what does the summary information in logfile for callback means

Answered

Amir Ghiasi Noughaby

• April 08, 2021 22:20

Hi, 

I use a callback in my model and call another solver in it. After my problem is solved by Gurobi solver, at log output it mentions that 

"User-callback calls 62, time in user-callback 2.76 sec"

I didn't understand what does it mean. I was wondering to know the interpretation of these numerical values. 

Thanks 

0

• 

  Jaromił Najman

  Gurobi Staff

  • April 12, 2021 11:27
  • Edited

  Hi Amir,

  The number of calls tells you how many times your callback function has been called during the optimization. The reported time is the summed time spent over all calls of your callback function.

  For example, you have a callback function \(\texttt{myCallback}\) which you pass to Gurobi via

  m.optimize(myCallback)

  The mentioned LOG line then tells you that your function \(\texttt{myCallback}\) has been called 62 times and Gurobi spent 2.76s in it during the whole optimization process. If you are interested in the number of calls to a particular part of your callback function, you can define a private counter variable

  model._counter = 0

  and increase it in your callback function

  def myCallback(model,where):
    # some code 
    model._counter += 1

  Best regards,
  Jaromił

  0
• 

  Amir Ghiasi Noughaby

  • April 16, 2021 02:11

  

  

  

  Thanks for your response 

  I have added this model._counter to the callback, and when the model is solved, I print the value of it. However, it is not as same as the LOG value. Why it could happen?

   

   

  0
• 

  Jaromił Najman

  Gurobi Staff

  • April 16, 2021 07:10
  • Edited

  Hi Amir,

  Gurobi does not count every call, e.g., the POLLING callbacks are not counted. Thus, it is best to work with your own counter to determine the exact number of calls to your callback function.

  Best regards,
  Jaromił

  0
• 

  Jacob Jin

  • February 22, 2022 12:10

  

  

  

  Hi,I want to know to define a private variable with Gurobi in Julia, just like:

  model._counter = 0

  0
• 

  Simon Bowly

  Gurobi Staff

  • February 23, 2022 06:00

  Hi Jacob,

  I assume you are using JuMP and are looking for a way to track some information across callbacks as we do in some of the Python example codes? Julia does not support dynamic attribute assignment the way Python does, so you'll need to define this variable in an enclosing scope. Here is a short example:

  using JuMP, Gurobi
  
  function solve()
  
      # build your model
      model = Model(Gurobi.Optimizer)
      @variable(model, 0 <= x <= 10, Int)
      @variable(model, 0 <= y <= 10, Int)
      @constraint(model, x + y <= 5)
      @objective(model, Max, x)
  
      # create 'counter' in the outer scope so that the
      # callback can access it
      counter = 0
      function callback(cb_data)
          counter += 1
          println("Lazy constraint callback")
      end
      MOI.set(model, MOI.LazyConstraintCallback(), callback)
  
      # solve
      optimize!(model)
      println("Calls: ", counter)
  end
  
  solve()

  There is some further explanation given in this forum post (although, be aware that this is a few years old, so the JuMP API has changed) https://discourse.julialang.org/t/jump-solver-callback-arguments/13936/3

  Is this what you are after?

  Best regards,
  Simon

  0

Please sign in to leave a comment.