提出统一框架，让模型量化突破单层限制，提升大模块精度。
LPCD: Unified Framework from Layer-Wise to Submodule Quantization
- 用松弛目标优化任意子模块，再投影到层级量化器上
- 在多种大模型架构中，比特率更低时仍保持更高精度
- 适合追求高精度低比特量化的研究者与工程师
训练后量化（PTQ）旨在保留模型整体行为，但现有方法多聚焦于单一线性层。即使近期方法如QEP和LoaQ通过缓解误差传播或针对特定子模块改进，仍依赖层级形式化，难以捕捉更大子模块的行为。本文提出层投影坐标下降（LPCD），一种统一框架，将PTQ扩展至任意子模块，通过优化松弛目标并用层级量化器投影解。LPCD推广了现有方法，为复杂子模块量化提供理论严谨的方案，同时保持层级PTQ流程的效率与兼容性。在多种LLM架构及比特率下，基于LPCD的子模块量化持续优于传统层级量化方法和现有子模块方法。
原文摘要 · Abstract (English)
Post-training quantization (PTQ) aims to preserve model-level behavior; however, most methods focus on individual linear layers. Even recent extensions, such as QEP and LoaQ, which mitigate error propagation or target specific submodules, still rely on layer-wise formulations and fail to capture the behavior of larger submodules. We introduce Layer-Projected Coordinate Descent (LPCD), a unified framework that extends PTQ beyond layers by optimizing relaxed objectives across arbitrary submodules and projecting the solutions with layer-wise quantizers. LPCD generalizes existing methods and provides a principled approach to quantizing complex submodules while maintaining the efficiency and compatibility of layer-wise PTQ pipelines. Across diverse LLM architectures and bit-widths, LPCD-based submodule quantization consistently enhances both layer-wise PTQ methods and existing submodule approaches.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。