Skip to main content

Dual value Pi and QCPi for specified constraint - python

Answered
Patricio Burdiles
Gurobi-versary
First Comment
First Question

Hello,

I am using Gurobi for Python. I am trying to obtain Pi or QCPi for a specific constraint in my model, similarly as it is asked in the following post:

Referencing constraints and retrieving their dual (Pi) values with C++ – Gurobi Help Center

Could you please extend the explanation for Python as well. I have tried searching for constraint by name and then getting the QCPi attribute, without success:

c0 = model.getConstrByName('dem_%s_%s' % (k,n))
sh_pr[k,n] = c0.get(GRB.Attr.QCPi) 

 

Also, is there a way to see which constraint is associated to Pi when doing the general call: model.getAttr(GRB.Attr.QCPi)

Related to

0

Comments

5 comments

Date Votes
  • Patricio Burdiles
    • Edited
    Gurobi-versary
    First Comment
    First Question

    I have also tried the following without succes:

     

    fixed = model.fixed()
    fixed.Params.QCPDual = 1
    fixed.optimize()

    constrs = fixed.getConstrs()

    shadow_price = fixed.getAttr(GRB.Attr.QCPi)  #this works but there is no information about associated constraints to QCPi values.

    sh_pr = np.zeros((len(shadow_price),2))
    sh_p_i = 0
    for ci in range(0,len(constrs)+1):
        try:
            sh_pr[sh_p_i,0] = constrs[ci].getAttr(GRB.Attr.QCPi) #this does not work
            sh_pr[sh_p_i,1] = constrs[ci].index
            sh_p_i = sh_p_i + 1
        except Exception:
            pass


    thanks

    0
  • Eli Towle
    Gurobi Staff Gurobi Staff

    Model.getConstrs() returns a list of all linear constraints in the model. Model.getQConstrs() returns a list of all quadratic constraints in the model.

    The code

    model.getAttr(GRB.Attr.QCPi)

    is equivalent to

    model.getAttr(GRB.Attr.QCPi, model.getQConstrs())
    0
  • Patricio Burdiles
    Gurobi-versary
    First Comment
    First Question

    Thanks Eli for the clarification. Now it works.

    0
  • Panagiotis Vasilopoulos
    • Edited
    Gurobi-versary
    First Comment
    First Question

    Hello, 

     

    as I'm writing the above after my initial model , (model=ConcreteModel(name=".......")), has been solved/optimized i get the below message:

    'ConcreteModel' object has no attribute 'fixed' . 

     

    Could you please advise why I'm getting this? Should i put the code before solving the initial quadtratic MILP model? 

    0
  • Eli Towle
    Gurobi Staff Gurobi Staff
    • Edited

    The code included in previous posts applies to models built using \(\texttt{gurobipy}\), Gurobi's native Python interface. Specifically, \(\texttt{model}\) is a Model object, and the user calls the Model.fixed() method.

    In contrast, you are building your model using Pyomo. This is a separate, third-party modeling framework. As far as I can tell, the ConcreteModel class has no analogous method to fix every non-continuous variable. However, the Pyomo documentation suggests using the following code snippet to fix all non-continuous variables to their current values:

    import pyomo.environ as pyo

    for var in instance.component_data_objects(pyo.Var, active=True):
    if not var.is_continuous():
        var.fixed = True

    The Fixing Variables and Re-solving section of the Pyomo documentation might also be useful.

    0

Please sign in to leave a comment.

Didn't find what you were looking for?

New post