[
https://issues.apache.org/jira/browse/SYSTEMML-845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15408671#comment-15408671
]
Mike Dusenberry edited comment on SYSTEMML-845 at 8/5/16 12:19 AM:
-------------------------------------------------------------------
[~niketanpansare] A full run of {{mnist_lenet-train.dml}} in Spark local mode
with 50GB driver memory (and other settings as seen in {{perf.sh}}) had the
following output:
{code}
Total execution time: 8230.825 sec.
Number of executed Spark inst: 137923.
{code}
However, {{lenet-train.dml}} run with the same settings ran as follows:
{code}
Total execution time: 2927.089 sec.
Number of executed Spark inst: 4.
{code}
So, these two scripts have the same performance in forced singlenode mode, but
different performance when run with Spark (in local mode), even with an
excessive amount of memory (50GB -- will run in 10GB or less on a laptop).
Awesome chance to make an optimizer improvement for major gains.
cc [~mboehm7]
was (Author: [email protected]):
[~niketanpansare] A full run of {{mnist_lenet-train.dml}} in Spark local mode
with 50GB driver memory (and other settings as seen in {{perf.sh}}) had the
following output:
{code}
Total execution time: 8230.825 sec.
Number of executed Spark inst: 137923.
{code}
However, {{lenet-train.dml}} run with the same settings ran as follows:
{code}
Total execution time: 2927.089 sec.
Number of executed Spark inst: 4.
{code}
So, these two scripts have the same performance in forced singlenode mode, but
different performance when run with Spark (in local mode). Awesome chance to
make an optimizer improvement for major gains.
cc [~mboehm7]
> Compare Performance of LeNet Scripts With & Without Using SystemML-NN
> ---------------------------------------------------------------------
>
> Key: SYSTEMML-845
> URL: https://issues.apache.org/jira/browse/SYSTEMML-845
> Project: SystemML
> Issue Type: Improvement
> Reporter: Mike Dusenberry
> Attachments: convert.dml, lenet-train-spark-explain.log,
> log08.03.16-1470268602.txt, mnist_lenet-train-spark-explain.log, perf.sh,
> run.sh
>
>
> This JIRA issue tracks the comparison of the performance of the LeNet scripts
> with & without using SystemML-NN. The goal is that they should have equal
> performance in terms of both accuracy and time. Any difference will be
> indicate areas of engine improvement.
> Scripts:
> * [mnist_lenet-train.dml |
> https://github.com/apache/incubator-systemml/blob/master/scripts/staging/SystemML-NN/examples/mnist_lenet-train.dml]
> - LeNet script that *does* use the SystemML-NN library.
> * [lenet-train.dml |
> https://github.com/apache/incubator-systemml/blob/master/scripts/staging/lenet-train.dml]
> - LeNet script that *does not* use the SystemML-NN library.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
