singa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SINGA-349) Create layer operations for autograd
Date Fri, 18 May 2018 04:53:00 GMT

    [ https://issues.apache.org/jira/browse/SINGA-349?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16480139#comment-16480139
] 

ASF subversion and git services commented on SINGA-349:
-------------------------------------------------------

Commit 5abcc6e0dd803fa8a631d957765891cf5e836ec3 in incubator-singa's branch refs/heads/master
from xuewanqi
[ https://git-wip-us.apache.org/repos/asf?p=incubator-singa.git;h=5abcc6e ]

SINGA-349 Create layer operations for autograd

1. layer operations works well.
2. following steps :
  - to change the API to Pytorch style.
  - add flags in forward and backward function
  - realize changeable initialization methods.


> Create layer operations for autograd
> ------------------------------------
>
>                 Key: SINGA-349
>                 URL: https://issues.apache.org/jira/browse/SINGA-349
>             Project: Singa
>          Issue Type: New Feature
>            Reporter: Xue Wanqi
>            Priority: Major
>
> Add more larger autograd operations, such as convolution_2d(), pooling().
> These operations can be implemented in two ways:
> 1.create cpp operations and export them to python vis swig
> 2. by calling some existing layers, package the corresponding layers into new operations.
> The second method is preferred as it is much more easier.



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

Mime
View raw message