Tuning error: Unable to improve on baseline parameter set
AnsweredDear Gurobi community,
I have a MILP model coded in python and I use Gurobi solver. It has 360 SOS constraints, 8869 continuous variables and 0 integer (0 binary) variable. It runs for hours without any solutions. So, I used the Gurobi tuning tool to find the best settings to solve this problem. However, I got this error:
"Tested 10 parameter sets in 108083.22s
Unable to improve on baseline parameter set."
Can you please give me any insights to solve this error?
Here is the tuning log file: https://www.dropbox.com/s/x9fe65obljhy16i/log_12dt14dx_pipe.txt?dl=0

Hi!
This seems to be a pretty difficult problem. The tuning tool works best if the baseline runs already find a solution  in your case, all runs were unsuccessful and so there are no hints on how to improve the performance. As a rule of thumb, you should try testing around 100 parameter sets to see some significant improvements coming from the tuner  10 is definitely not enough.
I suggest dialing up heuristics and also try the NoRelHeurWork or NoRelHeurTime parameters. Another useful parameter might be MIPFocus=1 to try to find a feasible solution at all.
It might also be that this instance is just extremely hard and you need to spend significantly more time to solve it.
Cheers,
Matthias
Please sign in to leave a comment.
Comments
1 comment