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

Posted to dev@singa.apache.org by "SHEYUJIAN (JIRA)" <ji...@apache.org> on 2018/05/30 06:25:00 UTC

[jira] [Updated] (SINGA-340) Support Keras APIs

     [ https://issues.apache.org/jira/browse/SINGA-340?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

SHEYUJIAN updated SINGA-340:
----------------------------
    Attachment: Keras_backend.xlsx

> Support Keras APIs
> ------------------
>
>                 Key: SINGA-340
>                 URL: https://issues.apache.org/jira/browse/SINGA-340
>             Project: Singa
>          Issue Type: New Feature
>          Components: PySINGA
>            Reporter: wangwei
>            Assignee: wangwei
>            Priority: Major
>         Attachments: Keras_backend.xlsx
>
>
> Keras provides simple [functional APIs|https://keras.io/models/model/] for beginners. Initially, it was developed using Theano as the backend. Two more backends have been added, [Tensorflow and CNTK|https://keras.io/backend/].
> It would be good to support the Keras APIs by SINGA. In other words, we add SINGA as another backend of Keras. Then Keras users can use SINGA's optimization for memory and speed without changing their code at all.
> To support Keras APIs, we mainly need to implement the functional APIs by calling SINGA's layer functions. Sub-tasks (and details) will be added in other tickets. 



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