Fix Dataset builder pattern. Decouple Dataset and reader.
I agree that design for this part could be improved a lot. Could you please give details about the PR? Also all tests seem to be failing?
- What do you mean by "the PR"?
- Yes, I will fix tests.
PR = Pull Request. The question is: what changes are you proposing?
Fastfix reader call in PredefinedKFold.split.
Actually, reader should not be called from PredefinedKFold.split because it leads to extra coupling. So I think a bit more redesign is needed. I will check it in a bt later.
If the changes only affect Dataset.split() then it's probably not worth spending too much time on it: the split() method is deprecated and replaced by the use of CV iterators.
- Move data preprocessing for building data sets into the dedicated functions (
load_from_dfandload_from_file) away from initializer (soDatasetAutoFolds.__init__will become trivial). - Move
read_ratingstoReaderbecause it isReaderfunctionality to read rating, notDataSet.