Very Long Processing Time for Gurobi Simulation
AnsweredI am trying to run a non-convex quadratically constrained Gurobi simulation with following output:
Yet I am unable to get any result, do I need more cores to process this program?
I set the time limit to 1000 sec - 17 mins approx, but even running for 24 hours gives no result and the program keeps going on. Currently I am using 8 cores for the processing.
Set parameter TimeLimit to value 1000
Set parameter NonConvex to value 2
Gurobi Optimizer version 10.0.1 build v10.0.1rc0 (win64)
Optimize a model with 57226 rows, 2095090 columns and 84051 nonzeros
Model fingerprint: 0xd7cbbdd9
Model has 1922 quadratic objective terms
Model has 8908 quadratic constraints
Model has 31 general constraints
Variable types: 2095090 continuous, 0 integer (0 binary)
Coefficient statistics:
Matrix range [2e-02, 6e+01]
QMatrix range [3e-04, 1e+02]
QLMatrix range [1e+00, 1e+00]
Objective range [1e+00, 1e+09]
QObjective range [4e-02, 4e-01]
Bounds range [1e+03, 5e+03]
RHS range [3e-03, 1e+03]
QRHS range [1e+00, 1e+00]
Presolve removed 52019 rows and 2079412 columns
Presolve time: 1.37s
Presolved: 56708 rows, 23045 columns, 148140 nonzeros
Presolved model has 1770 quadratic objective terms
Presolved model has 11650 bilinear constraint(s)
Variable types: 23045 continuous, 0 integer (0 binary)
Root simplex log...
Iteration Objective Primal Inf. Dual Inf. Time
25042 4.3889986e+01 3.038022e-02 0.000000e+00 5s
29570 4.3890182e+01 0.000000e+00 0.000000e+00 7s
Root relaxation: objective 4.389018e+01, 29570 iterations, 4.70 seconds (4.78 work units)
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 43.89018 0 3609 - 43.89018 - - 6s
0 0 3494.30254 0 3239 - 3494.30254 - - 24s
0 0 3789.74850 0 3210 - 3789.74850 - - 33s
0 0 3789.74850 0 3212 - 3789.74850 - - 33s
0 0 4044.12630 0 3368 - 4044.12630 - - 39s
0 0 4044.12738 0 3406 - 4044.12738 - - 44s
0 0 4268.17678 0 3568 - 4268.17678 - - 49s
0 0 4268.23903 0 3606 - 4268.23903 - - 50s
0 0 4473.65195 0 3657 - 4473.65195 - - 118s
0 0 4473.74061 0 3647 - 4473.74061 - - 119s
0 0 4681.40157 0 3670 - 4681.40157 - - 174s
0 0 4681.40157 0 3672 - 4681.40157 - - 175s
0 0 4844.95060 0 3647 - 4844.95060 - - 178s
0 0 4844.95060 0 3651 - 4844.95060 - - 178s
0 0 4987.02980 0 3642 - 4987.02980 - - 179s
0 0 4987.02980 0 3655 - 4987.02980 - - 179s
0 0 5109.23314 0 3694 - 5109.23314 - - 180s
0 0 5109.23314 0 3695 - 5109.23314 - - 180s
0 0 5218.74339 0 3691 - 5218.74339 - - 181s
0 0 5218.74339 0 3696 - 5218.74339 - - 181s
0 0 5316.40107 0 3705 - 5316.40107 - - 181s
0 0 5316.40107 0 3711 - 5316.40107 - - 182s
0 0 5404.33983 0 3756 - 5404.33983 - - 182s
0 0 5404.33983 0 3760 - 5404.33983 - - 182s
0 0 5483.29147 0 3767 - 5483.29147 - - 183s
0 0 5483.29147 0 3774 - 5483.29147 - - 183s
0 0 5554.39418 0 3828 - 5554.39418 - - 184s
0 0 5554.39418 0 3833 - 5554.39418 - - 184s
0 0 5618.43294 0 3802 - 5618.43294 - - 184s
0 0 5618.43294 0 3812 - 5618.43294 - - 184s
0 0 5676.19955 0 3897 - 5676.19955 - - 185s
0 0 5676.19955 0 3902 - 5676.19955 - - 185s
0 0 5676.30222 0 3981 - 5676.30222 - - 186s
0 0 5676.30222 0 3981 - 5676.30222 - - 186s
0 0 5676.35427 0 3979 - 5676.35427 - - 186s
0 0 5676.35427 0 3979 - 5676.35427 - - 188s
0 2 5676.35427 0 3979 - 5676.35427 - - 228s
3 8 5676.35508 2 3977 - 5676.35427 - 4.7 283s
19 22 5676.35508 4 3985 - 5676.35427 - 10912 305s
30 36 5676.35508 6 3983 - 5676.35427 - 9731 333s
37 68 5676.35508 7 3981 - 5676.35427 - 10060 372s
69 201 5676.35508 13 3978 - 5676.35427 - 7729 400s
202 273 5676.36237 45 3954 - 5676.35427 - 3486 472s
274 351 5676.35508 58 3951 - 5676.35427 - 3905 555s
352 465 5676.35508 73 3954 - 5676.35427 - 4250 664s
468 736 5676.35508 92 3956 - 5676.35427 - 4403 829s
747 1314 5676.35508 156 3956 - 5676.35427 - 3593 987s
1331 1440 5676.35508 302 3887 - 5676.35427 - 2438 1000s
Cutting planes:
RLT: 1278
BQP: 1116
Explored 1459 nodes (3551353 simplex iterations) in 1000.14 seconds (1228.97 work units)
Thread count was 8 (of 8 available processors)
Solution count 0
Time limit reached
Best objective -, best bound 5.676354272431e+03, gap -
-
Can you try using Gurobi 11.0.0 (the latest release)?
