RepQuant: Towards Accurate Post-Training Quantization of Large Transformer Models via Scale Reparameterization

TL;DR

RepQuant introduces a novel post-training quantization framework for large transformer models that employs complex quantizers during calibration and simplified ones during inference, achieving high accuracy and efficiency through scale reparameterization.

Contribution

The paper proposes RepQuant, a new PTQ method using scale reparameterization to decouple quantization and inference, enabling accurate and hardware-friendly quantization of large transformers.

Findings

Significant performance improvements over existing PTQ methods.

Effective quantization of LayerNorm and Softmax activations.

Versatile application across vision, language, and multi-modal transformers.

Abstract

Large transformer models have demonstrated remarkable success. Post-training quantization (PTQ), which requires only a small dataset for calibration and avoids end-to-end retraining, is a promising solution for compressing these large models. Regrettably, existing PTQ methods typically exhibit non-trivial performance loss. We find that the performance bottleneck stems from over-consideration of hardware compatibility in the quantization process, compelling them to reluctantly employ simple quantizers, albeit at the expense of accuracy. With the above insights, we propose RepQuant, a novel PTQ framework with quantization-inference decoupling paradigm to address the above issues. RepQuant employs complex quantizers in the quantization process and simplified quantizers in the inference process, and performs mathematically equivalent transformations between the two through quantization scale reparameterization, thus ensuring both accurate quantization and efficient inference. More specifically, we focus on two components with extreme distributions: LayerNorm activations and Softmax activations. Initially, we apply channel-wise quantization and log$\sqrt{2}$ quantization, respectively, which are tailored to their distributions. In particular, for the former, we introduce a learnable per-channel dual clipping scheme, which is designed to efficiently identify outliers in the unbalanced activations with fine granularity. Then, we reparameterize the scales to hardware-friendly layer-wise quantization and log2 quantization for inference. Moreover, quantized weight reconstruction is seamlessly integrated into the above procedure to further push the performance limits. Extensive experiments are performed on different large-scale transformer variants on multiple tasks, including vision, language, and multi-modal transformers, and RepQuant encouragingly demonstrates significant performance advantages.