Google TensorFlow :: Remember, remember, the velvet November


node:	Google TensorFlow
template:	4
parent:	Toolkits
owner:	ulkas
viewed by:
created:	01.09.2016 - 10:22:41

cwbe coordinatez:
101
63533
8106902
8207635
8207639
8207634

ABSOLUT
KYBERIA

permissions
you:	r,
system:	public
net:	yes

maniac	0
popo	0
dan0	0
studenyRadiator	0

https://www.tensorflow.org/

TensorFlow™ is an open source software library for numerical computation using data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. The flexible architecture allows you to deploy computation to one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. TensorFlow was originally developed by researchers and engineers working on the Google Brain Team within Google's Machine Intelligence research organization for the purposes of conducting machine learning and deep neural networks research, but the system is general enough to be applicable in a wide variety of other domains as well.

00000101000635330810690208207635082076390820763408603308

https://ai.googleblog.com/2019/03/introducing-gpipe-open-source-library.html

00000101000635330810690208207635082076390820763408603297

https://www.tensorflow.org/federated/

Google on Wednesday released TensorFlow Federated, open-source software that incorporates federated learning, an AI training system. It works by using data that's spread out across a lot of devices, such as smartphones and tablets, to teach itself new tricks. But rather than send the data back to a central server for study, it learns on your phone or tablet itself and sends only the lesson back to the app maker.
https://www.cnet.com/news/google-ai-tool-lets-outside-apps-get-smart-without-taking-all-your-data/

00000101000635330810690208207635082076390820763408460686

mate niekto skusenost s vyuzitim externej (thunderbolt3) NVIDIA grafiky na OSX s CUDA/Tensorflow?

prekompiloval som si tf pre ntb moj CPU (AVX2 / SSE4.2 etc) ale stale je to pomale (ofc), tak rozmyslam zohnat externu grafiku a hrat na s tym...

00000101000635330810690208207635082076390820763408351197

zdravicko,

mam problem/otazku, potrebujem v praci rozbehat tensorflow s podporou GPU na pocitaci kde mame 2 graficke karty: 1x Tesla C2050 na vypocty a Quadro FX580 na render na displej.. bezime tam debian jessie

problem:
tensorflow tvrdi, ze potrebuje CUDA 8.0, realne mu staci 7.0 a cuDNN 5.1, realne mu staci 4.0.. skusil som najprv nainstalovat CUDA 8.0, to islo v pohode, avsak po restarte sa mi stazoval ze nevie nastartovat lightdm.. vo /var/log/Xorg.0.log som sa dozvedel, ze problem ma, lebo Quadro (render karta) podporuje maximalne nvidia drivers 420, kdezto s CUDA 8.0 sa nainstalovali 473 (tusim, proste novsie).. tak ze ok, pozriem ktora CUDA naposledy podporuje 420 a to je bohuzial 6.5.. skusil som tu, ale tensorflow nerozchodim pod nou.. fakt potrebuje aspon 7.0.. na CUDA 7.0 vsak potrebujem novsie drivery na nvidia a tym padom mi uz Quadro nechce nic zobrazit.. (ked som mal tu chvilku nainstalovanu CUDA 8.0, tak cez TTY som zistil ze funguje v pohode, cize Tesla si to podporuje, problem je ale to Quadro.. nemozme tam bohuzial fungovat len cez cmd pristup.. chodi tam par ludi co potrebuje desktop a X server..)

mozne riesenie:
napadlo ma, ze ci sa da mat dva separe drivery na dve separe karty... nikdy som nic podobne nerobil.. napad je na Teslu mat CUDA 8.0, cize nividia-473 a na Quadro mat starsie 420.. alebo na Quadro kludne bezat pod xserver-nouveau driverami (tie boli predtym a no problemo)... pri googleni som nasiel moznost ako v xorg.conf nastavit cez ktoru kartu sa ma renderovat displej (skrz jej PCI adresu), to je ok, ale nikde som nevidel ako by sa dalo mat dva rozne drivery...
neviete, je to mozne? nejake skusenosti? pripadne aj ine riesenie co vas napada?

dikicko

00000101000635330810690208207635082076390820763408344919

Druhá generácia sa od prvej výrazne odlišuje a je univerzálnejšia. Nová verzia je hlavne určená už aj na trénovanie neurónových sietí, pričom už podporuje aj výpočty s číslami v pohyblivej rádovej čiarke a výkon TPU dosahuje až 180 teraflopov / s. Detailnú špecifikáciu novej generácie zatiaľ Google nezverejnil.

Kým prvú generáciu využíval výlučne Google na implementáciu svojich služieb, druhú generáciu sprístupňuje pod označením Cloud TPU ľubovoľným záujemcom s možnosťou využitia virtuálnych strojov s TPU v rámci jeho služby Google Compute Engine. TPU sa dajú programovať pomocou frameworku TensorFlow.

00000101000635330810690208207635082076390820763408327292

Sonnet is a new open source library announced by Alphabet’s DeepMind. It is built on top of their existing machine learning library TensorFlow along with extra features that fit DeepMind’s research requirements. Sonnet is designed to make it easier to create complex neural networks using TensorFlow.
https://github.com/deepmind/sonnet

00000101000635330810690208207635082076390820763408304338

TensorFlowOnSpark brings TensorFlow programs onto Apache Spark clusters.

yahoo zrobilo.
https://github.com/yahoo/TensorFlowOnSpark

00000101000635330810690208207635082076390820763408304337

00000101000635330810690208207635082076390820763408302344

At Google's inaugural TensorFlow Dev Summit in Mountain View, California, today, Google announced the release of version 1.0 of its TensorFlow open source framework for deep learning, a trendy type of artificial intelligence. Google says the release is now production-ready by way of its application programing interface (API). But there are also new tools that will be part of the framework, which includes artificial neural networks that can be trained on data and can then make inferences about new data. Now there are more traditional machine learning tools, including K-means and support vector machines (SVMs), TensorFlow's engineering director, Rajat Monga, said at the conference. And there's an integration with the Python-based Keras library, which was originally meant to ease the use of the Theano deep learning framework. And there are now "canned estimators," or models, Monga said, including simple neural networks to start using quickly.
http://venturebeat.com/2017/02/15/google-releases-tensorflow-1-0-with-new-machine-learning-tools/

00000101000635330810690208207635082076390820763408220830

https://research.googleblog.com/2016/09/show-and-tell-image-captioning-open.html
https://github.com/tensorflow/models/tree/master/im2txt

A TensorFlow implementation of the image-to-text model described in the paper:

"Show and Tell: Lessons learned from the 2015 MSCOCO Image Captioning Challenge."

Oriol Vinyals, Alexander Toshev, Samy Bengio, Dumitru Erhan.

IEEE transactions on pattern analysis and machine intelligence (2016).

Full text available at: http://arxiv.org/abs/1609.06647