blnicho
commented
1 month ago Looks like the Jenkins test failure is real. Here is the stack-trace:
15:25:17 =================================== FAILURES ===================================
15:25:17 _________________ TestCyIpoptSolver.test_solve_13arg_callback __________________
15:25:17
15:25:17 self = <pyomo.contrib.pynumero.algorithms.solvers.tests.test_cyipopt_solver.TestCyIpoptSolver testMethod=test_solve_13arg_callback>
15:25:17
15:25:17 @unittest.skipIf(
15:25:17 not cyipopt_available or not cyipopt_ge_1_3, "cyipopt version < 1.3.0"
15:25:17 )
15:25:17 def test_solve_13arg_callback(self):
15:25:17 m = create_model1()
15:25:17
15:25:17 iterate_data = []
15:25:17
15:25:17 def intermediate(
15:25:17 nlp,
15:25:17 problem,
15:25:17 alg_mod,
15:25:17 iter_count,
15:25:17 obj_value,
15:25:17 inf_pr,
15:25:17 inf_du,
15:25:17 mu,
15:25:17 d_norm,
15:25:17 regularization_size,
15:25:17 alpha_du,
15:25:17 alpha_pr,
15:25:17 ls_trials,
15:25:17 ):
15:25:17 iterate = problem.get_current_iterate()
15:25:17 x = iterate["x"]
15:25:17 y = iterate["mult_g"]
15:25:17 iterate_data.append((x, y))
15:25:17
15:25:17 x_sol = np.array([3.85958688, 4.67936007, 3.10358931])
15:25:17 y_sol = np.array([-1.0, 53.90357665])
15:25:17
15:25:17 solver = pyo.SolverFactory("cyipopt", intermediate_callback=intermediate)
15:25:17 res = solver.solve(m, tee=True)
15:25:17 > pyo.assert_optimal_termination(res)
15:25:17
15:25:17 pyomo/pyomo/contrib/pynumero/algorithms/solvers/tests/test_cyipopt_solver.py:406:
15:25:17 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
15:25:17
15:25:17 results = {'Problem': [{'Name': 'unknown', 'Lower bound': -inf, 'Upper bound': -504.0, 'Number of objectives': 1, 'Number of con... termination of the optimization.', 'Wallclock time': 0.003202378051355481, 'Termination condition': 'userInterrupt'}]}
15:25:17
15:25:17 def assert_optimal_termination(results):
15:25:17 """
15:25:17 This function checks if the termination condition for the solver
15:25:17 is 'optimal', 'locallyOptimal', or 'globallyOptimal', and the status is 'ok'
15:25:17 and it raises a RuntimeError exception if this is not true.
15:25:17
15:25:17 Parameters
15:25:17 ----------
15:25:17 results : Pyomo results object returned from solver.solve
15:25:17 """
15:25:17 if not check_optimal_termination(results):
15:25:17 msg = (
15:25:17 'Solver failed to return an optimal solution. '
15:25:17 'Solver status: {}, Termination condition: {}'.format(
15:25:17 results.solver.status, results.solver.termination_condition
15:25:17 )
15:25:17 )
15:25:17 > raise RuntimeError(msg)
15:25:17 E RuntimeError: Solver failed to return an optimal solution. Solver status: aborted, Termination condition: userInterrupt
15:25:17
15:25:17 pyomo/pyomo/opt/results/solver.py:178: RuntimeError
15:25:17 ----------------------------- Captured stdout call -----------------------------
15:25:17 This is Ipopt version 3.13.2, running with linear solver ma27.
