dataeval 0.76.0__tar.gz → 0.81.0__tar.gz

{dataeval-0.76.0 → dataeval-0.81.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dataeval
-Version: 0.76.0
+Version: 0.81.0
 Summary: DataEval provides a simple interface to characterize image data and its impact on model performance across classification and object-detection tasks
 Home-page: https://dataeval.ai/
 License: MIT
@@ -21,8 +21,12 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3 :: Only
 Classifier: Topic :: Scientific/Engineering
 Provides-Extra: all
-Requires-Dist: matplotlib ; extra == "all"
+Requires-Dist: defusedxml (>=0.7.1)
+Requires-Dist: fast_hdbscan (==0.2.0)
+Requires-Dist: matplotlib (>=3.7.1) ; extra == "all"
+Requires-Dist: numba (>=0.59.1)
 Requires-Dist: numpy (>=1.24.2)
+Requires-Dist: pandas (>=2.0) ; extra == "all"
 Requires-Dist: pillow (>=10.3.0)
 Requires-Dist: requests
 Requires-Dist: scikit-learn (>=1.5.0)
@@ -38,13 +42,17 @@ Description-Content-Type: text/markdown
 
 # DataEval
 
-To view our extensive collection of tutorials, how-to's, explanation guides, and reference material, please visit our documentation on **[Read the Docs](https://dataeval.readthedocs.io/)**
+To view our extensive collection of tutorials, how-to's, explanation guides,
+and reference material, please visit our documentation on
+**[Read the Docs](https://dataeval.readthedocs.io/)**
 
 ## About DataEval
 
 <!-- start tagline -->
 
-DataEval curates datasets to train and test performant, robust, unbiased and reliable AI models and monitors for data shifts that impact performance of deployed models. 
+DataEval curates datasets to train and test performant, robust, unbiased and
+reliable AI models and monitors for data shifts that impact performance of
+deployed models.
 
 <!-- end tagline -->
 
@@ -52,22 +60,33 @@ DataEval curates datasets to train and test performant, robust, unbiased and rel
 
 <!-- start needs -->
 
-DataEval is an effective, powerful, and reliable set of tools for any T&E engineer. Throughout all stages of the machine learning lifecycle, DataEval supports model development, data analysis, and monitoring with state-of-the-art algorithms to help you solve difficult problems. With a focus on computer vision tasks, DataEval provides simple, but effective metrics for performance estimation, bias detection, and dataset linting.
+DataEval is an effective, powerful, and reliable set of tools for any T&E
+engineer. Throughout all stages of the machine learning lifecycle, DataEval
+supports model development, data analysis, and monitoring with state-of-the-art
+algorithms to help you solve difficult problems. With a focus on computer
+vision tasks, DataEval provides simple, but effective metrics for performance
+estimation, bias detection, and dataset linting.
 
 <!-- end needs -->
 
 <!-- start JATIC interop -->
-DataEval is easy to install, supports a wide range of Python versions, and is compatible with many of the most popular packages in the scientific and T&E communities.
-DataEval also has native interopability between JATIC's suite of tools when using MAITE-compliant datasets and models.
+DataEval is easy to install, supports a wide range of Python versions, and is
+compatible with many of the most popular packages in the scientific and T&E
+communities.
+
+DataEval also has native interopability between JATIC's suite of tools when
+using MAITE-compliant datasets and models.
 <!-- end JATIC interop -->
 
 ## Getting Started
 
 **Python versions:** 3.9 - 3.12
 
-**Supported packages**: *NumPy*, *Pandas*, *Sci-kit learn*, *MAITE*, *NRTK*, *Gradient*
+**Supported packages**: *NumPy*, *Pandas*, *Sci-kit learn*, *MAITE*, *NRTK*,
+*Gradient*
 
-Choose your preferred method of installation below or follow our [installation guide](https://dataeval.readthedocs.io/en/v0.74.2/installation.html).
+Choose your preferred method of installation below or follow our
+[installation guide](https://dataeval.readthedocs.io/en/v0.74.2/installation.html).
 
 * [Installing with pip](#installing-with-pip)
 * [Installing with conda/mamba](#installing-with-conda)
@@ -75,7 +94,8 @@ Choose your preferred method of installation below or follow our [installation g
 
 ### **Installing with pip**
 
-You can install DataEval directly from pypi.org using the following command.  The optional dependencies of DataEval are `all`.
+You can install DataEval directly from pypi.org using the following command.
+The optional dependencies of DataEval are `all`.
 
 ```bash
 pip install dataeval[all]
@@ -83,8 +103,9 @@ pip install dataeval[all]
 
 ### **Installing with conda**
 
-DataEval can be installed in a Conda/Mamba environment using the provided `environment.yaml` file.  As some dependencies
-are installed from the `pytorch` channel, the channel is specified in the below example.
+DataEval can be installed in a Conda/Mamba environment using the provided
+`environment.yaml` file.  As some dependencies are installed from the `pytorch`
+channel, the channel is specified in the below example.
 
 ```bash
 micromamba create -f environment\environment.yaml -c pytorch
@@ -92,7 +113,9 @@ micromamba create -f environment\environment.yaml -c pytorch
 
 ### **Installing from GitHub**
 
-To install DataEval from source locally on Ubuntu, you will need `git-lfs` to download larger, binary source files and `poetry` for project dependency management.
+To install DataEval from source locally on Ubuntu, you will need `git-lfs` to
+download larger, binary source files and `poetry` for project dependency
+management.
 
 ```bash
 sudo apt-get install git-lfs
@@ -112,7 +135,9 @@ Install DataEval with optional dependencies for development.
 poetry install --all-extras --with dev
 ```
 
-Now that DataEval is installed, you can run commands in the poetry virtual environment by prefixing shell commands with `poetry run`, or activate the virtual environment directly in the shell.
+Now that DataEval is installed, you can run commands in the poetry virtual
+environment by prefixing shell commands with `poetry run`, or activate the
+virtual environment directly in the shell.
 
 ```bash
 poetry shell
@@ -131,7 +156,11 @@ If you have any questions, feel free to reach out to the people below:
 
 ### CDAO Funding Acknowledgement
 
-This material is based upon work supported by the Chief Digital and Artificial Intelligence Office under Contract No. W519TC-23-9-2033. The views and conclusions contained herein are those of the author(s) and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the U.S. Government.
+This material is based upon work supported by the Chief Digital and Artificial
+Intelligence Office under Contract No. W519TC-23-9-2033. The views and
+conclusions contained herein are those of the author(s) and should not be
+interpreted as necessarily representing the official policies or endorsements,
+either expressed or implied, of the U.S. Government.
 
 <!-- end acknowledgement -->
 

{dataeval-0.76.0 → dataeval-0.81.0}/README.md

@@ -1,12 +1,16 @@
 # DataEval
 
-To view our extensive collection of tutorials, how-to's, explanation guides, and reference material, please visit our documentation on **[Read the Docs](https://dataeval.readthedocs.io/)**
+To view our extensive collection of tutorials, how-to's, explanation guides,
+and reference material, please visit our documentation on
+**[Read the Docs](https://dataeval.readthedocs.io/)**
 
 ## About DataEval
 
 <!-- start tagline -->
 
-DataEval curates datasets to train and test performant, robust, unbiased and reliable AI models and monitors for data shifts that impact performance of deployed models. 
+DataEval curates datasets to train and test performant, robust, unbiased and
+reliable AI models and monitors for data shifts that impact performance of
+deployed models.
 
 <!-- end tagline -->
 
@@ -14,22 +18,33 @@ DataEval curates datasets to train and test performant, robust, unbiased and rel
 
 <!-- start needs -->
 
-DataEval is an effective, powerful, and reliable set of tools for any T&E engineer. Throughout all stages of the machine learning lifecycle, DataEval supports model development, data analysis, and monitoring with state-of-the-art algorithms to help you solve difficult problems. With a focus on computer vision tasks, DataEval provides simple, but effective metrics for performance estimation, bias detection, and dataset linting.
+DataEval is an effective, powerful, and reliable set of tools for any T&E
+engineer. Throughout all stages of the machine learning lifecycle, DataEval
+supports model development, data analysis, and monitoring with state-of-the-art
+algorithms to help you solve difficult problems. With a focus on computer
+vision tasks, DataEval provides simple, but effective metrics for performance
+estimation, bias detection, and dataset linting.
 
 <!-- end needs -->
 
 <!-- start JATIC interop -->
-DataEval is easy to install, supports a wide range of Python versions, and is compatible with many of the most popular packages in the scientific and T&E communities.
-DataEval also has native interopability between JATIC's suite of tools when using MAITE-compliant datasets and models.
+DataEval is easy to install, supports a wide range of Python versions, and is
+compatible with many of the most popular packages in the scientific and T&E
+communities.
+
+DataEval also has native interopability between JATIC's suite of tools when
+using MAITE-compliant datasets and models.
 <!-- end JATIC interop -->
 
 ## Getting Started
 
 **Python versions:** 3.9 - 3.12
 
-**Supported packages**: *NumPy*, *Pandas*, *Sci-kit learn*, *MAITE*, *NRTK*, *Gradient*
+**Supported packages**: *NumPy*, *Pandas*, *Sci-kit learn*, *MAITE*, *NRTK*,
+*Gradient*
 
-Choose your preferred method of installation below or follow our [installation guide](https://dataeval.readthedocs.io/en/v0.74.2/installation.html).
+Choose your preferred method of installation below or follow our
+[installation guide](https://dataeval.readthedocs.io/en/v0.74.2/installation.html).
 
 * [Installing with pip](#installing-with-pip)
 * [Installing with conda/mamba](#installing-with-conda)
@@ -37,7 +52,8 @@ Choose your preferred method of installation below or follow our [installation g
 
 ### **Installing with pip**
 
-You can install DataEval directly from pypi.org using the following command.  The optional dependencies of DataEval are `all`.
+You can install DataEval directly from pypi.org using the following command.
+The optional dependencies of DataEval are `all`.
 
 ```bash
 pip install dataeval[all]
@@ -45,8 +61,9 @@ pip install dataeval[all]
 
 ### **Installing with conda**
 
-DataEval can be installed in a Conda/Mamba environment using the provided `environment.yaml` file.  As some dependencies
-are installed from the `pytorch` channel, the channel is specified in the below example.
+DataEval can be installed in a Conda/Mamba environment using the provided
+`environment.yaml` file.  As some dependencies are installed from the `pytorch`
+channel, the channel is specified in the below example.
 
 ```bash
 micromamba create -f environment\environment.yaml -c pytorch
@@ -54,7 +71,9 @@ micromamba create -f environment\environment.yaml -c pytorch
 
 ### **Installing from GitHub**
 
-To install DataEval from source locally on Ubuntu, you will need `git-lfs` to download larger, binary source files and `poetry` for project dependency management.
+To install DataEval from source locally on Ubuntu, you will need `git-lfs` to
+download larger, binary source files and `poetry` for project dependency
+management.
 
 ```bash
 sudo apt-get install git-lfs
@@ -74,7 +93,9 @@ Install DataEval with optional dependencies for development.
 poetry install --all-extras --with dev
 ```
 
-Now that DataEval is installed, you can run commands in the poetry virtual environment by prefixing shell commands with `poetry run`, or activate the virtual environment directly in the shell.
+Now that DataEval is installed, you can run commands in the poetry virtual
+environment by prefixing shell commands with `poetry run`, or activate the
+virtual environment directly in the shell.
 
 ```bash
 poetry shell
@@ -93,6 +114,10 @@ If you have any questions, feel free to reach out to the people below:
 
 ### CDAO Funding Acknowledgement
 
-This material is based upon work supported by the Chief Digital and Artificial Intelligence Office under Contract No. W519TC-23-9-2033. The views and conclusions contained herein are those of the author(s) and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the U.S. Government.
+This material is based upon work supported by the Chief Digital and Artificial
+Intelligence Office under Contract No. W519TC-23-9-2033. The views and
+conclusions contained herein are those of the author(s) and should not be
+interpreted as necessarily representing the official policies or endorsements,
+either expressed or implied, of the U.S. Government.
 
 <!-- end acknowledgement -->

{dataeval-0.76.0 → dataeval-0.81.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "dataeval"
-version = "0.76.0" # dynamic
+version = "0.81.0" # dynamic
 description = "DataEval provides a simple interface to characterize image data and its impact on model performance across classification and object-detection tasks"
 license = "MIT"
 readme = "README.md"
@@ -42,6 +42,9 @@ packages = [
 [tool.poetry.dependencies]
 # required
 python = ">=3.9,<3.13"
+defusedxml = {version = ">=0.7.1"}
+fast_hdbscan = {version = "0.2.0"}  # 0.2.1 hits a bug in condense_tree comparing float to none
+numba = {version = ">=0.59.1"}
 numpy = {version = ">=1.24.2"}
 pillow = {version = ">=10.3.0"}
 requests = {version = "*"}
@@ -54,10 +57,11 @@ typing-extensions = {version = ">=4.12", python = "^3.9"}  # ParamSpec
 xxhash = {version = ">=3.3"}
 
 # optional
-matplotlib = {version = "*", optional = true}
+matplotlib = {version = ">=3.7.1", optional = true}
+pandas = {version = ">=2.0", optional = true}
 
 [tool.poetry.extras]
-all = ["matplotlib"]
+all = ["matplotlib", "pandas"]
 
 [tool.poetry.group.dev]
 optional = true
@@ -81,20 +85,20 @@ coverage = {version = "*", extras = ["toml"]}
 pyright = {version = "*", extras = ["nodejs"]}
 # prototype
 maite = {version = "*"}
-pandas = {version = "*"}
 seaborn = {version = "*"}
 # docs
 certifi = {version = ">=2024.07.04"}
 enum_tools = {version = ">=0.12.0", extras = ["sphinx"]}
 ipykernel = {version = ">=6.26.0"}
 ipywidgets = {version = ">=8.1.1"}
-jinja2 = {version = ">=3.1.5"}
+jinja2 = {version = ">=3.1.6"}
 jupyter-client = {version = ">=8.6.0"}
 jupyter-cache = {version = "*"}
 myst-nb = {version = ">=1.0.0"}
-sphinx-immaterial = {version = "*"}
 sphinx-autoapi = {version = "*"}
 sphinx-design = {version = "*"}
+sphinx-immaterial = {version = "*"}
+sphinx-new-tab-link = {version = "*"}
 sphinx-tabs = {version = "*"}
 Sphinx = {version = ">=7.2.6"}
 torchmetrics = {version = ">=1.0.0", source = "pytorch"}
@@ -128,6 +132,11 @@ reportMissingImports = false
 norecursedirs = ["prototype"]
 testpaths = ["tests"]
 addopts = ["--pythonwarnings=ignore::DeprecationWarning", "--verbose", "--durations=20", "--durations-min=1.0"]
+markers = [
+  "required: marks tests for required features",
+  "optional: marks tests for optional features",
+  "requires_all: marks tests that require the all extras",
+]
 
 [tool.coverage.run]
 source = ["src/dataeval"]
@@ -142,8 +151,9 @@ exclude_also = [
 ]
 include = ["*/src/dataeval/*"]
 omit = [
-  "*/torch/blocks.py",
-  "*/torch/utils.py",
+  "*/torch/_blocks.py",
+  "*/_clusterer.py",
+  "*/_fast_mst.py",
 ]
 fail_under = 90
 
