Originally posted by: josh7

Hello,
I'm currently trying to solve a second-order cone problem of around 500 variables using CPLEX and AMPL. Smaller instances are fine, but larger instances are reported as infeasible (they should not be). Eventually some variables will be discrete and I will use some sort of relaxed optimality or integrality condition, but there are feasibility problems even in the continuous case. I can provide code .mod and .dat files if helpful. Thanks in advance.

Josh
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Sorry for the late reply.

For us it would be the most convenient if you could send a .sav file of your model. I think that AMPL should be able to produce this.
My guess is that your issue arises from numerical problems, but of course I can only tell if I had the model.
Tobias
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

> Tobias Achterberg wrote:
> Sorry for the late reply.
>
> For us it would be the most convenient if you could send a .sav file of your model. I think that AMPL should be able to produce this.
> My guess is that your issue arises from numerical problems, but of course I can only tell if I had the model.
>
>
> Tobias

Yes, AMPL can export SAV files using the 'file' or 'writeprob' directives among your cplex_options. For example

ampl: model myfile.mod;
ampl: data myfile.dat;
ampl: option cplex_options 'file myfile.sav';
ampl: solve;
#CPLEXOptimizers
#DecisionOptimization