def test_ipopt_solver(model_name):
current_dir = pathlib.Path(__file__).parent
osil_file = current_dir / 'models' / (model_name + '.osil')
pyomo_model = read_pyomo_model(osil_file)
problem = problem_from_pyomo_model(pyomo_model)
atol = rtol = 1e-4
galini = Galini()
galini.update_configuration({
'galini': {
'constraint_violation_tol': 1e-2,
},
'logging': {
'stdout': True,
},
'branch_and_cut': {
'tolerance': atol,
'relative_tolerance': rtol,
'root_node_feasible_solution_search_timelimit': 0,
'cuts': {
'maxiter': 100,
}
},
'cuts_generator': {
'generators': ['outer_approximation'],
},
'ipopt': {
'ipopt': {
'acceptable_constr_viol_tol': 1e-3
},
},
})
set_timelimit(30)
start_timelimit()
solver = BranchAndBoundSolver(galini)
solver.before_solve(pyomo_model, problem)
solution = solver.solve(problem)
assert solution.status.is_success()
sol_file = current_dir / 'solutions' / (model_name + '.sol')
expected_solution = read_solution(sol_file)
expected_objective = expected_solution['objective']
assert solution.objective is not None
assert is_close(expected_objective,
solution.objective.value,
atol=atol,
rtol=rtol)
expected_variables = expected_solution['variables']
for var_sol in solution.variables:
assert is_close(
expected_variables[var_sol.name],
var_sol.value,
atol=atol,
rtol=rtol,
)
def problem():
Q = [[28.0, 23.0, 0.0, 0.0, 0.0, 2.0, 0.0, 24.0],
[23.0, 0.0, -23.0, -44.0, 10.0, 0.0, 7.0, -7.0],
[0.0, -23.0, 18.0, 41.0, 0.0, -3.0, -5.0, 2.0],
[0.0, -44.0, 41.0, -5.0, 5.0, -1.0, 16.0, -50.0],
[0.0, 10.0, 0.0, 5.0, 0.0, -2.0, -4.0, 21.0],
[2.0, 0.0, -3.0, -1.0, -2.0, 34.0, -9.0, 20.0],
[0.0, 7.0, -5.0, 16.0, -4.0, -9.0, 0.0, 0.0],
[24.0, -7.0, 2.0, -50.0, 21.0, 20.0, 0.0, -45.0]]
C = [-44, -48, 10, 45, 0, 2, 3, 4, 5]
Qc = [
[-28, 13, 5],
[13, 0, 0],
[0, 0, 0],
]
m = aml.ConcreteModel("model_1")
m.I = range(8)
m.x = aml.Var(m.I, bounds=(0, 1))
m.f = aml.Objective(expr=sum(-Q[i][j] * m.x[i] * m.x[j] for i in m.I
for j in m.I) + sum(-C[i] * m.x[i]
for i in m.I))
m.c = aml.Constraint(expr=sum(Qc[i][j] * m.x[i] * m.x[j] for i in m.I[0:3]
for j in m.I[0:3]) >= -10)
return problem_from_pyomo_model(m)
def test_relaxed_problem():
m = pe.ConcreteModel()
m.I = range(10)
m.x = pe.Var(m.I, bounds=(0, 1))
m.obj = pe.Objective(expr=sum(m.x[i] * m.x[i] for i in m.I))
m.c0 = pe.Constraint(expr=sum(m.x[i] * m.x[i] for i in m.I) >= 0)
dag = problem_from_pyomo_model(m)
relaxed = RelaxedProblem(_linear_relaxation(dag), dag)
assert len(relaxed.relaxed.constraints) == 1 + 1 + 4 * 10
linear_constraint = LinearExpression([dag.variable(i) for i in m.I],
[i for i in m.I], 0.0)
relaxed.add_constraint('test_linear', linear_constraint, None, 0.0)
assert len(relaxed.relaxed.constraints) == 43
quadratic_constraint = QuadraticExpression(
[dag.variable(0)],
[dag.variable(1)],
[-2.0],
)
relaxed.add_constraint('test_quadratic', quadratic_constraint, 0.0, 0.0)
assert len(relaxed.relaxed.constraints) == 43 + 1 + 4
relaxed.add_constraint(
'test_mixed', SumExpression([linear_constraint, quadratic_constraint]),
0.0, None)
assert len(relaxed.relaxed.constraints) == 49
def test_linear_relaxation_with_quadratic_and_linear():
m = pe.ConcreteModel()
m.I = range(10)
m.x = pe.Var(m.I, bounds=(0, 1))
m.obj = pe.Objective(
expr=pe.quicksum(m.x[i]*m.x[i] for i in m.I) + pe.quicksum(m.x[i] for i in m.I)
)
m.c0 = pe.Constraint(
expr=pe.quicksum(2 * m.x[i] * m.x[i] for i in m.I) + pe.quicksum(m.x[i] for i in m.I) >= 0
