LoRaQ: Optimized Low Rank Approximation for 4-bit Quantization

TL;DR

LoRaQ introduces a data-free, low-rank approximation method for 4-bit quantization, enabling fully sub-16 bit models that outperform existing approaches in resource-constrained deployment of diffusion transformers.

Contribution

LoRaQ removes the need for high-precision auxiliary branches and data-dependent calibration, achieving fully quantized models with improved performance over state-of-the-art methods.

Findings

LoRaQ outperforms existing methods on Pixart-Σ and SANA datasets.

It enables a fully sub-16 bit pipeline for diffusion transformers.

Mixed-precision configurations like W8A8, W6A6, and W4A8 yield superior results.

Abstract

Post-training quantization (PTQ) is essential for deploying large diffusion transformers on resource-constrained hardware, but aggressive 4-bit quantization significantly degrades generative performance. Low-rank approximation methods have emerged as a promising solution by appending auxiliary linear branches to restore performance. However, current state-of-the-art approaches assume these branches must retain high precision (W16A16) and rely on heavy, data-dependent calibration for initialization. We challenge both limitations with LoRaQ (Low-Rank Approximated Quantization), a simple, data-free calibration approach that optimizes quantization error compensation. By overcoming the need for high-precision branches, LoRaQ enables the first fully sub-16 bit pipeline, allowing the low-rank branch itself to be quantized. We demonstrate that, at equal memory overhead, LoRaQ outperforms the state-of-the-art methods in their native implementations on Pixart-$\Sigma$ and SANA. We also analyze mixed-precision configurations, showing that setups such as W8A8, W6A6, and W4A8 for the low-rank branch, alongside a W4 main layer, yield superior results while maintaining a fully quantized architecture compatible with modern mixed-precision hardware.