QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908832066699264 |
|---|---|
| author | Xu, Ke Wang, Yixin Li, Zhongcheng Cui, Hao Hu, Jinshui Zhang, Xingyi |
| author_facet | Xu, Ke Wang, Yixin Li, Zhongcheng Cui, Hao Hu, Jinshui Zhang, Xingyi |
| contents | 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 |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_12609 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching Xu, Ke Wang, Yixin Li, Zhongcheng Cui, Hao Hu, Jinshui Zhang, Xingyi Computer Vision and Pattern Recognition Artificial Intelligence 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 |
| title | QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2602.12609 |