---
source: https://iinventedlight.com/library/notes/hf-datasets-create-dataset.md
title: "Create a Hugging Face dataset from your own files"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
source_license: "Apache-2.0 (huggingface/datasets docs); restated in our own words"
attribution: "Summary by iinventedlight (AI-assisted). Sources: Hugging Face (Datasets docs), Create a dataset · Hugging Face, Apache-2.0 (huggingface/datasets docs); Hugging Face, Load, Apache-2.0 (huggingface/datasets docs)."
author: "iinventedlight (synthesis of cited sources); text drafted with AI assistance"
date: 2026-11-04
status: published
checked_on: 2026-11-05
as_of: 2026-11-05
confidence: medium
untested: "No code was run; only the fetched Hugging Face documentation pages were read."
category: huggingface-datasets
topic_area: "datasets: creating"
tags: huggingface, datasets, imagefolder, audiofolder
sources:
  - id: S1
    title: "Create a dataset · Hugging Face"
    publisher: Hugging Face (Datasets docs)
    url: https://huggingface.co/docs/datasets/create_dataset
    accessed: 2026-11-05
    license: "Apache-2.0 (huggingface/datasets docs)"
  - id: S2
    title: "Load"
    publisher: Hugging Face
    url: https://huggingface.co/docs/datasets/loading
    accessed: 2026-11-05
    license: "Apache-2.0 (huggingface/datasets docs)"
    license: "Apache-2.0 (huggingface/datasets docs)"
---
# Hugging Face Datasets: create a dataset

## Short answer
Build datasets from local folders (CSV/JSON/Parquet/text), from generators, or with ImageFolder/AudioFolder builders. [S1][S2]

## Points
- Folder builders are low-code ways to create image or audio sets. [S1]
- `Dataset.from_generator` supports data larger than memory. [S2]
- You can also load from the Hub, local files, or in-memory objects. [S2]

## Claims
- id: C1
  claim: Folder builders include ImageFolder and AudioFolder.
  source: S1
  evidence: "ImageFolder and AudioFolder"
- id: C2
  claim: Common formats include csv, json/jsonl, parquet, txt.
  source: S1
  evidence: "formats such as csv , json/jsonl , parquet , txt"
- id: C3
  claim: from_generator supports data larger than memory.
  source: S2
  evidence: "This approach supports loading data larger than available memory"
- id: C4
  claim: Datasets can load from the Hub or local files.
  source: S2
  evidence: "load a dataset from: The Hugging Face Hub Local files"

## Sources
- [S1] Create a dataset · Hugging Face. Hugging Face (Datasets docs). https://huggingface.co/docs/datasets/create_dataset . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-05.
- [S2] Load. Hugging Face. https://huggingface.co/docs/datasets/loading . License: Apache-2.0 (huggingface/datasets docs). Accessed 2026-11-05.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).

Source notice: Hugging Face Datasets documentation, Apache License 2.0 (https://github.com/huggingface/datasets/blob/main/LICENSE).