Everything open source
May 16–17, 2016: Training & Tutorials
May 18–19, 2016: Conference
Austin, TX

Diving into machine learning through TensorFlow

Julia Ferraioli (Google), Amy Unruh (Google), Eli Bixby (Google)
9:00am–12:30pm Tuesday, 05/17/2016
Data
Location: Ballroom B Level: Intermediate
Average rating: ***..
(3.11, 27 ratings)

Prerequisite knowledge

Attendees must have a basic understanding of Python.

Materials or downloads needed in advance

This workshop will be pretty hands-on if you would like it to be -- though that's not necessary to attend.
If you want to play along, then please visit: http://bit.ly/tf-install-oscon prior to the workshop, and follow the installation instructions there.
There will also be a bit of time at the start of the tutorial for working through setup issues, but it will be smoother if you can do this ahead of time.

Description

TensorFlow is an open source software library from Google for numerical computation using data flow graphs. It provides a flexible platform for defining and running machine-learning algorithms and is particularly suited for neural net applications. Julia Ferraioli, Amy Unruh, and Eli Bixby demonstrate how to use TensorFlow to define, train, and utilize a variety of machine-learning algorithms on a number of datasets.

Julia, Amy, and Eli start by providing some background and motivation for problems in machine learning, as well as a brief history of the field, from the perspective of both Google and the machine-learning community as a whole. They then give a brief overview of how Google uses TensorFlow in its services before diving into an in-depth, hands-on exploration of TensorFlow.

Topics include:

  • Using pretrained models for classification and regression on a variety of devices
  • Scalably training models on a cluster, using algorithms bundled with the TensorFlow library or defined by machine-learning experts in the community, such as stochastic gradient descent
  • Implementing key machine-learning algorithms in TensorFlow
Photo of Julia Ferraioli

Julia Ferraioli

Google

Julia Ferraioli is a senior developer advocate with Google Cloud Platform. Julia is a polyglot, though in code only, and is excited about machine learning, containers, whiskey, and sprinkles (in roughly that order). Her super powers are finding ways to incorporate her interests into her work and estimating how much stuff can fit inside a container.

Photo of Amy Unruh

Amy Unruh

Google

Amy Unruh is a developer programs engineer at Google for the Google Cloud Platform, where she works with TensorFlow as well as many other Cloud Platform technologies. Amy has a PhD in CS/AI, has worked in academia, at several startups, and in industrial R&D, and has published a book on App Engine.

Photo of Eli Bixby

Eli Bixby

Google

Eli Bixby is a developer programs engineer at Google currently developing on Google Cloud Platform’s DevOps distributed systems, machine-learning, and big data offerings. He joined Google as a developer programs engineer. Previously, Eli dabbled in several research areas, with papers in biophysics, algorithmic game theory, and most recently computational biology.

Comments on this page are now closed.

Comments

Picture of Julia Ferraioli
Julia Ferraioli
05/17/2016 2:09am CDT

Ah, thanks for letting us know, Timothy! We’ll keep an eye out for that one.

Timothy Witham
05/16/2016 7:41pm CDT

I saw a similar failure on Linux. https://github.com/tensorflow/tensorflow/issues/1990 says it is due to cp34 in the .whl filename not matching the python 3.5 version. The file rename trick didn’t work for me (complained BadZipFile) but revert to python 3.4 did:

conda create -n tensorflow34 python=3.4
source activate tensorflow34

Satish Kesiboyana
05/16/2016 6:03pm CDT

Thanks Julia, this helped resolve this issue. See you tomorrow.

Picture of Julia Ferraioli
Julia Ferraioli
05/16/2016 8:50am CDT

Hi Satish! Thanks for running through the setup ahead of time.

I ran into that error as well when I was setting it up, so I’m pretty confident that the problem is that you should be using the following command to install:

$ pip install —ignore-installed —upgrade https://storage.googleapis.com/tensorflow/mac/tensorflow-0.8.0-py3-none-any.whl

The reason is that the instructions are for Python 3, and the TensorFlow whl you were trying to use is for Python 2.

See you tomorrow. Cheers!

Satish Kesiboyana
05/16/2016 7:48am CDT

Hello, I’m having issue during setup. After activating Conda environment, I get following error when trying to install TensorFlow:

tensorflow-0.8.0-py2-none-any.whl is not a supported wheel on this platform.

I upgraded my pip to 8.1.2.

Thanks