Dynamic resizing of Horovod-based ring architecture training jobs is feasible, and a simulator using measured run times cuts average job completion time by more than half under moderate contention.
Open source high performance computing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Dynamic Scheduling of MPI-based Distributed Deep Learning Training Jobs
Dynamic resizing of Horovod-based ring architecture training jobs is feasible, and a simulator using measured run times cuts average job completion time by more than half under moderate contention.