[Ipopt] Re： A small bug
Stefan Vigerske
stefan at math.hu-berlin.de
Sat Nov 16 07:19:54 EST 2013
Hi,
I had asserts and Ipopt's DBG_ASSERT enabled.
If you can produce a simple driver that reproduces the problem, that
would be good. Just open a ticket in the bugtracking system for it, so
we can discuss this there.
I'm still not sure which problem you are exactly passing to Ipopt.
You wrote:
y = x1 * x2 + 0 * x3
subject to:
x1 * x2 - x3 = 0
x3 = 0
x1, x2 = 0; x3: unbound
However, the statistics output in your last mail says:
Number of nonzeros in equality constraint Jacobian...: 1
Number of nonzeros in inequality constraint Jacobian.: 0
Number of nonzeros in Lagrangian Hessian.............: 0
Total number of variables............................: 1
variables with only lower bounds: 0
variables with lower and upper bounds: 0
variables with only upper bounds: 0
Total number of equality constraints.................: 1
Total number of inequality constraints...............: 0
inequality constraints with only lower bounds: 0
inequality constraints with lower and upper bounds: 0
inequality constraints with only upper bounds: 0
So Ipopt sees only 1 out of your 3 variables and only 1 out of your 2
constraints.
I can reproduce the same statistics by solving
min 0*x3
s.t. x3 = 0 [as equation]
but nothing suspicious is going on then.
Stefan
On 11/13/2013 04:17 AM, 八重雾中渡 wrote:
> Hi,
>
> it is all right if you don't open the assert in debug version.
> In release version, no one will initialize the memory of the
> member "homogeneous_" in DenseVector if the DenseVector
> is uninitialized, so the issue in my first mail will happen.
>
> I don't use any interface and I think the difference
> between 3.11.1 and 3.11.5 is not important.
>
>
> List of options:
>
> Name Value # times used
> accept_every_trial_step = no 1
> acceptable_compl_inf_tol = 1e-006 1
> acceptable_constr_viol_tol = 1e-006 1
> acceptable_obj_change_tol = 1e+020 1
> acceptable_tol = 0.0001 1
> bound_mult_init_method = mu-based 1
> bound_relax_factor = 0 2
> compl_inf_tol = 1e-008 2
> constr_viol_tol = 1e-008 2
> diverging_iterates_tol = 1e+020 1
> dual_inf_tol = 1e-008 1
> file_print_level = 6 1
> hessian_approximation = limited-memory 4
> max_cpu_time = 1.79769e+308 1
> max_iter = 2147483647 1
> max_soc = 4 1
> mu_max = 1e+020 1
> mu_max_fact = 1e+008 1
> mu_min = 1e-020 1
> mu_oracle = quality-function 2
> mu_strategy = adaptive 2
> nlp_lower_bound_inf = -2e+024 1
> nlp_scaling_method = gradient-based 1
> nlp_upper_bound_inf = 2e+024 1
> obj_scaling_factor = -1 1
> output_file = ipopt.out 1
> tol = 1e-006 2
> watchdog_shortened_iter_trigger = 10 1
> watchdog_trial_iter_max = 3 1
> This is Ipopt version 3.11.1, running with linear solver mumps.
>
> Number of nonzeros in equality constraint Jacobian...: 1
> Number of nonzeros in inequality constraint Jacobian.: 0
> Number of nonzeros in Lagrangian Hessian.............: 0
>
> Hessian approximation will be done in the space of all 1 x variables.
>
> Scaling parameter for objective function = 1.000000e+000
> objective scaling factor = -1
> No x scaling provided
> No c scaling provided
> No d scaling provided
> Initial values of x sufficiently inside the bounds.
> Initial values of s sufficiently inside the bounds.
