[BUG] standgub.json from netlib/lp fails if zero-only final column is left in the constraint matrix #13
Description
The standgub dataset from the netlib/lp suite has a final column in the constraint matrix which is all zeros. If that column is removed from the dataset, cuopt will find an optimal solution. If that column is left in the dataset, cuopt will return dual infeasible. Is this behavior correct?
The attached zip file contains a Python runner, a JSON file with the standgub dataset from netlib/lp( converted from MPS using the cuOpt MPS parser to generate a dictionary), and a log of outputs using the runner in both modes.
Run the problem with the zero column included
python cuopt_json.py standgub.jsonRun the problem with the zero column removed
python cuopt_json.py standgub.json -z