mteb 2.7.16__py3-none-any.whl → 2.7.18__py3-none-any.whl

mteb/_create_dataloaders.py

@@ -30,7 +30,7 @@ logger = logging.getLogger(__name__)
 def _create_dataloader_from_texts(
     text: list[str],
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
     **kwargs: Any,
 ) -> DataLoader[TextInput]:
     """Create a dataloader from a list of text.
@@ -48,7 +48,7 @@ def _create_dataloader_from_texts(
     return DataLoader(
         dataset,
         batch_size=batch_size,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )
 
 
@@ -74,7 +74,7 @@ def _corpus_to_dict(
 def _create_dataloader_for_retrieval_corpus(
     dataset: Dataset,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[CorpusInput]:
     """Create a dataloader from a corpus.
 
@@ -94,7 +94,7 @@ def _create_dataloader_for_retrieval_corpus(
     return DataLoader(
         new_ds,
         batch_size=batch_size,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )
 
 
@@ -111,7 +111,7 @@ def _combine_queries_with_instruction_text(row: dict[str, str]) -> dict[str, str
 def _create_text_dataloader_for_queries(
     queries: QueryDatasetType,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[QueryInput]:
     """Create a dataloader from a list of queries.
 
@@ -131,7 +131,7 @@ def _create_text_dataloader_for_queries(
     return DataLoader(
         queries,
         batch_size=batch_size,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )
 
 
@@ -200,7 +200,7 @@ def _convert_conv_history_to_query(
 def _create_dataloader_for_queries_conversation(
     queries: QueryDatasetType,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[QueryInput]:
     """Create a dataloader from a list of queries.
 
@@ -220,7 +220,7 @@ def _create_dataloader_for_queries_conversation(
         ),
         collate_fn=_custom_collate_fn,
         batch_size=batch_size,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )
 
 
@@ -265,7 +265,7 @@ def _prepare_image_dataset(
     dataset: Dataset,
     image_column_name: str | None = None,
     transform: Callable[[Any], Any] | None = None,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> Dataset:
     """Prepare the image dataset by converting images to RGB and applying transformations."""
     if (
@@ -315,7 +315,7 @@ def _create_image_dataloader(
     batch_size: int = 32,
     transform: Callable[[Any], Any] | None = None,
     collate_fn: Callable[[list[dict[str, Any]]], dict[str, Any]] = _custom_collate_fn,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[ImageInput]:
     """Creates a DataLoader with the image dataset prepared using the explicit transformation.
 
@@ -341,14 +341,14 @@ def _create_image_dataloader(
         batch_size=batch_size,
         collate_fn=collate_fn,
         shuffle=False,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )
 
 
 def _create_text_queries_dataloader(
     dataset: Dataset,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[QueryInput]:
     if not isinstance(dataset["text"][0], list):
         return _create_text_dataloader_for_queries(
@@ -368,7 +368,7 @@ def _create_queries_dataloader(
     task_metadata: TaskMetadata,
     input_column: str | None = None,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[QueryInput | ImageInput]:
     """Create a dataloader for queries."""
     queries_type = task_metadata.get_modalities(PromptType.query)
@@ -393,7 +393,7 @@ def _create_document_dataloader(
     task_metadata: TaskMetadata,
     input_column: str | None = None,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
 ) -> DataLoader[CorpusInput | ImageInput]:
     """Create a dataloader for documents.
 
@@ -430,7 +430,7 @@ def create_dataloader(
     prompt_type: PromptType | None = None,
     input_column: str | None = None,
     batch_size: int = 32,
-    num_proc: int = 1,
+    num_proc: int | None = None,
     **kwargs: Any,
 ) -> DataLoader[BatchedInput]:
     """Create a dataloader from a dataset.
@@ -482,5 +482,5 @@ def create_dataloader(
     return DataLoader(
         dataset,
         batch_size=batch_size,
-        num_workers=num_proc if num_proc > 1 else 0,
+        num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
     )

mteb/_evaluators/any_sts_evaluator.py

@@ -66,7 +66,7 @@ class AnySTSEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> STSEvaluatorScores:
         logger.info("Running semantic similarity - Encoding samples (1/2)")
         embeddings1 = model.encode(

mteb/_evaluators/classification_metrics.py

@@ -1,7 +1,16 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
 import numpy as np
 
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
 
-def hamming_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
+def hamming_score(
+    y_true: NDArray[np.integer], y_pred: NDArray[np.integer | np.floating]
+) -> float:
     """Compute the Hamming score (a.k.a. label-based accuracy) for multilabel classification.
 
