Create a conda environment and test on binder

tonyfast commented 8 years ago

This is a really cool tutorial @parsing-science. Thank you for helping me use TensorFlow and PyMC for the first time! I was able to reproduce the notebooks on python=3 if I removed the byte strings; that is part of this pull request.

I also replaced requirements.txt with environment.yaml to create conda environments. conda environments can use pip; pymc3 is installed with git+.

To run your notebooks, simply download Anaconda for free and follow the instructions in the readme.md. The workflow is very similar to using virtualenv.

I suggest trying Anaconda for a few reasons:

At PyData events, most folks are using python through Anaconda.
conda installs for the numpy and scipy stack are more reliable than pip,
- It is easiest to access the bleeding edge versions of packages through conda-forge.
On reliability, conda installs will save time in your tutorial. It removes many potential installation problems across Mac, Windows, and Linux.

Another option too is Binder; free on demand notebook compute.

Clicking the badge above will launch a working version of your notebooks in the browser; I can qualify that the logistic notebook does indeed work. I have used Binder for a few tutorials; zero install issues. The only install issue is if Binder is working.

Good luck with your tutorial!

parsing-science commented 8 years ago

Thanks @tonyfast! I'm testing it now. I noticed that the environment installs python 3.5. Is there a way to allow the user to choose which python to use? I've only ever used Python 2.7. Some of the errors in my tutorial (e.g. the bytestrings) arise from using Python 2.7, and I wanted to be able to reproduce them during the tutorial.

tonyfast commented 8 years ago

you just have to add python=2.7 to the dependencies.

name: pymc3_quickstart_guide
dependencies: 
- python=2.7
- anaconda
- pip
- pip:
  - unidecode
  - git+https://github.com/pymc-devs/pymc3

Also, you can change environments within the notebook if you are using Anaconda,

parsing-science commented 8 years ago

@tonyfast: thanks for all of the help; I really appreciate it. I put in some instructions to that effect in my README and got the yml file in.

tonyfast commented 8 years ago

Nice one! Good luck during your tutorial.

parsing-science / pymc3_quickstart_guide

Create a conda environment and test on binder #1