Eager-SGD relaxes global synchronization in distributed SGD with solo and majority partial allreduces, speeding up training under load imbalance while preserving accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Taming Unbalanced Training Workloads in Deep Learning with Partial Collective Operations
Eager-SGD relaxes global synchronization in distributed SGD with solo and majority partial allreduces, speeding up training under load imbalance while preserving accuracy.