GRACE-MoE integrates expert grouping, dynamic replication, and locality-aware routing with hierarchical sparse communication to reduce end-to-end latency in distributed SMoE inference.
The design and implementation of all algorithms and experiments, as well as the analysis of results and conclusions, were independently conducted and written by the authors
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GRACE-MoE: Grouping and Replication with Locality-Aware Routing for Efficient Distributed MoE Inference
GRACE-MoE integrates expert grouping, dynamic replication, and locality-aware routing with hierarchical sparse communication to reduce end-to-end latency in distributed SMoE inference.