Ipopt.jl 0.7 contains breaking changes.

Access to the binary via amplexe and amplexefun has been removed because it required an initialization that caused problems with other packages.

• On Julia 1.3 or later, use Ipopt_jll.amplexe instead.
• On Julia 1.0, install Ipopt.jl version 0.6.x via ] add [email protected]. Make sure to restart Julia for the changes to take effect.

Ipopt.jl is a Julia interface to the COIN-OR nonlinear solver Ipopt.

Note: This wrapper is maintained by the JuMP community and is not a COIN-OR project.

Install Ipopt.jl using the Julia package manager:

import Pkg; Pkg.add("Ipopt")

In addition to installing the Ipopt.jl package, this will also download and install the Ipopt binaries. You do not need to install Ipopt separately.

If you require a custom build of Ipopt, see the instructions below.

For details on using a different linear solver, see the Linear Solvers section below.

You can use Ipopt with JuMP as follows:

using JuMP, Ipopt
model = Model(Ipopt.Optimizer)
set_optimizer_attribute(model, "max_cpu_time", 60.0)
set_optimizer_attribute(model, "print_level", 0)

Supported options are listed in the Ipopt documentation.

Ipopt provides a callback that can be used to log the status of the optimization during a solve. It can also be used to terminate the optimization by returning false. Here is an example:

using JuMP, Ipopt, Test
model = Model(Ipopt.Optimizer)
set_silent(model)
@variable(model, x >= 1)
@objective(model, Min, x + 0.5)
x_vals = Float64[]
function my_callback(
   prob::IpoptProblem,
   alg_mod::Cint,
   iter_count::Cint,
   obj_value::Float64,
   inf_pr::Float64,
   inf_du::Float64,
   mu::Float64,
   d_norm::Float64,
   regularization_size::Float64,
   alpha_du::Float64,
   alpha_pr::Float64,
   ls_trials::Cint,
)
   c = Ipopt.column(index(x))
   push!(x_vals, prob.x[c])
   @test isapprox(obj_value, 1.0 * x_vals[end] + 0.5, atol = 1e-1)
   # return `true` to keep going, or `false` to terminate the optimization.
   return iter_count < 1
end
MOI.set(model, Ipopt.CallbackFunction(), my_callback)
optimize!(model)
@test MOI.get(model, MOI.TerminationStatus()) == MOI.INTERRUPTED
@test length(x_vals) == 2

See the Ipopt documentation for an explanation of the arguments to the callback. They are identical to the output contained in the logging table printed to the screen.

Full documentation for the Ipopt C wrapper is available. However, we strongly encourage you to use Ipopt with JuMP instead.

If you get a termination status MOI.INVALID_MODEL, it is probably because you have some undefined value in your model, e.g., a division by zero. Fix this by removing the division, or by imposing variable bounds so that you cut off the undefined region.

Instead of

model = Model(Ipopt.Optimizer)
@variable(model, x)
@NLobjective(model, 1 / x)

do

model = Model(Ipopt.Optimizer)
@variable(model, x >= 0.0001)
@NLobjective(model, 1 / x)

Note: it is not necessary to compile a custom version of Ipopt to use a different linear solver. See the Linear Solvers section below.

To install custom built Ipopt binaries, you must compile the shared library ( e.g., libipopt.dylib, libipopt.so, or libipopt.dll) and the AMPL executable (e.g., ipopt or ipopt.exe).

If you cannot compile the AMPL executable, you can download an appropriate version from AMPL.

Next, set the environmental variables JULIA_IPOPT_LIBRARY_PATH and JULIA_IPOPT_EXECUTABLE_PATH to point the the shared library and AMPL executable repspectively. Then call import Pkg; Pkg.build("Ipopt").

For instance, given /Users/oscar/lib/libipopt.dylib and /Users/oscar/bin/ipopt, run:

ENV["JULIA_IPOPT_LIBRARY_PATH"] = "/Users/oscar/lib"
ENV["JULIA_IPOPT_EXECUTABLE_PATH"] = "/Users/oscar/bin"
import Pkg
Pkg.build("Ipopt")

Very important note: you must set these environment variables before calling using Ipopt in every Julia session.

For example:

ENV["JULIA_IPOPT_LIBRARY_PATH"] = "/Users/oscar/lib"
ENV["JULIA_IPOPT_EXECUTABLE_PATH"] = "/Users/oscar/bin"
using Ipopt

Alternatively, you can set these permanently through your operating system.

To switch back to the default binaries, run

delete!(ENV, "JULIA_IPOPT_LIBRARY_PATH")
delete!(ENV, "JULIA_IPOPT_EXECUTABLE_PATH")
import Pkg
Pkg.build("Ipopt")

To improve performance, Ipopt supports a number of linear solvers. Installing these can be tricky, however, the following instructions should work. If they don't, or are not explicit enough, please open an issue.

Tested on a clean install of Ubuntu 20.04.

1. Install lapack and libomp:

   sudo apt install liblapack3 libomp-dev
   

2. Download Pardiso from https://www.pardiso-project.org
3. Rename the file libpardiso-XXXXX.so to libpardiso.so
4. Place the libpardiso.so library somewhere on your load path

   • Alternatively, if the library is located at /full/path/libpardiso.so, start Julia with export LD_LIBRARY_PATH=/full/path; julia

     To make this permanent, modify your .bashrc to include:

     export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/full/path/"
     

5. Set the option linear_solver to pardiso:

   using Libdl
   # Note: these filenames may differ. Check `/usr/lib/x86_64-linux-gnu` for the
   # specific extension.
   Libdl.dlopen("/usr/lib/x86_64-linux-gnu/liblapack.so.3", RTLD_GLOBAL)
   Libdl.dlopen("/usr/lib/x86_64-linux-gnu/libomp.so.5", RTLD_GLOBAL)
   
   using JuMP, Ipopt
   model = Model(Ipopt.Optimizer)
   set_optimizer_attribute(model, "linear_solver", "pardiso")

Tested on a MacBook Pro, 10.15.7.

1. Download Pardiso from https://www.pardiso-project.org
2. Rename the file libpardiso-XXXXX.dylib to libpardiso.dylib.
3. Place the libpardiso.dylib library somewhere on your load path.
   • Alternatively, if the library is located at /full/path/libpardiso.dylib, start Julia with export DL_LOAD_PATH=/full/path; julia
4. Set the option linear_solver to pardiso:

   using JuMP, Ipopt
   model = Model(Ipopt.Optimizer)
   set_optimizer_attribute(model, "linear_solver", "pardiso")

Currently untested. If you have instructions that work, please open an issue.

Tested on a clean install of Ubuntu 20.04.

1. Install Fortran compiler if necessary

   sudo apt install gfortran
   

2. Download the appropriate version of HSL.
   • MA27: HSL for IPOPT from HSL
   • MA86: HSL_MA86 from HSL
   • Other: https://www.hsl.rl.ac.uk/catalogue/
3. Unzip the download, cd to the directory, and run the following:

   ./configure --prefix=</full/path/somewhere>
   make
   make install
   

   where </full/path/somewhere> is replaced as appropriate.
4. Rename the resutling HSL library to /full/path/somewhere/lib/libhsl.so.
   • For ma27, the file is /full/path/somewhere/lib/libcoinhsl.so
   • For ma86, the file is /full/path/somewhere/lib/libhsl_ma86.so
5. Place the libhsl.so library somewhere on your load path.
   • Alternatively, start Julia with export LD_LIBRARY_PATH=/full/path/somewhere/lib; julia
6. Set the option linear_solver to ma27 or ma86 as appropriate:

   using JuMP, Ipopt
   model = Model(Ipopt.Optimizer)
   set_optimizer_attribute(model, "linear_solver", "ma27")
   # or
   set_optimizer_attribute(model, "linear_solver", "ma86")

Tested on a MacBook Pro, 10.15.7.

1. Download the appropriate version of HSL.
   • MA27: HSL for IPOPT from HSL
   • MA86: HSL_MA86 from HSL
   • Other: https://www.hsl.rl.ac.uk/catalogue/
2. Unzip the download, cd to the directory, and run the following:

   ./configure --prefix=</full/path/somewhere>
   make
   make install
   

   where </full/path/somewhere> is replaced as appropriate.
3. Rename the resutling HSL library to /full/path/somewhere/lib/libhsl.dylib.
   • For ma27, the file is /full/path/somewhere/lib/libcoinhsl.dylib
   • For ma86, the file is /full/path/somewhere/lib/libhsl_ma86.dylib
4. Place the libhsl.dylib library somewhere on your load path.
   • Alternatively, start Julia with export DL_LOAD_PATH=/full/path/somewhere/lib; julia
5. Set the option linear_solver to ma27 or ma86 as appropriate:

   using JuMP, Ipopt
   model = Model(Ipopt.Optimizer)
   set_optimizer_attribute(model, "linear_solver", "ma27")
   # or
   set_optimizer_attribute(model, "linear_solver", "ma86")

Currently untested. If you have instructions that work, please open an issue.

Currently untested on all platforms. If you have instructions that work, please open an issue.