Flink-JPMML - Streaming Machine Learning Model Serving on Flink

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Flink-JPMML - Streaming Machine Learning Model Serving on Flink

Roberto Bentivoglio
Hi Everyone,

As promised we've published as open source flink-jpmml, a library that
we've developed internally to serve PMML models within Flink using
JPMML-Evaluator[1]. We've completed a refactoring of the library during the
last weeks, so we're proud to share with the Flink community our effort.

The library is available on our Github page[2] and we published also a blog
post containing further details about the library API and its internals[3].

We've decided finally to stay with the same licence coming from
JPMML-Evaluator (GNU AGPL 3.0 license) accordingly to the guidelines of the
Apache Software Foundation[4].

I want to mention the names of the guys that ideated the latest version of
the library, Andrea Spina and Francesco Frontera, and the guys that
developed the first prototype more than a year ago (Simone Robutti and
Stefano Baghino). Thanks for the work! Happy to mention also that Francesco
has worked widely on this project during his BSc thesis.

Looking forward to receive feedback from the community, feel free to open
bugs, enhancement proposals and contribute opening PRs, we will be more
than happy to improve the library with the community.

Thanks,
Radicalbit Team

[1] - Java Evaluator API for PMML - https://github.com/jpmml/jpmml-evaluator
[2] - flink-jpmml - https://github.com/radicalbit/flink-jpmml
[3] - Flink-JPMML - Streaming Machine Learning Model Serving on Flink -
http://radicalbit.io/flink-jpmml-streaming-machine-
learning-model-serving-flink-part-1/
[4] - Apache License and GPL compatibility - https://www.apache.org/
licenses/GPL-compatibility.html

--
Roberto Bentivoglio
CTO
e. [hidden email]

Radicalbit S.r.l.
Via Pietro Borsieri 41,
20159 Milan - Italy
radicalbit.io
Reply | Threaded
Open this post in threaded view
|

Re: Flink-JPMML - Streaming Machine Learning Model Serving on Flink

Fabian Hueske-2
Hi Roberto,

thanks for sharing Flink-JPMML!
IMO, model serving is a very important use case that many users will have
to deal with.
It's good to have a library to support this task.

Cheers, Fabian

2017-05-25 16:07 GMT+02:00 Roberto Bentivoglio <
[hidden email]>:

> Hi Everyone,
>
> As promised we've published as open source flink-jpmml, a library that
> we've developed internally to serve PMML models within Flink using
> JPMML-Evaluator[1]. We've completed a refactoring of the library during the
> last weeks, so we're proud to share with the Flink community our effort.
>
> The library is available on our Github page[2] and we published also a blog
> post containing further details about the library API and its internals[3].
>
> We've decided finally to stay with the same licence coming from
> JPMML-Evaluator (GNU AGPL 3.0 license) accordingly to the guidelines of the
> Apache Software Foundation[4].
>
> I want to mention the names of the guys that ideated the latest version of
> the library, Andrea Spina and Francesco Frontera, and the guys that
> developed the first prototype more than a year ago (Simone Robutti and
> Stefano Baghino). Thanks for the work! Happy to mention also that Francesco
> has worked widely on this project during his BSc thesis.
>
> Looking forward to receive feedback from the community, feel free to open
> bugs, enhancement proposals and contribute opening PRs, we will be more
> than happy to improve the library with the community.
>
> Thanks,
> Radicalbit Team
>
> [1] - Java Evaluator API for PMML - https://github.com/jpmml/
> jpmml-evaluator
> [2] - flink-jpmml - https://github.com/radicalbit/flink-jpmml
> [3] - Flink-JPMML - Streaming Machine Learning Model Serving on Flink -
> http://radicalbit.io/flink-jpmml-streaming-machine-
> learning-model-serving-flink-part-1/
> [4] - Apache License and GPL compatibility - https://www.apache.org/
> licenses/GPL-compatibility.html
>
> --
> Roberto Bentivoglio
> CTO
> e. [hidden email]
>
> Radicalbit S.r.l.
> Via Pietro Borsieri 41,
> 20159 Milan - Italy
> radicalbit.io
>
Reply | Threaded
Open this post in threaded view
|

Re: Flink-JPMML - Streaming Machine Learning Model Serving on Flink

Ted Yu
Roberto:
I enjoyed reading part 1 of the blog.
Any plan on publishing the part 2 of the blog ?

