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

Posted to dev@ignite.apache.org by Denis Magda <dm...@apache.org> on 2018/09/26 21:19:45 UTC

ML examples wrap logic in IgniteThread. Why?

Yury, ML folks,

I've mentioned a strange thing. Looks like every example we have wraps up
its logic in the following block

IgniteThread igniteThread = new
IgniteThread(ignite.configuration().getIgniteInstanceName(),
    KMeansClusterizationExample.class.getSimpleName(), () -> {


//ML specific stuff (training, predicting, calculations, etc.)

});

igniteThread.start();
igniteThread.join();


Why do we do that?

Denis

Re: ML examples wrap logic in IgniteThread. Why?

Posted by Denis Magda <dm...@apache.org>.

Thanks Yury, good to know about that!

--
Denis

On Wed, Sep 26, 2018 at 3:49 PM Юрий Бабак <y....@gmail.com> wrote:

> Denis,
>
> Thanks for this notice, actually this is some kind of atavism. Run this
> code inside IgniteThread was a requirement when we had a distributed
> matrices. But now all our algorithms builds over distributed datasets and
> we don't need it anymore.
>
> I created JIRA ticket <https://issues.apache.org/jira/browse/IGNITE-9711>
> for this.
>
> Thanks,
> Yuriy
>
> чт, 27 сент. 2018 г. в 0:20, Denis Magda <dm...@apache.org>:
>
> > Yury, ML folks,
> >
> > I've mentioned a strange thing. Looks like every example we have wraps up
> > its logic in the following block
> >
> > IgniteThread igniteThread = new
> > IgniteThread(ignite.configuration().getIgniteInstanceName(),
> >     KMeansClusterizationExample.class.getSimpleName(), () -> {
> >
> >
> > //ML specific stuff (training, predicting, calculations, etc.)
> >
> > });
> >
> > igniteThread.start();
> > igniteThread.join();
> >
> >
> > Why do we do that?
> >
> > Denis
> >
>

Re: ML examples wrap logic in IgniteThread. Why?

Posted by Юрий Бабак <y....@gmail.com>.

Denis,

Thanks for this notice, actually this is some kind of atavism. Run this
code inside IgniteThread was a requirement when we had a distributed
matrices. But now all our algorithms builds over distributed datasets and
we don't need it anymore.

I created JIRA ticket <https://issues.apache.org/jira/browse/IGNITE-9711>
for this.

Thanks,
Yuriy

чт, 27 сент. 2018 г. в 0:20, Denis Magda <dm...@apache.org>:

> Yury, ML folks,
>
> I've mentioned a strange thing. Looks like every example we have wraps up
> its logic in the following block
>
> IgniteThread igniteThread = new
> IgniteThread(ignite.configuration().getIgniteInstanceName(),
>     KMeansClusterizationExample.class.getSimpleName(), () -> {
>
>
> //ML specific stuff (training, predicting, calculations, etc.)
>
> });
>
> igniteThread.start();
> igniteThread.join();
>
>
> Why do we do that?
>
> Denis
>