     The Hamming score is the fraction of labels that are correctly predicted for each sample,

mteb/_evaluators/clustering_evaluator.py

@@ -45,7 +45,7 @@ class ClusteringEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> list[int]:
         data_loader = create_dataloader(
             self.dataset,

mteb/_evaluators/image/imagetext_pairclassification_evaluator.py

@@ -91,7 +91,7 @@ class ImageTextPairClassificationEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> list[torch.Tensor]:
         images = []
         if isinstance(self.images_column_names, str):
@@ -139,7 +139,7 @@ class ImageTextPairClassificationEvaluator(Evaluator):
             DataLoader(
                 CustomImageDataset(images),
                 collate_fn=_image_collate_fn,
-                num_workers=num_proc if num_proc > 1 else 0,
+                num_workers=num_proc if num_proc is not None and num_proc > 1 else 0,
             ),
             task_metadata=self.task_metadata,
             hf_subset=self.hf_subset,

mteb/_evaluators/pair_classification_evaluator.py

@@ -16,6 +16,7 @@ from mteb.similarity_functions import compute_pairwise_similarity
 
 if TYPE_CHECKING:
     from datasets import Dataset
+    from numpy.typing import NDArray
 
     from mteb.abstasks.task_metadata import TaskMetadata
     from mteb.models import EncoderProtocol
@@ -91,7 +92,7 @@ class PairClassificationEvaluator(Evaluator):
         self,
         model: EncoderProtocol,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> PairClassificationDistances:
         logger.info("Running pair classification - Encoding samples (1/2)")
         embeddings1 = model.encode(
@@ -155,7 +156,7 @@ class PairClassificationEvaluator(Evaluator):
         hf_split: str,
         hf_subset: str,
         **encode_kwargs: Any,
-    ) -> np.ndarray:
+    ) -> NDArray[np.floating]:
         index_map = {}
         all_unique_texts: list[str] = []
         all_texts_indexes = []

mteb/_evaluators/retrieval_evaluator.py

@@ -55,7 +55,7 @@ class RetrievalEvaluator(Evaluator):
         self,
         search_model: SearchProtocol,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> RetrievalOutputType:
         logger.info("Running retrieval task - Indexing corpus...")
         search_model.index(

mteb/_evaluators/retrieval_metrics.py

@@ -15,6 +15,8 @@ from mteb.types import RetrievalEvaluationResult
 if TYPE_CHECKING:
     from collections.abc import Mapping
 
+    from numpy.typing import NDArray
+
     from mteb.types import RelevantDocumentsType
 
 logger = logging.getLogger(__name__)
@@ -273,9 +275,9 @@ def confidence_scores(sim_scores: list[float]) -> dict[str, float]:
 
 
 def nauc(
-    conf_scores: np.ndarray,
-    metrics: np.ndarray,
-    abstention_rates: np.ndarray = np.linspace(0, 1, 11)[:-1],
+    conf_scores: NDArray[np.floating],
+    metrics: NDArray[np.floating],
+    abstention_rates: NDArray[np.floating] = np.linspace(0, 1, 11)[:-1],
 ) -> float:
     """Computes normalized Area Under the Curve (nAUC) on a set of evaluated instances as presented in the paper https://arxiv.org/abs/2402.12997
 
@@ -295,10 +297,10 @@ def nauc(
     """
 
     def abstention_curve(
-        conf_scores: np.ndarray,
-        metrics: np.ndarray,
-        abstention_rates: np.ndarray = np.linspace(0, 1, 11)[:-1],
-    ) -> np.ndarray:
+        conf_scores: NDArray[np.floating],
+        metrics: NDArray[np.floating],
+        abstention_rates: NDArray[np.floating] = np.linspace(0, 1, 11)[:-1],
+    ) -> NDArray[np.floating]:
         """Computes the raw abstention curve for a given set of evaluated instances and corresponding confidence scores
 
         Args:

mteb/_evaluators/sklearn_evaluator.py

@@ -10,6 +10,7 @@ from .evaluator import Evaluator
 if TYPE_CHECKING:
     import numpy as np
     from datasets import Dataset
+    from numpy.typing import NDArray
     from torch.utils.data import DataLoader
     from typing_extensions import Self
 
