jump-dev
How to settle down the "auto-reset" of programming-code using Gurobi.jl
Hello,
I am attempting to calculate the optimal result of a mixed integer linearized programming (MILP) using JuMP/Gurobi packages. My source code is a classical unit commitment, which is also a mathematical MILP problem.
Some code snippets list below:
scuc = JuMP.Model(Gurobi.Optimizer)
set_optimizer_attribute(scuc, "MIPGap", 0.002)
set_optimizer_attribute(scuc, "IterationLimit", 100)
balabalabala # this is a MILP problem including 600 lines
JuMP.optimize!(scuc)
println("callback gurobisolver\t\t\t\t\t\t\t done")
@test JuMP.termination_status(scuc) == MOI.OPTIMAL
@test JuMP.primal_status(scuc) == MOI.FEASIBLE_POINT
When I run this code, I get the following error:
Set parameter Username
Academic license - for non-commercial use only - expires 2023-04-07
Set parameter MIPGap to value 0.002
start...
objective_function
MILP_type objective_function done
subject to.
constraints: 1) minimum shutup/shutdown time limits done
constraints: 2) binary variable logic done
constraints: 3) shutup/shutdown cost done
constraints: 4) loadcurtailments and spoliedwinds done
constraints: 5) generatos power limits done
constraints: 6) system reserves limits done
constraints: 7) power balance constraints done
constraints: 8) ramp-up/ramp-down constraints done
constraints: 9) piece linearization constraints done
Model has been loaded
Step-4: calculation...
Set parameter MIPGap to value 0.002
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (linux64)
Thread count: 36 physical cores, 72 logical processors, using up to 32 threads
Optimize a model with 1861700 rows, 1271520 columns and 57974450 nonzeros
Model fingerprint: 0xf2ca8ff2
Variable types: 1252080 continuous, 19440 integer (19440 binary)
Coefficient statistics:
Matrix range [3e-02, 4e+02]
Objective range [4e-01, 2e+06]
Bounds range [0e+00, 0e+00]
RHS range [6e-02, 4e+02]
Presolve removed 19410 rows and 12690 columns (presolve time = 5s) ...
Presolve removed 100230 rows and 12690 columns (presolve time = 11s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 17s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 21s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 29s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 32s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 36s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 42s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 46s) ...
Presolve removed 294030 rows and 84745 columns (presolve time = 50s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 55s) ...
Presolve removed 304575 rows and 85325 columns (presolve time = 60s) ...
Presolve removed 305025 rows and 85375 columns (presolve time = 68s) ...
Presolve removed 305025 rows and 86155 columns (presolve time = 70s) ...
Presolve removed 308635 rows and 94225 columns (presolve time = 75s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 80s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 86s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 90s) ...
Presolve removed 310785 rows and 94555 columns (presolve time = 95s) ...
Presolve removed 312113 rows and 94555 columns (presolve time = 100s) ...
Presolve removed 312113 rows and 94555 columns (presolve time = 105s) ...
Presolve removed 312113 rows and 94555 columns
Presolve time: 106.95s
Presolved: 1549587 rows, 1176965 columns, 6653655 nonzeros
Variable types: 1159200 continuous, 17765 integer (17765 binary)
Deterministic concurrent LP optimizer: primal simplex, dual simplex, and barrier
Showing barrier log only...
Root barrier log...
Elapsed ordering time = 7s
Julia has exited.
Press Enter to start a new session.
I am not sure if my code is incorrect or if there is a potential bug here. Anyone helps me would be much appreciated.
To clarify, does
Julia has exited. Press Enter to start a new session.
mean that Julia quit? Is there anything else that gets printed?
If so, that looks like a bug in Gurobi!
We can diagnose things going wrong if you print out the MPS file. Change these two lines:
# scuc = JuMP.Model(Gurobi.Optimizer)
scuc = JuMP.direct_model(Gurobi.Optimizer())
# JuMP.optimize!(scuc)
Gurobi.GRBwrite(scuc, "issue_477.mps")
And then upload the MPS file. (You can upload files to Github through the "Attach files by dragging & dropping" dialog at the bottom of the comment field.)
If you get an error about constraints not supported, only change one line:
# JuMP.optimize!(scuc)
JuMP.write_to_file(scuc, "issue477.mps")
Thank you for your response.
I run my source code in a Atom/JunPro environment. And Julia will be auto-reset and quite after printing this message.
