Meta

Intelligent Elastic Feature Fading: Enabling Model Retrain-Free Feature Efficiency Rollouts at Scale

智能弹性特征衰减：实现无模型重训练特征效率大规模 rollout

Jieming Di, Xiaoyu Chen, Ying She, Siyu Wang, Lizzie Liu, Fenggang Wu, Jiaoying Mu, Tony Tsui, Amr Elroumy, Hsing Tang, Zewei Jiang, Qiao Yang, Lin Qi, Haibo Lin, Weifeng Cui, Daniel Li, Kapil Gupta, Shivendra Pratap Singh, Jie Zheng, Arnold Overwijk, Ling Leng, Sri Reddy, Robert Malkin, Rocky Liu

机构 * Meta 纠错

AI总结 本文提出IEFF系统，通过弹性控制特征覆盖与分布，实现无重训练的特征效率rollout，提升效率5倍，减少GPU开销，防止性能下降。

Comments 8 pages, 2 figures, 3 tables

Elimination Templates in Macaulay2

Macaulay2中的消元模板

Manav Batavia, Cheng Chen, Anna Natalie Chlopecki, Timothy Duff, William Huang, Aolong Li, Wanchun Shen

机构 * Department of Mathematics, Purdue University, 150 N University St., West Lafayette, IN 47907, USA ； Department of Mathematics, University of Wisconsin, 480 Lincoln Dr., Madison, WI 53706, USA ； Department of Mathematics, University of Missouri, 810 Rollins St, Columbia MO 65203, USA ； Meta Platforms, Inc., 1 Meta Way, Menlo Park, California 94025, USA ； Department of Mathematics, Harvard University, 1 Oxford St., Cambridge, MA 02138, USA 纠错

AI总结 本文介绍Macaulay2中的消元模板包，用于构造依赖于代数无关参数的零维素理想自动求解器，并展示其在计算机视觉中的应用。

Comments 13 pages with references

Learning Rate Transfer in Normalized Transformers

归一化变换器中的学习率转移

Boris Shigida, Boris Hanin, Andrey Gromov

机构 * Meta Superintelligence Labs（Meta超智能实验室） ； Princeton University（普林斯顿大学） 纠错

AI总结 本文提出νGPT模型，通过结合数值实验和对齐指数，改进了超参数转移方法，实现了模型宽度、深度和token horizon上的学习率转移。