 |
Google OR-Tools v9.11
a fast and portable software suite for combinatorial optimization
|
|
Google OR-Tools v9.11
a fast and portable software suite for combinatorial optimization
|
#include <parameters.h>
Public Member Functions | |
| SolveParametersProto | Proto () const |
Static Public Member Functions | |
| static absl::StatusOr< SolveParameters > | FromProto (const SolveParametersProto &proto) |
Public Attributes | |
| bool | enable_output = false |
| absl::Duration | time_limit = absl::InfiniteDuration() |
| std::optional< int64_t > | iteration_limit |
| std::optional< int64_t > | node_limit |
| std::optional< double > | cutoff_limit |
| std::optional< double > | objective_limit |
| std::optional< double > | best_bound_limit |
| std::optional< int32_t > | solution_limit |
| std::optional< int32_t > | threads |
| If unset, use the solver default. If set, it must be >= 1. | |
| std::optional< int32_t > | random_seed |
| std::optional< double > | absolute_gap_tolerance |
| std::optional< double > | relative_gap_tolerance |
| std::optional< int32_t > | solution_pool_size |
| std::optional< LPAlgorithm > | lp_algorithm |
| std::optional< Emphasis > | presolve |
| std::optional< Emphasis > | cuts |
| std::optional< Emphasis > | heuristics |
| std::optional< Emphasis > | scaling |
| GScipParameters | gscip |
| GurobiParameters | gurobi |
| glop::GlopParameters | glop |
| sat::SatParameters | cp_sat |
| pdlp::PrimalDualHybridGradientParams | pdlp |
| GlpkParameters | glpk |
| HighsOptionsProto | highs |
Parameters to control a single solve.
Contains both parameters common to all solvers, e.g. time_limit, and parameters for a specific solver, e.g. gscip. If a value is set in both common and solver specific fields, the solver specific setting is used.
The common parameters that are optional and unset indicate that the solver default is used.
Solver specific parameters for solvers other than the one in use are ignored.
Parameters that depends on the model (e.g. branching priority is set for each variable) are passed in ModelSolveParametersProto.
Definition at line 265 of file parameters.h.
|
static |
Definition at line 270 of file parameters.cc.
| SolveParametersProto operations_research::math_opt::SolveParameters::Proto | ( | ) | const |
Definition at line 215 of file parameters.cc.
| std::optional<double> operations_research::math_opt::SolveParameters::absolute_gap_tolerance |
An absolute optimality tolerance (primarily) for MIP solvers.
The absolute GAP is the absolute value of the difference between:
Must be >= 0 if set.
See also relative_gap_tolerance.
Definition at line 360 of file parameters.h.
| std::optional<double> operations_research::math_opt::SolveParameters::best_bound_limit |
The solver stops early as soon as it proves the best bound is at least this good, with termination reason kFeasible or kNoSolutionFound and limit kObjective.
See the user guide for a comparison with cutoff_limit.
Definition at line 317 of file parameters.h.
| sat::SatParameters operations_research::math_opt::SolveParameters::cp_sat |
Definition at line 418 of file parameters.h.
| std::optional<double> operations_research::math_opt::SolveParameters::cutoff_limit |
The solver stops early if it can prove there are no primal solutions at least as good as cutoff.
On an early stop, the solver returns termination reason kNoSolutionFound and with limit kCutoff and is not required to give any extra solution information. Has no effect on the return value if there is no early stop.
It is recommended that you use a tolerance if you want solutions with objective exactly equal to cutoff to be returned.
See the user guide for more details and a comparison with best_bound_limit.
Definition at line 306 of file parameters.h.
| std::optional<Emphasis> operations_research::math_opt::SolveParameters::cuts |
Effort on getting a stronger LP relaxation (MIP only) or the solver default effort level if unset.
Definition at line 404 of file parameters.h.
| bool operations_research::math_opt::SolveParameters::enable_output = false |
Enables printing the solver implementation traces. These traces are sent to the standard output stream.
Definition at line 271 of file parameters.h.
| glop::GlopParameters operations_research::math_opt::SolveParameters::glop |
Definition at line 417 of file parameters.h.
| GlpkParameters operations_research::math_opt::SolveParameters::glpk |
Definition at line 421 of file parameters.h.
| GScipParameters operations_research::math_opt::SolveParameters::gscip |
Definition at line 415 of file parameters.h.
| GurobiParameters operations_research::math_opt::SolveParameters::gurobi |
Definition at line 416 of file parameters.h.
| std::optional<Emphasis> operations_research::math_opt::SolveParameters::heuristics |
Effort in finding feasible solutions beyond those encountered in the complete search procedure (MIP only), or the solver default effort level if unset.