Now, I have being testing these source codes after modifying several input parameters. So, I may submit the MPS file several minus later. Please wait for me. Big thanks.
Another thing I thought might be interesting. When I corrected some input parameters for reducing the UC model complexity, this message is disappeared and the error is gone. From this point, it may not be a problem with my source code itself.
The below lists another successful information for my new case after corrected input boundary.
Set parameter Username
Academic license - for non-commercial use only - expires 2023-04-07
Set parameter MIPGap to value 0.02
start...
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
objective_function
MILP_type objective_function done
subject to.
constraints: 1) minimum shutup/shutdown time limits done
constraints: 2) binary variable logic done
┌ Warning: Axis contains one element: 1. If intended, you can safely ignore this warning. To explicitly pass the axis with one element, pass `[1]` instead of `1`.
└ @ JuMP.Containers ~/.julia/packages/JuMP/0C6kd/src/Containers/DenseAxisArray.jl:166
┌ Warning: Axis contains one element: 1. If intended, you can safely ignore this warning. To explicitly pass the axis with one element, pass `[1]` instead of `1`.
└ @ JuMP.Containers ~/.julia/packages/JuMP/0C6kd/src/Containers/DenseAxisArray.jl:166
constraints: 3) shutup/shutdown cost done
constraints: 4) loadcurtailments and spoliedwinds done
constraints: 5) generatos power limits done
constraints: 6) system reserves limits done
constraints: 7) hidden flexbility done
constraints: 8) power balance constraints done
constraints: 9) ramp-up/ramp-down constraints done
constraints: 10) piece linearization constraints done
Model has been loaded
Step-4: calculation...
Set parameter MIPGap to value 0.02
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (linux64)
Thread count: 36 physical cores, 72 logical processors, using up to 32 threads
Optimize a model with 559376 rows, 448080 columns and 1797299 nonzeros
Model fingerprint: 0x7fd4c5a9
Variable types: 440880 continuous, 7200 integer (6480 binary)
Coefficient statistics:
Matrix range [3e-02, 4e+02]
Objective range [4e-01, 2e+06]
Bounds range [0e+00, 0e+00]
RHS range [6e-02, 4e+02]
Presolve removed 82789 rows and 52235 columns (presolve time = 5s) ...
Presolve removed 102666 rows and 52886 columns (presolve time = 37s) ...
Presolve removed 210809 rows and 156403 columns (presolve time = 45s) ...
Presolve removed 210809 rows and 156403 columns (presolve time = 45s) ...
Presolve removed 210839 rows and 156404 columns
Presolve time: 47.25s
Presolved: 348537 rows, 291676 columns, 1152693 nonzeros
Variable types: 286840 continuous, 4836 integer (4508 binary)
Deterministic concurrent LP optimizer: primal and dual simplex
Showing first log only...
Root simplex log...
Iteration Objective Primal Inf. Dual Inf. Time
0 3.0462200e+10 3.980513e+00 2.857066e+09 51s
39359 1.0323864e+11 0.000000e+00 1.347848e+11 51s
144153 3.8065407e+10 0.000000e+00 1.245023e+12 55s
193617 3.7185028e+10 0.000000e+00 2.503942e+13 60s
208057 3.6270555e+10 0.000000e+00 1.635002e+12 65s
219076 3.6058880e+10 0.000000e+00 1.275159e+13 70s
230995 3.5532924e+10 0.000000e+00 5.772214e+12 75s
241599 3.5039462e+10 0.000000e+00 2.488863e+12 80s
251958 3.4653348e+10 0.000000e+00 1.863989e+12 85s
258421 3.4205817e+10 0.000000e+00 1.720968e+12 90s
270117 3.3534444e+10 0.000000e+00 1.241077e+13 95s
285159 3.3027940e+10 0.000000e+00 1.879681e+13 100s
292118 3.1911418e+10 0.000000e+00 3.372964e+12 105s
300674 3.1281612e+10 0.000000e+00 1.840619e+12 110s
307697 3.1072317e+10 0.000000e+00 1.892933e+12 115s
312816 3.0813234e+10 0.000000e+00 4.462104e+14 120s
323027 3.0622645e+10 0.000000e+00 2.892969e+13 125s
328238 2.9283110e+10 0.000000e+00 4.872103e+14 130s
332968 2.8813960e+10 0.000000e+00 7.943459e+12 135s
Concurrent spin time: 14.58s
Solved with dual simplex
Root relaxation: objective 1.926049e+09, 196599 iterations, 89.45 seconds (109.62 work units)
Total elapsed time = 140.25s
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 1.9260e+09 0 852 - 1.9260e+09 - - 181s
H 0 0 1.926551e+09 1.9260e+09 0.03% - 183s
Explored 1 nodes (197456 simplex iterations) in 183.53 seconds (153.99 work units)
Thread count was 32 (of 72 available processors)
Solution count 1: 1.92655e+09
Optimal solution found (tolerance 2.00e-02)
Best objective 1.926551391554e+09, best bound 1.926049488311e+09, gap 0.0261%
User-callback calls 40009, time in user-callback 0.05 sec
callback gurobisolver done
#TEST: termination_status pass
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
3.1806715919758853e7
the calculation_result has been saved into | calculation_result.txt | done
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Hi, GitHub seems to have a file size limitation of uploaded files.
