Accessing the values of Mvars during optimization

Answered

Harsh Gangwar

• July 06, 2024 00:45
• Edited

Hi everyone,

I am currently working on a project where I need to solve optimization models defined using PyCUTEst with Gurobi. My primary objective is to transfer PyCUTEst models to Gurobi models.

I am facing a challenge in that the objective function provided by PyCUTEst accepts input arguments as a NumPy array. Therefore, I need a method to extract the values of Gurobi decision variables (mvars) at any given instant to calculate the objective function. Essentially, I need to:

1. Transfer the values of mvars to the PyCUTEst objective function during the optimization process.
2. Alternatively, incorporate the PyCUTEst objective function directly into the Gurobi model.

So far I have tried Mvar.tolist(), model.m.getVarByName and model.cbGetSolution(x) methods which didn't work.

I would greatly appreciate any suggestions or guidance on achieving this, including relevant materials or documentation that could assist in this process.

Thank you for your help!

 

Related to

• Matrix API

0

• 

  Quin Ackerman

  • July 06, 2024 09:51

  

  

   Try using this:

  from gurobipy import Model, GRB
  import numpy as np
  
  # Create Gurobi model and define variables
  model = Model()
  mvars = model.addVars(n, vtype=GRB.CONTINUOUS, name="vars")
  
  # Define your PyCUTEst objective function
  def pycutest_objective_function(x):
      return np.sum(x**2)
  
  # Optimize the model and extract values
  model.optimize()
  if model.status == GRB.OPTIMAL:
      values = np.array(model.getAttr('X', mvars.values()))
      objective_value = pycutest_objective_function(values)
      print(f"Objective Value: {objective_value}")

  0
• 

  Harsh Gangwar

  • July 06, 2024 17:31

  

  

  "'gurobipy.MVar' object has no attribute 'values'", I also checked Gurobi python API documentation the method values() is not listed there for the Mvars.

  0
• 

  Riley Clement

  Gurobi Staff

  • July 07, 2024 11:41

  Hi Harsh,

  I think it may be worthwhile familiarizing yourself with how models are formulated in our solver (and mixed integer programs more generally).

  Some quick examples can be found here: Functional Code Examples

  More thorough examples here: Resouce Center: Demos

  I need a method to extract the values of Gurobi decision variables (mvars) at any given instant to calculate the objective function...  Transfer the values of mvars to the PyCUTEst objective function during the optimization process.

  It is not possible to transfer values to an external function to be evaluated as the objective during the optimization.

  Alternatively, incorporate the PyCUTEst objective function directly into the Gurobi model.

  This is the right way to do it.  You need to examine what the PyCUTEst objective function is doing with the inputs and formulate it as an expression in Gurobi (before the optimization is run).

  - Riley

  0

Please sign in to leave a comment.