Implementation:Huggingface Datasets Value

Knowledge Sources	Huggingface Datasets HF Datasets Docs
Domains	Data_Engineering, NLP
Last Updated	2026-02-14 18:00 GMT

Overview

Concrete tool for representing scalar data types in dataset schemas provided by the HuggingFace Datasets library.

Description

Value is a dataclass that represents a scalar feature value of a particular data type. It accepts a dtype string at construction and converts it to the corresponding PyArrow type. Supported dtypes include: null, bool, int8-int64, uint8-uint64, float16-float64, string, large_string, string_view, binary, large_binary, binary_view, date32, date64, time32, time64, timestamp, duration, decimal128, and decimal256. Aliases "float" and "double" are normalized to "float32" and "float64" respectively.

Usage

Use Value to define the type of any scalar column in a Features schema. It is the most commonly used feature type.

Code Reference

Source Location

• Repository: datasets
• File: src/datasets/features/features.py
• Lines: 490-573

Signature

@dataclass
class Value:
    dtype: str
    id: Optional[str] = field(default=None, repr=False)
    # Automatically constructed
    pa_type: ClassVar[Any] = None
    _type: str = field(default="Value", init=False, repr=False)

Import

from datasets import Value

I/O Contract

Inputs

Name	Type	Required	Description
dtype	str	Yes	Name of the data type (e.g., "int32", "string", "float64").
id	str	No	Optional identifier for the feature.

Outputs

Name	Type	Description
instance	Value	A Value feature with the corresponding PyArrow type set in pa_type.

Usage Examples

Basic Usage

from datasets import Features, Value

features = Features({
    "stars": Value("int32"),
    "text": Value("string"),
    "score": Value("float64"),
})
print(features)
# {'stars': Value('int32'), 'text': Value('string'), 'score': Value('float64')}

Related Pages

Implements Principle

• Principle:Huggingface_Datasets_Scalar_Value_Types