@@ -177,6 +187,9 @@ per-file-ignores = { "*.ipynb" = ["E402"] }
 [tool.ruff.lint.isort]
 known-first-party = ["dataeval"]
 
+[tool.ruff.lint.flake8-builtins]
+builtins-strict-checking = false
+
 [tool.ruff.format]
 quote-style = "double"
 indent-style = "space"
@@ -186,7 +199,7 @@ docstring-code-format = true
 docstring-code-line-length = "dynamic"
 
 [tool.codespell]
-skip = './*env*,./prototype,./output,./docs/build,./docs/source/.jupyter_cache,CHANGELOG.md,poetry.lock,*.html'
+skip = './*env*,./prototype,./output,./docs/build,./docs/source/.jupyter_cache,CHANGELOG.md,poetry.lock,*.html,./docs/source/*/data'
 ignore-words-list = ["Hart"]
 
 [build-system]

{dataeval-0.76.0 → dataeval-0.81.0}/src/dataeval/__init__.py

@@ -7,12 +7,12 @@ shifts that impact performance of deployed models.
 
 from __future__ import annotations
 
-__all__ = ["detectors", "log", "metrics", "utils", "workflows"]
-__version__ = "0.76.0"
+__all__ = ["config", "detectors", "log", "metrics", "typing", "utils", "workflows"]
+__version__ = "0.81.0"
 
 import logging
 
-from dataeval import detectors, metrics, utils, workflows
+from dataeval import config, detectors, metrics, typing, utils, workflows
 
 logging.getLogger(__name__).addHandler(logging.NullHandler())
 

dataeval-0.76.0/src/dataeval/output.py → dataeval-0.81.0/src/dataeval/_output.py

@@ -32,9 +32,23 @@ class Output:
         return f"{self.__class__.__name__}: {str(self.dict())}"
 
     def dict(self) -> dict[str, Any]:
+        """
+        Output attributes as a dictionary.
+
+        Returns
+        -------
+        dict[str, Any]
+        """
         return {k: v for k, v in self.__dict__.items() if not k.startswith("_")}
 
     def meta(self) -> dict[str, Any]:
+        """
+        Execution metadata as a dictionary.
+
+        Returns
+        -------
+        dict[str, Any]
+        """
         return {k.removeprefix("_"): v for k, v in self.__dict__.items() if k.startswith("_")}
 
 

dataeval-0.81.0/src/dataeval/config.py

@@ -0,0 +1,77 @@
+"""
+Global configuration settings for DataEval.
+"""
+
+from __future__ import annotations
+
+__all__ = ["get_device", "set_device", "get_max_processes", "set_max_processes"]
+
+import torch
+from torch import device
+
+_device: device | None = None
+_processes: int | None = None
+
+
+def set_device(device: str | device | int) -> None:
+    """
+    Sets the default device to use when executing against a PyTorch backend.
+
+    Parameters
+    ----------
+    device : str or int or `torch.device`
+        The default device to use. See `torch.device <https://pytorch.org/docs/stable/tensor_attributes.html#torch.device>`_
+        documentation for more information.
+    """
+    global _device
+    _device = torch.device(device)
+
+
+def get_device(override: str | device | int | None = None) -> torch.device:
+    """
+    Returns the PyTorch device to use.
+
+    Parameters
+    ----------
+    override : str or int or `torch.device` or None, default None
+        The user specified override if provided, otherwise returns the default device.
+
+    Returns
+    -------
+    `torch.device`
+    """
+    if override is None:
+        global _device
+        return torch.get_default_device() if _device is None else _device
+    else:
+        return torch.device(override)
+
+
+def set_max_processes(processes: int | None) -> None:
+    """
+    Sets the maximum number of worker processes to use when running tasks that support parallel processing.
+
+    Parameters
+    ----------
+    processes : int or None
+        The maximum number of worker processes to use, or None to use
+        `os.process_cpu_count <https://docs.python.org/3/library/os.html#os.process_cpu_count>`_
+        to determine the number of worker processes.
+    """
+    global _processes
+    _processes = processes
+
+
+def get_max_processes() -> int | None:
+    """
+    Returns the maximum number of worker processes to use when running tasks that support parallel processing.
+
+    Returns
+    -------
+    int or None
+        The maximum number of worker processes to use, or None to use
+        `os.process_cpu_count <https://docs.python.org/3/library/os.html#os.process_cpu_count>`_
+        to determine the number of worker processes.
+    """
+    global _processes
+    return _processes

{dataeval-0.76.0 → dataeval-0.81.0}/src/dataeval/detectors/__init__.py

@@ -4,4 +4,4 @@ Detectors can determine if a dataset or individual images in a dataset are indic
 
 __all__ = ["drift", "linters", "ood"]
 
-from dataeval.detectors import drift, linters, ood
+from . import drift, linters, ood

dataeval-0.81.0/src/dataeval/detectors/drift/__init__.py

@@ -0,0 +1,22 @@
+"""
+:term:`Drift` detectors identify if the statistical properties of the data has changed.
+"""
+
+__all__ = [
+    "DriftCVM",
+    "DriftKS",
+    "DriftMMD",
+    "DriftMMDOutput",
+    "DriftOutput",
+    "DriftUncertainty",
+    "preprocess_drift",
+    "updates",
+]
+
+from dataeval.detectors.drift import updates
+from dataeval.detectors.drift._base import DriftOutput
+from dataeval.detectors.drift._cvm import DriftCVM
+from dataeval.detectors.drift._ks import DriftKS
+from dataeval.detectors.drift._mmd import DriftMMD, DriftMMDOutput
+from dataeval.detectors.drift._torch import preprocess_drift
+from dataeval.detectors.drift._uncertainty import DriftUncertainty

dataeval-0.76.0/src/dataeval/detectors/drift/base.py → dataeval-0.81.0/src/dataeval/detectors/drift/_base.py

@@ -10,16 +10,18 @@ from __future__ import annotations
 
 __all__ = []
 
+import math
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
 from functools import wraps
 from typing import Any, Callable, Literal, TypeVar
 
 import numpy as np
-from numpy.typing import ArrayLike, NDArray
+from numpy.typing import NDArray
 
-from dataeval.interop import as_numpy
-from dataeval.output import Output, set_metadata
+from dataeval._output import Output, set_metadata
+from dataeval.typing import Array, ArrayLike
+from dataeval.utils._array import as_numpy, to_numpy
 
 R = TypeVar("R")
 
@@ -46,16 +48,9 @@ class UpdateStrategy(ABC):
 class DriftBaseOutput(Output):
     """
     Base output class for Drift Detector classes
-
-    Attributes
-    ----------
-    is_drift : bool
-        Drift prediction for the images
-    threshold : float
-        Threshold after multivariate correction if needed
     """
 
-    is_drift: bool
+    drifted: bool
     threshold: float
     p_val: float
     distance: float
@@ -64,14 +59,18 @@ class DriftBaseOutput(Output):
 @dataclass(frozen=True)
 class DriftOutput(DriftBaseOutput):
     """
-    Output class for :class:`DriftCVM`, :class:`DriftKS`, and :class:`DriftUncertainty` drift detectors.
+    Output class for :class:`.DriftCVM`, :class:`.DriftKS`, and :class:`.DriftUncertainty` drift detectors.
 
     Attributes
     ----------
-    is_drift : bool
+    drifted : bool
         :term:`Drift` prediction for the images
     threshold : float
         Threshold after multivariate correction if needed
+    p_val : float
+        Instance-level p-value
+    distance : float
+        Instance-level distance
     feature_drift : NDArray
         Feature-level array of images detected to have drifted
     feature_threshold : float
@@ -82,7 +81,7 @@ class DriftOutput(DriftBaseOutput):
         Feature-level distances
     """
 
-    # is_drift: bool
+    # drifted: bool
     # threshold: float
     # p_val: float
     # distance: float
@@ -196,7 +195,7 @@ class BaseDrift:
         if correction not in ["bonferroni", "fdr"]:
             raise ValueError("`correction` must be `bonferroni` or `fdr`.")
 