)
dag = problem_from_pyomo_model(m)
relaxation = _linear_relaxation(dag)
relaxed = relaxation.relax(dag)
print_problem(relaxed)
assert relaxed.objective
# 1 objective, 1 c0, 4 * 10 x^2 (3 mccormick, 1 midpoint)
assert len(relaxed.constraints) == 1 + 1 + 4*10
objective = relaxed.objective
constraint = relaxed.constraint('c0')
assert isinstance(objective.root_expr, LinearExpression)
assert isinstance(constraint.root_expr, SumExpression)
# Root is only objvar
assert len(objective.root_expr.children) == 1
c0, c1 = constraint.root_expr.children
assert isinstance(c0, LinearExpression)
assert isinstance(c1, LinearExpression)
assert len(c0.children) == 10
assert len(c1.children) == 10
def test_convex_relaxation_with_quadratic_only():
m = pe.ConcreteModel()
m.I = range(10)
m.x = pe.Var(m.I)
m.obj = pe.Objective(expr=pe.quicksum(m.x[i]*m.x[i] for i in m.I))
m.c0 = pe.Constraint(expr=pe.quicksum(2 * m.x[i] * m.x[i] for i in m.I) >= 0)
dag = problem_from_pyomo_model(m)
relaxation = _convex_relaxation(dag)
relaxed = relaxation.relax(dag)
assert relaxed.objective
# original constraint + 10 for disaggregated squares
assert len(relaxed.constraints) == 1 + 10
objective = relaxed.objective
constraint = relaxed.constraints[0]
assert isinstance(objective.root_expr, LinearExpression)
assert len(objective.root_expr.children) == 10
for constraint in relaxed.constraints[:-1]:
assert isinstance(constraint.root_expr, SumExpression)
children = constraint.root_expr.children
assert len(children) == 2
assert isinstance(children[0], LinearExpression)
assert isinstance(children[1], QuadraticExpression)
constraint = relaxed.constraints[-1]
assert len(constraint.root_expr.children) == 10
assert isinstance(constraint.root_expr, LinearExpression)
def test_ipopt_solver(model_name):
current_dir = pathlib.Path(__file__).parent
osil_file = current_dir / 'models' / (model_name + '.osil')
pyomo_model = read_pyomo_model(osil_file)
problem = problem_from_pyomo_model(pyomo_model)
galini = Galini()
galini.update_configuration({
'galini': {
'constraint_violation_tol': 1e-2,
},
'ipopt': {
'ipopt': {
'acceptable_constr_viol_tol': 1e-3
},
},
})
solver = IpoptNLPSolver(galini)
solution = solver.solve(problem)
assert solution.status.is_success()
sol_file = current_dir / 'solutions' / (model_name + '.sol')
expected_solution = read_solution(sol_file)
expected_objective = expected_solution['objective']
assert solution.objective is not None
assert np.isclose(expected_objective, solution.objective.value)
if False:
expected_variables = expected_solution['variables']
assert len(expected_variables) == len(solution.variables)
for variable, expected in zip(solution.variables, expected_variables):
assert np.isclose(expected, variable.value)
def create_problem():
m = aml.ConcreteModel()
m.I = range(10)
m.x = aml.Var(m.I, bounds=(-1, 2))
m.obj = aml.Objective(expr=sum(m.x[i] for i in m.I))
m.cons = aml.Constraint(
m.I[1:], rule=lambda m, i: aml.cos(m.x[0]) * aml.sin(m.x[i]) >= 0)
return problem_from_pyomo_model(m)
def execute(self, args):
assert args.problem
pyomo_model = read_pyomo_model(
args.problem,
objective_prefix=args.objective_prefix,
)
problem = problem_from_pyomo_model(pyomo_model)
return self.execute_with_problem(pyomo_model, problem, args)
def test_osil_model(order, model_name):
if model_name in ['hatfldf.py']:
pytest.skip('Known derivative fail.')
