Save optimizer state and resume.

Answered

Milin Kodnongbua

July 16, 2023 16:47

• 
• 
• 

Is it possible for the state of the optimizer to be saved? I want to run Gurobi on an preemptible node and want to resume from the saved state.

Related to

• Search Tree
• MIPStart

0

• 

  Simranjit Kaur

  • Gurobi Staff

  July 17, 2023 03:32 Edited

  Hi Milin,

  Calling the optimize() method again on a model will continue the optimisation process from the last state where the optimizer had stopped. Please see a sample code below.

  import gurobipy as gp
  
  #construct your model
  ...
  
  # your custom terminating criteria
  model.params.MipGap=0.05
  ...
  
  # optimize
  model.optimize()
  
  # do something with the solution/s
  ...
  
  #redefine termination criteria
  model.params.MipGap=0.02
  ...
  
  #resume optimisation
  model.optimize()

  Best regards,
  Simran

  0
• 

  Milin Kodnongbua

  July 17, 2023 04:00

  • 
  • 
  • 

  Thanks for the response.

  However, I mean load the state to a new instance of solver in a new process. The same solver is not in memory anymore. Is there a way to save/load from disk?

  0
• 

  Simranjit Kaur

  • Gurobi Staff

  July 17, 2023 04:36

  Hi Milin, 

  It is not possible to save a snapshot of the optimizer state to a disk. One cannot store the branch-and-bound tree or the presolve transformations. However, you can store the solution at the end of a solve by writing the SOL file and use it to warm-start the next solve.  

  Best regards,
  Simran

  0

Please sign in to leave a comment.