You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@predictionio.apache.org by anand bilgaiyan <an...@gmail.com> on 2019/01/17 11:22:02 UTC

Query w.r.t. Prediction IO

Hi ,

This is anand. I was evaluatingg the capabilities of prediction IO
framework building and deploying the ML models and expose them over REST
service.

I have few queries :

1) can prediction IO be installed or run on window server / machine ?

2) if we want to build our ML models using python then can we remove the
scala templates dependencies ? I means DASE implementation of engine I do
not want to do in scala , instead can i implement this thing entirely in
python and deploy this under engine server ?

There are one or two templates on prediction IO page which are using python
as language but they also has scala code templates so not sure on this part
and no documentation as well which cover this .

please help me on these two because these will be deciding factor for me to
go or not go for prediction IO.

Regards,
Anand B

Re: Query w.r.t. Prediction IO

Posted by anand bilgaiyan <an...@gmail.com>.
Thanks Naoki for your Response. It was very helpful.

Regards,
Anand B

On Fri, Jan 18, 2019 at 9:55 AM Naoki Takezoe <ta...@gmail.com> wrote:

> The enhanced Python support allows only SparkML in the next release.
>
> I'm not sure if PredictionIO will be able to support 3rd party
> libraries in the near future.
> We will need consider how to support them in PredictionIO.
>
> On Fri, Jan 18, 2019 at 2:24 AM anand bilgaiyan
> <an...@gmail.com> wrote:
> >
> > Hi Nakoi,
> >
> > Thanks a lot for your response . it is certainly very helpful.
> >
> > one last question/query is what is we want to use some other ML libs
> instead of using spark Mlib which is default in prediction IO ? can we use
> the other 3rd party ML algos and use inside prediction IO engine ?
> >
> > Regards,
> > Anand B
> >
> > On Thu, Jan 17, 2019 at 6:12 PM Naoki Takezoe <ta...@gmail.com> wrote:
> >>
> >> Hi Anand,
> >>
> >> On Thu, Jan 17, 2019 at 8:22 PM anand bilgaiyan
> >> <an...@gmail.com> wrote:
> >> > 1) can prediction IO be installed or run on window server / machine ?
> >>
> >> No, PredictionIO doesn't work on Windows.
> >>
> >> > 2) if we want to build our ML models using python then can we remove
> the scala templates dependencies ? I means DASE implementation of engine I
> do not want to do in scala , instead can i implement this thing entirely in
> python and deploy this under engine server ?
> >>
> >> It's impossible to remove for now. But Python support is enhanced a
> >> lot in the next version. It makes possible to use Python without
> >> Scala.
> >> https://jira.apache.org/jira/browse/PIO-192
> >>
> >> --
> >> Naoki Takezoe
>
>
>
> --
> Naoki Takezoe
>

Re: Query w.r.t. Prediction IO

Posted by Naoki Takezoe <ta...@gmail.com>.
The enhanced Python support allows only SparkML in the next release.

I'm not sure if PredictionIO will be able to support 3rd party
libraries in the near future.
We will need consider how to support them in PredictionIO.

On Fri, Jan 18, 2019 at 2:24 AM anand bilgaiyan
<an...@gmail.com> wrote:
>
> Hi Nakoi,
>
> Thanks a lot for your response . it is certainly very helpful.
>
> one last question/query is what is we want to use some other ML libs instead of using spark Mlib which is default in prediction IO ? can we use the other 3rd party ML algos and use inside prediction IO engine ?
>
> Regards,
> Anand B
>
> On Thu, Jan 17, 2019 at 6:12 PM Naoki Takezoe <ta...@gmail.com> wrote:
>>
>> Hi Anand,
>>
>> On Thu, Jan 17, 2019 at 8:22 PM anand bilgaiyan
>> <an...@gmail.com> wrote:
>> > 1) can prediction IO be installed or run on window server / machine ?
>>
>> No, PredictionIO doesn't work on Windows.
>>
>> > 2) if we want to build our ML models using python then can we remove the scala templates dependencies ? I means DASE implementation of engine I do not want to do in scala , instead can i implement this thing entirely in python and deploy this under engine server ?
>>
>> It's impossible to remove for now. But Python support is enhanced a
>> lot in the next version. It makes possible to use Python without
>> Scala.
>> https://jira.apache.org/jira/browse/PIO-192
>>
>> --
>> Naoki Takezoe



-- 
Naoki Takezoe

Re: Query w.r.t. Prediction IO

Posted by anand bilgaiyan <an...@gmail.com>.
Hi Nakoi,

Thanks a lot for your response . it is certainly very helpful.

one last question/query is what is we want to use some other ML libs
instead of using spark Mlib which is default in prediction IO ? can we use
the other 3rd party ML algos and use inside prediction IO engine ?

Regards,
Anand B

On Thu, Jan 17, 2019 at 6:12 PM Naoki Takezoe <ta...@gmail.com> wrote:

> Hi Anand,
>
> On Thu, Jan 17, 2019 at 8:22 PM anand bilgaiyan
> <an...@gmail.com> wrote:
> > 1) can prediction IO be installed or run on window server / machine ?
>
> No, PredictionIO doesn't work on Windows.
>
> > 2) if we want to build our ML models using python then can we remove the
> scala templates dependencies ? I means DASE implementation of engine I do
> not want to do in scala , instead can i implement this thing entirely in
> python and deploy this under engine server ?
>
> It's impossible to remove for now. But Python support is enhanced a
> lot in the next version. It makes possible to use Python without
> Scala.
> https://jira.apache.org/jira/browse/PIO-192
>
> --
> Naoki Takezoe
>

Re: Query w.r.t. Prediction IO

Posted by Naoki Takezoe <ta...@gmail.com>.
Hi Anand,

On Thu, Jan 17, 2019 at 8:22 PM anand bilgaiyan
<an...@gmail.com> wrote:
> 1) can prediction IO be installed or run on window server / machine ?

No, PredictionIO doesn't work on Windows.

> 2) if we want to build our ML models using python then can we remove the scala templates dependencies ? I means DASE implementation of engine I do not want to do in scala , instead can i implement this thing entirely in python and deploy this under engine server ?

It's impossible to remove for now. But Python support is enhanced a
lot in the next version. It makes possible to use Python without
Scala.
https://jira.apache.org/jira/browse/PIO-192

-- 
Naoki Takezoe