feat: expose random seed, small typing ix

Author: stephantul

model2vec/train/classifier.py

@@ -23,7 +23,7 @@
 from model2vec.train.base import FinetunableStaticModel, TextDataset
 
 logger = logging.getLogger(__name__)
-_RANDOM_SEED = 42
+_DEFAULT_RANDOM_SEED = 42
 
 LabelType = TypeVar("LabelType", list[str], list[list[str]])
 
@@ -158,6 +158,7 @@ def fit(  # noqa: C901  # Complexity is bad.
         X_val: list[str] | None = None,
         y_val: LabelType | None = None,
         class_weight: torch.Tensor | None = None,
+        seed: int = _DEFAULT_RANDOM_SEED,
     ) -> StaticModelForClassification:
         """
         Fit a model.
@@ -187,14 +188,14 @@ def fit(  # noqa: C901  # Complexity is bad.
         :param y_val: The labels to be used for validation.
         :param class_weight: The weight of the classes. If None, all classes are weighted equally. Must
             have the same length as the number of classes.
+        :param seed: The random seed to use. Defaults to 42.
         :return: The fitted model.
         :raises ValueError: If either X_val or y_val are provided, but not both.
         """
-        pl.seed_everything(_RANDOM_SEED)
+        pl.seed_everything(seed)
         logger.info("Re-initializing model.")
 
         # Determine whether the task is multilabel based on the type of y.
-
         self._initialize(y)
 
         if (X_val is not None) != (y_val is not None):
@@ -380,7 +381,7 @@ def to_pipeline(self) -> StaticModelPipeline:
         """Convert the model to an sklearn pipeline."""
         static_model = self.to_static_model()
 
-        random_state = np.random.RandomState(_RANDOM_SEED)
+        random_state = np.random.RandomState(_DEFAULT_RANDOM_SEED)
         n_items = len(self.classes)
         X = random_state.randn(n_items, static_model.dim)
         y = self.classes