You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@systemml.apache.org by "Niketan Pansare (JIRA)" <ji...@apache.org> on 2017/04/30 19:26:04 UTC
[jira] [Created] (SYSTEMML-1569) Test MLContext for robustness and
scalability
Niketan Pansare created SYSTEMML-1569:
-----------------------------------------
Summary: Test MLContext for robustness and scalability
Key: SYSTEMML-1569
URL: https://issues.apache.org/jira/browse/SYSTEMML-1569
Project: SystemML
Issue Type: Test
Affects Versions: SystemML 1.0
Reporter: Niketan Pansare
As more APIs are getting built on top of MLContext and with large-scale demos using MLContext and notebooks, we should test MLContext for robustness and scalability. The goal is that using MLContext should have constant overhead compared to commandline execution (both using similar formats).
As an example: we should check for potential OOM in Script History logic: https://github.com/apache/incubator-systemml/blob/master/src/main/java/org/apache/sysml/api/mlcontext/MLContextUtil.java#L902
If we uncomment https://github.com/apache/incubator-systemml/blob/master/src/main/java/org/apache/sysml/api/mlcontext/MLContextUtil.java#L897-L901, then you should get an OOM when passing large Numpy array with Python MLContext. This is because toString() method on MatrixBlock converts double [] into String.
[~deron] [~mwdusenb@us.ibm.com] [~reinwald] [~mboehm7]
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)