horn-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Edward J. Yoon" <edward.y...@samsung.com>
Subject RE: [DISCUSS] community development plan
Date Thu, 21 Jan 2016 02:43:31 GMT
Hmm, if google decide to donate tensorflow to the Apache too, I think we
seriously need to think about what we want to do.


--
Best Regards, Edward J. Yoon


-----Original Message-----
From: Edward J. Yoon [mailto:edward.yoon@samsung.com]
Sent: Wednesday, January 20, 2016 4:19 PM
To: dev@horn.incubator.apache.org
Subject: RE: [DISCUSS] community development plan

So, our strengths can be only:

1) more suitable for large neural networks simulation.
2) more flexible for analysis of evolving neural networks.
3) existing CPU cluster can be leveraged (easiness of configuration).

I personally think we can't compete with GPU systems (in term of
training/recognition speed), but 2) is quite interesting topic to me. Maybe
that's why I still work on this.

--
Best Regards, Edward J. Yoon


-----Original Message-----
From: Zachary Jaffee [mailto:zij@case.edu]
Sent: Wednesday, January 20, 2016 1:35 PM
To: Unknown
Subject: Re: [DISCUSS] community development plan

The takeaway is the the multi machine version of tensorflow is likely still
not fully developed. If anything, what they referred to on github is likely
just that they are using what was once distbelief for specific
applications, rather than the more general version. They very much have
gone for a GPU model of computation, where the developers mentioned that
their idea way to boost performance would be to have a server with a large
number of GPUs attached, while also noting how that isn't something you
would likely see anywhere now.

On Tue, Jan 19, 2016 at 7:06 PM, Zachary Jaffee <zij@case.edu> wrote:

> Quick note, they mentioned one of the bottle necks for releasing the
> tensorflow distributed version was they wanted it to run with gRPC and be
> able to run within kubernetes. From the looks of things tensorflow as is,
> is just something that they wanted to get out there, they commented issues
> with performance, and now with the most recent release is comparable in
> speed to torch. I'll update with more information later.
> On Jan 17, 2016 6:50 PM, "Edward J. Yoon" <edward.yoon@samsung.com> wrote:
>
>> > If we haven't done so already, we need a twitter account, I can manage
>> it
>>
>> When we entering Incubator, I've created Twitter account. I'll share the
>> passwd with PPMC members.
>>
>> > Another thing is that we can contact the newly formed openAI
>> organization
>> > and see if they want to work with us. I'll be attending this (>
>> > http://www.meetup.com/OpenAI-SV/events/227430778/) meetup, and I can>
>> > discuss apache horn. If you want to go over any talking points you want
>> me
>>
>> That's great. Please share trends with us.
>>
>> --
>> Best Regards, Edward J. Yoon
>>
>>
>> -----Original Message-----
>> From: Zachary Jaffee [mailto:zij@case.edu]
>> Sent: Friday, January 15, 2016 12:58 AM
>> To: Unknown
>> Subject: Re: [DISCUSS] community development plan
>>
>> Yeah, I've been very busy the past few weeks with the holidays and over
>> the
>> past two weeks moving, but I'll have time to do development over the next
>> week.
>>
>> If we haven't done so already, we need a twitter account, I can manage it
>> if you want, and should also work on making a logo for said account. We
>> are
>> in a difficult position, as there are a lot of other open source projects
>> in this space with communities already behind them, and I think it would
>> be
>> unlikely that we can pull those developers to work on horn. However, we
do
>> have the advantage of the apache name, maybe we can reach out to other
>> project teams and ask if people want to help with the project.
>>
>> Another thing is that we can contact the newly formed openAI organization
>> and see if they want to work with us. I'll be attending this (
>> http://www.meetup.com/OpenAI-SV/events/227430778/)
>> meetup, and I candiscuss
>> apache horn. If you want to go over any talking points you want me
>> to address, please let me know. One of us could also post a comment in
the
>> discussion below, and that will SEND an email to be sent to everyone who
>> is
>> in the event group (200+ people), all of who have an interest in AI. We
>> can
>> also try and reach out to the openAI organization directly, if thats
>> something your interested in.
>>
>> Also if you have any insight into the percentages of active email list
>> members for other projects, that might be interesting to hear.
>>
>> On Wed, Jan 13, 2016 at 4:11 AM, Edward J. Yoon <edwardyoon@apache.org>
>> wrote:
>>
>> > First of all, happy new year!
>> >
>> > It seems our subscribers continuously increasing (+4 increased from
>> > last month) but development activities and community momentum is quite
>> > low (maybe with long holidays, and others).
>> >
>> > Of course, we need to release 0.1 that shows our own strength, but do
>> > you have any ideas for developing our community?
>> >
>> > --
>> > Best Regards, Edward J. Yoon
>> >
>>
>>
>>
>> --
>> Zach Jaffee
>> B.S. Computer Science
>> Case Western Reserve University Class of 2017
>> Operations Director | WRUW FM 91.1 Cleveland
>> Secretary | Recruitment Chair | Phi Kappa Theta Fraternity
>> (917) 881-0646
>> zjaffee.com
>> github.com/ZJaffee
>>
>>
>>


--
Zach Jaffee
B.S. Computer Science
Case Western Reserve University Class of 2017
Operations Director | WRUW FM 91.1 Cleveland
Secretary | Recruitment Chair | Phi Kappa Theta Fraternity
(917) 881-0646
zjaffee.com
github.com/ZJaffee





Mime
View raw message