Deepspeed ZeRO系列算法原理+通信开销详解

原文: https://sumanthrh.com/post/distributed-and-efficient-finetuning/#zero-powered-data-parallelism

posted @ 2024-06-26 23:39 marsggbo 阅读(154) 评论(0) 编辑收藏举报

刷新页面返回顶部