Dataset Viewer
Auto-converted to Parquet Duplicate
text
string
id
string
dump
string
url
string
date
string
file_path
string
offset
int64
token_count
int64
language
string
page_average_lid
string
page_average_lid_score
float64
full_doc_lid
string
full_doc_lid_score
float64
per_page_languages
list
is_truncated
bool
extractor
string
page_ends
list
fw_edu_scores
list
fw_edu_v2_scores
list
dclm_scores
list
ocr_quality_scores
list
minhash_cluster_size
int64
duplicate_count
int64
embedding_tokens
int32
embedding_source_file
string
embedding_source_row
int64
selection_row_id
int64
selection_cluster
int32
selection_max_cosine
float32
selection_nearest_row_id
int64
canonical_tokens
int64
mixture_id
string
mixture_component
string
mixture_source_dataset
string
mixture_source_revision
string
mixture_source_file
string
mixture_source_row
int64
mixture_original_row_id
int64
mixture_cluster
int32
mixture_text_sha256
string
COLONEL HARRY An Original Screenplay by Johnny B. Campbell FADE IN INT. AG PLANE – DAY We fly over some trees, just missing them, and then drop down almost to the ground and find power lines and a fence right in front of us. Just missing the fence, we drop down again as if to land on the crops but we're going way...
<urn:uuid:b846080c-89f0-4e7f-b6c2-11f7cf68e599>
CC-MAIN-2017-13
http://www.jbcampbellextremismonline.com/uploads/8/5/1/3/8513891/colonel_harry.pdf
2017-03-27T10:37:30Z
crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00225-ip-10-233-31-227.ec2.internal.warc.gz
553,812,196
29,557
eng_Latn
eng_Latn
0.997798
eng_Latn
0.998836
[ "unknown", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn", "eng_Latn...
false
docling
[ 59, 1778, 3095, 4190, 5130, 6425, 7743, 8732, 9700, 10597, 11684, 12909, 14063, 14985, 16211, 17398, 18348, 20178, 21273, 22500, 23505, 24786, 26031, 27314, 28808, 29702, 30654, 31965, 32934, 33898, 35024, 36076, 37083, 38167, 39330, 40494, 415...
[ 0.53125, 0.62890625 ]
[ 0.38671875, 0.5625 ]
[ 0.83203125, 0.85546875 ]
[ 2.4375, 2.703125 ]
2
2
30,591
data/train/source-000001/part-000.parquet
10
30
476
0.886382
2,981,099
30,591
4c2de01beafadc8b9145e87e0e3c60b1a13c23bb6fe172047b35adfe0ae0d659
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
0
30
476
043bbbe8f054db0040dbe602cde8fb27ddcfcc594c75197ff1d3cafbbc1e981f
"RESEARCH INTO MID-SIZE BUSINESS GROWTH\n\nNOVEMBER 2011\n\nResearch into Mid-Size Business Growth\n(...TRUNCATED)
<urn:uuid:e41cce9f-7467-49d7-981e-d07843cdbcf9>
CC-MAIN-2017-13
https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/32109/11-1409-research-mid-size-business-growth.pdf
2017-03-27T10:45:30Z
"crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00224-ip-10-233-31(...TRUNCATED)
894,442,336
15,319
eng_Latn
eng_Latn
0.976038
eng_Latn
0.999085
["unknown","eng_Latn","unknown","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","e(...TRUNCATED)
false
docling
[53,453,463,2381,5113,7621,9881,11944,14612,16884,19478,22856,24690,28062,31024,32297,35430,38512,41(...TRUNCATED)
[ 0.79296875, 0.77734375 ]
[ 3.609375, 3.78125 ]
[ 2.828125, 2.921875 ]
[ 2.265625, 1.953125 ]
2
9
16,048
data/train/source-000001/part-000.parquet
15
35
509
0.909773
7,411,749
16,048
f85d668440e2378278ac2de18adadaeb8560198a24b1cb8b72eed9845b7bb8f7
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
1
35
509
93acbd7b8e703538485e7a09c64d25cd58c11f4d9adcfc5e83bf1a5c82d33c52
"Pittsburg State University\n\nPittsburg State University Digital Commons\n\nElectronic Thesis Colle(...TRUNCATED)
<urn:uuid:556529a4-1558-4d4b-8cab-8a0f0d6a27ce>
CC-MAIN-2017-13
http://digitalcommons.pittstate.edu/cgi/viewcontent.cgi?article=1039&context=etd
2017-03-27T10:46:14Z
"crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00229-ip-10-233-31(...TRUNCATED)
94,352,974
32,616
eng_Latn
eng_Latn
0.981032
eng_Latn
0.997317
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","unknown","eng_Latn","eng_Latn","eng_Latn","eng_Latn","(...TRUNCATED)
