Distributed Horovod

Description

The current support for Horovod is limited to one host.
Before Horovod can be supported on multiple hosts the hops-util-py library allreduce module must support it. In addition to changes on HopsWorks. This Jira tracks both these issues.

Status

Assignee

Ermias Gebremeskel

Reporter

Robin

Labels

None

Affects versions

Priority

Highest
Configure