Hy-Fi: Hybrid Five-Dimensional Parallel DNN Training on High-Performance GPU Clusters.
Arpan JainAamir ShafiQuentin AnthonyPouya KoushaHari SubramoniDhabaleswar K. PandaPublished in: ISC (2022)
Keyphrases
- graphics processing units
- training process
- parallel implementation
- parallel processing
- parallel computation
- parallel computing
- general purpose
- parallel programming
- clustering algorithm
- pc cluster
- distributed memory machines
- distributed memory
- graphics hardware
- highly parallel
- training algorithm
- parallel architectures
- parallel computers
- parallel hardware
- multi dimensional
- ibm sp
- cluster of workstations
- training phase
- online learning
- training set
- massively parallel
- test set
- shared memory
- parallel genetic algorithm
- scientific computing
- training data
- computing systems
- cluster analysis
- real time
- array processor
- supervised learning
- learning algorithm
- hierarchical clustering
- computer architecture
- floating point
- fuzzy clustering
- data clustering