Introduction

Gurobi can be accessed through our own API in many programming languages, including gurobipy for Python. Although we recommend using the official APIs, many of our users come from a situation where they already have implemented a model in an open-source modeling framework and wish to solve that model with Gurobi without rewriting their code. This article describes how to do that for the most popular modeling frameworks

Comparison with official Gurobi APIs

There are important advantages to using the Gurobi APIs:

• They are officially supported by our R&D and Experts team
• They expose all Gurobi functionality, while much of that is hidden by modeling frameworks
• They avoid any overhead (runtime and memory) caused by handing over the model from framework to Gurobi
• They always work with the latest version of Gurobi

Using an existing model in an open-source modeling framework

This article won't go into detail on each modeling framework, but provides a high-level overview of the steps required to get started using Gurobi in each framework.

Pyomo

• Install the gurobipy package; you don't need the full Gurobi installation
• Create or download your Gurobi license file
• Create a solver instance using opt = SolverFactory("gurobi_direct"); you may need to import SolverFactory from pyomo.opt. Note that you can choose to either let Pyomo manage the creation of a Gurobi environment, or use the default environment. The first option is useful in case you want to provide environment parameter programmatically (e.g. WLS or Instant Cloud keys) without relying on a license file.
• Use that solver to optimize your model using results = opt.solve(model, ...)
• If you plan on solving the model more than once (making changes in between), it is recommended to use either SolverFactory("gurobi_persistent") or the APPSI interface. When using the latter, make sure to release the Gurobi license with opt.release_license() after you've solved your model.

PuLP

• Install the gurobipy package; you don't need the full Gurobi installation
• Create or download your Gurobi license file
• Assuming you have import pulp as pl and your model is defined in variable prob, solve your model using prob.solve(pl.GUROBI())
• From PuLP v2.8.0 you can use gurobipy environments via the use of the parameters env or manageEnv:
  • Option 1 (manage and provide an environment yourself):

import gurobipy as gp
import pulp as pl

options = {
    "WLSACCESSID": "access-id",
    "WLSSECRET": "secret",
    "LICENSEID": "license-id",
}

with gp.Env(params=options) as env:
    # Pass environment as a parameter
    solver = pl.GUROBI(env=env)
    prob.solve(solver)
    solver.close() # dispose the model

• • Option 2 (let PuLP manage the Gurobi environment):

import pulp as pl

options = {
    "WLSACCESSID": "access-id",
    "WLSSECRET": "secret",
    "LICENSEID": "license-id",
}

solver = pl.GUROBI(manageEnv=True, envOptions=options)
prob.solve(solver) 
solver.close()

Google OR-Tools

• Install the full Gurobi software; OR-Tools will not work if you only have gurobipy available
• Create or download your Gurobi license file
• At runtime, OR-Tools will search for the Gurobi shared library in the default install path of the Gurobi installers, or by using the GUROBI_HOME environment variable.
• Import the required package using from ortools.linear_solver import pywraplp
• Create your model solver using solver = pywraplp.Solver.CreateSolver("GUROBI") 
• Once the model is defined properly, call status = solver.Solve() to solve it with Gurobi

PythonMIP

• Install the full Gurobi software; PythonMIP will not work if you only have gurobipy available
• During installation, make sure to define the GUROBI_HOME environment variable
• Create or download your Gurobi license file
• PythonMIP will use Gurobi automatically

CVXPY

• Install the gurobipy package; you don't need the full Gurobi installation
• Create or download your Gurobi license file
• Assuming you have import cvxpy as cp and your model is defined in variable prob, solve your model using prob.solve(solver=cp.GUROBI())

Other resources

• How do I write an MPS model file from a third-party API?
• How do I use the token server with modeling frameworks?