Datasets:

Butanium
/

femto-fineweb

license: odc-by
language:
  - en
tags:
  - debug
  - fineweb
  - sample
size_categories:
  - n<10
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
dataset_info:
  features:
    - name: text
      dtype: string
    - name: id
      dtype: string
    - name: dump
      dtype: string
    - name: url
      dtype: string
    - name: date
      dtype: string
    - name: file_path
      dtype: string
    - name: language
      dtype: string
    - name: language_score
      dtype: float64
    - name: token_count
      dtype: int64
  splits:
    - name: train
      num_bytes: 25549
      num_examples: 10
    - name: validation
      num_bytes: 14712
      num_examples: 8
  download_size: 42891
  dataset_size: 40261

femto-fineweb

A tiny 10-sample subset of FineWeb designed specifically for debugging purposes.

Purpose

This dataset contains only the first 10 samples from the FineWeb dataset, making it ideal for:

Quick debugging of data pipelines
Testing code without downloading large datasets
Rapid prototyping and development
CI/CD testing

Dataset Structure

The dataset has the same structure as the original FineWeb dataset:

text: The text content
id: Unique identifier
dump: CommonCrawl dump identifier
url: Source URL
date: Dump date
file_path: Path in the dump
language: Language code
language_score: Language detection confidence
token_count: Number of tokens

Usage

from datasets import load_dataset

dataset = load_dataset("Butanium/femto-fineweb", split="train")
print(f"Dataset size: {len(dataset)} samples")

Source

This dataset is derived from FineWeb (sample-10BT) and inherits its ODC-BY license.

Citation

If you use FineWeb in your research, please cite:

@software{penedo2024fineweb,
  author = {Penedo, Guilherme and Kydlíček, Hynek and Cappelli, Anton and Wolf, Thomas and Sasko, Mario},
  title = {FineWeb: decanting the web for the finest text data at scale},
  month = May,
  year = 2024,
  url = {https://huggingface.co/datasets/HuggingFaceFW/fineweb}
}