We are excited to announce that SolverStudio now supports developing and solving models developed using the Pyomo Python modelling environment. To quote the Pyomo web site:

*Python Optimization Modeling Objects (Pyomo) package is an open source tool for modeling optimization applications in Python. Pyomo can be used to define symbolic problems, create concrete problem instances, and solve these instances with standard solvers. Pyomo provides a capability that is commonly associated with algebraic modeling languages such as AMPL, AIMMS, and GAMS, but Pyomo’s modeling o**bjects are embedded within a full-featured high-level programming language with a rich set of supporting libraries. Pyomo leverages the capabilities of the Coopr software library, which integrates Python packages for defining optimizers, modeling optimization applications, and managing computational experiments.*

To use Pyomo under SolverStudio, you need to install both Python and then Pyomo; please see the Pyomo install information. (In the future, we hope to run Pyomo directly within SolverStudio using its internal IronPython. It is great to have support from the COOPR/Pyomo team to help make this happen.)

To install and test Pyomo with SolverStudio:

- Make sure you have Python installed in your machine. If not, you can download Python from https://www.python.org/download . (Note that you need CPython, not the IronPython that is built into SolverStudio.)
- Run the pip.exe installer in your Python Scripts folder, as detailed in these instructions.
- Download and install SolverStudio
- Launch Excel
- To open the Pyomo Demo:
- In Excel, select the
*Data*menu tab. Under the SolverStudio menu, choose Examples then Pyomo Examples.xlsx. - Choose one of the model sheets, and click
*Solve*to run your new Pyomo SolverStudio solver.

- In Excel, select the

You may wish to visit the main the Pyomo site, the older COOPR/Pyomo site, seek help on the Pyomo forum, or file a Pyomo bug report.

**Pyomo 4.1.10527 SolverStudio Incompatibility (Sept 2015)**: As detailed in this Pyomo Forum post, Pyomo 4.1.10527 removed the –json option that SolverStudio uses, leading to the following error (as reported here – thanks mbanco):

1 2 3 4 5 |
## Running: C:\Python27\Scripts\Pyomo.exe --solver=cbc --save-results="C:\Users\amas008\AppData\Local\Temp\SolverStudio 2pdv1xei\Sheet" "C:\Users\amas008\AppData\Local\Temp\SolverStudio 2pdv1xei\model.py" "C:\Users\amas008\AppData\Local\Temp\SolverStudio 2pdv1xei\SheetData.dat" --json WARNING: converting to the 'pyomo solve' subcommand usage: Pyomo-script.py solve [options] <model_or_config_file> [<data_files>] Pyomo-script.py solve: error: argument --json: expected one argument ## Error: The Pyomo output file did not contain any valid data. |

The Pyomo development team has released a new version, Pyomo-4.2.10784, on 21/9/2015, which reinstates this –json option (thanks Bill). This restores compatibility with SolverStudio.

Hi, I’m having trouble with defining the relative gap tolerance. how can I do it? I’ve already set the time limit writing “solver-options | timelimit=3600” in the pyomo options’ table. thanks in advance

You need to find the Pyomo option to set the gap (which I don’t remember, sorry), and then add it to the PyomoOptions defined range (as shown for another option in the “Transportation-Pyomo” example sheet). Andrew

Thank you Andrew. I’ve found that I should write “options.parameter2009=0.05” somewhere to get a 5% relative gap, but I dont know in which column I should write it and what should I write in the other one. I tried with “solver-options” but it didnt work. I get an error executing the model code. Any help will be truthfully appreciated. Augusto.

You can use Python commands to set your Pyomo solver options; see Section 18.12 of the Pyomo documentation. Hope this help. Andrew

I am testing pyomo with gurobi. But it fails with error

C:\Python27\Scripts\Pyomo.exe –solver=Gurobi –save-results=”C:\Users\sam.walker\AppData\Local\Temp\SolverStudio knxky2fg\Sheet” “C:\Users\sam.walker\AppData\Local\Temp\SolverStudio knxky2fg\model.py” “C:\Users\sam.walker\AppData\Local\Temp\SolverStudio knxky2fg\SheetData.dat” –json

The SolverFactory was unable to create the solver “Gurobi”

and returned an UnknownSolver object. This error is raised at the point

where the UnknownSolver object was used as if it were valid (by calling

method “config_block”).

The original solver was created with the following parameters:

type: Gurobi

_args: ()

options: {}

## Error: The Pyomo output file did not contain any valid data.

## Pyomo did not complete; no solution is available. The sheet has not been changed.

## Done

When I take the command line and change it to lower case –solver=gurobi , it runs ok.

So, is it just a small bug?

