ConcatTask have the `feature_set` name wrong

[00sapo]

music_symbolic_features/symbolic_features/data.py

Lines 532 to 556 in 158cd0c

	def __init__(self, tasks: List[Task]):
	self.tasks = tasks
	self.__loaded = False
	assert len(self.tasks) >= 2, "ConcatTask must have at least 2 tasks"
	extensions = [task.extension for task in self.tasks]
	assert all(ext == extensions[0] for ext in extensions), "Extensions must match"
	friendly_names = [task.dataset.friendly_name for task in self.tasks]
	assert all(
	name == friendly_names[0] for name in friendly_names
	), "Datasets must match"
	super().__init__(
	self.tasks[0].dataset, self.tasks[0].feature_set, extensions[0]
	)
	# self.dataset = self.tasks[0].dataset
	# self.extension = extensions[0]
	feature_set_names = [task.feature_set.name for task in self.tasks]
	self.name = (
	friendly_names[0]
	+ "-"
	+ "-".join(feature_set_names)
	+ "-"
	+ extensions[0][1:]
	)

The feature_set attribute of ConcatTask is initialized to the first feature set. It should be initialized to ''.join(feature_set_names), as in the task.name attribute.

Right now, the task using feature-set A as first feature-set will also include all the concatenations such as A-B, A-C, A-B-C (e.g. task including musif_native is actually including musif_native, musif_native-music21_native, musif_native-jsymbolic, musif_native-music21_native-jsymbolic, for a total of 5 tasks)

[00sapo]

Solving this issue would also allow for longer times in the AutoML when jobs are run on shared clusters where computing powers is scheduled according to the duration of the job

[00sapo]

Fixed in 9150c5f