ReSpinQuant: Efficient Layer-Wise LLM Quantization via Subspace Residual Rotation Approximation
This work addresses the trade-off between accuracy and efficiency in LLM quantization, offering a practical solution for deploying quantized models without sacrificing performance.
ReSpinQuant introduces a layer-wise quantization framework for LLMs that fuses activation rotations offline to avoid online computation overhead, achieving state-of-the-art W4A4 and W3A3 quantization performance with minimal inference cost.
Rotation-based Post-Training Quantization (PTQ) has emerged as a promising solution for mitigating activation outliers in the quantization of Large Language Models (LLMs). Global rotation methods achieve inference efficiency by fusing activation rotations into attention and FFN blocks, but suffer from limited expressivity as they are constrained to use a single learnable rotation matrix across all layers. To tackle this, layer-wise transformation methods emerged, achieving superior accuracy through localized adaptation. However, layer-wise methods cannot fuse activation rotation matrices into weights, requiring online computations and causing significant overhead. In this paper, we propose ReSpinQuant, a quantization framework that resolves such overhead by leveraging offline activation rotation fusion and matching basis using efficient residual subspace rotation. This design reconciles the high expressivity of layer-wise adaptation with only negligible inference overhead. Extensive experiments on W4A4 and W3A3 quantization demonstrate that ReSpinQuant achieves state-of-the-art performance, outperforming global rotation methods and matching the accuracy of computationally expensive layer-wise methods with minimal overhead.