README.md

metadata

language:
  - en
license: cc-by-sa-3.0
size_categories:
  - n<1K
  - 1K<n<10K
  - 10K<n<100K
dataset_info:
  - config_name: '100'
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: partial
        num_bytes: 98031
        num_examples: 100
      - name: full
        num_bytes: 315241.0851032817
        num_examples: 100
    download_size: 839250
    dataset_size: 413272.0851032817
  - config_name: 100k
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: partial
        num_bytes: 171032873
        num_examples: 100000
      - name: full
        num_bytes: 315241085.10328174
        num_examples: 100000
    download_size: 580606890
    dataset_size: 486273958.10328174
  - config_name: 10k
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: partial
        num_bytes: 17059273
        num_examples: 10000
      - name: full
        num_bytes: 31524108.51032817
        num_examples: 10000
    download_size: 58371936
    dataset_size: 48583381.51032817
  - config_name: 1k
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: partial
        num_bytes: 1007863
        num_examples: 1000
      - name: full
        num_bytes: 3152410.8510328173
        num_examples: 1000
    download_size: 8616768
    dataset_size: 4160273.8510328177
  - config_name: 50k
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: partial
        num_bytes: 85492041
        num_examples: 50000
      - name: full
        num_bytes: 157620542.55164087
        num_examples: 50000
    download_size: 289381080
    dataset_size: 243112583.55164087
  - config_name: 5k
    features:
      - name: id
        dtype: string
      - name: url
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: full
        num_bytes: 15762054.255164085
        num_examples: 5000
      - name: partial
        num_bytes: 5082253
        num_examples: 5000
    download_size: 32285884
    dataset_size: 20844307.255164087
configs:
  - config_name: '100'
    data_files:
      - split: full
        path: 100/full-*
      - split: partial
        path: 100/partial-*
  - config_name: 100k
    data_files:
      - split: full
        path: 100k/full-*
      - split: partial
        path: 100k/partial-*
  - config_name: 10k
    data_files:
      - split: full
        path: 10k/full-*
      - split: partial
        path: 10k/partial-*
  - config_name: 1k
    data_files:
      - split: full
        path: 1k/full-*
      - split: partial
        path: 1k/partial-*
  - config_name: 50k
    data_files:
      - split: full
        path: 50k/full-*
      - split: partial
        path: 50k/partial-*
  - config_name: 5k
    data_files:
      - split: full
        path: 5k/full-*
      - split: partial
        path: 5k/partial-*

mini_wiki

This is a sampled version of the wikimedia/wikipedia dataset. This repository provides scripts to generate the 100, 1k, 5k, 10k, 50k, 100k samples of the dataset, based on the "20231101.en" version.

Usage

There are two possible splits: full, which loads the entire article, or partial, which is a version that only contains the first 500 words of each article. It is recommended to use partial if you are performing retrieval and are only interested in the first paragraphs of the Wikipedia dataset.

Now, to run:

from datasets import load_dataset

# Load the 100-sample, full version of the dataset:
data = load_dataset('xhluca/mini_wiki', name="100", split="full")

print(data)

# Load partial version with 1k, 5k, 10k, 50k, 100k samples
data = load_dataset('xhluca/mini_wiki', name="1k", split="partial")
data = load_dataset('xhluca/mini_wiki', name="5k", split="partial")
data = load_dataset('xhluca/mini_wiki', name="10k", split="partial")
data = load_dataset('xhluca/mini_wiki', name="50k", split="partial")
data = load_dataset('xhluca/mini_wiki', name="100k", split="partial")