Fix pandas3 compat: use pd.Series(...).value_counts() instead of pd.value_counts(...)

[cakedev0]

Description

pandas 3 stopped exposing pd.value_counts so a couple of places were breaking. I changed to using pd.Series(...).value_counts() instead.

---

Checklist:

Completeness and readability

• Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
• I have resolved any merge conflicts that might occur with the base branch.

Testing

• I have run it locally and tested the changes extensively.
• All CI jobs are green or I have provided justification why they aren't.

[cakedev0]

I'm not sure why the CI is red but I think it's unrelated to my changes, so I'm marking this PR as ready for review.

[david-cortes-intel]

LGTM. CI errors are unrelated to this change.

But note that there's still other compatibility issues between the latest openml package and pandas>=3 which lead to errors:

  File "/localdisk2/mkl/dcortes/repos/scikit-learn_bench/sklbench/datasets/downloaders.py", line 99, in fetch_and_correct_openml
    x, y, _, _ = dataset.get_data(
                 ^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/openml/datasets/dataset.py", line 820, in get_data
    data, categorical, attribute_names = self._load_data()
                                         ^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/openml/datasets/dataset.py", line 611, in _load_data
    return self._cache_compressed_file_from_file(Path(file_to_load))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/openml/datasets/dataset.py", line 556, in _cache_compressed_file_from_file
    attribute_names, categorical, data = self._parse_data_from_file(data_file)
                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/openml/datasets/dataset.py", line 584, in _parse_data_from_file
    data, categorical, attribute_names = self._parse_data_from_arff(data_file)
                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/openml/datasets/dataset.py", line 472, in _parse_data_from_arff
    pd.factorize(type_)[0]
    ^^^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/pandas/core/algorithms.py", line 791, in factorize
    values = _ensure_arraylike(values, func_name="factorize")
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/localdisk2/mkl/dcortes/miniforge3/envs/sklbench12/lib/python3.12/site-packages/pandas/core/algorithms.py", line 239, in _ensure_arraylike
    raise TypeError(
TypeError: factorize requires a Series, Index, ExtensionArray, np.ndarray or NumpyExtensionArray got list.

(CC @avolkov-intel )

[cakedev0]

there's still other compatibility issues between the latest openml package and pandas>=3 which lead to errors

I don't see those. Do you have a config that reproduces it?

Note: I'm using pixi and I'm installing openml from pypi.

[david-cortes-intel]

there's still other compatibility issues between the latest openml package and pandas>=3 which lead to errors

I don't see those. Do you have a config that reproduces it?

Note: I'm using pixi and I'm installing openml from pypi.

It could be triggered like this, if you are interested in contributing to OpenML:

python -m sklbench --config configs/regular/ensemble.json --result-file try.json --filters algorithm:library=sklearn algorithm:device=cpu --prefetch

The openml package in conda-forge requires numpy<2, so right now PyPI is the only option.

[cakedev0]

Thanks, I will use pandas 2 then.