Zhongzhu's Blog
Keep
Home
About
Tags
Archives
0%
AF-Pipe
Tag
2026
05-14
DisagMoE：用解耦 Attention 和 FFN 打通 MoE 训练的 all-to-all 瓶颈
05-14
DisagMoE: Disaggregating Attention and FFN to Beat the MoE All-to-All Bottleneck