Removed version restriction for scikit-learn in examples dependencies

[brownbaerchen]

Due Diligence

• General:
  • title of the PR is suitable to appear in the Release Notes
• Implementation:
  • unit tests: all split configurations tested
  • unit tests: multiple dtypes tested
  • NEW unit tests: MPS tested (1 MPI process, 1 GPU)
  • benchmarks: created for new functionality
  • benchmarks: performance improved or maintained
  • documentation updated where needed

Description

See the discussion in #2084 for the problem and solution. Also, I refactored how the paths to the datasets are set up as pkg_resources is deprecated and the old solution actually did not work with my editable installation of heat. I did make sure that this works with an installation directly from PyPI. That is to say pip install heat[h5py]; python examples/classification/demo_knn.py does work.

Unrelated to the issue that this PR addresses: I see that the IO tests rely on being executed in a certain directory. As far as I can tell, this could be fixed by applying the same refactor from here, i.e. replacing

cls.HDF5_PATH = os.path.join(os.getcwd(), "heat/datasets/iris.h5")

by

cls.HDF5_PATH = os.path.join(os.path.dirname(ht.__file__), "datasets/iris.h5")

let me know if I should do this or if there is any reason not to.

Issue/s resolved:

#2084

Changes proposed:

• Remove version restriction on scikit-learn in examples dependencies
• Refactor generation of paths to datasets in tutorials

Type of change

Bug fix

[mtar]

In the examples a HDF5 file is loaded. Perhaps, we should add the library to the examples optional dependencies, too.

[mtar]

Thank you. Looks fine to me as we are only using the dataset in scikit-learn now.

[github-actions]

Successfully created backport PR for stable:

• [Backport stable] Removed version restriction for scikit-learn in examples dependencies #2104