> Total number of variables............................: 1
> variables with only lower bounds: 0
> variables with lower and upper bounds: 0
> variables with only upper bounds: 0
> Total number of equality constraints.................: 1
> Total number of inequality constraints...............: 0
> inequality constraints with only lower bounds: 0
> inequality constraints with lower and upper bounds: 0
> inequality constraints with only upper bounds: 0
>
>
>
> **************************************************
> *** Summary of Iteration: 0:
> **************************************************
>
> iter objective inf_pr inf_du lg(mu) ||d|| lg(rg) alpha_du alpha_pr ls
> 0 0.0000000e+000 0.00e+000 0.00e+000 0.0 0.00e+000 - 0.00e+000 0.00e+000 0 s
>
> **************************************************
> *** Beginning Iteration 0 from the following point:
> **************************************************
>
> Current barrier parameter mu = 1.0000000000000000e+000
> Current fraction-to-the-boundary parameter tau = 0.0000000000000000e+000
>
> ||curr_x||_inf = 0.0000000000000000e+000
> ||curr_s||_inf = 0.0000000000000000e+000
> ||curr_y_c||_inf = 0.0000000000000000e+000
> ||curr_y_d||_inf = 0.0000000000000000e+000
> ||curr_z_L||_inf = 0.0000000000000000e+000
> ||curr_z_U||_inf = 0.0000000000000000e+000
> ||curr_v_L||_inf = 0.0000000000000000e+000
> ||curr_v_U||_inf = 0.0000000000000000e+000
>
>
> ***Current NLP Values for Iteration 0:
>
> (scaled) (unscaled)
> Objective...............: -0.0000000000000000e+000 0.0000000000000000e+000
> Dual infeasibility......: 0.0000000000000000e+000 0.0000000000000000e+000
> Constraint violation....: 0.0000000000000000e+000 0.0000000000000000e+000
> Complementarity.........: 0.0000000000000000e+000 -0.0000000000000000e+000
> Overall NLP error.......: 0.0000000000000000e+000 -0.0000000000000000e+000
>
>
>
> MUMPS used permuting_scaling 5 and pivot_order 2.
> scaling will be 77.
> Number of doubles for MUMPS to hold factorization (INFO(9)) = 4
> Number of integers for MUMPS to hold factorization (INFO(10)) = 16
> Factorization successful.
>
> Number of Iterations....: 0
>
> (scaled) (unscaled)
> Objective...............: -0.0000000000000000e+000 0.0000000000000000e+000
> Dual infeasibility......: 0.0000000000000000e+000 0.0000000000000000e+000
> Constraint violation....: 0.0000000000000000e+000 0.0000000000000000e+000
> Complementarity.........: 0.0000000000000000e+000 -0.0000000000000000e+000
> Overall NLP error.......: 0.0000000000000000e+000 -0.0000000000000000e+000
>
>
> Number of objective function evaluations = 1
> Number of objective gradient evaluations = 1
> Number of equality constraint evaluations = 1
> Number of inequality constraint evaluations = 0
> Number of equality constraint Jacobian evaluations = 1
> Number of inequality constraint Jacobian evaluations = 0
> Number of Lagrangian Hessian evaluations = 0
> Total CPU secs in IPOPT (w/o function evaluations) = 0.026
> Total CPU secs in NLP function evaluations = 0.007
>
> EXIT: Optimal Solution Found.
>
>> ------------------ 原始邮件 ------------------
>> 发件人: "Stefan Vigerske";<stefan at math.hu-berlin.de>;
>> 发送时间: 2013年11月12日(星期二) 晚上6:35
>> 收件人: "八重雾中渡"<darknesswind at qq.com>;
>> 抄送: "ipopt"<ipopt at list.coin-or.org>;
>> 主题: Re: [Ipopt] A small bug
>>
>> Hi,
>>
>> sorry, I can't reproduce this.
>> If I run Ipopt 3.11 (current stable branch, Linux 64bit) through the
>> GAMS interface on your instance,
>> Variables x1, x2, x3, y;
>> Equations obj, e1, e2;
>> obj.. y =e= x1 * x2 + eps*x3;
>> e1.. x1 * x2 =e= x3;
>> e2.. x3 =e= 0;
>> x1.fx = 0; x2.fx = 0;
>> model m / all /;
>> solve m min y using nlp;
>>
>> then I get
>>
>>
>> List of user-set options:
>>
>> Name Value used
>> bound_mult_init_method = mu-based yes
>>
>> ******************************************************************************
>> This program contains Ipopt, a library for large-scale nonlinear
>> optimization.
