CPLEX ran out of memory even after nodefile=3!!!
188 views
Skip to first unread message
Amir Lm
Nov 10, 2016, 1:25:56 PM11/10/16
to AMPL Modeling Language
Hi all,
CPLEX is giving me the memory error as follows.
I have put the command below.
ampl: model IntM-S.mod;
ampl: data IntM-S.dat;
ampl: options solver cplex;
ampl: option show_stats 1;
ampl: option cplex_options 'mipdisplay=2 nodefile=3';
ampl: solve;
and CPLEX gave the the presolve of:
Presolve eliminates 277 constraints and 955 variables.
Adjusted problem:
2695 variables:
2340 binary variables
25 integer variables
330 linear variables
902 constraints, all linear; 21926 nonzeros
385 equality constraints
517 inequality constraints
1 linear objective; 2180 nonzeros.
CPLEX 12.6.1.0: mipdisplay=2
nodefile=3
MIP Presolve eliminated 228 rows and 180 columns.
MIP Presolve modified 445 coefficients.
Reduced MIP has 519 rows, 2360 columns, and 13704 nonzeros.
Reduced MIP has 2335 binaries, 13 generals, 0 SOSs, and 0 indicators.
Probing time = 0.03 sec. (3.30 ticks)
MIP Pre solved modified 30 coefficients.
Reduced MIP has 519 rows, 2360 columns, and 13704 nonzeros.
Reduced MIP has 2335 binaries, 13 generals, 0 SOSs, and 0 indicators.
Probing time = 0.03 sec. (3.20 ticks)
Clique table members: 8227.
MIP emphasis: balance optimality and feasibility.
MIP search method: dynamic search.
Parallel mode: deterministic, using up to 4 threads.
Root relaxation solution time = 0.08 sec. (17.02 ticks)
Any idea why do I still get this?!
There may be further error information in the clone logs.
GUB cover cuts applied: 13
Clique cuts applied: 48
Cover cuts applied: 256
Implied bound cuts applied: 3
Flow cuts applied: 2
Mixed integer rounding cuts applied: 111
Zero-half cuts applied: 26
Lift and project cuts applied: 3
Gomory fractional cuts applied: 13
Root node processing (before b&c):
Real time = 3.35 sec. (1476.59 ticks)
Parallel b&c, 4 threads:
Real time = 404.18 sec. (223642.10 ticks)
Sync time (average) = 36.63 sec.
Wait time (average) = 37.01 sec.
------------
Total (root+branch&cut) = 407.54 sec. (225118.69 ticks)
Freeing branch-and-bound tree with 252652 nodes
211000 nodes freed
223000 nodes freed
248000 nodes freed
CPLEX 12.6.1.0: ran out of memory.
0 MIP simplex iterations
0 branch-and-bound nodes
No basis.
Amir Lm
Nov 11, 2016, 12:11:06 PM11/11/16
to AMPL Modeling Language
Is there anyone that can give me some advice how to get rid of this problem?!
Robert Fourer
Nov 11, 2016, 9:36:49 PM11/11/16
to am...@googlegroups.com
Even with nodefile=3, there is a lot of information that CPLEX has to keep in memory. To reduce memory use try memoryemphasis=1, which compresses some data kept in memory; and threads=1, which will reduce the amount of memory needed though it will also slow the processing. Also in cplex_options, set mipdisplay=2 and mipinterval=100 to get some more useful information as to what is happening in CPLEX. (If there are too many lines then replace 100 by a larger interval.)
Bob Fourer
am...@googlegroups.com
=======
Bob Fourer
am...@googlegroups.com
=======
Amir Lm
Nov 12, 2016, 12:29:59 PM11/12/16
to AMPL Modeling Language, 4...@ampl.com
Thank you so much for your reply, Bob.
It was insightful as always.
Despite the fact that my model is considered a fairly small model with following characteristics, I was still not able to get the optimal results.
It looks like CPLEX can not prove optimality for the obtained solution.
============================================
ampl: reset;
ampl: model IntM-S.mod;
ampl: data IntM-S.dat;
ampl: options solver cplex;
ampl: option show_stats 1;
ampl: option cplex_options 'mipdisplay=2 nodefile=3 memoryemphasis=1 threads=1 mipinterval=100';
ampl: solve;
Presolve eliminates 281 constraints and 957 variables.
Adjusted problem:
2693 variables:
2340 binary variables
23 integer variables
330 linear variables
911 constraints, all linear; 21777 nonzeros
385 equality constraints
526 inequality constraints
memoryemphasis=1
threads=1
mipinterval=100
MIP Presolve eliminated 263 rows and 200 columns.
MIP Presolve modified 3559 coefficients.
Reduced MIP has 503 rows, 2348 columns, and 13413 nonzeros.
Reduced MIP has 2325 binaries, 11 generals, 0 SOSs, and 0 indicators.
Probing time = 0.02 sec. (3.19 ticks)
MIP Presolve modified 5 coefficients.
Reduced MIP has 503 rows, 2348 columns, and 13413 nonzeros.
Reduced MIP has 2325 binaries, 11 generals, 0 SOSs, and 0 indicators.
