Login / Signup
Lancet: Accelerating Mixture-of-Experts Training via Whole Graph Computation-Communication Overlapping.
Chenyu Jiang
Ye Tian
Zhen Jia
Shuai Zheng
Chuan Wu
Yida Wang
Published in:
MLSys (2024)
Keyphrases
</>
graph theory
communities in social networks
training process
subgraph isomorphism
training set
graph structure
information sharing
overlapping communities
communication systems
random walk
training examples
connected components
mixture model
subject matter experts
graph theoretic
graph representation
directed acyclic graph
data sets
domain experts
structured data
test set
communication cost
graph matching
directed graph
graph model
communication technologies
parallel algorithm
graph based algorithm
training samples
neural network