>> Ipopt is released as open source code under the Eclipse Public License
>> (EPL).
>> For more information visit http://projects.coin-or.org/Ipopt
>> ******************************************************************************
>>
>> This is Ipopt version 3.11, running with linear solver ma27.
>>
>> Too few degrees of freedom (n_x = 1, n_c = 2).
>> Trying fixed_variable_treatment = RELAX_BOUNDS
>>
>> Number of nonzeros in equality constraint Jacobian...: 4
>> Number of nonzeros in inequality constraint Jacobian.: 0
>> Number of nonzeros in Lagrangian Hessian.............: 1
>>
>> Total number of variables............................: 3
>> variables with only lower bounds: 0
>> variables with lower and upper bounds: 2
>> variables with only upper bounds: 0
>> Total number of equality constraints.................: 2
>> Total number of inequality constraints...............: 0
>> inequality constraints with only lower bounds: 0
>> inequality constraints with lower and upper bounds: 0
>> inequality constraints with only upper bounds: 0
>>
>> iter objective inf_pr inf_du lg(mu) ||d|| lg(rg) alpha_du
>> alpha_pr ls
>> 0 0.0000000e+00 0.00e+00 0.00e+00 0.0 0.00e+00 - 0.00e+00
>> 0.00e+00 0
>> 1 0.0000000e+00 0.00e+00 0.00e+00 -7.0 0.00e+00 - 1.00e+00
>> 1.00e+00 0
>>
>> Number of Iterations....: 1
>>
>> (scaled) (unscaled)
>> Objective...............: 0.0000000000000000e+00 0.0000000000000000e+00
>> Dual infeasibility......: 0.0000000000000000e+00 0.0000000000000000e+00
>> Constraint violation....: 0.0000000000000000e+00 0.0000000000000000e+00
>> Complementarity.........: 1.0000000000000001e-07 1.0000000000000001e-07
>> Overall NLP error.......: 1.0000000000000000e-08 1.0000000000000001e-07
>>
>>
>> Number of objective function evaluations = 2
>> Number of objective gradient evaluations = 2
>> Number of equality constraint evaluations = 2
>> Number of inequality constraint evaluations = 0
>> Number of equality constraint Jacobian evaluations = 2
>> Number of inequality constraint Jacobian evaluations = 0
>> Number of Lagrangian Hessian evaluations = 1
>> Total CPU secs in IPOPT (w/o function evaluations) = 2.637
>> Total CPU secs in NLP function evaluations = 0.052
>>
>> EXIT: Optimal Solution Found.
>>
>>
>>
>> And this is under valgrind.
>>
>> Stefan
>>
>> On 11/11/2013 04:16 AM, 八重雾中渡 wrote:
>>> I am very sorry, I had made a mistake on describing the problem.
>>> The problem we actually used is difference form the user input.
>>>
>>> Here is the problem to run the solver:
>>> y = x1 * x2 + 0 * x3
>>> subject to:
>>> x1 * x2 - x3 = 0
>>> x3 = 0
>>>
>>> x1, x2 = 0; x3: unbound
>>>
>>> the Ipopt version is 3.11.1
>>> bound_mult_init_method = mu-based
>>>
>>> ------------------ Original ------------------
>>> From: "Stefan Vigerske";<stefan at math.hu-berlin.de>;
>>> Date: Sun, Nov 10, 2013 00:48 AM
>>> To: "八重雾中渡"<darknesswind at qq.com>;
>>> Cc: "ipopt"<ipopt at list.coin-or.org>;
>>> Subject: Re: [Ipopt] A small bug
>>>
>>> Hi,
>>>
>>> I cannot reproduce this issue.
>>> If all variables are fixed, the code should not even get to the place
>>> you show in your stacktrace.
