arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.02538cs.LGcs.CLcs.CV
收藏

通过未来激活意识增强后训练量化

Enhancing Post-Training Quantization via Future Activation Awareness

Zhejiang University(浙江大学);Tencent(腾讯)

Zheqi Lv, Zhenxuan Fan, Qi Tian, Wenqiao Zhang, Yueting Zhuang

AI总结:

通过利用未来层激活信息,提出Future-Aware Quantization方法,有效提升后训练量化效果,适用于边缘部署。

摘要:

后训练量化(PTQ)是一种广泛用于压缩大型语言模型(LLMs)而无需微调的方法。它通常根据当前层的激活设置量化超参数(例如,缩放因子)。尽管这种方法高效,但会受到量化偏差和误差累积的影响,导致量化效果不佳且不稳定,尤其是在校准数据存在偏差时。为解决这些问题,我们提出了Future-Aware Quantization(FAQ),该方法利用未来层的激活来指导量化。这使能够更好地识别和保留重要权重,同时减少对校准噪声的敏感性。我们进一步引入了基于窗口的预览机制,以温和地聚合多个未来层的激活,缓解对单一层的过度依赖。为了避免昂贵的贪心搜索,我们使用预搜索的配置来最小化开销。实验表明,FAQ在几乎不增加额外成本的情况下,始终优于先前的方法,不需要反向传递、数据重建或调优,使其非常适合边缘部署。

英文摘要:

Post-training quantization (PTQ) is a widely used method to compress large language models (LLMs) without fine-tuning. It typically sets quantization hyperparameters (e.g., scaling factors) based on current-layer activations. Although this method is efficient, it suffers from quantization bias and error accumulation, resulting in suboptimal and unstable quantization, especially when the calibration data is biased. To overcome these issues, we propose Future-Aware Quantization (FAQ), which leverages future-layer activations to guide quantization. This allows better identification and preservation of important weights, while reducing sensitivity to calibration noise. We further introduce a window-wise preview mechanism to softly aggregate multiple future-layer activations, mitigating over-reliance on any single layer. To avoid expensive greedy search, we use a pre-searched configuration to minimize overhead. Experiments show that FAQ consistently outperforms prior methods with negligible extra cost, requiring no backward passes, data reconstruction, or tuning, making it well-suited for edge deployment.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