@@ -21,11 +22,15 @@ logger = logging.getLogger(__name__)
 
 
 class SklearnModelProtocol(Protocol):
-    def fit(self, X: Array, y: np.ndarray | list[int]) -> None: ...  # noqa: N803
-    def predict(self, X: Array) -> np.ndarray: ...  # noqa: N803
+    def fit(
+        self, X: Array, y: NDArray[np.integer | np.floating] | list[int | float]
+    ) -> None: ...
+    def predict(self, X: Array) -> NDArray[np.integer | np.floating]: ...
     def get_params(self) -> dict[str, Any]: ...
     def set_params(self, random_state: int, **kwargs: dict[str, Any]) -> Self: ...
-    def score(self, X: Array, y: np.ndarray | list[int]) -> float: ...  # noqa: N803
+    def score(
+        self, X: Array, y: NDArray[np.integer | np.floating] | list[int | float]
+    ) -> float: ...
 
 
 class SklearnEvaluator(Evaluator):
@@ -54,7 +59,9 @@ class SklearnEvaluator(Evaluator):
         self.evaluator_model = evaluator_model
 
     def create_dataloaders(
-        self, encode_kwargs: EncodeKwargs, num_proc: int
+        self,
+        encode_kwargs: EncodeKwargs,
+        num_proc: int | None,
     ) -> tuple[DataLoader[BatchedInput], DataLoader[BatchedInput]]:
         dataloader_train = create_dataloader(
             self.train_dataset,
@@ -78,8 +85,8 @@ class SklearnEvaluator(Evaluator):
         *,
         encode_kwargs: EncodeKwargs,
         test_cache: Array | None = None,
-        num_proc: int = 1,
-    ) -> tuple[np.ndarray, Array]:
+        num_proc: int | None = None,
+    ) -> tuple[NDArray[np.integer | np.floating], Array]:
         """Classification evaluation by training a sklearn classifier on the embeddings of the training set and evaluating on the embeddings of the test set.
 
         Args:

mteb/_evaluators/text/bitext_mining_evaluator.py

@@ -41,7 +41,7 @@ class BitextMiningEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> dict[str, list[dict[str, float]]]:
         pair_elements = {p for pair in self.pairs for p in pair}
         if isinstance(self.sentences, Dataset):

mteb/_evaluators/text/summarization_evaluator.py

@@ -100,7 +100,7 @@ class SummarizationEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> SummarizationDistances:
         # Get the human & machine summaries for the text in one go for all
         human_lens = [len(human_summaries) for human_summaries in self.human_summaries]

mteb/_evaluators/zeroshot_classification_evaluator.py

@@ -48,7 +48,7 @@ class ZeroShotClassificationEvaluator(Evaluator):
         model: EncoderProtocol,
         *,
         encode_kwargs: EncodeKwargs,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> Array:
         dataloader = create_dataloader(
             self.dataset,

mteb/abstasks/_stratification.py

@@ -38,21 +38,26 @@ Bibtex:
     }
 """
 
+from __future__ import annotations
+
 import itertools
-from typing import Any
+from typing import TYPE_CHECKING, Any
 
 import numpy as np
 import scipy.sparse as sp
 from sklearn.model_selection._split import _BaseKFold
 from sklearn.utils import check_random_state
 
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
 
 def _iterative_train_test_split(
-    X: np.ndarray,  # noqa: N803
-    y: np.ndarray,
+    X: NDArray[np.integer],
+    y: NDArray[np.integer],
     test_size: float,
     random_state: int | None = None,
-) -> tuple[np.ndarray, np.ndarray]:
+) -> tuple[NDArray[np.integer], NDArray[np.integer]]:
     """Iteratively stratified train/test split
 
     Slighltly modified from:
@@ -79,8 +84,8 @@ def _iterative_train_test_split(
 
 
 def _fold_tie_break(
-    desired_samples_per_fold: np.ndarray,
-    M: np.ndarray,  # noqa: N803
+    desired_samples_per_fold: NDArray[np.floating],
+    M: NDArray[np.integer],  # noqa: N803
     random_state: np.random.RandomState,
 ):
     """Helper function to split a tie between folds with same desirability of a given sample
@@ -179,7 +184,7 @@ class IterativeStratification(_BaseKFold):
             ]
 
     def _prepare_stratification(
-        self, y: np.ndarray
+        self, y: NDArray[np.integer]
     ) -> tuple[
         list[list[int]],
         dict[int, bool],
@@ -301,7 +306,7 @@ class IterativeStratification(_BaseKFold):
             self.desired_samples_per_fold[fold_selected] -= 1
             folds[fold_selected].append(row)
 
