Hyper-Connections

We present hyper-connections, a simple yet effective method that can serve as an alternative to residual connections. This approach specifically addresses common drawbacks observed in residual connection variants, such as the seesaw effect between gradient vanishing and representation collapse. Theo...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-11
Main Authors: Zhu, Defa, Huang, Hongzhi, Huang, Zihao, Zeng, Yutao, Mao, Yunyao, Wu, Banggu, Min, Qiyang, Zhou, Xun
Format: Article
Language:English
Subjects:
Large language models
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We present hyper-connections, a simple yet effective method that can serve as an alternative to residual connections. This approach specifically addresses common drawbacks observed in residual connection variants, such as the seesaw effect between gradient vanishing and representation collapse. Theoretically, hyper-connections allow the network to adjust the strength of connections between features at different depths and dynamically rearrange layers. We conduct experiments focusing on the pre-training of large language models, including dense and sparse models, where hyper-connections show significant performance improvements over residual connections. Additional experiments conducted on vision tasks also demonstrate similar improvements. We anticipate that this method will be broadly applicable and beneficial across a wide range of AI problems.
ISSN:2331-8422