Open fkiraly opened 1 year ago
@fkiraly can you confirm this was fixed by #115.
It's not released yet, is it? Since 0.3.0 predates #115.
I've opened a version of the refactor branch with install from current skbase
main
branch, if the tests pass here, it should tell us whether it's fixed: https://github.com/sktime/sktime/pull/4208
@fkiraly looks like sktime/sktime#4208 is failing, but not for the same reasons as the prior test. I think they might be un-related to BaseObject
, but I haven't looked to closely (seems like errors had to do with missing parameters in construction in a few estimators and a test case).
yes, indeed. I'm not sure what causes these, seems like default parameter generation.
I would put the functions that retrieve the set of parameters (like get_param_names
) under suspicion.
This is the most recent change to this kind of function, in skbase
, that I remember:
https://github.com/sktime/skbase/issues/65
From this refactor PR in
sktime
: https://github.com/sktime/sktime/pull/3777It would appear that
all_objects
returns the variable name, not the class name, as the string (first tuple element) when the two are unequal.I'm not entirely sure about the "should" here and whether that's indeed the root cause of the test failure, but it seems to be discrepant with behaviour in
sktime
.