metadata
dataset_info:
- config_name: ArXiv
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 105393748
num_examples: 2212
- name: test
num_bytes: 33002660
num_examples: 745
download_size: 128240714
dataset_size: 138396408
- config_name: BookCorpus2
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 92009676
num_examples: 257
- name: test
num_bytes: 2769179
num_examples: 8
download_size: 112489690
dataset_size: 94778855
- config_name: Books3
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 108880516
num_examples: 197
- name: test
num_bytes: 41526969
num_examples: 87
download_size: 177202552
dataset_size: 150407485
- config_name: DM Mathematics
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 104858342
num_examples: 12735
- name: test
num_bytes: 4941242
num_examples: 601
download_size: 55363453
dataset_size: 109799584
- config_name: Enron Emails
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 16701351
num_examples: 9378
- name: test
num_bytes: 541500
num_examples: 291
download_size: 19423900
dataset_size: 17242851
- config_name: EuroParl
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 94200121
num_examples: 1334
- name: test
num_bytes: 3629223
num_examples: 42
download_size: 106611978
dataset_size: 97829344
- config_name: FreeLaw
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 106892914
num_examples: 6519
- name: test
num_bytes: 25162182
num_examples: 1588
download_size: 136641886
dataset_size: 132055096
- config_name: Github
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 103649709
num_examples: 19216
- name: test
num_bytes: 31333794
num_examples: 5651
download_size: 100372522
dataset_size: 134983503
- config_name: Gutenberg (PG-19)
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 102759663
num_examples: 256
- name: test
num_bytes: 8678842
num_examples: 21
download_size: 137454348
dataset_size: 111438505
- config_name: HackerNews
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 79788458
num_examples: 15917
- name: test
num_bytes: 2540857
num_examples: 493
download_size: 103554428
dataset_size: 82329315
- config_name: NIH ExPorter
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 39126793
num_examples: 18002
- name: test
num_bytes: 1222934
num_examples: 557
download_size: 43770700
dataset_size: 40349727
- config_name: OpenSubtitles
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 104735849
num_examples: 3389
- name: test
num_bytes: 6108121
num_examples: 199
download_size: 130099164
dataset_size: 110843970
- config_name: OpenWebText2
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 106039804
num_examples: 26421
- name: test
num_bytes: 39817479
num_examples: 10125
download_size: 178103310
dataset_size: 145857283
- config_name: PhilPapers
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 47970203
num_examples: 647
- name: test
num_bytes: 1538027
num_examples: 21
download_size: 52345194
dataset_size: 49508230
- config_name: Pile-CC
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 103171504
num_examples: 23300
- name: test
num_bytes: 72105538
num_examples: 16421
download_size: 208522136
dataset_size: 175277042
- config_name: PubMed Abstracts
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 104936477
num_examples: 76172
- name: test
num_bytes: 12600783
num_examples: 9185
download_size: 129554148
dataset_size: 117537260
- config_name: PubMed Central
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 102994643
num_examples: 3273
- name: test
num_bytes: 56626731
num_examples: 1779
download_size: 148298044
dataset_size: 159621374
- config_name: StackExchange
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 103761034
num_examples: 46841
- name: test
num_bytes: 20466176
num_examples: 9247
download_size: 129854838
dataset_size: 124227210
- config_name: UPSTO Backgrounds
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 104750819
num_examples: 24632
- name: test
num_bytes: 14799000
num_examples: 3484
download_size: 113457922
dataset_size: 119549819