current_dir = pathlib.Path(__file__).parent
osil_file = current_dir / 'models' / model_name
pyomo_model = read_pyomo_model(osil_file)
dag = problem_from_pyomo_model(pyomo_model)
derivative_check(model_name, dag, order)
def problem():
m = aml.ConcreteModel()
m.x = aml.Var()
m.y = aml.Var()
m.z = aml.Var()
m.obj = aml.Objective(expr=aml.exp(m.x) + aml.exp(m.y))
m.c0 = aml.Constraint(expr=2.0 * m.x*m.x + 3.0 * m.x*m.y + 4.0 * m.y*m.y >= 0)
return problem_from_pyomo_model(m)
def problem():
m = aml.ConcreteModel()
m.I = range(10)
m.x = aml.Var(m.I, bounds=(-2, 2))
m.y = aml.Var(bounds=(-3, 3))
m.obj = aml.Objective(expr=m.y)
m.linear = aml.Constraint(expr=sum(m.x[i] for i in m.I) >= 0)
m.not_linear = aml.Constraint(expr=m.y * sum(m.x[i] for i in m.I) >= 0)
return problem_from_pyomo_model(m)
def problem():
m = aml.ConcreteModel()
m.x = aml.Var(bounds=(-2, 2))
m.y = aml.Var(bounds=(-3, 3))
m.z = aml.Var(bounds=(-4, 2))
m.obj = aml.Objective(expr=m.x)
m.bilinear = aml.Constraint(expr=m.x * m.y >= 0)
m.bilinear_coef = aml.Constraint(expr=5.0 * m.x * m.y >= 0)
m.bilinear_coef_2 = aml.Constraint(expr=m.x * m.y * 6.0 >= 0)
m.bilinear_sum = aml.Constraint(
expr=2 * m.x * m.y + 3 * m.x * m.z + 4 * m.y * m.z >= 0)
m.trilinear = aml.Constraint(expr=m.x * m.y * m.z >= 0)
m.power = aml.Constraint(expr=m.x**2 >= 0)
return problem_from_pyomo_model(m)
def problem():
m = pe.ConcreteModel()
m.I = range(2)
m.x = pe.Var(m.I, domain=pe.Integers, bounds=(0, 200))
m.x[0].setub(4.0)
m.x[1].setub(7.0)
@m.Objective()
def f(m):
return m.x[0]**2 - 16 * m.x[0] + m.x[1]**2 - 4 * m.x[1] + 68
@m.Constraint()
def g0(m):
return m.x[0]**2 + m.x[1] >= 0
@m.Constraint()
def g1(m):
return -0.33 * m.x[0] - m.x[1] + 4.5 >= 0
return problem_from_pyomo_model(m)
def test_convex_relaxation_with_quadratic_and_linear():
m = pe.ConcreteModel()
m.I = range(10)
m.x = pe.Var(m.I)
m.obj = pe.Objective(
expr=pe.quicksum(m.x[i]*m.x[i] for i in m.I) + pe.quicksum(m.x[i] for i in m.I)
)
m.c0 = pe.Constraint(
expr=pe.quicksum(2 * m.x[i] * m.x[i] for i in m.I) + pe.quicksum(m.x[i] for i in m.I) >= 0
)
dag = problem_from_pyomo_model(m)
relaxation = _convex_relaxation(dag)
relaxed = relaxation.relax(dag)
print_problem(relaxed)
assert relaxed.objective
assert len(relaxed.constraints) == 1 + 10
objective = relaxed.objective
assert isinstance(objective.root_expr, SumExpression)
assert all(
isinstance(c, LinearExpression)
for c in objective.root_expr.children
)
for constraint in relaxed.constraints[:-1]:
assert isinstance(constraint.root_expr, SumExpression)
children = constraint.root_expr.children
assert len(children) == 2
assert isinstance(children[0], LinearExpression)
assert isinstance(children[1], QuadraticExpression)
constraint = relaxed.constraints[-1]
assert all(
isinstance(c, LinearExpression)
for c in constraint.root_expr.children
)
def problem():
m = aml.ConcreteModel()
m.x = aml.Var(bounds=(2, 20))
m.y = aml.Var(bounds=(3, 30))
m.z = aml.Var(bounds=(4, 20))
m.obj = aml.Objective(expr=m.x)
m.univariate_concave = aml.Constraint(expr=aml.sqrt(m.x) >= 0)
m.univariate_concave_2 = aml.Constraint(
expr=aml.sqrt(3.0 * m.x + 2.0) >= 0)
m.univariate_concave_3 = aml.Constraint(
expr=aml.sqrt(aml.log(3 * m.x + 2)) >= 0)
m.univariate_nonconcave = aml.Constraint(expr=aml.sin(m.x) >= 0)
m.univariate_nonconcave_2 = aml.Constraint(expr=m.x**3 >= 0)
m.univariate_nonconcave_3 = aml.Constraint(expr=3**m.x >= 0)
m.nonunivariate = aml.Constraint(expr=m.x**m.y >= 0)
m.nonunivariate_2 = aml.Constraint(expr=m.x + m.y + m.z >= 0)
m.nonunivariate_3 = aml.Constraint(expr=aml.sin(m.x + aml.log(m.y)) >= 0)
return problem_from_pyomo_model(m)
def test_convex_relaxation_of_linear_problem():
m = pe.ConcreteModel()
m.I = range(10)
m.x = pe.Var(m.I)
m.obj = pe.Objective(expr=pe.quicksum(m.x[i] for i in m.I) + 2.0)
m.c0 = pe.Constraint(expr=pe.quicksum(2 * m.x[i] for i in m.I) - 4.0 >= 0)
dag = problem_from_pyomo_model(m)
relaxation = _convex_relaxation(dag)
relaxed = relaxation.relax(dag)
assert relaxed.objective
assert len(relaxed.constraints) == 1
objective = relaxed.objective
constraint = relaxed.constraints[0]
assert isinstance(objective.root_expr, LinearExpression)
assert len(objective.root_expr.children) == 10
assert np.isclose(objective.root_expr.constant_term, 2.0)
assert isinstance(constraint.root_expr, LinearExpression)
assert len(constraint.root_expr.children) == 10
assert np.isclose(constraint.root_expr.constant_term, -4.0)