Cheers

On Fri, May 26, 2017 at 3:11 PM, Fabian Hueske <[hidden email]> wrote:

> Hi Roberto,
>
> thanks for sharing Flink-JPMML!
> IMO, model serving is a very important use case that many users will have
> to deal with.
> It's good to have a library to support this task.
>
> Cheers, Fabian
>
> 2017-05-25 16:07 GMT+02:00 Roberto Bentivoglio <
> [hidden email]>:
>
> > Hi Everyone,
> >
> > As promised we've published as open source flink-jpmml, a library that
> > we've developed internally to serve PMML models within Flink using
> > JPMML-Evaluator[1]. We've completed a refactoring of the library during
> the
> > last weeks, so we're proud to share with the Flink community our effort.
> >
> > The library is available on our Github page[2] and we published also a
> blog
> > post containing further details about the library API and its
> internals[3].
> >
> > We've decided finally to stay with the same licence coming from
> > JPMML-Evaluator (GNU AGPL 3.0 license) accordingly to the guidelines of
> the
> > Apache Software Foundation[4].
> >
> > I want to mention the names of the guys that ideated the latest version
> of
> > the library, Andrea Spina and Francesco Frontera, and the guys that
> > developed the first prototype more than a year ago (Simone Robutti and
> > Stefano Baghino). Thanks for the work! Happy to mention also that
> Francesco
> > has worked widely on this project during his BSc thesis.
> >
> > Looking forward to receive feedback from the community, feel free to open
> > bugs, enhancement proposals and contribute opening PRs, we will be more
> > than happy to improve the library with the community.
> >
> > Thanks,
> > Radicalbit Team
> >
> > [1] - Java Evaluator API for PMML - https://github.com/jpmml/
> > jpmml-evaluator
> > [2] - flink-jpmml - https://github.com/radicalbit/flink-jpmml
> > [3] - Flink-JPMML - Streaming Machine Learning Model Serving on Flink -
> > http://radicalbit.io/flink-jpmml-streaming-machine-
> > learning-model-serving-flink-part-1/
> > [4] - Apache License and GPL compatibility - https://www.apache.org/
> > licenses/GPL-compatibility.html
> >
> > --
> > Roberto Bentivoglio
> > CTO
> > e. [hidden email]
> >
> > Radicalbit S.r.l.
> > Via Pietro Borsieri 41,
> > 20159 Milan - Italy
> > radicalbit.io
> >
>
Reply | Threaded
Open this post in threaded view
|

Re: Flink-JPMML - Streaming Machine Learning Model Serving on Flink

Roberto Bentivoglio
Hi Guys,

Many thanks for your feedback!

@Ted Andrea Spina will probably publish the second part of the blog post
within a couple of weeks.

Best,
Roberto

Il giorno sab 27 mag 2017 alle 01:15 Ted Yu <[hidden email]> ha
scritto:

> Roberto:
> I enjoyed reading part 1 of the blog.
> Any plan on publishing the part 2 of the blog ?
>
> Cheers
>
> On Fri, May 26, 2017 at 3:11 PM, Fabian Hueske <[hidden email]> wrote:
>
> > Hi Roberto,
> >
> > thanks for sharing Flink-JPMML!
> > IMO, model serving is a very important use case that many users will have
> > to deal with.
> > It's good to have a library to support this task.
> >
> > Cheers, Fabian
> >
> > 2017-05-25 16:07 GMT+02:00 Roberto Bentivoglio <
> > [hidden email]>:
> >
> > > Hi Everyone,
> > >
> > > As promised we've published as open source flink-jpmml, a library that
> > > we've developed internally to serve PMML models within Flink using
> > > JPMML-Evaluator[1]. We've completed a refactoring of the library during
> > the
> > > last weeks, so we're proud to share with the Flink community our
> effort.
> > >
> > > The library is available on our Github page[2] and we published also a
> > blog
> > > post containing further details about the library API and its
> > internals[3].
> > >
> > > We've decided finally to stay with the same licence coming from
> > > JPMML-Evaluator (GNU AGPL 3.0 license) accordingly to the guidelines of
> > the
> > > Apache Software Foundation[4].
> > >
> > > I want to mention the names of the guys that ideated the latest version
> > of
> > > the library, Andrea Spina and Francesco Frontera, and the guys that
> > > developed the first prototype more than a year ago (Simone Robutti and
> > > Stefano Baghino). Thanks for the work! Happy to mention also that
> > Francesco
> > > has worked widely on this project during his BSc thesis.
> > >
> > > Looking forward to receive feedback from the community, feel free to
> open
> > > bugs, enhancement proposals and contribute opening PRs, we will be more
> > > than happy to improve the library with the community.
> > >
> > > Thanks,
> > > Radicalbit Team
> > >
> > > [1] - Java Evaluator API for PMML - https://github.com/jpmml/
> > > jpmml-evaluator
> > > [2] - flink-jpmml - https://github.com/radicalbit/flink-jpmml
> > > [3] - Flink-JPMML - Streaming Machine Learning Model Serving on Flink -
> > > http://radicalbit.io/flink-jpmml-streaming-machine-
> > > learning-model-serving-flink-part-1/
> > > [4] - Apache License and GPL compatibility - https://www.apache.org/
> > > licenses/GPL-compatibility.html
> > >
> > > --
> > > Roberto Bentivoglio
> > > CTO
> > > e. [hidden email]
> > >
> > > Radicalbit S.r.l.
> > > Via Pietro Borsieri 41,
> > > 20159 Milan - Italy
> > > radicalbit.io
> > >
> >
>