メインコンテンツへスキップ

Numerical Issues with Large QIP

回答済み

コメント

6件のコメント

  • 正式なコメント
    Simranjit Kaur
    • Gurobi Staff
    This post is more than two 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 try Gurobot, our chatbot interface offering instant, expert-level support.
  • Jaromił Najman
    • Gurobi Staff

    Hi Derek,

    Could you please share the model such that we can have a closer look? You can use the write method to generate an MPS file.

    Note that uploading files in the Community Forum is not possible but we discuss an alternative in Posting to the Community Forum.

    Best regards, 
    Jaromił

    0
  • Derek Long
    • Gurobi-versary
    • First Comment
    • First Question

    Hi Jaromił,

    Thanks for your response. Here's the model with t=1,...,100 and the model with t=1,...,177 (which is the actual problem I am aiming to solve but cannot currently optimize due to insufficient RAM). If you have a chance to run the latter model, would you be able to tell me how much RAM I would need for it?

    Kind regards,

    Derek

    0
  • Jaromił Najman
    • Gurobi Staff

    Hi Derek,

    I was able to successfully run model with t=1,..,100 on a machine with 4 cores and 32GB of RAM. I ran the model for ~2 hours with default settings and it used at most ~8GB of RAM.

    I was also able to successfully run model with t=1,..,177 on the same machine, i.e., a machine with 4 cores and 32GB of RAM. I ran the model for ~1 hour with default settings and it used at most ~28GB of RAM.

    If you are planning to solve these models, I would recommend switching to a powerful machine, maybe a cloud machine.

    Best regards, 
    Jaromił

    0
  • Derek Long
    • Gurobi-versary
    • First Comment
    • First Question

    Great, were you able to find a feasible solution to those models within that time?

    Kind regards,
    Derek

    0
  • Jaromił Najman
    • Gurobi Staff

    No, unfortunately not. Because each LP relaxation solve takes a very long time, I think that one good approach would be to run the NoRelHeurTime for a very long time, probably at least a few hours and hope that it finds a feasible solution.

    An alternative would be to compute a feasible solution via a custom heuristic. However, this is most often very difficult and maybe even impossible.

    Best regards, 
    Jaromił

    0

投稿コメントは受け付けていません。