-    def _iter_test_indices(self, X, y=None, groups=None):  # noqa: N803
+    def _iter_test_indices(self, X, y=None, groups=None):
         """Internal method for providing scikit-learn's split with folds
 
         Args:

mteb/abstasks/abstask.py

@@ -116,7 +116,7 @@ class AbsTask(ABC):
             logger.warning(msg)
             warnings.warn(msg)
 
-    def dataset_transform(self, num_proc: int = 1, **kwargs: Any) -> None:
+    def dataset_transform(self, num_proc: int | None = None, **kwargs: Any) -> None:
         """A transform operations applied to the dataset after loading.
 
         This method is useful when the dataset from Huggingface is not in an `mteb` compatible format.
@@ -136,7 +136,7 @@ class AbsTask(ABC):
         *,
         encode_kwargs: EncodeKwargs,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> Mapping[HFSubset, ScoresDict]:
         """Evaluates an MTEB compatible model on the task.
@@ -219,7 +219,7 @@ class AbsTask(ABC):
         hf_subset: str,
         encode_kwargs: EncodeKwargs,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> ScoresDict:
         raise NotImplementedError(
@@ -324,7 +324,7 @@ class AbsTask(ABC):
             )  # only take the specified test split.
         return dataset_dict
 
-    def load_data(self, num_proc: int = 1, **kwargs: Any) -> None:
+    def load_data(self, num_proc: int | None = None, **kwargs: Any) -> None:
         """Loads dataset from HuggingFace hub
 
         This is the main loading function for Task. Do not overwrite this, instead we recommend using `dataset_transform`, which is called after the

mteb/abstasks/classification.py

@@ -31,6 +31,8 @@ from .abstask import AbsTask
 if TYPE_CHECKING:
     from pathlib import Path
 
+    from numpy.typing import NDArray
+
     from mteb._evaluators.sklearn_evaluator import SklearnModelProtocol
     from mteb.models import MTEBModels
     from mteb.types import EncodeKwargs, HFSubset, ScoresDict
@@ -136,7 +138,7 @@ class AbsTaskClassification(AbsTask):
         *,
         encode_kwargs: EncodeKwargs,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> dict[HFSubset, ScoresDict]:
         """Evaluate a model on the classification task.
@@ -199,7 +201,7 @@ class AbsTaskClassification(AbsTask):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> FullClassificationMetrics:
         if not isinstance(model, EncoderProtocol):
@@ -270,8 +272,8 @@ class AbsTaskClassification(AbsTask):
 
     def _calculate_scores(
         self,
-        y_test: np.ndarray | list[int],
-        y_pred: np.ndarray,
+        y_test: NDArray[np.integer] | list[int],
+        y_pred: NDArray[np.integer | np.floating] | list[int],
     ) -> ClassificationMetrics:
         scores = ClassificationMetrics(
             accuracy=accuracy_score(y_test, y_pred),

mteb/abstasks/clustering.py

@@ -169,7 +169,7 @@ class AbsTaskClustering(AbsTask):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> ScoresDict:
         if not isinstance(model, EncoderProtocol):

mteb/abstasks/clustering_legacy.py

@@ -95,7 +95,7 @@ class AbsTaskClusteringLegacy(AbsTask):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> ScoresDict:
         if not isinstance(model, EncoderProtocol):

mteb/abstasks/image/image_text_pair_classification.py

@@ -134,7 +134,7 @@ class AbsTaskImageTextPairClassification(AbsTask):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> ImageTextPairClassificationMetrics:
         if not isinstance(model, EncoderProtocol):

mteb/abstasks/multilabel_classification.py

@@ -23,6 +23,8 @@ from .classification import AbsTaskClassification
 if TYPE_CHECKING:
     from pathlib import Path
 
