Python/ Running time: Why do I get different running time each time I solve my model?

Ongoing

Mahdis Bayani

• September 01, 2020 18:09

Hi,

I have a simple BQP model coded in python and solved by gurobi. But each time I run it, although the solution and obj value are the same, the running time varies even to 25% in some instances. I also have a simple LP model which results in the same problem. I check out the time with both model.Runtime and time.time() and both of them change in iterations of running the code for each specific instance. When I check the gurobi log, I see the logs are quite different each time I run the model. I wonder if I need to fix something to get consistent running time. I would add I have the same problem either running on my computer or in a shared cluster.

I appreciate any assistance.  

0

• 

  Jaromił Najman

  Gurobi Staff

  • September 02, 2020 07:56

  Hi,

  Could you share the logs of a run with different runtimes on your computer?

  Please note that when running on a shared cluster, differences in runtimes are to be expected except for the case when you are able to reserve a cluster node for your computation only.

  Please also have a look at our knowledge base articles on determinism, performance on different machines, and fingerprint as these may hold the answer for your question.

  Best regards,
  Jaromił

  0
• 

  Matthias Miltenberger

  Gurobi Staff

  • September 02, 2020 09:39

  duplicate of this Stackoverflow question

  0

Please sign in to leave a comment.