[Ipopt] PyOmo/Ipopt fails with “can't evaluate pow”

[Stefan Vigerske]

Hi,

that might be an issue with Pyomos interface to Ipopt.
If a point cannot be evaluated, the implementation of the eval callback 
should return false and not stop.

If Pyomo uses the AMPL interface of Ipopt, then check whether the 
halt_on_ampl_error option has been set.

Further, you can try whether you can work around this by setting option 
bound_relax_factor to 0 
(https://coin-or.github.io/Ipopt/OPTIONS.html#OPT_NLP).

Stefan

On 6/29/19 6:28 PM, Zhisheng Xu wrote:
> Hi,
> 
> I'm using pyomo/ipopt and got the following error message. Is there an
> option setting to avoid this and get pow() evaluated to be 0?
> 
> iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du
> alpha_pr  ls
>     0 -1.7091136e+003 5.24e+002 1.00e+000  -6.0 0.00e+000    -  0.00e+000
> 0.00e+000   0
> Error evaluating constraint 56: can't evaluate pow(-3.50288e-10,2.5)
> 
> 
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> https://list.coin-or.org/mailman/listinfo/ipopt
>