1 paper
Guowei Liu, Hongming Li, Yaning Guo +5
Deploying large-scale MoE models presents challenges in memory capacity and bandwidth for expert activation. While Attention-FFN Disaggregation (AFD) has emerged as a potential arc…