重新审视非对称量化：视觉模型权重中的隐藏对称性
Rethinking Asymmetric Quantization: Hidden Symmetry in Vision Model Weights
对需要在边缘/FPGA上部署视觉模型的读者，DASQ 给出了一条可复用的权重分解思路：先用稀疏分量剥离少量离群值，使剩余稠密权重近似对称，从而用无零点的对称量化替代非对称量化，在分类、检测、实例分割上同时降低 BOPs 与硬件开销。
原文依据 after removing a few sparse outliers, the distribution becomes nearly symmetric
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论训练后量化（PTQ）能够实现深度预训练模型的快速部署。在低比特场景下，近期面向视觉模型的 PTQ 方法采用非对称量化（AsymQ），引入零点偏移以缓解量化误差。然而，这些偏移带来可观的硬件开销，且无法充分刻画预训练权重分布的非对称结构，导致大量量化等级未被使用。本文揭示了预训练权重中存在的一种隐藏对称性：在去除少量稀疏离群值之后，权重分布变得近似对称。据此，我们提出稠密与加性稀疏量化（DASQ），将权重分解为稠密矩阵与稀疏矩阵两部分。稠密分量刻画零附近的对称结构，稀疏分量对去除的离群值建模，二者可并行处理，并能够以高效的无零点计算实现。在图像分类、目标检测与实例分割上的实验表明，DASQ 以更低的 BOPs 超越最优的 PTQ 方法。在 FPGA 上，DASQ 在相近吞吐下也展现出比 AsymQ 更高的精度与更低的功耗。
英文原文摘要 1212 字符
Post-training quantization (PTQ) enables rapid deployment of deep pretrained models. In the low-bit regime, recent PTQ methods for vision models adopt asymmetric quantization (AsymQ), introducing zero-point offsets to mitigate quantization errors. However, these offsets impose substantial hardware overhead and fail to fully capture the non-symmetric structure of pretrained weight distributions, leaving many quantization levels unused.In this paper, we reveal a hidden symmetry in the pretrained weights: after removing a few sparse outliers, the distribution becomes nearly symmetric.Accordingly, we propose Dense and Additive Sparse Quantization (DASQ), which decomposes the weights into dense and sparse matrices.The dense component captures the symmetric structure around zero, while the sparse component models the removed outliers, and both can be processed in parallel and can be implemented with efficient zero-point-free computation.Experiments on image classification, object detection, and instance segmentation show that DASQ surpasses state-of-the-art PTQ methods with lower BOPs. On an FPGA, DASQ also demonstrates higher accuracy and lower power consumption than AsymQ at comparable throughput.