Julian Hall
> My test instance is the LP-relaxation of a 0-1 ILP and the test is performed on a machine without dedicated GPU. Could you share the instance with me?
Thanks: the better way is to write the instance as an MPS file by calling writeModel("foo. mps")
Thanks. I've reproduced your observations.
@galabovaa These memory leaks are fixed now, aren't they?
The Gurobi observation reiterates what I said but, as we discussed, after solving the LP corresponding to fixing the integer variables at their optimal values, constraints that are active with...
Hmm... that does look a bit alarming. Are you able to reproduce the behaviour from a .lp or .mps file?
No problem, I can work with.mps
The two file formats will induce a reordering of the variables and constraints in the problem, and that will lead to the MIP solver behaving differently.
Having looked in more detail, it's the nature of MIP that errors like this can happen, without preventing the correct solution from being found. I'll still try to identify what's...
Closed by #2462?