15:25:17
15:25:17 Number of nonzeros in equality constraint Jacobian...: 2
15:25:17 Number of nonzeros in inequality constraint Jacobian.: 2
15:25:17 Number of nonzeros in Lagrangian Hessian.............: 4
15:25:17
15:25:17 Total number of variables............................: 3
15:25:17 variables with only lower bounds: 2
15:25:17 variables with lower and upper bounds: 0
15:25:17 variables with only upper bounds: 0
15:25:17 Total number of equality constraints.................: 1
15:25:17 Total number of inequality constraints...............: 1
15:25:17 inequality constraints with only lower bounds: 0
15:25:17 inequality constraints with lower and upper bounds: 0
15:25:17 inequality constraints with only upper bounds: 1
15:25:17
15:25:17 iter objective inf_pr inf_du lg(mu) ||d|| lg(rg) alpha_du alpha_pr ls
15:25:17 0 -5.0400000e+02 5.00e+00 2.17e+01 -1.0 0.00e+00 - 0.00e+00 0.00e+00 0
15:25:17
15:25:17 Number of Iterations....: 0
15:25:17
15:25:17 (scaled) (unscaled)
15:25:17 Objective...............: -8.7500000000000000e+01 -5.0400000000000000e+02
15:25:17 Dual infeasibility......: 2.1710422009792488e+01 1.2505203077640472e+02
15:25:17 Constraint violation....: 5.0000000000000000e+00 5.0000000000000000e+00
15:25:17 Complementarity.........: 4.0000000099999999e+00 2.3040000057600000e+01
15:25:17 Overall NLP error.......: 2.1710422009792488e+01 1.2505203077640472e+02
15:25:17
15:25:17
15:25:17 Number of objective function evaluations = 1
15:25:17 Number of objective gradient evaluations = 1
15:25:17 Number of equality constraint evaluations = 1
15:25:17 Number of inequality constraint evaluations = 1
15:25:17 Number of equality constraint Jacobian evaluations = 1
15:25:17 Number of inequality constraint Jacobian evaluations = 1
15:25:17 Number of Lagrangian Hessian evaluations = 0
15:25:17 Total CPU secs in IPOPT (w/o function evaluations) = 0.002
15:25:17 Total CPU secs in NLP function evaluations = 0.000
15:25:17
15:25:17 EXIT: Stopping optimization at current point as requested by user.
15:25:17 ----------------------------- Captured stderr call -----------------------------
15:25:17 Exceptions caught in Qt event loop:
15:25:17 ________________________________________________________________________________
15:25:17 Traceback (most recent call last):
15:25:17 File ".../python310/pyomo/pyomo/contrib/pynumero/interfaces/cyipopt_interface.py", line 500, in intermediate
15:25:17 return self._intermediate_callback(
15:25:17 File ".../python310/pyomo/pyomo/contrib/pynumero/algorithms/solvers/tests/test_cyipopt_solver.py", line 396, in intermediate
15:25:17 iterate = problem.get_current_iterate()
15:25:17 File "cyipopt/cython/ipopt_wrapper.pyx", line 701, in ipopt_wrapper.Problem.get_current_iterate
15:25:17 RuntimeError: get_current_iterate only supports Ipopt version >=3.14.0 CyIpopt is compiled with version 3.13.2
15:25:17 =========================== short test summary info ============================
15:25:17 FAILED pyomo/pyomo/contrib/pynumero/algorithms/solvers/tests/test_cyipopt_solver.py::TestCyIpoptSolver::test_solve_13arg_callback - RuntimeError: Solver failed to return an optimal solution. Solver status: aborted, Termination condition: userInterrupt
codecov[bot]
Summary/Motivation:
Currently, users are unable to access the
get_current_iterateandget_current_violationsmethods during an intermediate callback as they don't have access to thecyipopt.Problemobject. We currently expect a user-provided callback to be a function with 12 arguments: the 11 arguments that Ipopt gives us, plus the NLP object that cyipopt is using. This PR adds support for a new callback API, where the user's callback accepts 13 arguments. The new argument is thecyipopt.Probleminstance (ourCyIpoptNLP), from which the user can access theget_current*methods.This is the first step towards implementing something like https://github.com/Pyomo/pyomo/pull/2860. With this PR, users can implement their own callbacks to track primal-dual iterates and infeasibilities. Eventually, I'd like to commit a basic debugging callback that automatically tracks useful information.
Changes proposed in this PR:
Legal Acknowledgement
By contributing to this software project, I have read the contribution guide and agree to the following terms and conditions for my contribution: