Keynote (TensorFlow Dev Summit 2017)

Jeff Dean, Rajat Monga, and Megan Kacholia delivered the keynote address at the inaugural TensorFlow Dev Summit. They discuss:
- The origins of TensorFlow
- Progress since TensorFlow’s open-source launch
- TensorFlow’s thriving open-source community
- TensorFlow performance and scalability
- TensorFlow applications around the world
... and share some exciting announcements!
Visit the TensorFlow website for all session recordings: goo.gl/bsYmza
Subscribe to the Google Developers channel at goo.gl/mQyv5L
event: TensorFlow Dev Summit 2017; re_ty: Publish;

[@chenjus]

Where can I get that tensorflow shirt?

[@rotichbill637]

The cucumber sorting system is a great idea especially for agriculture based countries like Kenya...

[@PatrickNgobiro]

true. It has many agricultural applications

[@explorermr30]

beautiful

[@CristianChaparroa]

What is the golang repository for Tensorflow?

[@PeterSodhi]

Awesome

[@caenorstfuji]

Can we have more details about the server configuration you use when you benchmark multi-GPU ? The numbers look very interesting, multi middle-end gpu may be now more interesting than 1 high-end now ? Have you done more test with differents configurations/GPUs ?

[@saanvisharma2081]

Caenorst what's the intentions behind making tensor open for everyone

[@Matt-yu8xc]

At 3:42, is that logo on the bottom right for the R programming language? I've never seen that one before.

[@Matt-yu8xc]

ah yes, thanks!

[@Brysett]

TensorFlow on as many platforms as possible? GPU being one of them? Sorta only half truth.
It supports CUDA. That's NVidia.
As far as I know there is still no support for Radeon GPU's, which is sad considering the upcoming Radeon Instinct (Vega), which supposedly is a slight bit faster (GFlop wise) than the GTX 1080 (and probably a heck of a lot cheaper).
I hope Radeon support will be added (hopefully sooner than later).

[@JialuTu]

Having Radeon support would be awesome for Tensorflow! However, I had always thought that the reason why most GPU processing is currently done on NVidia is more due to the libraries that they have provided for their GPU, which has done a fantastic job in supporting this area. So it would be more for AMD to play the catch up game rather than for Google to fit it in.

[@caenorstfuji]

Radeon GPU's are no use. Until now there is no CuDNN equivalent, Nvidia have built CuDNN, not Google or any open source developper. I don't see why Google should spend effort to built a CuDNN equivalent while everything is working perfectly well with Nvidia. It's AMD whom have to do the effort, so blame them.

[@complexinteger]

I know they are working on OpenCL support. github.com/tensorflow/tensorflow/issues/22
I haven't tried it out, but if you build from source there is an OpenCL option. It isn't officially supported so I'm assuming aspects of Tensorflow are broken with the OpenCL build.

[@motivationmusic4821]

Cant find information about "Neuron Architecture search" 19:23 ... Can somebody help?

[@andrewjohnrichardson2444]

openreview.net/pdf?id=r1Ue8Hcxg
I think thats the paper you are looking for :)

[@motivationmusic4821]

thank you!

[@alph4966]

ジェフディーン氏はエンジニア界の神

[@jackreeceejini2522]

I wish I could do Tensorflow in Wolfram language

[@FranckDernoncourt]

8:58 58x improvement compared to what? 1 GPU?

[@FranckDernoncourt]

1 GPU indeed, answer at 18:31

[@motivationmusic4821]

kind of marketing, I guess

[@WalterReade]

Franck Dernoncourt The point isn't really that TF is "fast" but rather that it scales efficiently.

[@johnhammer8668]

Jeff dean has grown backwards.

[@CariagaXIII]

he got younger?

[@AmCanTech]

So two cucumber farmers have a son who is an embeded systems engineer, sounds like a joke!

[@eldrenbahianlorcasr3204]

THIS IS ELDREN BAHIAN LORCA SR,162 WORLD BOOKS OF GENES RECORDS IN FACEBOOK AND 31 OTHER TABS IN THE PHILIPPINES,MINDANAO,ILIGAN CITY.SELF STUDY IN ADVANCED DOCTORATE IN ADVANCED DERIVATIONS OF GENERAL QUANTUM GEO-ASTROPHYSICS FORMULATIONS EQUILIBRIUM COMBINATIONS IN 7-8-9 UNTIL NOW.