You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you can solve most of the failing tests on mb-sandbox in a reasonable time, it's worth spending the effort to have a benchmark we can work with. However, @lilianschuster said that there is also quite some boiler-plate code that might hinder development itself, so if this gets too complicated maybe drop this step altogether.
Regardless of the outcome of the above, perhaps the easiest would still be to start from scratch on OGGM core itself, by:
starting a new module called massbalance_sandbox.py in oggm/core
creating a first class called DailyTIModel which mirrors MonthlyTIModel but uses @lilianschuster's code to deal with daily inputs
DailyTIModel can be tested the same way as MonthlyTIModel and it can be quite shallow, for example by copiing some of the tests in
Yep sounds good ... happy to start the script and the class and ask input when needed. The aim first would be to re-create that test_prepro.py::test_mb_calibration_from_scalar_mb with @lilianschuster's classes/code and go from there... to put together the new massbalance_sandbox.py script.
Lili's code has a different way of defining that though.. She has a TIModel_Parent class which does not look quite the same? The parent class deals with chosen different MB types within Lili's code (e.g. smb with surface type distinction etc)
The question is perhaps more for @lilianschuster:
The code that we want to keep and integrate is it both? the: TIModel_Parent(MassBalanceModel) and TIModel(TIModel_Parent): or just the TIModel(TIModel_Parent):?? I guess the main feature is everything related in both classes to the mb_type = mb_real_daily
@bearecinos after our meeting and some thinking with @lilianschuster, I'd suggest the following:
If you can solve most of the failing tests on mb-sandbox in a reasonable time, it's worth spending the effort to have a benchmark we can work with. However, @lilianschuster said that there is also quite some boiler-plate code that might hinder development itself, so if this gets too complicated maybe drop this step altogether.
Regardless of the outcome of the above, perhaps the easiest would still be to start from scratch on OGGM core itself, by:
DailyTIModel
which mirrors MonthlyTIModel but uses @lilianschuster's code to deal with daily inputsDailyTIModel
can be tested the same way asMonthlyTIModel
and it can be quite shallow, for example by copiing some of the tests inoggm/oggm/tests/test_prepro.py
Line 1410 in 9faee85
Thoughts?
The text was updated successfully, but these errors were encountered: