cuOpt Options#
ampl: option solver cuopt; # change the solver
ampl: option cuopt_options 'option1=value1 option2=value2'; # specify options
ampl: solve; # solve the problem
Solver options obtained with $ cuopt -=.
CUOPTLP Optimizer Options for AMPL
--------------------------------------------
To set these options, assign a string specifying their values to the AMPL
option "cuoptmp_options". For example:
ampl: option cuoptmp_options 'mipgap=1e-6';
'' Options:
acc:_all
Solver acceptance level for all constraints and expressions. Value
meaning: as described in the specific acc:... options.
Can be useful to disable all reformulations (acc:_all=2), or force
linearization (acc:_all=0.)
acc:lineq
Solver acceptance level for 'LinConEQ' as flat constraint, default 2:
0 - Not accepted natively, automatic redefinition will be attempted
1 - Accepted but automatic redefinition will be used where possible
2 - Accepted natively and preferred
acc:linge
Solver acceptance level for 'LinConGE' as flat constraint, default 2:
0 - Not accepted natively, automatic redefinition will be attempted
1 - Accepted but automatic redefinition will be used where possible
2 - Accepted natively and preferred
acc:linle
Solver acceptance level for 'LinConLE' as flat constraint, default 2:
0 - Not accepted natively, automatic redefinition will be attempted
1 - Accepted but automatic redefinition will be used where possible
2 - Accepted natively and preferred
alg:method (method)
Designate the method to solve the LP problem:
0 - Concurrent (default)
1 - Pdlp
2 - Dual simplex
alg:relax (relax)
0*/1: Whether to relax integrality of variables.
alg:solver_mode (solver_mode)
Designate the solver mode used by PDLP to solve the problem:
0 - Stable1
1 - Stable2 (default)
2 - Methodical1
3 - Fast1
cvt:bigM (cvt:bigm, cvt:mip:bigM, cvt:mip:bigm)
Default value of big-M for linearization of logical constraints. Not
used by default. Use with care (prefer tight bounds). Should be smaller
than (1.0 / [integrality tolerance])
cvt:dvelim (dvelim)
Eliminate AMPL defined variables by substitution into linear, quadratic,
and polynomial expressions:
0 - Do not eliminate, always instantiate the variables.
1 - Eliminate only those used 1x. This can increase model density but
greatly simplifies some models.
2 - Always substitute where possible, even if the variable needs to be
instantiated for use in other places. Can introduce redundancy, but
seems best for some models (default.)
See also AMPL options linelim and substout.
cvt:expcones (expcones)
0*/1: Recognize exponential cones.
cvt:mip:eps (cvt:cmp:eps, cmp:eps)
Tolerance for strict comparison of continuous variables for MIP. Applies
to <, >, and != operators. Also applies to negation of conditional
comparisons: b==1 <==> x<=5 means that with b==0, x>=5+eps. Default:
1e-4.
cvt:multoutcard (multoutcard)
Up to which (estimated) QP matrix cardinality should a product of 2
linear expressions be multiplied out. Default 1e9.
Can speed up model input, but prone to numerical issues.
cvt:names (names, modelnames)
Whether to read or generate variable / constraint / objective names:
0 - No names
1 - (Default) Only provide names if at least one of .col / .row name
files was written by AMPL (AMPL: `option [<solver>_]auxfiles rc;`)
2 - Read names from AMPL, but create generic names if not provided
3 - Create generic names.
cvt:plapprox:domain (plapprox:domain, plapproxdomain)
For piecewise-linear approximated functions, both arguments and result
are bounded to +-[pladomain]. Default 1e6.
cvt:plapprox:reltol (plapprox:reltol, plapproxreltol)
Relative tolerance for piecewise-linear approximation. Default 0.01.
cvt:pre:all
0/1*: Set to 0 to disable most presolve in the flat converter.
cvt:pre:eqbinary
0/1*: Preprocess reified equality comparison with a binary variable.
cvt:pre:eqresult
0/1*: Preprocess reified equality comparison's decidable cases.
cvt:pre:ineqresult
0/1*: Preprocess reified inequality comparison's decidable cases.
cvt:pre:ineqrhs
0/1*: Preprocess reified inequality comparison's right-hand sides.
cvt:pre:unnest
0/1*: Inline nested expressions, currently Ands/Ors.
cvt:prod (cvt:pre:prod)
Product preprocessing flags. Sum of a subset of the following bits:
1 - Quadratize higher-order products in the following order: integer
terms first, then real-valued ones; in each group, smaller-range terms
first.
2 - Logicalize products of 2 binary terms. Logicalizing means that the
product is converted to a conjunction. If the solver does not support it
natively (see acc:and), the conjunction is linearized.
4 - Logicalize products of >=3 binary terms.
Default: 7.
Bits 2 or 4 imply bit 1.
cvt:qp2passes (cvt:qp2pass, qp2passes, qp2pass)
Parse sums of QP expressions in 2 passes. Usually faster. Default 1.
cvt:quadcon (passquadcon)
Convenience option. Set to 0 to disable quadratic constraints. Synonym
for acc:quad..=0. Currently this disables out-multiplication of
quadratic terms, then they are linearized.
cvt:quadobj (passquadobj)
0*/1: Pass quadratic objective terms to the solver. When 0, if the
solver accepts quadratic constraints, such a constraint will be created
with those, otherwise linearly approximated.
cvt:socp (socpmode, socp)
Second-Order Cone recognition mode:
0 - Do not recognize SOCP forms
1 - Recognize from non-quadratic expressions only (sqrt, abs)
2 - Recognize from quadratic and non-quadratic SOCP forms. Helpful if
the solver does not recognize non-standard forms
Recognized SOCP forms can be further converted to (SOCP-standardized)
quadratic constraints, see cvt:socp2qc. Default: 0.
cvt:socp2qc (socp2qcmode, socp2qc)
Mode to convert recognized SOCP forms to SOCP-standardized quadratic
constraints:
0 - Do not convert
1 - Convert if no other cone types found, and not all original
quadratics could be recognized as SOC, in particular if the
objective is quadratic
2 - Always convert
Such conversion can be necessary if the solver does not accept a mix of
conic and quadratic constraints/objectives. Default: 2.
cvt:sos (sos)
0/1*: Whether to honor declared suffixes .sosno and .ref describing SOS
sets. Each distinct nonzero .sosno value designates an SOS set, of type
1 for positive .sosno values and of type 2 for negative values. The .ref
suffix contains corresponding reference values used to order the
variables.
cvt:sos2 (sos2)
0*/1: Whether to honor SOS2 constraints for nonconvex piecewise-linear
terms, using suffixes .sos and .sosref provided by AMPL. Currently under
rework.
cvt:uenc:negctx:max (uenc:negctx:max, uenc:negctx)
If cvt:uenc:ratio applies, max number of constants in comparisons
x==const in negative context (equivalently, x!=const in positive
context) to skip UEnc(x). Default 1.
cvt:uenc:ratio (uenc:ratio)
Min ratio (ub-lb)/Nvalues to skip unary encoding for a variable x, where
Nvalues is the number of constants used in conditional comparisons
x==const. Instead, indicator constraints (or big-Ms) are used, if
uenc:negctx also applies. Default 0.
lim:iterations (iterations)
Iteration limit after which the solver will stop and return the current
solution
lim:ncputhreads (ncputhreads)
Number of CPU threads used in the LP and MIP solvers
lim:timelimit (timelimit)
Time limit in seconds after which the solver will stop and return the
current solution
lp:absdualtol (absdualtol)
Absolute dual tolerance used in PDLP's dual feasibility check
lp:absgaptol (absgaptol)
Absolute gap tolerance used in PDLP's duality gap check
lp:absprimaltol (absprimaltol)
Absolute primal tolerance used in PDLP's primal feasibility check
lp:crossover (crossover)
Crossover to a basic solution after a optimal solution is found
lp:firstprimalfeas (firstprimalfeas)
Stop when the first primal feasible solution is found
lp:infeasdetect (infeasdetect)
Detect infeasibility PDLP
lp:perconsres (perconsres)
Compute the primal & dual residual per constraint instead of globally
lp:reldualtol (reldualtol)
Relative dual tolerance used in PDLP's dual feasibility check
lp:relgap (relgap)
Relative tolerance used to terminate the MIP solve
lp:relgaptol (relgaptol)
Relative gap tolerance used in PDLP's duality gap check
lp:relprimaltol (relprimaltol)
Relative primal tolerance used in PDLP's primal feasibility check
lp:savebestprimal (savebestprimal)
Save the best primal solution so far
lp:strictinfeas (strictinfeas)
Stop if current or the average solution is detected as infeasible
mip:absgap (absgap)
Absolute tolerance used to terminate the MIP solve
mip:abstol (abstol)
Absolute tolerance used in mip
mip:bestbound (bestbound, return_bound)
Whether to return suffix .bestbound for the best known MIP dual bound on
the objective value:
0 - No (default)
1 - Yes.
The suffix is on the objective and problem and is -Infinity for
minimization problems and +Infinity for maximization problems if there
are no integer variables or if a dual bound is not available.
mip:hueristicsonly (hueristicsonly)
Run only the GPU heuristics
mip:inttol (inttol)
Integrality tolerance used in mip
mip:reltol (reltol)
Relative tolerance used in mip
mip:return_gap (return_mipgap)
Whether to return mipgap suffixes or include mipgap values (|objectve -
.bestbound|) in the solve_message: sum of
1 - Return .relmipgap suffix (relative to |obj|)
2 - Return .absmipgap suffix (absolute mipgap)
4 - Suppress mipgap values in solve_message.
Default = 0. The suffixes are on the objective and problem. Returned
suffix values are +Infinity if no integer-feasible solution has been
found, in which case no mipgap values are reported in the solve_message.
mip:round (round)
Whether to round integer variables to integral values before returning
the solution, and whether to report that the solver returned noninteger
values for integer values: sum of
1 ==> Round nonintegral integer variables
2 ==> Modify solve_result
4 ==> Modify solve_message
Default = 0. Modifications that were or would be made are reported in
solve_result and solve_message only if the maximum deviation from
integrality exceeded mip:round_reptol.
mip:round_reptol (round_reptol)
Tolerance for reporting rounding of integer variables to integer values;
see "mip:round". Default = 1e-9.
mip:scale (scale)
Apply Scaling to MIP problems
obj:multi (multiobj)
Whether to use multi-objective optimization:
0 - Single objective, see option obj:no (default)
1 - Multi-objective, solver's native handling if available
2 - Multi-objective, force emulation
When obj:multi>0 and several objectives are present, suffixes
.objpriority, .objweight, .objreltol, and .objabstol on the objectives
are relevant. Objectives with greater .objpriority values (integer
values) have higher priority. Objectives with the same .objpriority are
weighted by .objweight, according to the option obj:multi:weight.
Objectives with positive .objabstol or .objreltol are allowed to be
degraded by lower priority objectives by amounts not exceeding the
.objabstol (absolute) and .objreltol (relative) limits.