Probing time = 0.02 sec. (3.14 ticks)
Clique table members: 8074.
MIP emphasis: balance optimality and feasibility.
MIP search method: dynamic search.
Parallel mode: none, using 1 thread.
Root relaxation solution time = 0.03 sec. (14.38 ticks)
Nodes Cuts/
Node Left Objective IInf Best Integer Best Bound ItCnt Gap
0 0 12921.7081 141 12921.7081 772
0 0 12923.5210 225 Cuts: 225 1228
0 0 12923.5740 245 Cuts: 225 1662
0 0 12924.1634 228 Cuts: 225 3183
0 0 12924.7409 288 Cuts: 225 4115
0 0 12925.4059 333 Cuts: 207 5542
0 0 12926.0790 246 Cuts: 189 6312
0 0 12926.1596 278 Cuts: 225 6967
0 2 12926.1596 278 12926.1596 6967
Nodes Cuts/
Node Left Objective IInf Best Integer Best Bound ItCnt Gap
0 0 12921.7081 141 12921.7081 772
0 0 12923.5210 225 Cuts: 225 1228
0 0 12923.5740 245 Cuts: 225 1662
0 0 12924.1634 228 Cuts: 225 3183
0 0 12924.7409 288 Cuts: 225 4115
0 0 12925.4059 333 Cuts: 207 5542
0 0 12926.0790 246 Cuts: 189 6312
0 0 12926.1596 278 Cuts: 225 6967
0 2 12926.1596 278 12926.1596 6967
Elapsed time = 2.42 sec. (1273.46 ticks, tree = 0.00 MB)
100 102 12932.3146 95 12926.1596 14249
200 202 12946.4404 37 12926.1596 15264
300 283 12926.3449 126 12926.1596 20245
400 383 12935.1371 62 12926.1596 25797
500 469 12954.8483 22 12926.1596 26632
600 556 12926.9074 151 12926.2037 34008
700 656 12928.9205 98 12926.2037 48189
800 754 12946.4989 47 12926.2037 49444
900 854 12926.6216 107 12926.2037 66528
1000 954 12936.0334 69 12926.2037 69176
.
.
.
Elapsed time = 3842.37 sec. (1645488.74 ticks, tree = 1755.55 MB)
1506100 1400957 12931.9093 57 12944.0000 12927.0000 9599439 0.13%
1506200 1401052 12925.9279 70 12944.0000 12927.0000 9600005 0.13%
1506300 1401150 12932.0634 52 12944.0000 12927.0000 9600540 0.13%
1506400 1401241 12932.5283 62 12944.0000 12927.0000 9601215 0.13%
1506500 1401338 12927.5025 72 12944.0000 12927.0000 9601856 0.13%
1506600 1401433 12925.3909 91 12944.0000 12927.0000 9602338 0.13%
1506700 1401533 12933.8317 52 12944.0000 12927.0000 9602811 0.13%
1506800 1401627 12927.8280 71 12944.0000 12927.0000 9603439 0.13%
1506900 1401724 12926.5709 68 12944.0000 12927.0000 9604222 0.13%
1507000 1401824 12933.6508 47 12944.0000 12927.0000 9604904 0.13%
Elapsed time = 3844.64 sec. (1646473.14 ticks, tree = 1756.74 MB)
1507100 1401918 12931.3799 73 12944.0000 12927.0000 9605519 0.13%
1507200 1402009 12926.9555 69 12944.0000 12927.0000 9605980 0.13%
1507300 1402103 12927.2025 90 12944.0000 12927.0000 9606381 0.13%
1507400 1402203 12934.2211 71 12944.0000 12927.0000 9606907 0.13%
1507500 1402297 12932.5559 71 12944.0000 12927.0000 9607657 0.13%
Root node processing (before b&c):
Real time = 194.64 sec. (1248.83 ticks)
Sequential b&c:
Real time = 3843.83 sec. (1647466.70 ticks)
------------
Total (root+branch&cut) = 4038.48 sec. (1648715.53 ticks)
=====================================================
Is there any way that I can obtain the optimal solution by further increasing the solving time or reducing the memory usage in AMPL?
Thank you so much in advance for your insights!
Amir
Robert Fourer
Nov 13, 2016, 10:21:32 AM11/13/16
to am...@googlegroups.com
Considering that after generating 1.5 million nodes, CPLEX still has 1.4 million to be examined, and the number of unexamined nodes is still increasing, a successful proof of optimality seems very unlikely. Check out the CPLEX options (http://ampl.com/products/solvers/solvers-we-sell/cplex/options/) for stopping after a given amount of time or number of nodes, which will cause CPLEX to return the solution with objective 12944 that it has found. This may be the optimal solution (just not proven optimal yet) and if there is a better solution (that CPLEX has not found despite all the searching so far) it will not have an objective better than 12927.
--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ampl+uns...@googlegroups.com.
To post to this group, send email to am...@googlegroups.com.
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ampl+uns...@googlegroups.com.
To post to this group, send email to am...@googlegroups.com.
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.
Reply all
Reply to author
Forward
0 new messages