如何在Python Gekko中设置求解器选项（例如容错）？

Question

There are two ways to set solver options in Python Gekko with m.options and m.solver_options . Which method takes precedence and when should one or the other be used?

For example, I would like to set the objective tolerance ( OTOL ) and equation residual tolerance ( RTOL ) for the solver. Which one does Gekko use ( 1e-7 or 1e-8 )?

from gekko import GEKKO
m = GEKKO() # Initialize gekko
m.options.SOLVER=1  # APOPT is an MINLP solver

m.options.OTOL = 1.0e-8
m.options.RTOL = 1.0e-8

# solver settings with APOPT
m.solver_options = ['objective_convergence_tolerance 1.0e-7', \
                    'constraint_convergence_tolerance 1.0e-7']

# Initialize variables
x1 = m.Var(value=1,lb=1,ub=5)
x2 = m.Var(value=5,lb=1,ub=5)
x3 = m.Var(value=5,lb=1,ub=5,integer=True)
x4 = m.Var(value=1,lb=1,ub=5)
# Equations
m.Equation(x1*x2*x3*x4>=25)
m.Equation(x1**2+x2**2+x3**2+x4**2==40)
m.Obj(x1*x4*(x1+x2+x3)+x3) # Objective
m.solve(disp=False) # Solve
print('Results')
print('x1: ' + str(x1.value))
print('x2: ' + str(x2.value))
print('x3: ' + str(x3.value))
print('x4: ' + str(x4.value))
print('Objective: ' + str(m.options.objfcnval))

This produces the solution:

Results
x1: [1.0]
x2: [4.5992789966]
x3: [4.0]
x4: [1.3589086474]
Objective: 17.044543237

Sometimes a problem needs more or less accuracy but there are also other options that I'd like to use for IPOPT or APOPT . I'd like to know which option Gekko is using.

Answer 1

Solvers such as APOPT or IPOPT use m.solver_options values if both m.options and m.solver_options are set. The Gekko m.options values are only a subset of all the solver options but also some of the most common configuration parameters that are adjustable for all solvers. Some of the common options are convergence tolerances ( RTOL and OTOL ), maximum iterations ( MAX_ITER ), and maximum time ( MAX_TIME ). Common solver results are also output such as objective function value ( OBJFCNVAL ), solve time ( SOLVETIME ), and solution status ( APPINFO ).

There are also specific options that are configurable by the solver type. For example, the APOPT solver is a Mixed Integer Nonlinear Programming (MINLP) solver. There are additional options that are configurable only from m.solver_options such as:

m.solver_options = ['minlp_maximum_iterations 500', \
                    # minlp iterations with integer solution
                    'minlp_max_iter_with_int_sol 10', \
                    # treat minlp as nlp
                    'minlp_as_nlp 0', \
                    # nlp sub-problem max iterations
                    'nlp_maximum_iterations 50', \
                    # 1 = depth first, 2 = breadth first
                    'minlp_branch_method 1', \
                    # maximum deviation from whole number
                    'minlp_integer_tol 0.05', \
                    # covergence tolerance
                    'minlp_gap_tol 0.01']

The IPOPT solver is a Nonlinear Programming (NLP) solver so it doesn't use the MINLP options.