1. Python Setup (Google Colab & Jupyter Notebook with Pyomo)
2. Set up for AMPL

1.1 Download Binaries & Install Pyomo

For Google Colab (Nightly Build):

!wget -q https://www.ieor.iitb.ac.in/files/faculty/amahajan/minotaur/nightly/origin/minotaur-nightly.tar.gz
!tar -zxf minotaur-nightly.tar.gz
!rm -f mbin
!ln -s minotaur-nightly mbin
!pip install -q pyomo

For Jupyter Notebook (Stable 0.4.1 Release):

!wget -nc https://www.ieor.iitb.ac.in/files/faculty/amahajan/minotaur/bin/minotaur-0.4.1-linux-x86_64.tar.gz
!tar -zxf minotaur-0.4.1-linux-x86_64.tar.gz
!ln -s minotaur-0.4.1-linux-x86_64/bin mbin
!pip install -q pyomo

1.2 Create Optimization Model in Pyomo

Define your mixed-integer optimization model using Pyomo environment:

from pyomo.environ import *

model = ConcreteModel()
model.x1 = Var(within=Integers)
model.x2 = Var(within=Integers)
model.obj = Objective(expr=5.5*model.x1 + 2.1*model.x2, sense=maximize)
model.con1 = Constraint(expr=-model.x1 + model.x2 <= 2)
model.con2 = Constraint(expr=8*model.x1 + 2*model.x2 <= 17)

1.3 Solve Model using Minotaur Solver

Use one of Minotaur's solvers (mbnb, mqg, mglob, mmultistart) to solve the model. Here, we use the mglob solver:

# In Colab use '/content/mbin/mglob', in local Jupyter use './mbin/mglob'
mntr = SolverFactory("mglob", executable='./mbin/mglob')

mntr.options['--log_level'] = 2
mntr.options['--time_limit'] = 60
# mntr.options['--qp_engine'] = 'None'
# mntr.options['--nlp_engine'] = 'IPOPT'

# Set tee to True if you want to see verbose solver log output
result = mntr.solve(model, tee=True)

1.4 View Solution & Solver Results

After solving the model, view termination status, optimal objective value, bounds, and runtime:

print("Solver termination status:", result.solver.status)
print("Solver termination condition:", result.solver.termination_condition)
print("Best solution value:", model.obj())
print("Best bound: ", result.problem.lower_bound)
print("Solver time:", result.solver.time)

To use Minotaur with AMPL, follow the steps below:

1. Ensure you have Minotaur and AMPL installed on your system. If Minotaur is not installed, download the precompiled binaries from the Installation page.
2. Extract the downloaded .tar.gz archive, open the extracted folder, and navigate to the bin/ subfolder.
3. Copy the four binaries/executables (mbnb, mqg, mglob, mmultistart) from the bin/ folder to your main AMPL directory that contains the ampl executable.

Create AMPL Model (example.mod)

# AMPL Model (example.mod)
var x1 integer;
var x2 integer;

maximize obj: 5.5*x1 + 2.1*x2;
subject to con1: -x1 + x2 <= 2;
subject to con2: 8*x1 + 2*x2 <= 17;

display x1, x2, obj;

AMPL Commands to Set Up & Solve

Use the following AMPL commands to set up and solve the model with Minotaur's mglob solver:

# Load model file
model example.mod;

# If Minotaur is built/installed in a specific directory:
option solver "/path/to/minotaur/build/bin/mglob";

# If binaries/executables are in main AMPL directory:
option solver mglob;

# Solve the optimization problem
solve;

# Display output solution
display x1, x2, obj;

Replace /path/to/minotaur/build/bin/mglob with the actual path to your Minotaur executable.