LoRaQ: Optimized Low Rank Approximation for 4-bit Quantization
For deploying large diffusion transformers on resource-constrained hardware, LoRaQ removes the need for high-precision branches and data-dependent calibration, enabling efficient 4-bit quantization.
LoRaQ introduces a data-free calibration method for low-rank approximation that enables fully sub-16-bit quantization of diffusion transformers, outperforming state-of-the-art methods at equal memory overhead on Pixart-Σ and SANA.
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-$Σ$ 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.