So, I have to export and sent the "issue 447.mps" file to you through Google Drive. Please check that. Thank you very much.
Oh wow pretty big!
Can you try calling Gurobi from the cmd prompt?
gurobi_cl issue477.mps
Can you post the link to the Google Drive file?
No Problem. Please check this link: https://drive.google.com/file/d/1ttQTnIYJfmn6O7tU_NjwWjSalSVB7fSs/view?usp=sharing. And I have also run "gurobi_cl issue477.mps" as you suggested.
The below shows the prompt.
Is that for the big original problem that crashes Julia? Hmm. It's going to be hard to find the problem if we can't reproduce via gurobi_cl. How much memory do you have on the machine?
Yes. I am trying to test my code through actual cases. Therefore, the size of model structure is a bit large. I run my code in a Linux-based Ubuntu server. The memory is about 125 GB.
Great! "gurobi_cl" work well.
I run "gurobi_cl issue477.mps" on my person PC desktop, and successfully calculate the optimal result. I'm curious about this issue. Why "JuMP.optimize!(scuc)" does not work, but "gurobi_cl" does.
PS D:\New folder> gurobi_cl issue477.mps
Set parameter Username
Set parameter LogFile to value "gurobi.log"
Academic license - for non-commercial use only - expires 2023-04-02
Using license file C:\Users\YipingYUAN\gurobi.lic
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (win64)
Copyright (c) 2022, Gurobi Optimization, LLC
Read MPS format model from file issue477.mps
Reading time = 41.49 seconds
: 1861700 rows, 1271520 columns, 57974450 nonzeros
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 1861700 rows, 1271520 columns and 57974450 nonzeros
Model fingerprint: 0xfe150f41
Variable types: 1252080 continuous, 19440 integer (19440 binary)
Coefficient statistics:
Matrix range [3e-02, 4e+02]
Objective range [4e-01, 2e+06]
Bounds range [1e+00, 1e+00]
RHS range [6e-02, 4e+02]
Presolve removed 100230 rows and 12690 columns (presolve time = 5s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 11s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 15s) ...
Presolve removed 294030 rows and 84630 columns (presolve time = 20s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 25s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 30s) ...
Presolve removed 304575 rows and 85325 columns (presolve time = 35s) ...
Presolve removed 305025 rows and 86155 columns (presolve time = 42s) ...
Presolve removed 308635 rows and 94225 columns (presolve time = 45s) ...
Presolve removed 310465 rows and 94255 columns (presolve time = 50s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 56s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 61s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 65s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 70s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 75s) ...
Presolve removed 312240 rows and 94555 columns
Presolve time: 77.01s
Presolved: 1549460 rows, 1176965 columns, 6259850 nonzeros
Variable types: 1159200 continuous, 17765 integer (17765 binary)
Deterministic concurrent LP optimizer: primal simplex, dual simplex, and barrier
Showing barrier log only...
Root barrier log...
