OmOptim crashes after model selection

[filip.jezek@…]

OMOptim is still not usable.

1) Start OMOptim
2) Load Modelica
3) Load Model (.mo) - I load some of my models
4) Select problem - optimization

Crash!

Win 8.1, 64b

[Christoph Buchner <buchner@…>]

I'm seeing the same thing

[sjoelund.se]

Milestone moved to 1.12.0 due to 1.11.0 already being released.

[casella]

Milestone moved to 1.13.0 due to 1.12.0 already being released.

[Christoph Buchner <buchner@…>]

Is there an obvious/quick fix for this? Otherwise, how far back does one have to go to get a working OmOptim to play around with?

[sjoelund.se]

We didn't even ship OMOptim as part of the 1.9.4 OM release on Linux (separate package then, but I think it was broken already at that time)? I think you need to go back to OM 1.8.1 for a tested OMOptim. It might be possible to set OPENMODELICAHOME and use some old OMOptim with a newer OpenModelica, but I think some of the internal API functions called by OMOptim and some parts might not fully work if you do that...

[sjoelund.se]

And yes, I think there was basically no real maintenance of the code since 2011 except just making it compile (there are no tests for OMOptim in the testsuite).

[casella]

Replying to sjoelund.se:

And yes, I think there was basically no real maintenance of the code since 2011 except just making it compile (there are no tests for OMOptim in the testsuite).

This is the main problem with OMOptim, that the main developer moved on to other things and the tool has not really been maintained ever since.

Is anybody interested at actively contributing to that?

[Christoph Buchner <buchner@…>]

If I understand correctly, OMOptim "just" called the compiled OM code with varying parameters determined by some algorithm and settings?

If that's the case, and we don't find a maintainer for OMOptim, maybe it would make sense to transfer this task outside, e.g. to the Python ecosystem. We would have to make sure to offer the relevant APIs/functionality in an accessible manner via OMPython, then advise users to use whatever optimization methods are out there in the scientific python stack. Ideally there is already some optimization GUI out there.

Alternatively, what's the status of ​https://github.com/PySimulator/PySimulator? Maybe optimization could be added as a plugin there?

Just to clarify, having an optimization suite would be really usefule for OM in my opinion, for example to do determination of unknown parameters using measurement data, but a broken/unmaintained solution is worse than none.

[Christoph Buchner <buchner@…>]

Has there been an update on the OMOptim maintenance situation in the meantime? I just tried using it again but did not get far/was not successful in completing an optimisation.

[sjoelund.se]

Yes, there was a change to the OMOptim situation: we no longer even compile it for the testing and builds...

[Christoph Buchner <buchner@…>]

I see, thanks.

[adrpo]

It is still built on Windows and we will try to resurect it when we have more resources and time.