>>> In DefaultIterateInitializer::SetInitialIterates(), the call to
>>> IpData().InitializeDataStructures() just at the beginning (line 194)
>>> should throw an exception, so that it doesn't get to line 278
>>> (iterates->z_L_NonConst()->ElementWiseDivide(*IpCq().trial_slack_x_L())).
>>>
>>> The exception is thrown by IpTNLPAdapter.cpp:684,
>>>
>>> if (status == SUCCESS) {
>>> jnlst_->Printf(J_WARNING, J_INITIALIZATION,
>>> "All variables are fixed and constraint
>>> violation %e\n is below tolerance %e. Declaring success.\n", max_viol,
>>> tol_);
>>>
>>> and catched again in IpIpoptApplication.cpp:1048 (end of
>>> IpoptApplication::call_optimize()).
>>>
>>> Can you specify more which Ipopt version you are using and which interface?
>>>
>>> Stefan
>>>
>>> On 11/08/2013 08:48 AM, darknesswind wrote:
>>>> When I run the solver on a problem with all fixed variables, I found the DenseVector "ip_data_->curr()->z_L()" is uninitialized in function IpoptCalculatedQuantities::trial_slack_x_L.It may cause a crash in release version because of access violation.
>>>>
>>>> The problem is:
>>>> y = x1 * x2
>>>> subject to:
>>>> x1 * x2 = 0
>>>> x1 = 0
>>>> x2 = 0
>>>>
>>>> Here is the call stack: The dense_x->homogeneous_ equal to false but the dense_x->values_ is null.
>>>>
>>>>> ipopt.dll!Ipopt::DenseVector::ElementWiseDivideImpl(const Ipopt::Vector & x) Line 240 C++
>>>> ipopt.dll!Ipopt::Vector::ElementWiseDivide(const Ipopt::Vector & x) Line 602 + 0x13 bytes C++
>>>> ipopt.dll!Ipopt::IpoptCalculatedQuantities::CalculateSafeSlack(Ipopt::SmartPtr<Ipopt::Vector> & slack, const Ipopt::SmartPtr<Ipopt::Vector const > & bound, const Ipopt::SmartPtr<Ipopt::Vector const > & curr_point, const Ipopt::SmartPtr<Ipopt::Vector const > & multiplier) Line 481 C++
>>>> ipopt.dll!Ipopt::IpoptCalculatedQuantities::trial_slack_x_L() Line 365 + 0x5b bytes C++
>>>> ipopt.dll!Ipopt::DefaultIterateInitializer::SetInitialIterates() Line 278 + 0x16 bytes C++
>>>> ipopt.dll!Ipopt::IpoptAlgorithm::InitializeIterates() Line 545 + 0x26 bytes C++
>>>> ipopt.dll!Ipopt::IpoptAlgorithm::Optimize(bool isResto) Line 274 C++
>>>> ipopt.dll!Ipopt::IpoptApplication::call_optimize() Line 860 + 0xd bytes C++
>>>> ipopt.dll!Ipopt::IpoptApplication::OptimizeNLP(const Ipopt::SmartPtr<Ipopt::NLP> & nlp, Ipopt::SmartPtr<Ipopt::AlgorithmBuilder> & alg_builder) Line 754 + 0x8 bytes C++
>>>> ipopt.dll!Ipopt::IpoptApplication::OptimizeNLP(const Ipopt::SmartPtr<Ipopt::NLP> & nlp) Line 717 + 0x17 bytes C++
>>>> ipopt.dll!Ipopt::IpoptApplication::OptimizeTNLP(const Ipopt::SmartPtr<Ipopt::TNLP> & tnlp) Line 696 + 0x16 bytes C++
>>>>
>>>> _______________________________________________
>>>> Ipopt mailing list
>>>> Ipopt at list.coin-or.org
>>>> http://list.coin-or.org/mailman/listinfo/ipopt
>>>>
>>>
>>> .
>>>
>>
>> .
More information about the Ipopt
mailing list