Join TensorFlow at Google I/O, May 11-12 Register now

c4

  • Description:

A colossal, cleaned version of Common Crawl's web crawl corpus.

Based on Common Crawl dataset: https://commoncrawl.org

To generate this dataset, please follow the instructions from t5.

Due to the overhead of cleaning the dataset, it is recommend you prepare it with a distributed service like Cloud Dataflow. More info at https://www.tensorflow.org/datasets/beam_datasets

FeaturesDict({
    'content-length': Text(shape=(), dtype=tf.string),
    'content-type': Text(shape=(), dtype=tf.string),
    'text': Text(shape=(), dtype=tf.string),
    'timestamp': Text(shape=(), dtype=tf.string),
    'url': Text(shape=(), dtype=tf.string),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
content-length Text tf.string
content-type Text tf.string
text Text tf.string
timestamp Text tf.string
url Text tf.string 
@article{2019t5,
  author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
  title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
  journal = {arXiv e-prints},
  year = {2019},
  archivePrefix = {arXiv},
  eprint = {1910.10683},
}

c4/en (default config)

  • Config description: English C4 dataset.

  • Download size: 12.28 MiB

  • Dataset size: 806.92 GiB

  • Splits:

Split Examples
'train' 364,868,901
'validation' 364,608

c4/en.noclean

  • Config description: Disables all cleaning (deduplication, removal based on bad words, etc.)

  • Download size: 12.25 MiB

  • Dataset size: 6.21 TiB

  • Splits:

Split Examples
'train' 1,063,805,324
'validation' 1,065,029

c4/realnewslike

  • Config description: Filters from the default config to only include content from the domains used in the 'RealNews' dataset (Zellers et al., 2019).

  • Download size: 12.41 MiB

  • Dataset size: 36.89 GiB

  • Splits:

Split Examples
'train' 13,799,838
'validation' 13,863

c4/webtextlike

  • Config description: Filters from the default config to only include content from the URLs in OpenWebText (https://github.com/jcpeterson/openwebtext).

  • Download size: 14.12 MiB

  • Dataset size: 18.00 GiB

  • Splits:

Split Examples
'train' 4,500,788
'validation' 4,493

c4/multilingual

  • Config description: Multilingual C4 (mC4) has 101 languages and is generated from 71 Common Crawl dumps.

  • Download size: 22.74 MiB

  • Dataset size: 26.76 TiB

  • Splits:

Split Examples
'af' 2,152,243
'af-validation' 2,118
'am' 162,870
'am-validation' 155
'ar' 53,256,040
'ar-validation' 52,978
'az' 5,285,720
'az-validation' 5,239
'be' 1,742,030
'be-validation' 1,712
'bg' 23,409,799
'bg-Latn' 162,461
'bg-Latn-validation' 144
'bg-validation' 23,503
'bn' 7,444,098
'bn-validation' 7,415
'ca' 14,492,899
'ca-validation' 14,489
'ceb' 351,894
'ceb-validation' 367
'co' 494,913
'co-validation' 565
'cs' 60,149,680
'cs-validation' 60,462
'cy' 4,131,915
'cy-validation' 4,103
'da' 28,777,331
'da-validation' 28,945
'de' 397,006,993
'de-validation' 398,583
'el' 41,753,736
'el-Latn' 449,943
'el-Latn-validation' 468
'el-validation' 42,358
'en' 3,079,081,989
'en-validation' 3,083,850
'eo' 500,048
'eo-validation' 496
'es' 416,057,992
'es-validation' 416,256
'et' 6,941,360
'et-validation' 6,848
'eu' 1,555,887
'eu-validation' 1,580
'fa' 53,927,287
'fa-validation' 53,685
'fi' 26,842,650
'fi-validation' 26,710
'fil' 2,102,197
'fil-validation' 2,158
'fr' 332,674,575
'fr-validation' 331,328
'fy' 1,104,359
'fy-validation' 1,094
'ga' 465,670
'ga-validation' 490
'gd' 322,404
'gd-validation' 338
'gl' 4,549,465
'gl-validation' 4,631
'gu' 631,600
'gu-validation' 651
'ha' 247,479
'ha-validation' 258
'haw' 84,312
'haw-validation' 86
'hi' 18,507,273
'hi-Latn' 626,154
'hi-Latn-validation' 638
'hi-validation' 18,392
'hmn' 295,549
'hmn-validation' 312
'ht' 269,174
'ht-validation' 281
'hu' 36,819,508
'hu-validation' 36,756
'hy' 2,401,949
'hy-validation' 2,410
'id' 69,625,551
'id-validation' 69,739
'ig' 92,909
'ig-validation' 87
'is' 2,069,293
'is-validation' 2,065
'it' 186,404,508
'it-validation' 186,030
'iw' 12,334,609
'iw-validation' 12,207
'ja' 87,337,884
'ja-Latn' 533,516
'ja-Latn-validation' 506
'ja-validation' 87,420
'jv' 581,528
'jv-validation' 609
'ka' 2,295,551
'ka-validation' 2,279
'kk' 2,392,401
'kk-validation' 2,400
'km' 756,612
'km-validation' 745
'kn' 1,056,849
'kn-validation' 1,039
'ko' 15,602,947
'ko-validation' 15,771
'ku' 298,389
'ku-validation' 298
'ky' 995,539
'ky-validation' 976
'la' 1,674,463
'la-validation' 1,654
'lb' 2,740,336
'lb-validation' 2,692
'lo' 141,776
'lo-validation' 145
'lt' 11,274,295
'lt-validation' 11,245
'lv' 6,414,223
'lv-validation' 6,598
'mg' 345,040
'mg-validation' 367
'mi' 101,169
'mi-validation' 106
'mk' 2,058,417
'mk-validation' 2,054
'ml' 2,044,981
'ml-validation' 2,002
'mn' 2,054,674
'mn-validation' 2,090
'mr' 7,774,331
'mr-validation' 7,928
'ms' 13,180,647
'ms-validation' 13,391
'mt' 2,261,303
'mt-validation' 2,322
'my' 813,530
'my-validation' 858
'ne' 2,942,785
'ne-validation' 2,951
'nl' 96,210,458
'nl-validation' 96,637
'no' 25,402,139
'no-validation' 25,766
'ny' 174,696
'ny-validation' 162
'pa' 363,399
'pa-validation' 346
'pl' 126,164,277
'pl-validation' 125,997
'ps' 335,452
'ps-validation' 318
'pt' 169,239,084
'pt-validation' 169,417
'ro' 45,738,857
'ro-validation' 45,512
'ru' 755,585,265
'ru-Latn' 745,491
'ru-Latn-validation' 753
'ru-validation' 756,418
'sd' 743,057
'sd-validation' 774
'si' 534,759
'si-validation' 509
'sk' 17,729,698
'sk-validation' 17,865
'sl' 8,499,456
'sl-validation' 8,504
'sm' 98,467
'sm-validation' 108
'sn' 326,392
'sn-validation' 306
'so' 893,012
'so-validation' 888
'sq' 4,113,147
'sq-validation' 4,086
'sr' 3,398,483
'sr-validation' 3,443
'st' 66,837
'st-validation' 88
'su' 280,719
'su-validation' 269
'sv' 48,570,979
'sv-validation' 48,633
'sw' 985,654
'sw-validation' 994
'ta' 3,514,561
'ta-validation' 3,510
'te' 1,188,243
'te-validation' 1,211
'tg' 1,280,757
'tg-validation' 1,259
'th' 15,463,131
'th-validation' 15,344
'tr' 87,595,290
'tr-validation' 87,596
'uk' 38,556,465
'uk-validation' 38,550
'und' 1,866,266,695
'und-validation' 1,867,450
'ur' 1,950,124
'ur-validation' 1,885
'uz' 796,416
'uz-validation' 847
'vi' 78,587,159
'vi-validation' 78,611
'xh' 69,048
'xh-validation' 62
'yi' 143,708
'yi-validation' 161
'yo' 46,214
'yo-validation' 42
'zh' 54,542,308
'zh-Latn' 373,664
'zh-Latn-validation' 387
'zh-validation' 54,656
'zu' 555,458
'zu-validation' 548