false
docling
[903,1222,1787,2911,2936,4239,6140,7880,9695,11528,12887,14046,15853,17563,19361,21148,23037,24841,2(...TRUNCATED)
[ 1.078125, 1.171875 ]
[ 4.71875, 4.5625 ]
[ 3.09375, 2.515625 ]
[ 2.578125, 2.703125 ]
3
1
33,403
data/train/source-000001/part-000.parquet
31
51
127
0.880666
1,356,464
33,403
e95bdce5fda56359a77f8c8276526f6cc851668b102b3d6b255ccb04f3c20322
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
2
51
127
f73aed85071bfd0c06c1e66c896d75579dc4cb621e06239fc0480c0651ded7d9
"ANNUAL REPORT 2011\n\nTABLE OF CONTENTS\n\nPublication date: 1 November 2012\n\n© 2012 Green Resou(...TRUNCATED)
<urn:uuid:565b5404-8e22-4226-87e8-c8fdb6b815bb>
CC-MAIN-2017-13
http://greenresources.no/portals/0/pdf/AnnualReport2011.pdf
2017-03-27T10:41:56Z
"crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00230-ip-10-233-31(...TRUNCATED)
163,011,206
29,064
eng_Latn
eng_Latn
0.825523
eng_Latn
0.997021
["unknown","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","(...TRUNCATED)
false
docling
[18,4102,11691,16624,23058,28354,31394,31635,31933,33924,34188,41345,46985,54075,61562,63094,69581,7(...TRUNCATED)
[ 1.0859375, 0.65625 ]
[ 2.890625, 1.28125 ]
[ 2.046875, 1.515625 ]
[ 2.390625, 1.8125 ]
2
1
31,096
data/train/source-000001/part-000.parquet
32
52
158
0.900639
3,812,606
31,096
94bd7fe61605e3c30a2120acf0638709f5bf39bba17a60913c2f3dd0af5beabd
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
3
52
158
6fc92d2aab86912e672b83848aed1c7f989f078c3ed5ede8f0d2c0e95ee73bf5
"Chairmanship of Parliamentary Committees in Some Selected Places\n\n13 May 2003\n\nPrepared by\n\nC(...TRUNCATED)
<urn:uuid:fccb294b-6d23-4a07-a14c-fc44aba255a7>
CC-MAIN-2017-13
http://www.legco.gov.hk/yr02-03/english/sec/library/0203rp06e.pdf
2017-03-27T10:39:31Z
"crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00234-ip-10-233-31(...TRUNCATED)
585,387,134
15,119
eng_Latn
eng_Latn
0.953187
eng_Latn
0.994887
["eng_Latn","unknown","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","(...TRUNCATED)
false
docling
[360,370,1158,5206,6771,7001,8698,11751,14085,16369,18760,21353,23698,25635,27352,28341,30293,32512,(...TRUNCATED)
[ 1.1328125, 1.0703125 ]
[ 3.8125, 3.65625 ]
[ 2.765625, 2.984375 ]
[ 2.4375, 2.390625 ]
1
3
16,160
data/train/source-000001/part-000.parquet
66
86
893
0.908717
5,538,472
16,160
a6b569d7ed4dae95243fb6cc4bd2c3363c127305c39b88f4d15e1b0d54c6b0c4
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
5
86
893
37adc4dba0f26f40bdf5980f0401650942acae7f88170b9d0462b302fa1fcb8c
"CANADIAN COUNCIL FOR REFUGEES\n\nREFUGEES AND SECURITY\n\nRevised February 2003 (original paper dat(...TRUNCATED)
<urn:uuid:de6616fb-31d5-4372-aef0-ec545fea0c4b>
CC-MAIN-2017-13
http://ccrweb.ca/sites/ccrweb.ca/files/security.pdf
2017-03-27T10:39:35Z
"crawl-data/CC-MAIN-2017-13/segments/1490218189471.55/warc/CC-MAIN-20170322212949-00237-ip-10-233-31(...TRUNCATED)
65,798,756
22,428
eng_Latn
eng_Latn
0.997589
eng_Latn
0.998152
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn",(...TRUNCATED)
false
docling
[3650,8121,12197,16459,20651,25208,29612,34737,37405,42472,46846,51616,56068,60710,65235,70046,74267(...TRUNCATED)
[ 0.88671875, 0.8984375 ]
[ 4.46875, 3.75 ]
[ 3.40625, 2.25 ]
[ 2.578125, 2.515625 ]
3
2
22,994
data/train/source-000001/part-000.parquet
78
98
291
0.918608
1,367,281
22,994
88fa92f7b5166e057f35074834cda7a81945139d79098e69767f6d0f5df29b6c
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
6
98
291
58bd06f397fdd89e04aec578fcb00a64ed2c2754056f55c0e593a4ce26e81a12
"2\n\nThis article lists tax changes that took effect in 2012, or were announced in 2012 to take eff(...TRUNCATED)
<urn:uuid:83549433-ec57-4a71-953d-90324ca74e86>
CC-MAIN-2018-47
http://robertleach.co.uk/wp-content/uploads/2011/05/Tax-changes-in-20121.pdf
2018-11-14T01:14:46Z
"crawl-data/CC-MAIN-2018-47/segments/1542039741569.29/warc/CC-MAIN-20181114000002-20181114022002-002(...TRUNCATED)
286,373,755
15,599
eng_Latn