Definition at line 409 of file parameters.h.
| HighsOptionsProto operations_research::math_opt::SolveParameters::highs |
Definition at line 422 of file parameters.h.
| std::optional<int64_t> operations_research::math_opt::SolveParameters::iteration_limit |
Limit on the iterations of the underlying algorithm (e.g. simplex pivots). The specific behavior is dependent on the solver and algorithm used, but often can give a deterministic solve limit (further configuration may be needed, e.g. one thread).
Typically supported by LP, QP, and MIP solvers, but for MIP solvers see also node_limit.
Definition at line 286 of file parameters.h.
| std::optional<LPAlgorithm> operations_research::math_opt::SolveParameters::lp_algorithm |
The algorithm for solving a linear program. If nullopt, use the solver default algorithm.
For problems that are not linear programs but where linear programming is a subroutine, solvers may use this value. E.g. MIP solvers will typically use this for the root LP solve only (and use dual simplex otherwise).
Definition at line 393 of file parameters.h.
| std::optional<int64_t> operations_research::math_opt::SolveParameters::node_limit |
Limit on the number of subproblems solved in enumerative search (e.g. branch and bound). For many solvers this can be used to deterministically limit computation (further configuration may be needed, e.g. one thread).
Typically for MIP solvers, see also iteration_limit.
Definition at line 293 of file parameters.h.
| std::optional<double> operations_research::math_opt::SolveParameters::objective_limit |
The solver stops early as soon as it finds a solution at least this good, with termination reason kFeasible and limit kObjective.
Definition at line 310 of file parameters.h.
| pdlp::PrimalDualHybridGradientParams operations_research::math_opt::SolveParameters::pdlp |
Definition at line 419 of file parameters.h.
| std::optional<Emphasis> operations_research::math_opt::SolveParameters::presolve |
Effort on simplifying the problem before starting the main algorithm, or the solver default effort level if unset.
Definition at line 397 of file parameters.h.
| std::optional<int32_t> operations_research::math_opt::SolveParameters::random_seed |
Seed for the pseudo-random number generator in the underlying solver. Note that all solvers use pseudo-random numbers to select things such as perturbation in the LP algorithm, for tie-break-up rules, and for heuristic fixings. Varying this can have a noticeable impact on solver behavior.
Although all solvers have a concept of seeds, note that valid values depend on the actual solver.
Definition at line 347 of file parameters.h.
| std::optional<double> operations_research::math_opt::SolveParameters::relative_gap_tolerance |
A relative optimality tolerance (primarily) for MIP solvers.
The relative GAP is a normalized version of the absolute GAP (defined on absolute_gap_tolerance), where the normalization is solver-dependent, e.g. the absolute GAP divided by the objective value of the best feasible solution found.
The solver can stop once the relative GAP is at most relative_gap_tolerance (when set), and return TerminationReason::kOptimal.
Must be >= 0 if set.
See also absolute_gap_tolerance.
Definition at line 375 of file parameters.h.
| std::optional<Emphasis> operations_research::math_opt::SolveParameters::scaling |
Effort in rescaling the problem to improve numerical stability, or the solver default effort level if unset.
Definition at line 413 of file parameters.h.
| std::optional<int32_t> operations_research::math_opt::SolveParameters::solution_limit |
The solver stops early after finding this many feasible solutions, with termination reason kFeasible and limit kSolution. Must be greater than zero if set. It is often used to get the solver to stop on the first feasible solution found. Note that there is no guarantee on the objective value for any of the returned solutions.
Solvers will typically not return more solutions than the solution limit, but this is not enforced by MathOpt, see also b/214041169.
Currently supported for Gurobi and SCIP, and for CP-SAT only with value 1.
Definition at line 329 of file parameters.h.
| std::optional<int32_t> operations_research::math_opt::SolveParameters::solution_pool_size |
Maintain up to solution_pool_size solutions while searching. The solution pool generally has two functions: (1) For solvers that can return more than one solution, this limits how many solutions will be returned. (2) Some solvers may run heuristics using solutions from the solution pool, so changing this value may affect the algorithm's path. To force the solver to fill the solution pool, e.g. with the n best solutions, requires further, solver specific configuration.
Definition at line 385 of file parameters.h.
| std::optional<int32_t> operations_research::math_opt::SolveParameters::threads |
If unset, use the solver default. If set, it must be >= 1.
Definition at line 332 of file parameters.h.
| absl::Duration operations_research::math_opt::SolveParameters::time_limit = absl::InfiniteDuration() |
Maximum time a solver should spend on the problem.
This value is not a hard limit, solve time may slightly exceed this value. Always passed to the underlying solver, the solver default is not used.
Definition at line 277 of file parameters.h.
1.12.0