You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@mxnet.apache.org by "Xinyu Chen (JIRA)" <ji...@apache.org> on 2018/04/29 01:24:00 UTC

[jira] [Commented] (MXNET-290) MKLDNN support for model quantization

    [ https://issues.apache.org/jira/browse/MXNET-290?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16457876#comment-16457876 ] 

Xinyu Chen commented on MXNET-290:
----------------------------------

Happy MayDay!


> MKLDNN support for model quantization
> -------------------------------------
>
>                 Key: MXNET-290
>                 URL: https://issues.apache.org/jira/browse/MXNET-290
>             Project: Apache MXNet
>          Issue Type: New Feature
>            Reporter: Xinyu Chen
>            Priority: Major
>          Time Spent: 13h 20m
>  Remaining Estimate: 0h
>
> This PR add MKLDNN support for model quantization. Accuracy gap between fp32 and int8 is less than 1% for resnet and inception. Further optimization will be submitted based on this PR.



--
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