CVXPY installation instructions for non standard platforms

Clemens edited this page Sep 6, 2016 · 4 revisions

These instructions are only for non-standard platforms! See cvxpy.org for the main instructions.

Ubuntu 12.04

The SciPy available through apt-get is out of date on Ubuntu 12.04. Follow these instructions to install CVXPY:

  1. Install Anaconda. Follow the instructions on the web page.
  2. Create a new conda environment for CVXPY.

    conda create -n cvxpy-env python=2.7 numpy=1.8 scipy=0.14 cvxopt pip nose
    
  3. Switch into the new conda environment.

    source activate cvxpy-env
    
  4. Install ATLAS and gfortran (needed for SCS).

    sudo apt-get install libatlas-base-dev gfortran
    
  5. Export the location of the ATLAS library so SCS can find it.

    export ATLAS="/usr/lib/atlas-base/libatlas.so"
    
  6. Install CVXPY.

    pip install cvxpy
    

Debian

As of 2014-05-14 the debian wheezy repositories do not meet the minimum cvxpy requirements. If on jessie or sid, the installation is the same as above.

Archlinux

A

  pip install cvxpy

will download and install appropriate solvers and finish the installation. If you do not have / want root access, you can issue

  pip install --user cvxpy

in the user's temrinal.

Windows

64 bit

  1. Install Python(x,y) - This will take care of just about ALL the standard dependencies, painlessly. I personally uninstalled my python 2.7 releases that existed beforehand, removed the directory (by renaming it) to get rid of any libs, removed any other python versions from the PATH environment variable, and then installed Python(x,y).

  2. Install ECOS (requires the steps outlined in https://github.com/ifa-ethz/ecos/issues/53)

    pip install ecos
    
  3. Install cvxpy.

    pip install cvxpy
    

32 bit

  1. Install the 32bit version of Anaconda: https://store.continuum.io/cshop/anaconda/
  2. Install cvxopt using: https://code.google.com/p/pythonxy/downloads/list?can=2&q=cvxopt (This is a 32bit version)
  3. Add "C:\Anaconda\MinGW\bin" to your path
  4. Install ecos with pip.

    pip install ecos
    
  5. Install cvxpy with pip.

    pip install cvxpy