Output solutions during solve

Answered

Aaron Slobodin

September 12, 2022 16:14

• 
• 
• 

Hi there.

When using a gurobi solver to solve an instance of an ILP, is it possible for it to output 'current' best possible solutions (in terms of variable assignment) during the solve? For instance, suppose that (during the solve) the current 'best' solution gives an objective value of 23. However, a new 'best' solution of 22 is found. While it continues to solve, it would be fantastic to have it also print out (in the log) the variable assignments for the current best. This would be particularly helpful for solves which require a long time.

Thanks!

Related to

• Pyomo

0

• 

  Eli Towle

  • Gurobi Staff

  September 12, 2022 17:50

  Here are two ways to view new incumbent solutions as Gurobi finds them:

  • Query the new incumbent solution in a MIPSOL callback. In Python, this can be done with Model.cbGetSolution(). You might find the callback examples (e.g., callback.py) helpful.
  • Set the SolFiles parameter. This directs Gurobi to write each new incumbent solution to disk as a .sol solution file. This is a bit easier than the previous approach, as it only requires you to set a parameter.

  0
• 

  Aaron Slobodin

  September 13, 2022 03:01

  • 
  • 
  • 

  Awesome. Thanks so much Eli. SolFile is perfect.

  0
• 

  Javier Salmeron

  May 13, 2025 17:52

  • 

  Hi Eli.  I'd like to follow up on this. 

  I want to use the SolFiles option because I am calling Gurobi from Python-Pyomo, that is, I'm not using the native Gurobi callable library. 

  The problem I see is that the .sol files contain the solution in the internal storage form x[1], x[2], … instead of the way in which I name the variables in Pyomo.  Is there a way to obtain those original names in the .sol files?  (I looked for other Gurobi name-related parameters but did not find anything).

  Thank you.

  0
• 

  Eli Towle

  • Gurobi Staff

  May 13, 2025 19:13

  The variable names in the .sol files match the names Pyomo gives Gurobi when it is creating the Gurobi model. You need to direct Pyomo to preserve the names of the Pyomo components when passing the model to Gurobi. For example, when using the GurobiPersistent interface:

  solver = pyo.SolverFactory("gurobi_persistent")
  solver.set_instance(model, symbolic_solver_labels=True)
  results = solver.solve(model, tee=True, options={"SolFiles": "foo"})

  0
• 

  Javier Salmeron

  May 14, 2025 00:01

  • 

  Thank you, Eli and Maliheh.  I am using the standard gurobi (not gurobi_persistent). 

  opt = SolverFactory("gurobi")
  opt.options["symbolic_solver_labels"] = True
  result = opt.solve(myModel)

  This causes an error that “symbolic_solver_labels” is not recognized by gurobipy.

  Is there another way to pass this request to gurobi or do I need to (install and) use gurobi_persistent to enable this feature?

  0
• 

  Eli Towle

  • Gurobi Staff

  May 14, 2025 01:25

  Can you try passing Pyomo's \(\texttt{symbolic_solver_labels}\) setting like this instead?

  opt = SolverFactory("gurobi")
  results = opt.solve(
      model, tee=True, options={"SolFiles": "foo"}, symbolic_solver_labels=True
  )

  0
• 

  Javier Salmeron

  May 14, 2025 16:20

  • 

  Hi Eli, 

  This definitely did the trick.  Thank you very much!

   

  0

Please sign in to leave a comment.