Hierarchical optimization: fix unnecessary simulation

[dweindl]

Currently model simulation is run twice, even when no sensitivities are requested and the optimal inner parameters could be applied directly.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.46%. Comparing base (2558763) to head (e2501a9).

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1327      +/-   ##
===========================================
+ Coverage    84.42%   84.46%   +0.03%     
===========================================
  Files          153      153              
  Lines        12529    12529              
===========================================
+ Hits         10577    10582       +5     
+ Misses        1952     1947       -5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Doresic]

Yes, nice change. Didn't think of this case in the implementation.
To confirm we're thinking the same: the only thing this changes is that if the sensitivity method is adjoint, then the sensi_orders = (0,) calls of the objective will be called through self.calculate_directly(...) instead of self.call_amici_twice(...), right?

[dweindl]

To confirm we're thinking the same: the only thing this changes is that if the sensitivity method is adjoint, then the sensi_orders = (0,) calls of the objective will be called through self.calculate_directly(...) instead of self.call_amici_twice(...), right?

Correct.

It was necessary because checking for sensitivity method alone is insufficient. sensitivity method and sensitivity order jointly decide whether sensitivities are to be computed.