Bug in dataset splits?

[puhsu]

I'm reading the dataset code and there is probably a bug here:

hopular/hopular/auxiliary/data.py

Lines 597 to 602 in 3e0c39f

	self.__splits = (split_training, split_validation, split_test)
	# Sort dataset according to splits.
	self.__data = np.concatenate((
	self.__data[split_training], self.__data[split_validation], self.__data[split_test]
	), axis=0)

After that, you are using the old indices to index into the shuffled/concatenated arrays. So the splits are different (not stratified, for example):

hopular/hopular/auxiliary/data.py

Lines 637 to 638 in 3e0c39f

	def split_train(self) -> torch.Tensor:
	return self.__splits[0]