eng_Latn
0.998976
eng_Latn
0.999164
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn",(...TRUNCATED)
false
docling
[1927,4569,7018,9369,12093,14771,17144,19605,22060,24172,26091,28633,30749,33124,35983,38233,40341,4(...TRUNCATED)
[ 0.52734375, 0.8203125 ]
[ 3.65625, 3.421875 ]
[ 2.890625, 2.609375 ]
[ 2.359375, 2.875 ]
2
4
17,131
data/train/source-000001/part-000.parquet
118
138
450
0.909718
163,715
17,131
41be7f4daef20c7872d6ca2b22e78f19eb5fd71e491678e2a6953d9017c774eb
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
9
138
450
d957fb7c195cc45833be78f32cc539efb4ca758607654420772df0a921e29be4
"Dennis J. Kucinich of Ohio\n\nIn the United States House of Representatives\n\nMonday, June 9th, 20(...TRUNCATED)
<urn:uuid:e1e4dc2a-ead7-4c90-bc90-7c96e3d3e4f6>
CC-MAIN-2018-47
http://arabvoices.net/docs/kucinich_bush-impeachment.pdf
2018-11-14T00:23:16Z
"crawl-data/CC-MAIN-2018-47/segments/1542039741569.29/warc/CC-MAIN-20181114000002-20181114022002-002(...TRUNCATED)
18,950,552
41,247
eng_Latn
eng_Latn
0.990873
eng_Latn
0.997175
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn",(...TRUNCATED)
false
docling
[1357,3113,4906,8669,12773,20550,23972,27945,31835,35754,39623,43070,47502,50004,51564,53707,55595,5(...TRUNCATED)
[ 0.734375, 0.81640625 ]
[ 2.9375, 3.71875 ]
[ 1.6640625, 2.390625 ]
[ 2.5, 2.734375 ]
8
18
42,343
data/train/source-000001/part-000.parquet
124
144
449
0.929849
8,283,971
42,343
75db32472c3a5e570d55bdd036dbb5d32b99cbee0a3a6721c4cf6c79f5b71f97
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
10
144
449
b30439ac71e98fcf561b4e6d8d204e08d8408480c138098d0f492ab30e52b69d
"Master of Business Administration Department of Business Administration\n\nOld name, new management(...TRUNCATED)
<urn:uuid:5b34aac0-3a14-4eb3-9589-48658d18bf1f>
CC-MAIN-2024-42
https://skemman.is/bitstream/1946/39944/1/Master%20thesis%20-%20Eydi%CC%81s%20O%CC%88gn%20Uffado%CC%81ttir.pdf
2024-10-08T19:21:14+00:00
"crawl-data/CC-MAIN-2024-42/segments/1727944253554.4/warc/CC-MAIN-20241008171352-20241008201352-0002(...TRUNCATED)
468,237,898
37,567
eng_Latn
eng_Latn
0.968711
eng_Latn
0.99819
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","unknown","unknown","eng_Latn","eng_Latn","eng_Latn","e(...TRUNCATED)
false
docling
[240,2067,2887,3294,3313,3319,5672,8151,10530,11372,13740,15569,17979,19647,21707,23675,25383,27326,(...TRUNCATED)
[ 0.85546875, 0.90625 ]
[ 3.59375, 2.21875 ]
[ 3.015625, 3.921875 ]
[ 2.375, 2.359375 ]
1
0
38,685
data/train/source-000001/part-000.parquet
130
150
431
0.898326
6,377,130
38,685
dc47dcb403ee90adea50d7a3e8e51661c0b09589f43cffed66231a78db929ce1
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
11
150
431
c43827bc69b5aedd7dc491c85eb3e5c0696db462a46bc93f49b3384867875454
"LTSS CHO ICES\nRespite Services : A Critical Support for Family Caregivers\nSusan Reinhard, Jane Ti(...TRUNCATED)
<urn:uuid:f4f7935e-0d6c-49a2-b995-33bd1d2e89bb>
CC-MAIN-2024-42
https://respiteforall.org/wp-content/uploads/2024/04/AARP1461_LTSSChoices_RespiteCare_508-RFA.pdf
2024-10-08T17:33:52+00:00
"crawl-data/CC-MAIN-2024-42/segments/1727944253554.4/warc/CC-MAIN-20241008171352-20241008201352-0002(...TRUNCATED)
425,589,096
17,071
eng_Latn
eng_Latn
0.990466
eng_Latn
0.996192
["eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn","eng_Latn",(...TRUNCATED)
false
docling
[2558,5972,9338,13117,15989,19408,23278,26702,30472,33734,36887,39764,43393,46550,50131,53567,57401,(...TRUNCATED)
[ 1.0234375, 0.98046875 ]
[ 3.78125, 4.15625 ]
[ 2.921875, 3.765625 ]
[ 2.328125, 2.484375 ]
1
0
17,643
data/train/source-000001/part-000.parquet
133
153
948
0.917078
5,960,587
17,643
709ceadf04baf9a33fed4cee9d5f4a7cf49802c5bf53c433706cc1956c1f4578
finepdf
storagestuffnobody/finepdf-semantic-long
635d36d3a18b24ddf6f88a32fd18c284bb4b6041
data/train/train-00001.parquet
12
153
948
6a1b4902ac6ad0a61bdde15ea263f2a46637f7c955a59fd62042a2887ad748bf
End of preview. Expand in Data Studio

