| 标题 |
Drop-Upcycling: Training Sparse Mixture of Experts with Partial Re-initialization |
| 网址 | |
| DOI |
10.48550/arxiv.2502.19261
doi
|
| 其它 |
期刊:International Conference on Learning Representations 作者:Taishi Nakamura; Takuya Akiba; Kazuki Fujii; Yusuke Oda; Rio Yokota; et al 出版日期:2025-02-26 |
| 求助人 | |
| 下载 | 求助已完成,仅限求助人下载。 |
PDF的下载单位、IP信息已删除
(2025-6-4)