Implementation:Huggingface Datasets Dataset From Dict

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

Overview

Concrete tool for creating a Dataset from a Python dictionary provided by the HuggingFace Datasets library.

Description

Dataset.from_dict is a class method that converts a Python dictionary of column-name-to-values mappings into an Apache Arrow backed Dataset. Each dictionary key becomes a column name and each value (a list or Arrow array) provides the column data. If an explicit Features schema is supplied, columns are encoded and cast accordingly; otherwise, types are inferred. The resulting dataset lives in memory and has no associated cache directory.

Usage

Use Dataset.from_dict when you have data organized as a dictionary of lists (columnar format) and want to create a Dataset without going through file I/O. This is the standard entry point for programmatic dataset construction.

Code Reference

Source Location

• Repository: datasets
• File: src/datasets/arrow_dataset.py
• Lines: 973-1034

Signature

@classmethod
def from_dict(
    cls,
    mapping: dict,
    features: Optional[Features] = None,
    info: Optional[DatasetInfo] = None,
    split: Optional[NamedSplit] = None,
) -> "Dataset":

Import

from datasets import Dataset

I/O Contract

Inputs

Name	Type	Required	Description
mapping	dict	Yes	Mapping of column names (strings) to Arrays or Python lists of values.
features	Features	No	Explicit dataset features schema. If provided, data is cast to match.
info	DatasetInfo	No	Dataset metadata such as description, citation, etc.
split	NamedSplit	No	Name of the dataset split (e.g., "train", "test").

Outputs

Name	Type	Description
return	Dataset	A new in-memory Dataset backed by an Arrow table.

Usage Examples

Basic Usage

from datasets import Dataset

ds = Dataset.from_dict({
    "text": ["Hello world", "Goodbye world"],
    "label": [1, 0],
})
print(ds)
# Dataset({
#     features: ['text', 'label'],
#     num_rows: 2
# })

With Explicit Features

from datasets import Dataset, Features, Value, ClassLabel

features = Features({
    "text": Value("string"),
    "label": ClassLabel(names=["negative", "positive"]),
})

ds = Dataset.from_dict(
    {"text": ["Hello", "World"], "label": [0, 1]},
    features=features,
)

Related Pages

Implements Principle

• Principle:Huggingface_Datasets_Dataset_From_Dict_Construction