Event ·Multi-source

Token-Energy Law:用活跃字节和解码时间预测 LLM 每 token 能耗

论文提出 Token-Energy Law,把 GPU 上稠密模型每生成一个 token 的能耗归结为两个变量:活跃字节数(区分权重共享读取与每请求私有的 KV 缓存读取)和实际解码时间。两者以仿射形式组合,系数按模型和硬件/运行时配置标定。实验覆盖三个模型在 NVIDIA Hopper 和 Blackwell GPU 上的 BF16 解码,包括 Hopper 上双卡张量并行,留出条件下能量估计误差不超过 3.73%。该公式可解释批处理如何摊薄权重读取成本、更长的注意力历史如何增加 KV 需求。

Current read

这篇论文把每 token 的电耗拆成活跃字节加解码时间两个量,误差 3.73% 以内,做推理部署省电的可以照着算。

11 sources31° AI heatLast updated 2026/10/20 17:32:08

Evidence

11 sources
Primary sourcearXiv cs.LG
View source
Related source 1Yuchen Jin
View source
Related source 2marktechpost
View source
Related source 3Teortaxes
View source
Related source 5掘金本周最热
View source
Related source 6TestingCatalog
View source
Related source 8LMSYS Org (SGLang)
View source
Related source 10@atomic_chat_hq
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.