How do you install glpk-solver along with pyomo in

Question

I want to use "pyomo" for my studies. I installed pyomo via easy_install coopr install instructions, Pyomo needs a solver to work so I wanted to install the (GNU Linear Programming Kit) glpk_webpage, pyomo seems to be installed just right because I can import it in spyder (i am using WinPython-64bit-2.7.5.3) with import coopr.pyomo

However, I cannot do anything without glpk I guess...

I downloaded glpk-4.52 (latest version) from the ftp server but I do not know what to do with the batch files I found in the "w64"-folder I should use(?) according to "Installing GLPK"

I do not have Visual Studio installed - Isn't it possible to work without it?

Answer 1

Better late than never: in order to use GLPK (executable glpsol.exe), it must be somewhere on your sytem environment variable "Path". For sake of an example, let's assume you put the GLPK executable into the folder C:\GLPK\bin. Then (steps copied from this answer by melhosseiny):

1. Hold Win and press Pause.
2. Click Advanced System Settings.
3. Click Environment Variables.
4. Append ;C:\GLPK\binto the Path variable.
5. Restart command prompt.

Now try to launch glpsol from any directory. If it is found, pyomo should now be able to use it.

Answer 2

I recently installed GLPK for use with python 3.5 and pyomo under windows 7 and would like to report how I succeeded. I installed pyomo via:

pip install pyomo

Then, download WinGLPK 4.55 from here: WinGLPK
This does not work for newer versions at the moment.
Unzip it and copy the whole w64 folder to C:\w64
Include folder C:\w64 in your system PATH (so that glpsol.exe is found).

Check your installation using the simple example from the official pyomo documentation:

pyomo solve abstract1.py abstract1.dat --solver=glpk 

The files abstract1.py and abstract1.dat can also be found in the pyomo documentation.

I hope this will help the next desperate GLPK installer.

Answer 3

If you are using Anaconda, both pyomo and glpk can be installed with conda install. In the Windows terminal, activate your conda environment, then:

conda install -c conda-forge pyomo
conda install -c conda-forge glpk

To test the glpk installation:

glpsol --help

Answer 4

FYI, you can now use Coopr without installing local solvers. The latest Coopr release supports an interface with the NEOS solver. For example, if your MILP model is in the file model.py, then the following command would optimize the model using CBC:

pyomo --solver-manager=neos --solver=cbc model.py