Thanks for the feedback. It looks like Pyomo requires solver names to be given in lower case. You can manually set the solver (with the name “gurobi”) by following the example in “Pyomo Examples.xlsx : Transportation-Pyomo” where we define a named range “PyomoOptions” with an entry of “solver” and “gurobi”. We will fix this capitalisation issue in the next SolverStudio release. Sorry for the hassle, but I hope this gets you going. Andrew

I am getting an error TypeError:expected index value,got str when I am running pyomo optimization problem from Solverstudio. I am a novice in using this and anyhelp will be greatly appreciated

from coopr.pyomo import *

model = AbstractModel()

#defining Sets

model.t=set()

model.budget=Param(initialize=30000)

#Defining Variables

model.Newspaper=Var(model.t,bounds=(0,1000),within=NonNegativeReals)

model.BSCS=Var(model.t,within=NonNegativeReals)

model.DDTB=Var(model.t,within=NonNegativeReals)

model.SEM=Var(model.t,within=NonNegativeReals)

model.DBIMP=Var(model.t,within=NonNegativeReals)

#Constraint

def totalConstraint(model):

return sum(model.Newspaper[k]+model.BSCS[k]+model.DDTB[k]+model.SEM[k]+model.DBIMP[k] for k in model.t)==model.budget

model.totconstraint=Constraint(rule=totalConstraint)

#Objective Function

def maximizeconversion(model):

ans=0

for i in model.t:

ans=ans+model.Newspaper[i]+model.BSCS[i]+model.DDTB[i]+model.SEM[i]+model.DBIMP[i]

return ans

model.SolverResult=Objective(rule=maximizeconversion,sense=maximize)

have already defined t in edit data option as a data range with values from 1 to 30.I am just experimenting with solverstudio and the constraints as such might not make any practical sense now.

Thanks for your interest in SolverStudio. Is this an error reported by SolverStudio in a pop-up dialog, or a Pyomo-reported error shown in the text output window? Andrew

I’m getting the same TypeError, and it’s in a pop-up dialog. Any help greatly appreciated!

Can you please tell us what versions of SolverStudio, Pyomo and Python you are using? Perhaps you could also email the workbook to email hidden; JavaScript is required Sorry for this hassle; we try to work across a wide range of scenarios whenever possible. Andrew

I have installed Python 2.78 & Coopr_3.4.7842 & while trying solving coopr example I receive following

[ 0.00] Setting up Pyomo environment

[ 0.00] Applying Pyomo preprocessing actions

[ 0.00] Creating model

[ 0.03] Applying solver

[ 0.09] Processing results

Number of solutions: 1

Solution Information

Gap: 0.0

Status: optimal

Function Value: 15.58

[ 0.17] Applying Pyomo postprocessing actions

[ 0.17] Pyomo Finished

## COOPR did not complete; no solution is available. The sheet has not been changed.

## Done

Sorry to hear it is not working for you. I have checked, and all the COOPR examples provided with SolverStudio work fine for me using Coopr 3.4.7842 (CPython 2.7.5 on Windows 7). I doubt that the slightly different Python would have any impact. What example are you trying to solve? Note that if you choose File… View Working Files, and then open the Sheet file (using a text editor), you will see what was produced by COOPR when solving your model. Andrew

Hello Andrew,

Im trying to use SolverStudio and COOPR ( or other languages) for a linear optimization model. Before trying to learn how to use COOPR and SolverStudio I wanted to know if it would have the capability of solving 2,000,000 variables and about 1,000,000 constraints.

Do you have any suggestions regarding what to use for my problem?

Thank you

1 million constraints is a large problem; I don’t know how any of the modelling languages would handle that sized problem. What are the 2 million variables – what do they represent? Is this an integer programme?

It is a maximization linear model. the variables are binary variables, which refers to having and not adding an item to a portfolio. The reason for the large number of variables is that my binary variables have 6 dimensions and the total number of variables is the total permutations that they can take. Any suggestion would be extremely helpful.

I can’t find useful info on the limitations of each of these solvers yet.

There are few limitations on these solvers other than memory and time. Predicting how much of these you will need is very hard, which is why you will not find any clear problem size specifications given. I suggest you test your problem using a smaller sized experiment, perhaps trying out AMPL on NEOS or PuLP with CBC. I look forward to hearing how you get on. Andrew

I have attempted to solve my problem with OpenSolver( Which I think uses CBC solver, please correct me if I’m wrong) and after a certain expansion I made to the model. The cbc ran for a 4 days and didn’t stop when I limit the time to make sure it ends it returns fractions instead of binary variables. So if I use PULP and then CBC would it be the same story? The same solver power?

Yes, OpenSolver uses CBC. PuLP can use CBC and other commercial solvers. Using PuLP with CBC should give the same answer as OpenSolver, so it won’t really help you. The commercial solvers can be much faster, so worth trying if you can find someone with one. You can also use Gurobi on NEOS with AMPL; see http://www.neos-server.org/neos/solvers/milp:Gurobi/AMPL.html Hope this helps, Andrew

Pingback: INFORMS SolverStudio Presentation MD08 « SolverStudio for Excel