add nan safe log&divide

[FFroehlich]

Add guards to log & divide functions to avoid nans during gradient computation

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 77.80%. Comparing base (851d389) to head (0c226e0).
Report is 1 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2611      +/-   ##
===========================================
+ Coverage    77.78%   77.80%   +0.01%     
===========================================
  Files          329      329              
  Lines        21724    21737      +13     
  Branches      1477     1477              
===========================================
+ Hits         16899    16912      +13     
  Misses        4814     4814              
  Partials        11       11              

Flag	Coverage Δ
cpp	74.14% <100.00%> (+0.01%)	⬆️
cpp_python	34.15% <69.23%> (+0.02%)	⬆️
petab	36.99% <69.23%> (+0.02%)	⬆️
python	72.48% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
python/sdist/amici/jax/model.py	76.34% <100.00%> (+1.63%)	⬆️
python/sdist/amici/jaxcodeprinter.py	95.00% <100.00%> (+2.69%)	⬆️

[dweindl]

Makes sense to have that, but I am slightly worried that this may cover up potential issues in the model.

Depending on how common these issues are, make it optional? Also fine as is for now.

[FFroehlich]

Makes sense to have that, but I am slightly worried that this may cover up potential issues in the model.

Depending on how common these issues are, make it optional? Also fine as is for now.

Rather common, happens with Bachmann, Laske, Lucarelli and Weber.

For jax we (currently) have to evaluate all possible observables for all timepoints and all loss functions as well as avoid nan values from ever occurring. There might be a smarter way of going about this, this seems reasonable to me. You will still end up with pretty big values (rather than infs/nans) anyways.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud