[
https://issues.apache.org/jira/browse/MXNET-158?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Olivier reassigned MXNET-158:
-----------------------------------
Assignee: Anirudh Subramanian (was: Chris Olivier)
Looks to be something with the engine or some other fundamental problem, since
the behavior can't be reproduced alone with the operators – but only with the
full resnet model (not a simpler model, either).
I spent several days on this, but I think it would be more efficient for Eric
to take a look.
> Training with the same parameters and seed gets significantly different
> results
> -------------------------------------------------------------------------------
>
> Key: MXNET-158
> URL: https://issues.apache.org/jira/browse/MXNET-158
> Project: Apache MXNet
> Issue Type: Bug
> Reporter: Rahul Huilgol
> Assignee: Anirudh Subramanian
> Priority: Critical
>
> https://github.com/apache/incubator-mxnet/issues/9410
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