- config_name: USPTO Backgrounds
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 104750819
num_examples: 24632
- name: test
num_bytes: 14799000
num_examples: 3484
download_size: 113457922
dataset_size: 119549819
- config_name: Ubuntu IRC
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 105150097
num_examples: 203
- name: test
num_bytes: 324716
num_examples: 7
download_size: 111839542
dataset_size: 105474813
- config_name: Wikipedia (en)
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 105309044
num_examples: 34309
- name: test
num_bytes: 16269639
num_examples: 5305
download_size: 140418662
dataset_size: 121578683
- config_name: YoutubeSubtitles
features:
- name: text
dtype: string
- name: meta
struct:
- name: pile_set_name
dtype: string
- name: domain
dtype: string
splits:
- name: train
num_bytes: 79086408
num_examples: 3322
- name: test
num_bytes: 1373528
num_examples: 103
download_size: 99103806
dataset_size: 80459936
configs:
- config_name: ArXiv
data_files:
- split: train
path: ArXiv/train-*
- split: test
path: ArXiv/test-*
- config_name: BookCorpus2
data_files:
- split: train
path: BookCorpus2/train-*
- split: test
path: BookCorpus2/test-*
- config_name: Books3
data_files:
- split: train
path: Books3/train-*
- split: test
path: Books3/test-*
- config_name: DM Mathematics
data_files:
- split: train
path: DM Mathematics/train-*
- split: test
path: DM Mathematics/test-*
- config_name: Enron Emails
data_files:
- split: train
path: Enron Emails/train-*
- split: test
path: Enron Emails/test-*
- config_name: EuroParl
data_files:
- split: train
path: EuroParl/train-*
- split: test
path: EuroParl/test-*
- config_name: FreeLaw
data_files:
- split: train
path: FreeLaw/train-*
- split: test
path: FreeLaw/test-*
- config_name: Github
data_files:
- split: train
path: Github/train-*
- split: test
path: Github/test-*
- config_name: Gutenberg (PG-19)
data_files:
- split: train
path: Gutenberg (PG-19)/train-*
- split: test
path: Gutenberg (PG-19)/test-*
- config_name: HackerNews
data_files:
- split: train
path: HackerNews/train-*
- split: test
path: HackerNews/test-*
- config_name: NIH ExPorter
data_files:
- split: train
path: NIH ExPorter/train-*
- split: test
path: NIH ExPorter/test-*
- config_name: OpenSubtitles
data_files:
- split: train
path: OpenSubtitles/train-*
- split: test
path: OpenSubtitles/test-*
- config_name: OpenWebText2
data_files:
- split: train
path: OpenWebText2/train-*
- split: test
path: OpenWebText2/test-*
- config_name: PhilPapers
data_files:
- split: train
path: PhilPapers/train-*
- split: test
path: PhilPapers/test-*
- config_name: Pile-CC
data_files:
- split: train
path: Pile-CC/train-*
- split: test
path: Pile-CC/test-*
- config_name: PubMed Abstracts
data_files:
- split: train
path: PubMed Abstracts/train-*
- split: test
path: PubMed Abstracts/test-*
- config_name: PubMed Central
data_files:
- split: train
path: PubMed Central/train-*
- split: test
path: PubMed Central/test-*
- config_name: StackExchange
data_files:
- split: train
path: StackExchange/train-*
- split: test
path: StackExchange/test-*
- config_name: UPSTO Backgrounds
data_files:
- split: train
path: UPSTO Backgrounds/train-*
- split: test
path: UPSTO Backgrounds/test-*
- config_name: USPTO Backgrounds
data_files:
- split: train
path: USPTO Backgrounds/train-*
- split: test
path: USPTO Backgrounds/test-*
- config_name: Ubuntu IRC
data_files:
- split: train
path: Ubuntu IRC/train-*
- split: test
path: Ubuntu IRC/test-*
- config_name: Wikipedia (en)
data_files:
- split: train
path: Wikipedia (en)/train-*
- split: test
path: Wikipedia (en)/test-*
- config_name: YoutubeSubtitles
data_files:
- split: train
path: YoutubeSubtitles/train-*
- split: test
path: YoutubeSubtitles/test-*
Smaller version of armelr's dataset, which is a a split version of the general text dataset, "The Pile".
Features of this version
- This version has a both a train + test set
- Is easily downloadable in ~2.3GB
- Can choose text split
If you want a small mixes set, look at minipile instead.