Cheers,
David1 -
Thank you David, I just pulled the update and new license!
Here's the result I got, (I added feasibility tolerance to bare minimum as well) :
Set parameter TimeLimit to value 1000
Set parameter FeasibilityTol to value 0.01
Set parameter NonConvex to value 2
Gurobi Optimizer version 11.0.0 build v11.0.0rc2 (win64 - Windows 10.0 (19045.2))
CPU model: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz, instruction set [SSE2|AVX|AVX2|AVX512]
Thread count: 4 physical cores, 8 logical processors, using up to 8 threads
Optimize a model with 57226 rows, 2095090 columns and 84051 nonzeros
Model fingerprint: 0xd7cbbdd9
Model has 1922 quadratic objective terms
Model has 8908 quadratic constraints
Model has 31 general constraints
Variable types: 2095090 continuous, 0 integer (0 binary)
Coefficient statistics:
Matrix range [2e-02, 6e+01]
QMatrix range [3e-04, 1e+02]
QLMatrix range [1e+00, 1e+00]
Objective range [1e+00, 1e+09]
QObjective range [4e-02, 4e-01]
Bounds range [1e+03, 5e+03]
RHS range [3e-03, 1e+03]
QRHS range [1e+00, 1e+00]
Presolve removed 48316 rows and 2074701 columns
Presolve time: 0.20s
Presolved: 68802 rows, 28974 columns, 166303 nonzeros
Presolved model has 1770 quadratic objective terms
Presolved model has 13534 bilinear constraint(s)
Warning: Model contains variables with very large bounds participating
in product terms.
Presolve was not able to compute smaller bounds for these variables.
Consider bounding these variables or reformulating the model.
Solving non-convex MIQCP
Variable types: 28974 continuous, 0 integer (0 binary)
Root simplex log...
Iteration Objective Primal Inf. Dual Inf. Time
16143 6.5447843e+03 9.490257e+01 0.000000e+00 5s
32185 1.5302265e+05 5.807236e+01 0.000000e+00 10s
45563 4.9409851e+03 0.000000e+00 1.322601e+01 15s
53269 4.9410334e+03 0.000000e+00 0.000000e+00 18s
Root relaxation: objective 4.941033e+03, 53269 iterations, 13.98 seconds (10.62 work units)
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 4941.03340 0 3388 - 4941.03340 - - 17s
0 0 4941.03340 0 3332 - 4941.03340 - - 18s
0 0 4941.03340 0 3280 - 4941.03340 - - 20s
0 0 4941.03340 0 3280 - 4941.03340 - - 20s
0 0 4941.03340 0 3280 - 4941.03340 - - 20s
0 0 4941.03340 0 3280 - 4941.03340 - - 20s
0 0 4941.03340 0 3280 - 4941.03340 - - 21s
0 2 4941.03340 0 3280 - 4941.03340 - - 559s
1 5 4941.03340 1 3280 - 4941.03340 - 6.0 605s
3 8 4941.03340 2 3280 - 4941.03340 - 35.7 805s
7 12 4941.03340 3 3280 - 4941.03340 - 18820 857s
11 16 4941.03340 4 3279 - 4941.03340 - 12421 897s
15 20 4941.03340 4 3279 - 4941.03340 - 9120 916s
19 24 4941.03340 5 3276 - 4941.03340 - 7207 945s
23 28 4941.03340 5 3278 - 4941.03340 - 5967 965s
27 32 4941.03340 6 3277 - 4941.03340 - 5089 985s
31 36 4941.03340 6 3276 - 4941.03340 - 4443 1000s
Cutting planes:
RLT: 2090
Explored 35 nodes (197142 simplex iterations) in 1000.06 seconds (501.66 work units)
Thread count was 8 (of 8 available processors)
Solution count 0
Time limit reached
Best objective -, best bound 4.941033396743e+03, gap -0 -
Please let me know what do you suggest further?
0 -
Looks like a tough model!
You can also try some of our heuristics, such as:They might help obtain feasible solutions.
If none of these help you might want to try:- An alternative formulation for your problem.
- The following: PreMIQCPForm, PreQLinearize
Cheers,
David1
Please sign in to leave a comment.
Comments
4 comments