Parameter identification wang

[Peter230655]

As discussed yesterday, I included additional eoms with added noise for every measurement.
number of measurements and number of 'noisy' eoms for every measurement may be set freely.

[Peter230655]

Whatever I did apparently worked.
Actually I could not find much difference in the performance - however it is interesting how few iterations are needed to find a solution.

[moorepants]

Is this any different than what I have in #250?

[Peter230655]

Is this any different than what I have in #250?

I think, same idea.
All I did was to allow the number of measurements and the number of 'repeats per measurements' to be freely settable.
No new idea.
What amazed me how few iteration it needs to find a solution.

[Peter230655]

Cosmetic changes. By avoiding some unnecessary integration reduced running time by around 5 sec on my PC

[Peter230655]

No idea why it failed all these tests. I only made cosmetic changes compared to the version which passed all 17 tests.
Also: test_direct_collocation.py for examle has many test functions. I would be helpful to get information which one(s) it failed.
I'd be happy to do this, if you could tell me how to do it.

[moorepants]

No idea why it failed all these tests.

You have to click the failures and inspect the error messages.

[Peter230655]

No idea why it failed all these tests.

You have to click the failures and inspect the error messages.

I did this, and got this:

This looks like it failed just about everything, and I only made small changes to the version which passed easily.
(This is why I made the suggestion, that maybe it could tell which test(s) it failed in, say, test_direct_collocation.py)

[Peter230655]

No idea why it failed all these tests.

You have to click the failures and inspect the error messages.

I did this, and got this:

[moorepants]

The failures are coming from a new version of scotch that was released on conda forge yesterday:

libscotch                 7.0.5                hab7c3ca_0    conda-forge

[moorepants]

If you scroll up you see the relevant error:

E   ImportError: dlopen(/Users/runner/miniconda3/envs/test/lib/python3.9/site-packages/ipopt_wrapper.cpython-39-darwin.so, 0x0002): symbol not found in flat namespace '_SCOTCH_errorPrint'

[Peter230655]

If you scroll up you see the relevant error:

E   ImportError: dlopen(/Users/runner/miniconda3/envs/test/lib/python3.9/site-packages/ipopt_wrapper.cpython-39-darwin.so, 0x0002): symbol not found in flat namespace '_SCOTCH_errorPrint'

I saw this, but of course did not understand it.
What can I do about it? My previous version would have failed, too, I suppose?
Thanks!

NB: I continue to be amazed how fast it converges!

[moorepants]

What can I do about it? My previous version would have failed, too, I suppose?

We can report the issue upstream on the scotch feedstock and we can add a dependency pin for scotch <7.0.5 to the CI configuration.

[Peter230655]

What can I do about it? My previous version would have failed, too, I suppose?

We can report the issue upstream on the scotch feedstock and we can add a dependency pin for scotch <7.0.5 to the CI configuration.

A bit beyond my capabilities :-(
At least nothing I can do on my simulation?

[moorepants]

Closed/reopened to trigger the CI again. The upstream scotch issue may be fixed.