Unexpected changes in Gurobi settings when using amplpy

I have noticed that when I use amplpy with Gurobi, some of Gurobi’s default settings seem to change. For example, the default feasibility tolerance in Gurobi is 1e-6, but when using amplpy, this tolerance appears to be much lower. Is this expected behavior? Does amplpy modify solver settings by default?

Hi @salvapinedam,

Thank you for reaching out. We don’t change the feasibility tolerance which is controlled by the option shown below:

alg:feastol (feastol)
      Primal feasibility tolerance (default 1e-6).

What is giving you that impression? Note that any options being changed are usually shown in the Gurobi log when verbose output is enabled.