pyomo using Gurobi with parameters

[xiwen]

Hi All,

I am using gurobi as a solver: 

self._opt_solver = coopr.opt.base.solvers.SolverFactory('gurobi')

I want to set the mipgap, and also the threads to use, can any one tell me how to do it?

I have tried:

self._opt_solver.options["mipgap"] = 0.05

self._opt_solver.options["Threads"]=12

But it looks not working.

thanks

xiwen

[Gabriel Hackebeil]

The following works for me.

options[“threads”]=4

If you add tee=True to the solve you should be able to verify this from Gurobi’s output. If that does not, work you might consider updating to the latest version of Pyomo.

Gabe

--
You received this message because you are subscribed to the Google Groups "Pyomo Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pyomo-forum...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[xiwen]

Hi Gabriel,

thanks

Below is my gurobi.log, looks like gurobi changes MIPGap = 0.05 and threads = 4.

But from gurobi.txt, i found

       message: Model was solved to optimality (subject to tolerances), and an optimal solution is available.

       objective: 33203254.0977

       gap: 0.0

To me, it looks gurobi stop when the gap = 0?

thanks

xiwen

~~~~~~~~~~~~~~~~~~~~log~~~~~~~~~~~~~~~~~~~~~~~~~

Solver command line: ['C:\\gurobi550\\win64\\bin\\gurobi.bat']

Read LP format model from file c:\\users\\richar~1\\appdata\\local\\temp\\2\\tmpivgz5z.pyomo.lp

Reading time = 0.51 seconds

(null): 135531 rows, 65811 columns, 271107 nonzeros

Optimize a model with 135531 rows, 65811 columns and 271107 nonzeros

Concurrent LP optimizer: dual simplex and barrier

Showing barrier log only...

Presolve removed 114111 rows and 57955 columns

Presolve time: 0.10s

Presolved: 21420 rows, 7856 columns, 46959 nonzeros

Ordering time: 0.00s

Barrier statistics:

 AA' NZ     : 8.333e+04

 Factor NZ  : 1.088e+05 (roughly 13 MBytes of memory)

 Factor Ops : 7.040e+05 (less than 1 second per iteration)

 Threads    : 1

                  Objective                Residual

Iter       Primal          Dual         Primal    Dual     Compl     Time

   0   1.14552733e+08 -6.74649395e+08  1.31e+05 5.56e+00  6.58e+05     0s

   1   9.58055676e+07 -3.54152176e+08  3.78e+04 2.45e+02  1.90e+05     0s

   2   9.58081766e+07 -1.91137390e+08  5.66e+03 3.94e+01  3.63e+04     0s

   3   8.73244174e+07 -6.96007313e+07  5.68e+02 1.28e+01  1.11e+04     0s

   4   6.63838418e+07  2.53867715e+06  9.89e+00 3.20e+00  3.03e+03     0s

   5   5.03466569e+07  1.50462602e+07  2.26e+00 1.80e+00  1.46e+03     0s

   6   4.49162210e+07  2.35089320e+07  1.22e+00 9.68e-01  8.00e+02     0s

   7   3.96912998e+07  2.79845813e+07  5.84e-01 4.93e-01  4.09e+02     0s

   8   3.61693123e+07  2.96189725e+07  2.34e-01 3.26e-01  2.24e+02     0s

   9   3.55666060e+07  3.01633116e+07  1.83e-01 2.73e-01  1.84e+02     0s

  10   3.45721150e+07  3.15842828e+07  1.01e-01 1.39e-01  9.96e+01     0s

  11   3.39459651e+07  3.21799681e+07  5.47e-02 8.35e-02  5.87e+01     0s

  12   3.35827505e+07  3.25999900e+07  2.75e-02 4.66e-02  3.26e+01     0s

  13   3.33817056e+07  3.28823269e+07  1.34e-02 2.29e-02  1.65e+01     0s

  14   3.33025291e+07  3.30333296e+07  7.66e-03 1.13e-02  8.82e+00     0s

  15   3.32565535e+07  3.31238257e+07  4.08e-03 5.15e-03  4.30e+00     0s

  16   3.32225926e+07  3.31693149e+07  1.52e-03 1.91e-03  1.72e+00     0s

  17   3.32103679e+07  3.31932602e+07  6.00e-04 4.60e-04  5.42e-01     0s

  18   3.32054229e+07  3.32008877e+07  1.77e-04 8.03e-05  1.41e-01     0s

  19   3.32035508e+07  3.32029814e+07  1.87e-05 1.67e-06  1.72e-02     0s

  20   3.32032975e+07  3.32031972e+07  2.46e-06 2.26e-07  3.04e-03     0s

  21   3.32032656e+07  3.32032412e+07  6.19e-07 3.94e-08  7.38e-04     0s

  22   3.32032544e+07  3.32032533e+07  1.01e-08 3.64e-10  3.34e-05     0s

  23   3.32032541e+07  3.32032541e+07  9.79e-11 1.72e-13  2.81e-09     0s

Barrier solved model in 23 iterations and 0.39 seconds

Optimal objective 3.32032541e+07

Crossover log...

       0 DPushes remaining with DInf 1.5813809e-12                 0s

      24 PPushes remaining with PInf 0.0000000e+00                 0s

       0 PPushes remaining with PInf 0.0000000e+00                 0s

  Push phase complete: Pinf 0.0000000e+00, Dinf 1.5813809e-12      0s

Iteration    Objective       Primal Inf.    Dual Inf.      Time

    1070    3.3203254e+07   0.000000e+00   0.000000e+00      0s

    1070    3.3203254e+07   0.000000e+00   0.000000e+00      0s

Solved with dual simplex

Solved in 13404 iterations and 0.57 seconds

Optimal objective  3.320325410e+07

Changed value of parameter QCPDual to 1

   Prev: 0   Min: 0   Max: 1   Default: 0

Changed value of parameter MIPGap to 0.05

   Prev: 0.0001   Min: 0.0   Max: 1e+100   Default: 0.0001

Parameter MIPGap unchanged

   Value: 0.05   Min: 0.0   Max: 1e+100   Default: 0.0001

Parameter MIPGap unchanged

   Value: 0.05   Min: 0.0   Max: 1e+100   Default: 0.0001

Changed value of parameter Threads to 4

   Prev: 0   Min: 0   Max: 16   Default: 0

[Watson, Jean-Paul]

At the bottom of the trace, you’l see that the options are actually being processed as you indicate. However, this is a linear (as opposed to mixed-integer) model – it was solved, apparently, by both barrier and dual simplex. Thus, there isn’t a gap; the mipgap parameter is not applicable.

jpw