Sending multiple small optimization problems at once to gurobipy to avoid communication overhead

回答済み

Madhushini Narayana Prasad

2022年12月15日 16:32

• 
• 
• 

Is there a quick and smart to construct several small optimization problems and send them all to Gurobi at once to avoid communication overhead?

関連

• Python
• Compute Server

0

• 

  Jaromił Najman

  • Gurobi Staff

  2022年12月19日 09:06

  Currently, there is no dedicated feature to support sending multiple models at once. However, we are aware of the usefulness of this feature and it is on our future roadmap.

  For now, as a workaround, you can combine multiple small models in one bigger model. For example, let's assume you have 1000 tiny models which are solved in split seconds such that the communication overhead already takes longer than the solution of each model. Then, you could combine 100 models each into 1 big model to have only 10 bigger models. The bigger models would consist of disconnected models which Gurobi will detect and solve quickly. You should experiment a bit to find a good amount of models to combine and send as one model.

  0

サインインしてコメントを残してください。