2018-01-24 JAN

Tensor Flow

1. Playground for topology optimization

http://playground.tensorflow.org

2. Conventional local Approach

Course at Lynda.com <-- remember this was supposed to be the bulk of this session. We'll explore some alternatives depending on your reaction.

TensorFlow 1.0 also comes with a dockerized version hosting of its own Jupyter Notebook (snapshot pict)!

Let's start by setting your docker environment (link above) and then Get Started.

For future reference - as soon as you have a good handle on using TF from the dockerized TF+Jupyter environment let's have a close look at the a) making sure Docker is taking advantage of GPUs, and b) using the same TF service directly on Google's TPU-backed hardware. For reaching GPUs from Docker in machines with NVIDIA boards see this link for more info.

3. Advanced web computing deployment of TF

Two options for consumer-facing deployment:

1. Training and deploying in the browser: deeplearnjs.org

2. Training locally/cloud, deploying in the browser: keras.io, keras.js