AMICI
AMICI copied to clipboard
Generate model code for all test models in Python
Currently AMICI C++ code is pre-generated from Matlab. For proper testing, models should be imported on the fly.
Requires:
- [x] Allow for import of raw ODEs (YAML) #654 #857 #732
- [x] Python support for events #757
- [ ] Python support for second order sensitivities #264
Then
- [ ] Specify models as YAML
- [ ] Setup import and use in tests (
tests/testModels.py
) - [ ] Remove pre-generated C++ files
Wanted to ask: For events, we basically have test cases already in the python tests, the same for Heaviside functions. What's now still missing is:
- Getting splines merged (would allow for the Swameye JAK/STAT model, which was a typical test case before)
- Getting second order code implemented
On the second order code, I wanted to ask how we want to proceed at all: The old trick of augmenting a model, or rather really implementig the additional total derivatives in C++-base of AMICI and performing some additional (very sparse) partial second order derivatives in python?
Pro first: Faster to be implemented Pro second: More scalability to larger models, as more structure can be exploited and expressions are way more sparse...
Pro second: More scalability to larger models, as more structure can be exploited and expressions are way more sparse...
Sounds more attractive. Although I can't really judge the extra effort.
Pro second: More scalability to larger models, as more structure can be exploited and expressions are way more sparse...
Sounds more attractive. Although I can't really judge the extra effort.
Always depends on the amount of free coffee... :D