EDDIS: Accelerating Distributed Data-Parallel DNN Training for Heterogeneous GPU Cluster.
Shinyoung AhnHooyoung AhnHyeonseong ChoiJaehyun LeePublished in: IPDPS (Workshops) (2024)
Keyphrases
- distributed data
- training process
- distributed data sources
- data sharing
- parallel implementation
- distributed data mining
- parallel computing
- parallel processing
- parallel computation
- data distribution
- parallel programming
- databases
- communication cost
- graphics processing units
- integrating heterogeneous
- data mining algorithms
- data points
- machine learning
- massively parallel
- data mining applications
- parallel algorithm
- training set
- high dimensional data
- peer to peer
- similarity measure
- semantically heterogeneous