[MNT] deduplicating utilities with scikit-base

[fkiraly]

Deduplicates utilities that were copy-pasted to avoid a scikit-base dependency early on.

Since the v1 testing framework will require scikit-base, it makes sense to now attempt a deduplication.

Adds scikit-base as a core dependency.

[phoeenniixx]

I think there are still some places left where _get_installed_packages are left in the tests folder. Like in test_nbeats_kan and test_metrics

[fkiraly]

yes, you are right - I forgot to commit some files

[fkiraly]

@phoeenniixx, could you kindly run pre-commit on the files failing linting? Something is not quite working for me locally, the files do not get changed. Not sure what I might be doing wrong. Did you change something recently, e.g., target python version?

(I think my virus scanner just put my precommit env into quarantine)

[fkiraly]

the rest should be fixed though

[phoeenniixx]

Did you change something recently, e.g., target python version?

Actually yes. I changed the target python version from 3.9 - 3.10. As mentioned in this comment #1856 (comment). I also looked at sktime pyproject, there also the target version was py310. I thought this would not cause an issue?

[phoeenniixx]

pre-commit seems to run fine in my env:

 [~/pytorch-forecasting]
 aryan   skbase-dep  pre-commit run --all-files                                  
trim trailing whitespace.................................................Passed
fix end of files.........................................................Passed
check yaml...............................................................Passed
check python ast.........................................................Passed
ruff.....................................................................Failed
- hook id: ruff
- exit code: 1
- files were modified by this hook

pytorch_forecasting/data/_tslib_data_module.py:23:14: UP007 Use `X | Y` for type annotations
   |
21 | from pytorch_forecasting.utils._coerce import _coerce_to_dict
22 | 
23 | NORMALIZER = Union[TorchNormalizer, EncoderNormalizer, NaNLabelEncoder]
   |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ UP007
   |
   = help: Convert to `X | Y`

pytorch_forecasting/data/data_module.py:26:14: UP007 Use `X | Y` for type annotations
   |
24 | from pytorch_forecasting.utils._coerce import _coerce_to_dict
25 | 
26 | NORMALIZER = Union[TorchNormalizer, NaNLabelEncoder, EncoderNormalizer]
   |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ UP007
   |
   = help: Convert to `X | Y`

pytorch_forecasting/data/data_module.py:96:89: E501 Line too long (99 > 88)
   |
94 |         add_target_scales: bool = False,
95 |         add_encoder_length: bool | str = "auto",
96 |         target_normalizer: NORMALIZER | str | list[NORMALIZER] | tuple[NORMALIZER] | None = "auto",
   |                                                                                         ^^^^^^^^^^^ E501
97 |         categorical_encoders: dict[str, NaNLabelEncoder] | None = None,
98 |         scalers: dict[str, StandardScaler | RobustScaler | TorchNormalizer | EncoderNormalizer] | None = None,
   |

pytorch_forecasting/data/data_module.py:98:89: E501 Line too long (110 > 88)
    |
 96 |         target_normalizer: NORMALIZER | str | list[NORMALIZER] | tuple[NORMALIZER] | None = "auto",
 97 |         categorical_encoders: dict[str, NaNLabelEncoder] | None = None,
 98 |         scalers: dict[str, StandardScaler | RobustScaler | TorchNormalizer | EncoderNormalizer] | None = None,
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^ E501
 99 |         randomize_length: None | tuple[float, float] | bool = False,
100 |         batch_size: int = 32,
    |

pytorch_forecasting/data/encoders.py:1418:89: E501 Line too long (94 > 88)
     |
1416 |         return_norm: bool = False,
1417 |         target_scale: list[torch.Tensor] = None,
1418 |     ) -> list[tuple[np.ndarray | torch.Tensor, np.ndarray]] | list[np.ndarray | torch.Tensor]:
     |                                                                                         ^^^^^^ E501
1419 |         """
1420 |         Scale input data.
     |

pytorch_forecasting/data/timeseries/_timeseries.py:150:14: UP007 Use `X | Y` for type annotations
    |
150 | NORMALIZER = Union[TorchNormalizer, NaNLabelEncoder, EncoderNormalizer]
    |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ UP007
151 | 
152 | Columns = list[str]
    |
    = help: Convert to `X | Y`

pytorch_forecasting/data/timeseries/_timeseries.py:466:89: E501 Line too long (99 > 88)
    |
464 |         add_target_scales: bool = False,
465 |         add_encoder_length: bool | str = "auto",
466 |         target_normalizer: NORMALIZER | str | list[NORMALIZER] | tuple[NORMALIZER] | None = "auto",
    |                                                                                         ^^^^^^^^^^^ E501
467 |         categorical_encoders: dict[str, NaNLabelEncoder] | None = None,
468 |         scalers: dict[str, StandardScaler | RobustScaler | TorchNormalizer | EncoderNormalizer] | None = None,
    |

pytorch_forecasting/data/timeseries/_timeseries.py:468:89: E501 Line too long (110 > 88)
    |
466 |         target_normalizer: NORMALIZER | str | list[NORMALIZER] | tuple[NORMALIZER] | None = "auto",
467 |         categorical_encoders: dict[str, NaNLabelEncoder] | None = None,
468 |         scalers: dict[str, StandardScaler | RobustScaler | TorchNormalizer | EncoderNormalizer] | None = None,
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^ E501
469 |         randomize_length: None | tuple[float, float] | bool = False,
470 |         predict_mode: bool = False,
    |

pytorch_forecasting/models/nn/embeddings.py:40:89: E501 Line too long (105 > 88)
   |
38 |     def __init__(
39 |         self,
40 |         embedding_sizes: dict[str, tuple[int, int]] | dict[str, int] | list[int] | list[tuple[int, int]],
   |                                                                                         ^^^^^^^^^^^^^^^^^ E501
41 |         x_categoricals: list[str] = None,
42 |         categorical_groups: dict[str, list[str]] | None = None,
   |

pytorch_forecasting/models/nn/rnn.py:12:15: UP007 Use `X | Y` for type annotations
   |
10 | from torch.nn.utils import rnn
11 | 
12 | HiddenState = Union[tuple[torch.Tensor, torch.Tensor], torch.Tensor]
   |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ UP007
   |
   = help: Convert to `X | Y`

pytorch_forecasting/models/xlstm/_xlstm.py:114:89: E501 Line too long (121 > 88)
    |
112 |         self,
113 |         x: dict[str, torch.Tensor],
114 |         hidden_states: tuple[torch.Tensor, torch.Tensor] | tuple[torch.Tensor, torch.Tensor, torch.Tensor] | None = None,
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ E501
115 |     ) -> dict[str, torch.Tensor]:
116 |         """Forward Pass for the model."""
    |

pytorch_forecasting/utils/_utils.py:442:89: E501 Line too long (132 > 88)
    |
441 | def move_to_device(
442 |     x: dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor],
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ E501
443 |     device: str | torch.DeviceObjType,
444 | ) -> dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]:
    |

pytorch_forecasting/utils/_utils.py:444:89: E501 Line too long (130 > 88)
    |
442 |     x: dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor],
443 |     device: str | torch.DeviceObjType,
444 | ) -> dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]:
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ E501
445 |     """
446 |     Move object to device.
    |

pytorch_forecasting/utils/_utils.py:479:89: E501 Line too long (132 > 88)
    |
478 | def detach(
479 |     x: dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor],
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ E501
480 | ) -> dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]:
481 |     """
    |

pytorch_forecasting/utils/_utils.py:480:89: E501 Line too long (130 > 88)
    |
478 | def detach(
479 |     x: dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor],
480 | ) -> dict[str, torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]] | torch.Tensor | list[torch.Tensor] | tuple[torch.Tensor]:
    |                                                                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ E501
481 |     """
482 |     Detach object
    |

Found 406 errors (391 fixed, 15 remaining).

ruff-format..............................................................Failed
- hook id: ruff-format
- files were modified by this hook

10 files reformatted, 197 files left unchanged

nbqa-black...............................................................Passed
nbqa-ruff................................................................Passed
nbqa-check-ast...........................................................Passed

Is this a windows issue?

[phoeenniixx]

py310 changes the formatting conventions drastically, that is why pre-commit is failing. Sorry, I didnot know that. I think I might need to raise a PR with just pre-commit run --all-files to solve this issue.

[fkiraly]

no problem - could you kindly run precommit on these failing files though? The reason being, even if you raise a precommit with all, it will now create merge conflicts, and it might be easier to get this PR in first (and/or release)

[phoeenniixx]

Done. I also updated the skbase version from <0.13.0 to <0.14.0 as 0.13.0 is the only version that supports python 3.14.