Repository files navigation

This repository contains code to accompanythe O'Reilly tutorial on generative adversarial networks written byJon Bruner andAdit Deshpande. Seethe original tutorial to run this code in a pre-built environment on O'Reilly's servers with cell-by-cell guidance, or run these files on your own machine.

There are three versions of our simple GAN model in this repository:

• gan-notebook.ipynb is identical to the interactive tutorial, available here so that you can run it on your own machine.
• gan-script.py is a straightforward Python script containing code drawn directly from the tutorial, to be run from the command line. Note that it doesn't print anything when it's executed, but it does send regular updates toTensorBoard so that you can track its progress.
• gan-script-fast.py is a modest refactoring of gan-script.py that runs slightly faster because more of its computations are contained in the TensorFlow graph.

In order to rungan-script.py organ-script-fast.py, you'll needTensorFlow version 1.0 or later andNumPy. In order to rungan-notebook.ipynb, you'll additionally needJupyter andmatplotlib.

If you've already got TensorFlow on your machine, then you've got NumPy and should be able to run the raw Python scripts.

The easiest way to install TensorFlow as well as NumPy, Jupyter, and matplotlib is to start with the Anaconda Python distribution.

1. Follow theinstallation instructions for Anaconda Python.We recommend using Python 3.6.

2. Follow the platform-specificTensorFlow installation instructions. Be sure to follow the "Installing with Anaconda" process, and create a Conda environment namedtensorflow.

3. If you aren't still inside your Conda TensorFlow environment, enter it by opening your terminal and typing

   source activate tensorflow

4. Download and unzipthis entire repository from GitHub, either interactively, or by entering

   git clone https://github.com/jonbruner/generative-adversarial-networks.git

5. Usecd to navigate into the top directory of the repo on your machine

6. Launch Jupyter by entering

   jupyter notebook

   and, using your browser, navigate to the URL shown in the terminal output (usuallyhttp://localhost:8888/)