[CI] Julia tests for the filtersqp preset #224
Conversation
…es + included one instance
|
Hi @odow @amontoison, Uno 2.0.0: Infeasible stationary point
Options
3
0
1
0
2
2
2
2
0
0
1.526378828812082e-09
0.5
objno 0 200
suffix 4 0 18 0 0
upper_bound_duals
suffix 4 0 18 0 0
lower_bound_duals200 is the error code for infeasible models, see top of p283 in the AMPL manual. solution_summary(; result = 1, verbose = false)
├ solver_name : AmplNLWriter
├ Termination
│ ├ termination_status : LOCALLY_INFEASIBLE
│ ├ result_count : 1
│ └ raw_status : Uno 2.0.0: Infeasible stationary point
├ Solution (result = 1)
│ ├ primal_status : UNKNOWN_RESULT_STATUS
│ ├ dual_status : NO_SOLUTION
│ ├ objective_value : 0.00000e+00
│ └ dual_objective_value : 0.00000e+00
└ Work counters
└ solve_time (sec) : 2.42654e-01so JuMP has no information related to the primal solution (the _check_status_subset(JuMP.primal_status(model), primal_target[problem_type])How is that possible, given that the AMPL solution contains all the information? Is it the wrong kind of test here? |
| @testset "MINLPTests" begin | ||
| primal_target = Dict( | ||
| MINLPTests.FEASIBLE_PROBLEM => MOI.FEASIBLE_POINT, | ||
| MINLPTests.INFEASIBLE_PROBLEM => MOI.NO_SOLUTION, |
There was a problem hiding this comment.
@cvanaret the issue is here. JuMP expects that if Uno has found an infeasible problem, it will not return a primal solution.
(See the comment in your filterslp above)
There was a problem hiding this comment.
Got it. I saw the comment and tried both MOI.NO_SOLUTION and MOI.INFEASIBLE_POINT but it made no difference. I'll just skip the test. Thanks!
There was a problem hiding this comment.
It didn't make a difference because the value was UNKNOWN_RESULT_STATUS
| "005_010", | ||
| # Okay to exclude forever: AmplNLWriter does not support | ||
| # user-defined functions. | ||
| "006_010", |
There was a problem hiding this comment.
| "006_010", | |
| "006_010", | |
| "007_010", |
| r"^test_quadratic_SecondOrderCone_basic$", | ||
| r"^test_quadratic_nonconvex_constraint_basic$", | ||
| r"^test_solve_DualStatus_INFEASIBILITY_CERTIFICATE_", | ||
| r"^test_solve_VariableIndex_ConstraintDual_MAX_SENSE$", |
There was a problem hiding this comment.
Why was this one skipped? It is a very suspicious failure
There was a problem hiding this comment.
test_quadratic_SecondOrderCone_basic and test_quadratic_nonconvex_constraint_basic are reported as locally infeasible. The latter is nonconvex so it makes sense. For the former, starting at
If we started from a feasible point, I think that by convexity all the QPs would be feasible.
There was a problem hiding this comment.
No I meant test_solve_VariableIndex_ConstraintDual_MAX_SENSE. It is not complicated:
There was a problem hiding this comment.
Let me try again. Perhaps I'm writing the AMPL solution with incorrect signs.
#228
ipoptandfilterslppresetsfiltersqppreset