Ordering time: 3.31s
Barrier statistics:
Dense cols : 1433
AA' NZ : 5.678e+06
Factor NZ : 4.914e+07 (roughly 600 MB of memory)
Factor Ops : 5.764e+10 (roughly 1 second per iteration)
Threads : 3
Objective Residual
Iter Primal Dual Primal Dual Compl Time
0 -3.34544413e+12 -5.89446833e+12 1.64e+03 7.64e+08 2.69e+09 107s
1 -2.84771993e+12 -6.38234443e+12 1.20e+03 4.55e+08 1.92e+09 108s
2 -1.94522095e+12 -7.62174646e+12 7.56e+02 1.01e+08 1.18e+09 109s
3 -7.19606302e+11 -7.99508070e+12 2.87e+02 4.09e+06 4.70e+08 110s
4 -5.07994690e+10 -6.64884375e+12 3.50e+01 2.15e+04 6.32e+07 111s
5 6.18493756e+09 -4.25346850e+12 1.46e+01 1.31e+03 2.54e+07 113s
6 3.14815192e+10 -2.55563044e+12 7.20e+00 2.20e+02 1.15e+07 115s
7 4.18017949e+10 -1.51462495e+12 4.47e+00 6.45e+01 6.05e+06 116s
8 4.34925852e+10 -1.08873215e+12 3.27e+00 3.61e+01 3.99e+06 118s
9 4.18602814e+10 -8.39104307e+11 2.28e+00 2.39e+01 2.73e+06 119s
10 3.67468169e+10 -5.68907713e+11 1.57e+00 1.32e+01 1.70e+06 120s
11 2.96258312e+10 -3.32230032e+11 1.06e+00 5.57e+00 9.52e+05 122s
12 1.94947395e+10 -1.74835351e+11 6.17e-01 2.21e+00 4.67e+05 124s
13 1.77579592e+10 -1.16874299e+11 4.25e-01 1.28e+00 3.03e+05 125s
14 1.40744586e+10 -7.41560173e+10 2.75e-01 7.27e-01 1.86e+05 127s
15 1.10957831e+10 -2.48885522e+10 1.75e-01 1.89e-01 7.60e+04 130s
16 1.04491781e+10 -1.19107065e+10 1.53e-01 1.46e-01 4.79e+04 132s
17 7.10684637e+09 -2.92472567e+09 4.17e-02 5.87e-02 1.85e+04 133s
18 6.67974447e+09 -2.20196144e+09 3.08e-02 5.31e-02 1.59e+04 135s
19 6.47759578e+09 2.34115567e+09 2.53e-02 2.07e-02 7.72e+03 136s
20 5.91352316e+09 4.55124474e+09 6.63e-03 5.43e-03 2.41e+03 138s
21 5.83704336e+09 5.36088327e+09 3.33e-03 1.55e-03 8.93e+02 140s
Barrier performed 21 iterations in 140.57 seconds (174.40 work units)
Barrier solve interrupted - model solved by another algorithm
Concurrent spin time: 3.74s
Solved with dual simplex
Root relaxation: objective 5.783519e+09, 154691 iterations, 50.87 seconds (27.02 work units)
Total elapsed time = 153.76s
Total elapsed time = 160.21s
Total elapsed time = 165.28s
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 5.7835e+09 0 1099 - 5.7835e+09 - - 193s
H 0 0 5.790956e+09 5.7835e+09 0.13% - 208s
H 0 0 5.787176e+09 5.7835e+09 0.06% - 225s
0 0 5.7837e+09 0 339 5.7872e+09 5.7837e+09 0.06% - 365s
0 0 5.7837e+09 0 91 5.7872e+09 5.7837e+09 0.06% - 442s
0 0 5.7837e+09 0 278 5.7872e+09 5.7837e+09 0.06% - 1037s
0 0 5.7837e+09 0 139 5.7872e+09 5.7837e+09 0.06% - 1089s
0 0 5.7837e+09 0 93 5.7872e+09 5.7837e+09 0.06% - 1132s
0 0 5.7837e+09 0 93 5.7872e+09 5.7837e+09 0.06% - 1135s
0 0 5.7837e+09 0 106 5.7872e+09 5.7837e+09 0.06% - 1169s
0 0 5.7837e+09 0 103 5.7872e+09 5.7837e+09 0.06% - 1181s
0 0 5.7837e+09 0 73 5.7872e+09 5.7837e+09 0.06% - 1230s
0 0 5.7837e+09 0 63 5.7872e+09 5.7837e+09 0.06% - 1251s
0 0 5.7837e+09 0 74 5.7872e+09 5.7837e+09 0.06% - 1277s
0 0 5.7837e+09 0 78 5.7872e+09 5.7837e+09 0.06% - 1291s
0 0 5.7837e+09 0 79 5.7872e+09 5.7837e+09 0.06% - 1307s
0 0 5.7837e+09 0 72 5.7872e+09 5.7837e+09 0.06% - 1322s
0 0 5.7837e+09 0 84 5.7872e+09 5.7837e+09 0.06% - 1334s
0 0 5.7837e+09 0 62 5.7872e+09 5.7837e+09 0.06% - 1367s
0 0 5.7837e+09 0 60 5.7872e+09 5.7837e+09 0.06% - 1382s
0 0 5.7837e+09 0 58 5.7872e+09 5.7837e+09 0.06% - 1400s
0 0 5.7837e+09 0 58 5.7872e+09 5.7837e+09 0.06% - 1435s
H 0 0 5.783749e+09 5.7837e+09 0.00% - 1459s
Cutting planes:
Gomory: 57
Implied bound: 9969
MIR: 6186
Flow cover: 2475
RLT: 137
Relax-and-lift: 2733
Explored 1 nodes (1170103 simplex iterations) in 1460.45 seconds (1045.78 work units)
Thread count was 12 (of 12 available processors)
Solution count 3: 5.78375e+09 5.78718e+09 5.79096e+09
Optimal solution found (tolerance 1.00e-04)
Best objective 5.783748865417e+09, best bound 5.783727145098e+09, gap 0.0004%
PS D:\New folder>
Why "JuMP.optimize!(scuc)" does not work, but "gurobi_cl" does.
I don't know. That's the problem!
Can you try with all of your options that you had set?
gurobi_cl MIPGap=0.002 IterationLimit=100 issue477.mps
Of course.
The below shows the calculating process.
PS D:\gurobi_test> gurobi_cl MIPGap=0.002 IterationLimit=100 issue477.mps
Set parameter Username
Set parameter IterationLimit to value 100
Set parameter MIPGap to value 0.002
Set parameter LogFile to value "gurobi.log"
Academic license - for non-commercial use only - expires 2023-04-02
Using license file C:\Users\YipingYUAN\gurobi.lic
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (win64)
Copyright (c) 2022, Gurobi Optimization, LLC
Read MPS format model from file issue477.mps
Reading time = 40.75 seconds
: 1861700 rows, 1271520 columns, 57974450 nonzeros
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 1861700 rows, 1271520 columns and 57974450 nonzeros
Model fingerprint: 0xfe150f41
Variable types: 1252080 continuous, 19440 integer (19440 binary)
Coefficient statistics:
Matrix range [3e-02, 4e+02]
Objective range [4e-01, 2e+06]
Bounds range [1e+00, 1e+00]
RHS range [6e-02, 4e+02]
Presolve removed 100230 rows and 12690 columns (presolve time = 5s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 11s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 15s) ...
Presolve removed 294030 rows and 84630 columns (presolve time = 20s) ...
Presolve removed 294030 rows and 84745 columns (presolve time = 25s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 30s) ...
Presolve removed 304575 rows and 85325 columns (presolve time = 35s) ...
Presolve removed 305025 rows and 85375 columns (presolve time = 42s) ...
Presolve removed 305025 rows and 86155 columns (presolve time = 45s) ...
Presolve removed 308635 rows and 94225 columns (presolve time = 50s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 55s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 61s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 66s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 70s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 75s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 80s) ...
Presolve removed 312240 rows and 94555 columns
Presolve time: 83.77s
Presolved: 1549460 rows, 1176965 columns, 6259850 nonzeros
Variable types: 1159200 continuous, 17765 integer (17765 binary)
Deterministic concurrent LP optimizer: primal simplex, dual simplex, and barrier
Showing barrier log only...
Root relaxation: iteration limit, 100 iterations, 9.76 seconds (3.24 work units)
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 postponed 0 - 5.5722e+09 - - 112s
Explored 1 nodes (100 simplex iterations) in 112.66 seconds (140.79 work units)
Thread count was 12 (of 12 available processors)
Solution count 0
Iteration limit reached
Best objective -, best bound 5.572193738095e+09, gap -
PS D:\gurobi_test>
Hmm. Did you use GRBwrite to make the MPS files? The fingerprints are different
Model fingerprint: 0xf2ca8ff2 Model fingerprint: 0xfe150f41
Is there any randomness in how you create the problem?
Oh....
During my case, there are some randomness input data representing the stochastic realization of uncertain resources.
Last night till now, I used this source code to continue running other cases. But I forget to comment this line “Gurobi.GRBwrite(scuc, "issue_447.mps")”, which generating issue447.mps files. As for this, the source code may overwrite the previous .mps file and regenerate a new issue477.mps file.
I have downloaded the issue447,mps file that I shared yesterday, and ran this command. But I found the model fingerprint is still 0xfe150f41. Please check it below.
Gurobi 9.5.1 (win64, gurobi_cl) logging started Sun May 22 14:13:36 2022
Set parameter Username
Set parameter IterationLimit to value 100
Set parameter MIPGap to value 0.002
Set parameter LogFile to value "gurobi.log"
Academic license - for non-commercial use only - expires 2023-04-02
Using license file C:\Users\YipingYUAN\gurobi.lic
Gurobi Optimizer version 9.5.1 build v9.5.1rc2 (win64)
Copyright (c) 2022, Gurobi Optimization, LLC
Read MPS format model from file issue477.mps
Reading time = 40.58 seconds
: 1861700 rows, 1271520 columns, 57974450 nonzeros
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 1861700 rows, 1271520 columns and 57974450 nonzeros
Model fingerprint: 0xfe150f41
Variable types: 1252080 continuous, 19440 integer (19440 binary)
Coefficient statistics:
Matrix range [3e-02, 4e+02]
Objective range [4e-01, 2e+06]
Bounds range [1e+00, 1e+00]
RHS range [6e-02, 4e+02]
Presolve removed 100230 rows and 12690 columns (presolve time = 5s) ...
Presolve removed 100230 rows and 12960 columns (presolve time = 11s) ...
Presolve removed 293910 rows and 84240 columns (presolve time = 15s) ...
Presolve removed 294030 rows and 84630 columns (presolve time = 20s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 25s) ...
Presolve removed 303910 rows and 85145 columns (presolve time = 30s) ...
Presolve removed 304725 rows and 85325 columns (presolve time = 35s) ...
Presolve removed 305025 rows and 86155 columns (presolve time = 42s) ...
Presolve removed 308635 rows and 94225 columns (presolve time = 45s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 50s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 55s) ...
Presolve removed 310765 rows and 94555 columns (presolve time = 60s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 65s) ...
Presolve removed 312240 rows and 94555 columns (presolve time = 70s) ...
Presolve removed 312240 rows and 94555 columns
Presolve time: 74.55s
Presolved: 1549460 rows, 1176965 columns, 6259850 nonzeros
Variable types: 1159200 continuous, 17765 integer (17765 binary)
Deterministic concurrent LP optimizer: primal simplex, dual simplex, and barrier
Showing barrier log only...
Root relaxation: iteration limit, 100 iterations, 7.83 seconds (3.24 work units)
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 postponed 0 - 5.5722e+09 - - 97s
Explored 1 nodes (100 simplex iterations) in 98.16 seconds (140.79 work units)
Thread count was 12 (of 12 available processors)
Solution count 0
Iteration limit reached
Best objective -, best bound 5.572193738095e+09, gap -
Maybe, I should remove any randomness of my input data, and rerun these processes. Is it right ?
To be able to debug this, we need an MPS file that can trigger the segfault when run via gurobi_cl.
Of course, if you can reliably trigger that via Gurobi.jl but not via gurobi_cl, then that's also interesting, but a lot harder to debug. If you can reliably trigger via Gurobi.jl, then try
using JuMP, Gurobi
env = Gurobi.Env(; start = false)
GRBsetintparam(env.ptr_env, GRB_INT_PAR_RECORD, 1)
GRBstartenv(env.ptr_env)
model = Model(() -> Gurobi.Optimizer(env))
You'll have to be on the latest release of Gurobi.jl for this to work: https://github.com/jump-dev/Gurobi.jl/releases/tag/v0.11.2
"GRB_INT_PAR_RECORD" is https://www.gurobi.com/documentation/9.5/refman/record.html#parameter:Record
OK, I will try that, and feedback to you ASAP. Thanks for your help. Please wait, I am very busy these days. I will upload these files in the next days.
..., I will upload related code and continue finding this prob. in the next days. Thanks for waiting.
Closing as stale and non-reproducible, and because this doesn't appear to be a bug in Gurobi.jl.
If you can reproduce this from the gurobi_cl, please contact Gurobi support and they'll be able to help you.