+    from numpy.typing import NDArray
+
     from mteb._evaluators.sklearn_evaluator import SklearnModelProtocol
     from mteb.models import MTEBModels
     from mteb.types import Array, EncodeKwargs
@@ -32,10 +34,10 @@ logger = logging.getLogger(__name__)
 
 def _evaluate_classifier(
     embeddings_train: Array,
-    y_train: np.ndarray,
+    y_train: NDArray[np.integer],
     embeddings_test: Array,
     classifier: SklearnModelProtocol,
-) -> tuple[np.ndarray, SklearnModelProtocol]:
+) -> tuple[NDArray[np.integer | np.floating], SklearnModelProtocol]:
     classifier_copy: SklearnModelProtocol = clone(classifier)
     classifier_copy.fit(embeddings_train, y_train)
     return classifier_copy.predict(embeddings_test), classifier_copy
@@ -93,7 +95,7 @@ class AbsTaskMultilabelClassification(AbsTaskClassification):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> FullMultilabelClassificationMetrics:
         if not isinstance(model, EncoderProtocol):
@@ -208,8 +210,8 @@ class AbsTaskMultilabelClassification(AbsTaskClassification):
 
     def _calculate_scores(  # type: ignore[override]
         self,
-        y_test: np.ndarray,
-        y_pred: np.ndarray,
+        y_test: NDArray[np.integer],
+        y_pred: NDArray[np.integer | np.floating],
         x_test_embedding: Array,
         current_classifier: SklearnModelProtocol,
     ) -> MultilabelClassificationMetrics:

mteb/abstasks/pair_classification.py

@@ -97,7 +97,7 @@ class AbsTaskPairClassification(AbsTask):
         hf_subset: str,
         encode_kwargs: EncodeKwargs,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs,
     ) -> dict[str, float]:
         if not isinstance(model, EncoderProtocol):

mteb/abstasks/regression.py

@@ -24,6 +24,7 @@ from .classification import AbsTaskClassification
 
 if TYPE_CHECKING:
     from datasets import Dataset
+    from numpy.typing import NDArray
 
     from mteb._evaluators.sklearn_evaluator import SklearnModelProtocol
     from mteb.types.statistics import (
@@ -123,8 +124,8 @@ class AbsTaskRegression(AbsTaskClassification):
 
     def _calculate_scores(  # type: ignore[override]
         self,
-        y_test: np.ndarray | list[int],
-        y_pred: np.ndarray,
+        y_test: NDArray[np.floating] | list[float],
+        y_pred: NDArray[np.floating] | list[float],
     ) -> RegressionMetrics:
         mse = mean_squared_error(y_test, y_pred)
         return RegressionMetrics(

mteb/abstasks/retrieval.py

@@ -148,7 +148,10 @@ class AbsTaskRetrieval(AbsTask):
             )
         )
 
-    def convert_v1_dataset_format_to_v2(self, num_proc: int) -> None:
+    def convert_v1_dataset_format_to_v2(
+        self,
+        num_proc: int | None,
+    ) -> None:
         """Convert dataset from v1 (from `self.queries`, `self.document`) format to v2 format (`self.dotaset`)."""
         # check if dataset is `v1` version
         if not hasattr(self, "queries"):
@@ -257,7 +260,7 @@ class AbsTaskRetrieval(AbsTask):
         if hasattr(self, "top_ranked"):
             del self.top_ranked
 
-    def load_data(self, num_proc: int = 1, **kwargs) -> None:
+    def load_data(self, num_proc: int | None = None, **kwargs) -> None:
         """Load the dataset for the retrieval task."""
         if self.data_loaded:
             return
@@ -301,7 +304,7 @@ class AbsTaskRetrieval(AbsTask):
         *,
         encode_kwargs: EncodeKwargs,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs: Any,
     ) -> Mapping[HFSubset, ScoresDict]:
         """Evaluate the model on the retrieval task.
@@ -342,7 +345,7 @@ class AbsTaskRetrieval(AbsTask):
         hf_split: str,
         hf_subset: str,
         prediction_folder: Path | None = None,
-        num_proc: int = 1,
+        num_proc: int | None = None,
         **kwargs,
     ) -> ScoresDict:
         """Evaluate a model on a specific subset of the data.
@@ -473,7 +476,7 @@ class AbsTaskRetrieval(AbsTask):
         split: str,
         hf_subset: str | None = None,
         compute_overall: bool = False,
-        num_proc: int = 1,
+        num_proc: int | None = None,
     ) -> RetrievalDescriptiveStatistics:
         self.convert_v1_dataset_format_to_v2(num_proc)
         if hf_subset and hf_subset in self.dataset: