QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching
This addresses efficiency challenges in deploying large language models across diverse quantization scenarios, though it appears incremental as it builds on existing elastic quantization and adapter techniques.
The paper tackles the high storage and optimization costs of elastic precision quantization for Transformers by proposing QuEPT, a post-training scheme that uses one-shot calibration and dynamic adapters to enable multi-bit switching without repeated optimization, achieving comparable or better performance than state-of-the-art methods in experiments.
Elastic precision quantization enables multi-bit deployment via a single optimization pass, fitting diverse quantization scenarios.Yet, the high storage and optimization costs associated with the Transformer architecture, research on elastic quantization remains limited, particularly for large language models.This paper proposes QuEPT, an efficient post-training scheme that reconstructs block-wise multi-bit errors with one-shot calibration on a small data slice. It can dynamically adapt to various predefined bit-widths by cascading different low-rank adapters, and supports real-time switching between uniform quantization and mixed precision quantization without repeated optimization. To enhance accuracy and robustness, we introduce Multi-Bit Token Merging (MB-ToMe) to dynamically fuse token features across different bit-widths, improving robustness during bit-width switching. Additionally, we propose Multi-Bit Cascaded Low-Rank adapters (MB-CLoRA) to strengthen correlations between bit-width groups, further improve the overall performance of QuEPT. Extensive experiments demonstrate that QuEPT achieves comparable or better performance to existing state-of-the-art post-training quantization methods.Our code is available at https://github.com/xuke225/QuEPT