ShamAN-Q: 用于亚1位LLM权重的洗发水增强纳米量化
研究人员开发了ShamAN-Q，这是一种新的子1位后训练量化方法，适用于大型语言模型。该方法在NanoQuant的基础上进行了改进，引入了密集曲率度量和马哈拉诺比斯重建损失。
01
影响到谁
- 1ShamAN-Q
- 属于 →事实
- 需要 →事实
- 使用 →事实
- 依赖 →事实
- 使用 →事实
02
证据
- AarXiv cs.LG研究机构2026-10-02 12:00
a sub-1-bit post-training quantization method
- AarXiv cs.LG研究机构2026-10-02 12:00
We introduce ShamAN-Q, a sub-1-bit post-training quantization method that extends NanoQuant by replacing each its diagonal reconstruction geometry with a tract…
- AarXiv cs.LG研究机构2026-10-02 12:00
Abstract: We introduce ShamAN-Q, a sub-1-bit post-training quantization method that extends NanoQuant by replacing each its diagonal reconstruction geometry wi…