Fixed AO3 text and the complete selected mixture

1,717,205 full, untruncated records. Every configuration has a text column. This is the text release; the companion packed dataset applies the 1,048,576-token training cap.

from datasets import load_dataset
stories = load_dataset("storagestuffnobody/mid-training-40b-fixed-ao3", "stories", split="train", streaming=True)
print(next(iter(stories))["text"])

Use finepdf, stack or glotcc for the other configurations. Their source-specific metadata schemas intentionally differ. FinePDF's inherited one-record test split stays separate and is not presented as a robust evaluation set.

AO3's 347,887 originally selected works were reparsed from archived HTML. text contains the clean work title, byline, available summaries/notes, tags grouped by category, word count and latest available updated/completed date (otherwise posted date), then numbered chapter sections and end notes. HTML template boundaries identify chapters; the old embedding extractor's chapter_ends are not used as chapter boundaries. Headings inside authors' notes remain note content.

Each completed chapter ends with the literal <|role_end|>, the native EOS spelling of inclusionAI/Ling-3.0-tiny (ID156895). Each Stack file receives the same marker after its exact original file contents. Stack boundaries use stored character offsets and content hashes; delimiter-looking text inside code does not create a file boundary. FinePDF and GlotCC Parquet files are unchanged.

The EOS marker is part of text; it encodes to one special token with the pinned Ling tokenizer. No BOS, chat roles or extra work/repository-level EOS are added. One work or repository remains one attention document: chapter/file EOS markers do not reset attention. EOS alone is not chat role formatting.

boundary_eos_char_offsets gives marker starts in Python Unicode characters. AO3 chapter_ends now gives exclusive ends after each chapter's EOS; original_chapter_ends preserves the old spans. original_title preserves the old browser-page title. Original AO3 metadata remains, with reparsed_ao3_metadata_json and archived HTML verification added. Stack files[].text_start/text_end still recover exact file contents from the new text; original_text_start/original_text_end preserve the old offsets.

For modified configurations, text_sha256 hashes the current UTF-8 text and original_text_sha256 hashes the previous text. The inherited mixture_text_sha256 and source fields still describe the original selected record. token-counts.parquet supplies per-record full/capped Ling counts and inserted-EOS counts without truncating this text release.

Provenance and verification

The selected mixture is unchanged from storagestuffnobody/mid-training-40b. AO3 HTML was retrieved from the original archived snapshot, using stored WARC offsets and verifying record IDs, target URLs and content digests. No live AO3 pages were substituted. Original selection, review and upstream attribution remain in the pinned source repository and the copied original dataset card. Source-specific author, repository and license metadata are retained; this release does not assign a new blanket license.

canonical_tokens retains the original selection's counts (39,999,916,670 total), not counts of the new serialization. The pinned Ling tokenizer produces 39,868,598,979 tokens before the training cap and 39,757,790,311 after it. The “40b” name identifies the original mixture.

Configuration Records Full formatted Ling tokens Retained training tokens
stories 347,887 12,452,781,552 12,391,560,040
stack 104,653 7,220,697,584 7,171,110,428
finepdf 514,578 11,140,390,837 11,140,390,837
glotcc 750,087 9,054,729,006 9,054,729,006

manifest.json lists every data-file checksum, row count and original canonical-token count. The final audit checks all source/token hashes, document identities, lengths and offsets, plus independently re-encodes deterministic first/last/interior samples from every shard. A separate deterministic archive sample checks prose preservation across 512 works and 4,090 chapters. Reports and preparation code are included. No training or quality evaluation of model weights is claimed.

Downloads last month
3,980