You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@singa.apache.org by "Luo Zhaojing (Jira)" <ji...@apache.org> on 2021/05/13 12:50:00 UTC

[jira] [Commented] (SINGA-515) Improve CNN training efficiency by moving augmentation outside of training iterations

    [ https://issues.apache.org/jira/browse/SINGA-515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17343848#comment-17343848 ] 

Luo Zhaojing commented on SINGA-515:
------------------------------------

For benchmark datasets, e.g., CIFAR-10 and ImageNet, the augmentation method is pre-defined before the model training, so the augmentation does not need to be carried out on the fly

> Improve CNN training efficiency by moving augmentation outside of training iterations
> -------------------------------------------------------------------------------------
>
>                 Key: SINGA-515
>                 URL: https://issues.apache.org/jira/browse/SINGA-515
>             Project: Singa
>          Issue Type: Improvement
>            Reporter: Luo Zhaojing
>            Assignee: Luo Zhaojing
>            Priority: Minor
>
> Improve CNN training efficiency by moving augmentation out of training iterations
>  
> In examples/largedataset_cnn/train_largedata.py, the augmentation should be moved outside of training iterations



--
This message was sent by Atlassian Jira
(v8.3.4#803005)