AMPL Options for IPOPT

NOTE: This is the documentation for the (old) Fortran version of Ipopt. The options listed here are called differently in the current C++ version of Ipopt!!!

You can set options for IPOPT from AMPL using the

option ipopt_options 'option1=value1 option2=value2 ...';

command in your AMPL model. For example, if you want to set the maximal number of iterations to 1000, and want the problem to be solved to an accuracy of 1e-6, you would add

option ipopt_options 'imaxiter=1000 dtol=1e-6';

(Note: Even if you rename your AMPL solver executable, you still need to use the ipopt_options keyword.)

The following table describes some of the available options. If you are looking for more options, please consult the README.IPOPT file.

Parameter	Default Value	Description
imaxiter	10000	Maximum number of iterations
imaxcpusec	999999999	Upper limit on computation time (CPU seconds)
dtol	1e-8	Overall desired (scaled) error tolerance
dcmaxtol	1e300	Absolute tolerance for (unscaled) primal infeasibility
dinfmaxtol	1e300	Absolute tolerance for (unscaled) dual infeasibility
iprint	0	Amount of output. The larger, the more detailed output (-1: minimal output)
ioutput	0	If set to 1, a longer information line is printed per iteration
iscale	2	Options for automatic scaling of the problem statement (0: no scaling)
dfscale	1.	Scaling factor for the objective function
dmu0	0.1	Initial value of the barrier parameter
dbndfrac	0.01	Relative distance of starting point from closest bound
dbndpush	0.01	Absolute distance of starting point from closest bound
dmovebounds	1e-8	Initial relative perturbation of the bounds
dpivtol	1e-8	Pivot tolerance for linear solver
dfillinfact	5.0	Factor for estimating memory requirement for linear solver (has to be >=1) If IPOPT runs out of memory, a smaller value (such as 1, 1.2, or 2) might correct the problem.