Skip to main content

'Python has stopped working' after called Gurobi to solve hundreds of large size MILP

Answered
Bo Lan
  • Gurobi-versary
  • First Comment
  • First Question

Greetings,

I  need to  generate 600 MILP instances and solve them by using Python calling Gurobi. Each of the MILP instance has a similar scale like this:

Optimize a model with 328593 rows, 485927 columns and 2346683 nonzeros
Variable types: 485640 continuous, 287 integer (286 binary)

I set 'model.Params.method = 2' to use the interior point method to reduce computing  time (rather than Simplex). For most instances they can be solved within 20~40 seconds. So I set a time limit=180 seconds. If an instance  cannot be solved to optimal in 180 seconds the code will skip that instance and randomly generate another instance and so on. But normally when Gurobi solved about 300 instances Python would say 'Python has stopped working' and 'Not Responding'. I googled and tried three things as the following:

(1) Move the  Python file to a large disk (215GB free space). Windows 7  Enterprise, Intel core i7-4790 CPU @ 3.60GHz, RAM 32GB, 64-bit operating System.

(2) model.Params.NodefileStart = 0.5  to save the node file to the disk.
(3) model.Params.Threads = 2  to reduce the utilization of cores from as many as possible (4) to 2.

But the same problem still happened, i.e. Python stopped working after solving 100~300 large size problems.

I'm not sure whether I should adjust Gurobi or Python. Shall I clean the memory after I calculate each of the large size instance something else?

When I solved 600 smaller sized similar instances like this:

Optimize a model with 16683 rows, 23501 columns and 113093 nonzeros
Variable types: 23436 continuous, 65 integer (64 binary)

There was no problem.

Hope someone can give  me some hints.

Thanks,

Larry

0

Comments

5 comments

Date Votes
  • Official comment
    Simranjit Kaur
    • Gurobi Staff Gurobi Staff
    This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum. Or why not try our AI Gurobot?.
  • Sonja Mars
    • Gurobi Staff Gurobi Staff

    Hi Larry,

    Are you generating these instances all at the same time (in parallel) or one after another?

    • If the first is the case you need a machine with lots of RAM. Disk space is not really interesting here.
    • If the second is the case and you don't need the models anymore you should make sure to delete them. It is possible that the automatic garbage collector of Python has not yet freed the memory and you are simply running out of memory.

    Best regards,

      Sonja

    0
  • Bo Lan
    • Gurobi-versary
    • First Comment
    • First Question

    Sonja,

    Thank you for your reply.

    I generated instances in series, i.e. generated one and solved it then generated and solved the next one. It means at anytime there's only one instance being processed.

    Every time when I use

    model = grb.Model()

    to generate  a new model I think the old one should have been deleted. Is that correct? Or is there any  simple code to delete the old model completely?

    Thanks,

    Larry

    0
  • Sonja Mars
    • Gurobi Staff Gurobi Staff
    Edited

    Hi,

    If you do it that way they should get deleted. However, you can also force a model model to be garbage collected with the statement

    del model

    Best regards,

    Sonja

     

    0
  • Bo Lan
    • Gurobi-versary
    • First Comment
    • First Question

    Thank you, Sonja. That's easy and helpful.

    Larry

    0

Post is closed for comments.

Didn't find what you were looking for?

New post