Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
in part 2.3,it says "Currently, we support the sequential and eventual consistency." however ,th kv-store implemetion in parameter server support bounded-delay consistency model.
I want to know whether mxnet support bounded-delay consistency model now?
And what method the distributred traning process deploy? BSP, SSP or other ?
Description
I read the paper of MXNet
in part 2.3,it says "Currently, we support the sequential and eventual consistency." however ,th kv-store implemetion in parameter server support bounded-delay consistency model.
I want to know whether mxnet support bounded-delay consistency model now? And what method the distributred traning process deploy? BSP, SSP or other ?