的反馈，替代 FLOPs 等代理指标。
• 使用兼顾模型精度、延迟、能效和模型大小的奖励函数训练强化学习智能体。
• 采用策略梯度方法（如 REINFORCE 或 PPO）基于硬件反馈优化量化策略。
• 通过与目标硬件加速器协同设计，实现端到端自动化量化策略。
• 通过在多样化模型上进行训练，实现所学策略在不同神经网络架构间的可迁移性。

实验结果

研究问题

• RQ1 通过强化学习实现硬件感知量化，相较于固定精度量化，如何提升推理效率？
• RQ2 在不同资源约束下，边缘与云硬件架构的最优量化策略有何差异？
• RQ3 来自模拟器的硬件反馈在多大程度上可替代 FLOPs 等代理指标来指导量化决策？
• RQ4 所提出的框架如何在降低延迟和能效的同时保持高模型精度？

主要发现

• HAQ 在多个模型和硬件平台上，相比 8 位固定精度量化，将推理延迟降低了 1.4 倍至 1.95 倍。
• 平均能效消耗降低了 1.9 倍，显著提升了效率，且精度损失极小。
• 最优量化策略在边缘与云硬件之间存在显著差异，表明“一刀切”式量化策略并非最优。
• 来自模拟器的硬件特定反馈使强化学习智能体发现了比基于代理指标更高效的位宽分配方案。
• 该框架揭示了混合精度策略对目标硬件特性的高度敏感性，为硬件与神经架构协同设计提供了可操作的洞察。
• 该方法在显著降低模型大小和计算成本的同时，实现了与全精度模型几乎相同的精度。