返回全部动态
复现研究揭示FLOPs估算公式在新硬件上的局限性
原标题：FLOPs vs Real Work: The Importance of Replication in AI Efficiency Assessment
AI 摘要
该论文旨在复现一项关于FLOPs与执行时间关系的研究，以验证其估算公式在新硬件上的适用性。复现结果证实了原始论文的论点，即FLOPs不能直接反映执行时间，但发现新硬件上执行时间存在不稳定性和不连续性，公式往往低估了实际时间。研究还指出了原始研究在复现材料方面的不足，并提供了完整的复现包。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
FLOPs vs Real Work: The Importance of Replication in AI Efficiency Assessment Abstract. AI efficiency has recently taken the spotlight in both academy and industry due to massive model scales, high energy demands, and environmental costs. While reporting Floating Point Operations (FLOPs) is a traditional approach for assessing computational costs, the relationship between FLOPs and execution time is not straightforward, as layers with the same number of FLOPs may not have the same execution time
发布时间：2026-08-18 12:00
抓取时间：2026-08-18 12:00
来源机构：arXiv