This behavior is unexpected, when I specify the search space, I don't want something to be silently attached to it. Also, rsm parameter for catboost is not supported when training on GPU and is present in the default parameter list, leading to crash.
Please consider no using fault parameters when there are user-passed ones here: https://github.com/yandexdataschool/modelgym/blob/master/modelgym/utils/model_space.py#L18
This behavior is unexpected, when I specify the search space, I don't want something to be silently attached to it. Also,
rsm
parameter for catboost is not supported when training on GPU and is present in the default parameter list, leading to crash.