[
https://issues.apache.org/jira/browse/SINGA-199?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15361362#comment-15361362
]
ASF subversion and git services commented on SINGA-199:
-------------------------------------------------------
Commit 7333517b492c49b9fcde22998063b84f98178fe3 in incubator-singa's branch
refs/heads/dev from [~flytosky]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-singa.git;h=7333517 ]
SINGA-199 Implement Python classes for SGD optimizers
Add regularizer and constraint python classes and all C++ optimizers.
Add swig files for optimzier and test_optimzier.py; add intializer.py
commit from @aaronwwf
- change module name from 'singa' to 'singa_wrap'
- fix model_optimizer.i naming mistake
Fix all bugs; Changed the CnMemPool's initialized and mtx fields to be
class members. Add a global counter to check the CnMemPool instance
numbers (which must be less than 2). CnMemPool must be used as a singleton,
because cnmemInit can only be called once. Python unittest may run test cases
in parallel, in which case we need to create cuda devices outside of test cases.
> Implement Python classes for SGD optimizers
> -------------------------------------------
>
> Key: SINGA-199
> URL: https://issues.apache.org/jira/browse/SINGA-199
> Project: Singa
> Issue Type: New Feature
> Reporter: wangwei
> Assignee: wangwei
>
> This ticket is going to add the python binding classes and functions for
> optimization (e.g, sgd) algorithms which are now implemented in C++.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