-        self._x_ref = as_numpy(x_ref)
+        self._x_ref = x_ref
         self.x_ref_preprocessed: bool = x_ref_preprocessed
 
         # Other attributes
@@ -204,25 +203,25 @@ class BaseDrift:
         self.update_x_ref = update_x_ref
         self.preprocess_fn = preprocess_fn
         self.correction = correction
-        self.n: int = len(self._x_ref)
+        self.n: int = len(x_ref)
 
         # Ref counter for preprocessed x
         self._x_refcount = 0
 
     @property
-    def x_ref(self) -> NDArray[Any]:
+    def x_ref(self) -> ArrayLike:
         """
         Retrieve the reference data, applying preprocessing if not already done.
 
         Returns
         -------
-        NDArray
+        ArrayLike
             The reference dataset (`x_ref`), preprocessed if needed.
         """
         if not self.x_ref_preprocessed:
             self.x_ref_preprocessed = True
             if self.preprocess_fn is not None:
-                self._x_ref = as_numpy(self.preprocess_fn(self._x_ref))
+                self._x_ref = self.preprocess_fn(self._x_ref)
 
         return self._x_ref
 
@@ -323,32 +322,44 @@ class BaseDriftUnivariate(BaseDrift):
         # lazy process n_features as needed
         if not isinstance(self._n_features, int):
             # compute number of features for the univariate tests
-            if not isinstance(self.preprocess_fn, Callable) or self.x_ref_preprocessed:
-                # infer features from preprocessed reference data
-                self._n_features = self.x_ref.reshape(self.x_ref.shape[0], -1).shape[-1]
-            else:
-                # infer number of features after applying preprocessing step
-                x = as_numpy(self.preprocess_fn(self._x_ref[0:1]))  # type: ignore
-                self._n_features = x.reshape(x.shape[0], -1).shape[-1]
+            x_ref = (
+                self.x_ref
+                if self.preprocess_fn is None or self.x_ref_preprocessed
+                else self.preprocess_fn(self._x_ref[0:1])
+            )
+            # infer features from preprocessed reference data
+            shape = x_ref.shape if isinstance(x_ref, Array) else as_numpy(x_ref).shape
+            self._n_features = int(math.prod(shape[1:]))  # Multiplies all channel sizes after first
 
         return self._n_features
 
     @preprocess_x
-    @abstractmethod
     def score(self, x: ArrayLike) -> tuple[NDArray[np.float32], NDArray[np.float32]]:
         """
-        Abstract method to calculate feature scores after preprocessing.
+        Calculates p-values and test statistics per feature.
 
         Parameters
         ----------
         x : ArrayLike
-            The batch of data to calculate univariate :term:`drift<Drift>` scores for each feature.
+            Batch of instances
 
         Returns
         -------
         tuple[NDArray, NDArray]
-            A tuple containing p-values and distance :term:`statistics<Statistics>` for each feature.
+            Feature level p-values and test statistics
         """
+        x_np = to_numpy(x)
+        x_np = x_np.reshape(x_np.shape[0], -1)
+        x_ref_np = as_numpy(self.x_ref)
+        x_ref_np = x_ref_np.reshape(x_ref_np.shape[0], -1)
+        p_val = np.zeros(self.n_features, dtype=np.float32)
+        dist = np.zeros_like(p_val)
+        for f in range(self.n_features):
+            dist[f], p_val[f] = self._score_fn(x_ref_np[:, f], x_np[:, f])
+        return p_val, dist
+
+    @abstractmethod
+    def _score_fn(self, x: NDArray[np.float32], y: NDArray[np.float32]) -> tuple[np.float32, np.float32]: ...
 
     def _apply_correction(self, p_vals: NDArray) -> tuple[bool, float]:
         """