Assigning solver options using Python.

[hbshrestha]

Hi. I am a complete beginner to using the HIGHS solver.
I am using it to solve a mps file using Python.

I have a very simple script, inspired from your documentation.

import highspy
import time

#read model
filename = 'abc.mps'
h.readModel(filename)

#run the model
h.run()
print('Model ', filename, ' has status ', h.getModelStatus())

solution = h.getSolution()
basis = h.getBasis()
info = h.getInfo()
model_status = h.getModelStatus()
print('Model status = ', h.modelStatusToString(model_status))
print()
print('Optimal objective = ', info.objective_function_value)
print('Iteration count = ', info.simplex_iteration_count)
print('Primal solution status = ', h.solutionStatusToString(info.primal_solution_status))
print('Dual solution status = ', h.solutionStatusToString(info.dual_solution_status))
print('Basis validity = ', h.basisValidityToString(info.basis_validity))

#write solution
h.writeSolution("abc_solution.sol",1)

I have a few questions in this regard:

1. I read here that it is possible to use the interior point method for solving the problem. Is the default method simplex? What is the syntax for it? I tried h.run(solver = "ipm"), but I get TypeError with it.

2. How can I set the parallel parameter in the above Python script? For ipm, is it off by default?

3. How can I set run_crossover parameter to off in the above Python script? I see the crossover is set to on as default.

4. How can I set infinite_cost parameter and infinite_bound parameter using Python?

[jajhall]

I have a few questions in this regard:

1. I read [here](https://ergo-code.github.io/HiGHS/dev/solvers/) that it is possible to use the interior point method for solving the problem. Is the default method simplex? What is the syntax for it? I tried `h.run(solver = "ipm")`, but I get TypeError with it.

You can't set options this way, although there is some argument for highspy's run to take the same arguments as the command line HiGHS executable.

You need to make an explicit call to setOptionValue("solver", "ipm"), as illustrated in

HiGHS/examples/call_highs_from_python_highspy.py

Line 29 in 4a7f24a

h.setOptionValue("log_to_console", False)

2. How can I set the `parallel` [parameter](https://ergo-code.github.io/HiGHS/dev/options/definitions/#parallel) in the above Python script? For ipm, is it off by default?

Call

setOptionValue("parallel", "on")

However the IPM solver is serial

3. How can I set `run_crossover` [parameter](https://ergo-code.github.io/HiGHS/dev/options/definitions/#run_crossover) to off in the above Python script? I see the crossover is set to on as default.

Call

setOptionValue("run_crossover", "off")

4. How can I set `infinite_cost` [parameter](https://ergo-code.github.io/HiGHS/dev/options/definitions/#infinite_cost) and `infinite_bound` [parameter](https://ergo-code.github.io/HiGHS/dev/options/definitions/#infinite_bound) using Python?

Call

• setOptionValue("infinite_cost", value)
• setOptionValue("infinite_bound", value)

However, if you're wanting to do this so that you can use costs or bounds greater than 1e15, I strongly advise you not to, as you increase the danger of the solver failing. You should rescale your variables and constraints so that the costs and bounds are rather less than 1e15. Remember that HiGHS will be solving your model using primal/dual tolerances of 1e-7 using double precision (16-digit arithmetic) arithmetic, so costs/bounds greater than 1e16*1e-7=1e9 are asking for trouble

[hbshrestha]

Thank you very much! The setOptionValue method worked for me to set the different solver options/settings.

Regarding scaling the variables and constraints, I understand it is the job of the user. Please flag if there are any parameters in HIGHS that can do such scaling.

[jajhall]

If you try to solve a badly-scaled problem, you'll see output like this

Coefficient ranges:
Matrix [1e+00, 3e+00]
Cost [1e+00, 1e+15]
Bound [0e+00, 0e+00]
RHS [8e+00, 1e+16]
WARNING: Problem has excessively large costs: consider scaling the costs by 1e+5 or less, or setting option user_cost_scale to -17 or less
WARNING: Problem has excessively large bounds: consider scaling the bounds by 1e+7 or less, or setting option user_bound_scale to -21 or less

If you don't want to do your own scaling, you can call

• setOptionValue("user_cost_scale", -17)
• setOptionValue("user_bound_scale", -21)

This will scale the all the costs down by 2^{17}, and all the variable and constraint bounds down by 2^{21}, possibly over-scaling some well-scaled variables and constraints, so that model precision is lost. [Some variables may be constrained to lie between bounds that are comparable with the primal feasibility tolerance, or some costs may be small comparable with the dual feasibility tolerance.

[hbshrestha]

Thank you for your response!

[hbshrestha]

I have started a new related discussion in a separate thread. I will close this one.