[QST] How do I get the values of the solution after calling .solve()

[araml]

I'm trying to understand the api and I found the following issue

The Problem class exposes the method getIncumbentSolution

cuopt/python/cuopt/cuopt/linear_programming/problem.py

Line 1694 in 0aa1b31

def getIncumbentValues(self, solution, vars):

which takes a solution and the variables as an input, you can get the variables from the method

cuopt/python/cuopt/cuopt/linear_programming/problem.py

Line 389 in 0aa1b31

def getVariables(self):

but there is no way to get a solution without using

cuopt/python/cuopt/cuopt/linear_programming/solver/solver.py

Line 13 in 0aa1b31

def Solve(data_model, solver_settings=None):

directly which I think defeats the purpose of using the Problem class

A possible fix would be for Problem.solve()

cuopt/python/cuopt/cuopt/linear_programming/problem.py

Line 1931 in 0aa1b31

def solve(self, settings=solver_settings.SolverSettings()):

to return the solution after populate_solution or save it in a member variable like self.solution

[araml]

I wasn't sure if this was a bug or a request, I can reopen if it needed but it seems at least a bit annoying. Also I can also create a PR if the proposed fix makes sense :).

[chris-maes]

I'm not sure I understand the issue. You can get solutions directly.

You can access the optimal solution values directly from the variables. If x is a variable you have created, you can use x.Value or x.getValue() to get the value of the variable in a an optimal solution.

[chris-maes]

These examples might help make things more clear
https://docs.nvidia.com/cuopt/user-guide/latest/cuopt-python/lp-qp-milp/lp-qp-milp-examples.html

[araml]

I'm not sure I understand the issue. You can get solutions directly.

You can access the optimal solution values directly from the variables. If x is a variable you have created, you can use x.Value or x.getValue() to get the value of the variable in a an optimal solution.

What I'm saying is that you cannot use getIncumbentValues without calling solver.Solve() yourself since the class Problem doesn't expose or store the solution returned by solver.Solve(...). So either getIncumbentValues is a free function or there is something missing in Problem :)

[araml]

These examples might help make things more clear https://docs.nvidia.com/cuopt/user-guide/latest/cuopt-python/lp-qp-milp/lp-qp-milp-examples.html

I don't see anything related to getIncumbentValues.

[araml]

So suppose you do something like

problem = Problem.readMPS(...)
problem.solve()

At this point I can't call getIncumbentValues since Problem doesn't expose the solution returned by solver.Solve(...).