Note that with solver's native handling (when obj:multi=1 and
supported), some solvers might have special rules for the tolerances,
especially for LP, and not allow quadratic objectives. See the solver
documentation.
obj:multi:weight (multiobjweight, obj:multi:weights, multiobjweights)
How to interpret each objective's weight sign:
1 - relative to the sense of the 1st objective
2 - relative to its own sense (default)
With the 1st option (legacy behaviour), negative .objweight for
objective i would make objective i's sense the opposite of the model's
1st objective. Otherwise, it would make objective i's sense the opposite
to its sense defined in the model.
obj:no (objno)
Objective to optimize:
0 - None
1 - First (default, if available)
2 - Second (if available), etc.
sol:chk:fail (chk:fail, checkfail)
Fail on MP solution check violations, with solve result 150.
sol:chk:feastol (sol:chk:eps, chk:eps, chk:feastol)
Absolute tolerance to check objective values, variable and constraint
bounds. Default 1e-6.
sol:chk:feastolrel (sol:chk:epsrel, chk:epsrel, chk:feastolrel)
Relative tolerance to check objective values, variable and constraint
bounds. Default 1e-6.
sol:chk:infeas (chk:infeas, checkinfeas)
Check even infeasible solution condidates, whenever solver reports them.
sol:chk:inttol (sol:chk:inteps, sol:inteps, chk:inttol)
Solution checking tolerance for variables' integrality. Default 1e-5.
sol:chk:mode (solcheck, checkmode, chk:mode)
Solution checking mode. Sum of a subset of the following bits:
1 - Check variable bounds and integrality.
2 - Check original model constraints, as well as any non-linear
expression values reported by the solver.
4 - Check intermediate auxiliary constraints (i.e., those which were
reformulated further).
8 - Check final auxiliary constraints sent to solver.
16 - Check objective values.
32, 64, 128, 256, 512 - similar, but non-linear expressions are
recomputed (vs using their values reported by the solver.)
*Experimental.* This is an idealistic check, because it does not
consider possible tolerances applied by the solver when computing
expression values.
Default: 1+2+512.
sol:chk:prec (chk:prec, chk:precision)
AMPL solution_precision option when checking: number of significant
digits.
sol:chk:round (chk:round, chk:rnd)
AMPL solution_round option when checking: round to this number of
decimals after comma (before comma if negative.)
sol:report_uncertain (report_uncertain_sol)
0/1*: whether to report objective value(s) in solve_message when
solve_result is '?' (unknown).
tech:consolelog (consolelog)
Log information to the console during a solve
tech:debug (debug)
0*/1: whether to assist testing & debugging, e.g., by outputting
auxiliary information (mostly via suffixes).
tech:optionfile (optionfile, option:file)
Name of an AMPL solver option file to read (surrounded by 'single' or
"double" quotes if the name contains blanks). Lines that start with #
are ignored. Otherwise, each nonempty line should contain "name=value",
e.g., "lim:iter=500".
tech:outlev (outlev)
Whether to log information to the console
tech:outlev_mp (outlev_mp)
0*/1: whether to print MP model information.
tech:timing (timing, tech:report_times, report_times)
0*/1/2: Whether to print and return timings for the run, all times are
wall times. If set to 1, return the solution times in the problem
suffixes 'time_solver', 'time_setup' and 'time', 'time'=
time_solver+time_setup+time_output is a measure of the total time spent
in the solver driver. If set to 2, return more granular times, including
'time_read', 'time_conversion' and 'time_output'.
tech:version (version)
Single-word phrase: report version details before solving the problem.
tech:wantsol (wantsol)
In a stand-alone invocation (no "-AMPL" on the command line), what
solution information to write. Sum of
1 - Write ".sol" file
2 - Primal variables to stdout
4 - Dual variables to stdout
8 - Suppress solution message.
tech:writegraph (cvt:writegraph, writegraph, exportgraph)
File to export conversion graph. Format: JSON Lines.
tech:writesolution (writesol, writesolution)
Specifies the names of files where to export the solution and/or other
result files in solver's native formats. Option can be repeated. File
name extensions can be ".sol[.tar.gz]", ".json", ".bas", ".ilp", etc.
threads (threads)
Number of CPU threads used in the LP and MIP solvers
timelim (timelim)
Time limit in seconds after which the solver will stop and return the
current solution