Ring allreduce achieves more scalable synchronous SGD training than parameter-server or peer-to-peer architectures because its per-step communication cost is nearly independent of worker count.
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
REJECT 1representative citing papers
citing papers explorer
-
Performance Analysis and Comparison of Distributed Machine Learning Systems
Ring allreduce achieves more scalable synchronous SGD training than parameter-server or peer-to-peer architectures because its per-step communication cost is nearly independent of worker count.