mxnet-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Rahul Huilgol (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (MXNET-22) Support for distributed training based on NCCL Kvstore
Date Tue, 27 Feb 2018 06:32:00 GMT

     [ https://issues.apache.org/jira/browse/MXNET-22?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Rahul Huilgol updated MXNET-22:
-------------------------------
    Component/s: MXNet Engine

> Support for distributed training based on NCCL Kvstore
> ------------------------------------------------------
>
>                 Key: MXNET-22
>                 URL: https://issues.apache.org/jira/browse/MXNET-22
>             Project: Apache MXNet
>          Issue Type: New Feature
>          Components: MXNet Engine
>            Reporter: Rahul Huilgol
>            Priority: Major
>
> NCCL kvstore is currently supported for a single machine only. When training is distributed,
there is no way to use NCCL for aggregation of gradients from different GPUs on a single machine.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org


Mime
View raw message