Re: [sage-support] Re: Binary variables in Integer Programs solved with CPLEX have wrong bounds
Sorry for the delay in responding, but I was traveling over the holidays. Happy New Year! On Sun, Dec 15, 2013 at 5:18 AM, Nathann Cohen wrote: > {0: 2.0, 1: 0.0} >> > Note that x[0] has the solution 2, which shouldn't happen for a binary > variable. >> >> When I ran the example of code you gave (removing the comment before > binary=True, I get the solution 0:1, 1:1, which is good. If you really get > a variable equal to 2 on this example, something is dead wrong indeed. > Something is indeed wrong, and I have experienced similar problems in the past. What version of CPLEX and which version of Sage are you using? I am using IBM ILOG CPLEX 12.5.1.0, and Sage 5.12 on Fedora Linux 19 and Sage 5.10 on Fedora 13. > Well, I would be glad to help but I would need to be able to reproduce one > of the bugs for a start :-P > What other information can I give to help track this down? Thanks, Stephen -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at http://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/groups/opt_out.
Re: [sage-support] Re: Binary variables in Integer Programs solved with CPLEX have wrong bounds
Hellooo !!! > Something is indeed wrong, and I have experienced similar problems in the past. What version of CPLEX and which version of Sage are you using? I am using IBM ILOG CPLEX 12.5.1.0, and Sage 5.12 on Fedora Linux 19 and Sage 5.10 on Fedora 13. > > What other information can I give to help track this down? HM... Well, to be honest I don't know. The truth is : I don't remember having met a nasty bug like that with Sage at any point, though it is possible that it may have been fixed since since you do not use the latest version of Sage.. Coud you try it with a more recent version of Sage, like the 6.0 ? Actually, there is a trac ticket waiting for review related to how LP solvers handle bounds, but I do not think it is related to this bug http://trac.sagemath.org/ticket/15622 (it does *not only* fix GLPK code) It would be cool to know if the bug still exists.. Which would be a good way out, for I do not really know how to debug this from a distance ^^; Nathann -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at http://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/groups/opt_out.
Re: [sage-support] Re: Binary variables in Integer Programs solved with CPLEX have wrong bounds
On Tue, Jan 7, 2014 at 1:58 AM, Nathann Cohen wrote: > > Something is indeed wrong, and I have experienced similar problems in > the past. What version of CPLEX and which version of Sage are you using? I > am using IBM ILOG CPLEX 12.5.1.0, and Sage 5.12 on Fedora Linux 19 and Sage > 5.10 on Fedora 13. > > HM... Well, to be honest I don't know. The truth is : I don't remember > having met a nasty bug like that with Sage at any point, though it is > possible that it may have been fixed since since you do not use the latest > version of Sage.. Coud you try it with a more recent version of Sage, like > the 6.0 ? > I upgraded to Sage 6.0, and the output is still wrong. I then checked IBM, and a new version of CPLEX was released in early December 2013. With CPLEX 12.6.0 and Sage 6.0, I now get the correct answers. So it seems that CPLEX 12.5.10 was the culprit. Though it is surprising to me that there would be such a basic bug. I am somewhat worried that there is something wrong with the CPLEX interface code in Sage, but it seems very straightforward. I recall having weird problems using CPLEX in Sage in the past; one specific instance from summer 2012 jumps to mind. I'll see if I can't dig up one of those instances and see if they have been fixed with new versions of Sage and CPLEX. > Actually, there is a trac ticket waiting for review related to how LP > solvers handle bounds, but I do not think it is related to this bug > http://trac.sagemath.org/ticket/15622 > (it does *not only* fix GLPK code) > Do you mean http://trac.sagemath.org/ticket/14581 ? Best wishes, Stephen -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at http://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/groups/opt_out.
Re: [sage-support] Re: Binary variables in Integer Programs solved with CPLEX have wrong bounds
Yo !!! > I upgraded to Sage 6.0, and the output is still wrong. I then checked IBM, > and a new version of CPLEX was released in early December 2013. With CPLEX > 12.6.0 and Sage 6.0, I now get the correct answers. HMmmm O_o > So it seems that CPLEX 12.5.10 was the culprit. Though it is surprising to > me that there would be such a basic bug. I am somewhat worried that there is > something wrong with the CPLEX interface code in Sage, but it seems very > straightforward. Well, at first we stored all the LP data, but now we really just forward every Sage command to the solver-specific backend. But if there was a bug like that in Cplex, it is highly unlikely that all tests passed in the graph/ folder. It is stuffed with binary problems. > I recall having weird problems using CPLEX in Sage in the past; one specific > instance from summer 2012 jumps to mind. I'll see if I can't dig up one of > those instances and see if they have been fixed with new versions of Sage and > CPLEX. Okay ! > Do you mean http://trac.sagemath.org/ticket/14581 ? Yep sorry >_< Nathann -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at http://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/groups/opt_out.