Environment:Cleanlab Cleanlab Datalab Dependencies

Knowledge Sources	cleanlab setup.py extras
Domains	Data_Centric_AI, Dataset_Auditing
Last Updated	2026-02-09 19:30 GMT

Overview

Optional dependency environment extending the core cleanlab install with the HuggingFace `datasets` package, required for Datalab automated dataset auditing.

Description

The Datalab module provides automated multi-issue detection across a dataset (label errors, outliers, duplicates, class imbalance, non-IID violations, null values, underperforming groups, data valuation). It requires the HuggingFace `datasets` library (>= 2.7.0) for its internal data storage and type system. When the `datasets` package is not installed, Datalab is replaced with a `DatalabUnavailable` stub that raises `ImportError` with installation instructions on any access attempt.

Usage

Use this environment when running the Datalab automated dataset audit workflow. Install via the `[datalab]` extras group. This is the prerequisite for the Datalab_Init, Datalab_Find_Issues, Datalab_Report, Datalab_Get_Issues, and Datalab_Get_Issue_Summary implementations.

System Requirements

Category	Requirement	Notes
OS	Linux, macOS, or Windows	Same as core cleanlab
Hardware	CPU	No GPU required
Python	>= 3.10	Same as core cleanlab
Disk	Moderate	HuggingFace datasets may cache data to disk

Dependencies

System Packages

No additional system-level packages required beyond core cleanlab.

Python Packages

All core cleanlab dependencies (see Environment:Cleanlab_Cleanlab_Python_Core_Environment)
`datasets` >= 2.7.0

Credentials

No credentials required. HuggingFace datasets can load local or public datasets without authentication.

Quick Install

pip install 'cleanlab[datalab]'

Code Evidence

Datalab import gating from `cleanlab/__init__.py:32-41`:

def _datalab_import_factory():
    try:
        from .datalab.datalab import Datalab as _Datalab
        return _Datalab
    except ImportError:
        return DatalabUnavailable(
            "Datalab is not available due to missing dependencies. "
            "To install Datalab, run `pip install 'cleanlab[datalab]'`."
        )

Datasets package import check from `cleanlab/datalab/internal/data.py:8-15`:

try:
    import datasets
except ImportError as error:
    raise ImportError(
        "Cannot import datasets package. "
        "Please install it and try again, or just install cleanlab with "
        "all optional dependencies via: `pip install 'cleanlab[all]'`"
    ) from error

Datasets 4.0.0+ compatibility handling from `cleanlab/datalab/internal/data.py:22-29`:

# Import Column types for compatibility with datasets 4.0.0+
try:
    from datasets.arrow_dataset import Column
    from datasets.iterable_dataset import IterableColumn
except ImportError:
    # For backwards compatibility with older datasets versions
    Column = None
    IterableColumn = None

Optional dependency definition from `setup.py:24-28`:

DATALAB_REQUIRE = [
    # Mainly for Datalab's data storage class.
    # Still some type hints that require datasets
    "datasets>=2.7.0",
]

Common Errors

Error Message	Cause	Solution
`Datalab is not available due to missing dependencies`	`datasets` package not installed	`pip install 'cleanlab[datalab]'`
`IssueManager is not available due to missing dependencies for Datalab`	`datasets` package not installed	`pip install 'cleanlab[datalab]'`
`Cannot import datasets package`	Attempting to use Datalab internals without `datasets`	`pip install 'cleanlab[all]'`

Compatibility Notes

datasets >= 4.0.0: Introduced new `Column` and `IterableColumn` types. Cleanlab handles both old and new versions gracefully via try/except import.
datasets < 2.7.0: Not supported. May cause subtle type or API errors.

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment