Gurobi & Python

Gurobi is mathematical optimization software, a suite of solvers for mixed-integer and linear programming problems.

Academic licensing exists for individuals and institutions; the clusters have an installation that uses an institutional license.

Jupyter Notebooks

In order to utilize Gurobi from a Jupyter notebook via Python, you need a complete installation of Gurobi¹, Python, Jupyter, and the gurobipy module in your Python environment, as well as a valid license file.

Setup

For software projects, it is good practice to use a tool to manage packages and dependencies. The conda command ships with Anaconda and Miniconda, among others, and is favored on the clusters.

We will use this tool to set up our project.

1. Access the cluster environment via a shell.²

2. Load Anaconda via the Environment Modules system. For this example, we will use version 2023.3.

    module load anaconda3/2023.3
   

   Anaconda prefixes your shell prompt with the name of the current Anaconda environment. The “base” environment is the default, but you should never work in the “base”. The next few steps address this.

3. Create a conda environment.

   This command downloads most of Gurobi and the gurobipy module from the gurobi conda channel, as well as the ipykernel module, which allows Jupyter to talk to Python. For this example, the environment is named gpyproject, which you can change to any name you’d like.

    conda create -y --name gpyproject --channel gurobi gurobi ipykernel
   

   The clusters have an existing Gurobi installation, so this is a little redundant, but it simplifies reproducibility of our development environment.

4. Visit the web interface for the cluster and select Jupyter under the Interactive Apps menu.

5. Select the proper Anaconda version for the “Anaconda3 version used for starting up jupyter interface” option, noting that our example above used 2023.3.

6. Click Launch. Once your session is allocated a button labeled “Connect to Jupyter” will appear.

7. Click the New button within Jupyter and select the option that includes the environment name. Our example used gpyproject.

8. Test Gurobi. I like formulating and solving a simple MIP model.

    from gurobipy import *
    try:
        # Create a new model
        m = Model("mip1")
        # Create variables
        x = m.addVar(vtype=GRB.BINARY, name="x")
        y = m.addVar(vtype=GRB.BINARY, name="y")
        z = m.addVar(vtype=GRB.BINARY, name="z")
        # Set objective
        m.setObjective(x + y + 2 * z, GRB.MAXIMIZE)
        # Add constraint: x + 2 y + 3 z <= 4
        m.addConstr(x + 2 * y + 3 * z <= 4, "c0")
        # Add constraint: x + y >= 1
        m.addConstr(x + y >= 1, "c1")
        m.optimize()
        for v in m.getVars():
            print('%s %g' % (v.varName, v.x))
        print('Obj: %g' % m.objVal)
    except GurobiError as e:
        print('Error code ' + str(e.errno) + ": " + str(e))
    except AttributeError:
        print('Encountered an attribute error')
   

   The output should look something like this.

    Restricted license - for non-production use only
    Gurobi Optimizer version 10.0.1 build v10.0.1rc0 (linux64)
    Optimize a model with 2 rows, 3 columns and 5 nonzeros
    Model fingerprint: 0x98886187
    Variable types: 0 continuous, 3 integer (3 binary)
    Coefficient statistics:
      Matrix range     [1e+00, 3e+00]
      Objective range  [1e+00, 2e+00]
      Bounds range     [1e+00, 1e+00]
      RHS range        [1e+00, 4e+00]
    Found heuristic solution: objective 2.0000000
    Presolve removed 2 rows and 3 columns
    Presolve time: 0.00s
    Presolve: All rows and columns removed
       
    Explored 0 nodes (0 simplex iterations) in 0.01 seconds (0.00 work units)
    Thread count was 1 (of 64 available processors)
   
    Solution count 2: 3 2 
       
    Optimal solution found (tolerance 1.00e-04)
    Best objective 3.000000000000e+00, best bound 3.000000000000e+00, gap 0.0000%
    x 1
    y 0
    z 1
    Obj: 3