Datasets:
marin_name stringlengths 3 51 | marin_tokens int64 1.52M 2,114B | category stringclasses 5
values | synthetic bool 2
classes | machine_translated bool 2
classes | pdf bool 2
classes | hf_repo stringlengths 12 50 ⌀ | hf_subset stringlengths 1 33 ⌀ | license stringlengths 3 165 ⌀ | transform_tldr stringclasses 5
values |
|---|---|---|---|---|---|---|---|---|---|
coderforge | 10,288,088,964 | code | true | false | false | togethercomputer/CoderForge-Preview | null | Per-file: MIT, BSD-3, Apache-2.0, BSD, BSD-2, BSD-4, ISC, PostgreSQL, PSF, CC0-1.0, MIT-NA, MIT-CMU, HPND, Dual MIT/Apache-2.0, Dual Apache-2.0/BSD-3, Dual BSD-3/MIT | null |
common_corpus/english | 1,015,389,781,525 | web | false | false | false | PleIAs/common_corpus | null | null | null |
cp/arxiv_abstracts | 536,806,535 | web | false | false | false | common-pile/arxiv_abstracts | null | CC0 1.0 | null |
cp/arxiv_papers | 6,632,151,543 | web | false | false | true | common-pile/arxiv_papers | null | CC BY (per-paper) | null |
cp/biodiversity | 8,601,082,415 | web | false | false | false | common-pile/biodiversity_heritage_library_books | null | Public Domain | OCR pages stitched into full books by item_id |
cp/caselaw | 17,549,302,777 | web | false | false | false | common-pile/caselaw_access_project | null | Public Domain | null |
cp/data_provenance | 824,372,627 | specialized | false | false | false | common-pile/data_provenance_initiative | null | Open (per-dataset) | null |
cp/doab | 2,928,940,287 | web | false | false | false | common-pile/doab | null | CC BY (per-book) | null |
cp/foodista | 21,072,461 | web | false | false | false | common-pile/foodista | null | CC BY | null |
cp/github_archive | 10,255,254,351 | code | false | false | false | common-pile/github_archive | null | Blue Oak Council (per-repo) | null |
cp/library_of_congress | 8,064,185,907 | web | false | false | false | common-pile/library_of_congress | null | Public Domain | null |
cp/libretexts | 84,269,653 | web | false | false | false | common-pile/libretexts | null | CC BY (per-text) | null |
cp/news | 54,022,046 | web | false | false | false | common-pile/news | null | CC BY / CC BY-SA | null |
cp/oercommons | 10,830,089 | web | false | false | false | common-pile/oercommons | null | CC BY (per-resource) | null |
cp/peS2o | 40,738,676,861 | web | false | false | false | common-pile/peS2o | null | CC BY (per-paper) | null |
cp/peps | 2,781,263 | code | false | false | false | common-pile/python_enhancement_proposals | null | Public Domain | null |
cp/pre_1929_books | 10,564,513,170 | web | false | false | false | common-pile/pre_1929_books | null | Public Domain | null |
cp/pressbooks | 125,754,068 | web | false | false | false | common-pile/pressbooks | null | CC BY (per-book) | null |
cp/project_gutenberg | 4,908,125,204 | web | false | false | false | common-pile/project_gutenberg | null | Public Domain | null |
cp/public_domain_review | 1,517,771 | web | false | false | false | common-pile/public_domain_review | null | Public Domain | null |
cp/pubmed | 38,076,110,051 | web | false | false | false | common-pile/pubmed | null | CC BY (per-article) | null |
cp/regulations | 1,280,878,078 | web | false | false | false | common-pile/regulations | null | Public Domain | null |
cp/stackexchange | 21,892,818,944 | web | false | false | false | common-pile/stackexchange | null | CC BY-SA | null |
cp/stackv2_code | 352,758,688,082 | code | false | false | false | common-pile/stackv2_code | null | Blue Oak Council (per-file) | Filtered to 40+ programming language extensions |
cp/ubuntu_irc | 1,760,185,881 | web | false | false | false | common-pile/ubuntu_irc | null | Public Domain | null |
cp/uk_hansard | 2,128,630,536 | web | false | false | false | common-pile/uk_hansard | null | Open Parliament License | null |
cp/usgpo | 7,779,046,590 | web | false | false | false | common-pile/usgpo | null | Public Domain | null |
cp/uspto | 142,408,263,704 | web | false | false | false | common-pile/uspto | null | Public Domain | null |
cp/wikiteam | 2,965,095,868 | web | false | false | false | common-pile/wikiteam | null | CC BY-SA (per-wiki) | null |
cp/youtube | 4,071,331,344 | web | false | false | false | common-pile/youtube | null | CC BY | null |
davinci-dev/ctx-native | 57,570,358,078 | code | true | false | false | GAIR/daVinci-Dev | ctx-native | other | null |
davinci-dev/env-native | 2,577,319,940 | code | true | false | false | GAIR/daVinci-Dev | env-native | other | null |
finepdfs | 1,186,467,124,567 | web | false | false | true | HuggingFaceFW/finepdfs | eng_Latn | odc-by | null |
finepdfs/arb_Arab | 29,718,852,033 | multilingual | false | false | true | HuggingFaceFW/finepdfs | arb_Arab | odc-by | null |
finepdfs/ces_Latn | 29,826,870,499 | multilingual | false | false | true | HuggingFaceFW/finepdfs | ces_Latn | odc-by | null |
finepdfs/cmn_Hani | 32,969,217,932 | multilingual | false | false | true | HuggingFaceFW/finepdfs | cmn_Hani | odc-by | null |
finepdfs/deu_Latn | 177,095,554,439 | multilingual | false | false | true | HuggingFaceFW/finepdfs | deu_Latn | odc-by | null |
finepdfs/fra_Latn | 164,750,999,806 | multilingual | false | false | true | HuggingFaceFW/finepdfs | fra_Latn | odc-by | null |
finepdfs/hun_Latn | 37,439,557,050 | multilingual | false | false | true | HuggingFaceFW/finepdfs | hun_Latn | odc-by | null |
finepdfs/ind_Latn | 20,318,872,038 | multilingual | false | false | true | HuggingFaceFW/finepdfs | ind_Latn | odc-by | null |
finepdfs/ita_Latn | 94,785,999,293 | multilingual | false | false | true | HuggingFaceFW/finepdfs | ita_Latn | odc-by | null |
finepdfs/jpn_Jpan | 115,867,714,407 | multilingual | false | false | true | HuggingFaceFW/finepdfs | jpn_Jpan | odc-by | null |
finepdfs/nld_Latn | 46,968,293,614 | multilingual | false | false | true | HuggingFaceFW/finepdfs | nld_Latn | odc-by | null |
finepdfs/pol_Latn | 54,398,138,012 | multilingual | false | false | true | HuggingFaceFW/finepdfs | pol_Latn | odc-by | null |
finepdfs/por_Latn | 94,689,170,875 | multilingual | false | false | true | HuggingFaceFW/finepdfs | por_Latn | odc-by | null |
finepdfs/ron_Latn | 22,607,091,268 | multilingual | false | false | true | HuggingFaceFW/finepdfs | ron_Latn | odc-by | null |
finepdfs/rus_Cyrl | 146,947,409,239 | multilingual | false | false | true | HuggingFaceFW/finepdfs | rus_Cyrl | odc-by | null |
finepdfs/spa_Latn | 216,737,892,859 | multilingual | false | false | true | HuggingFaceFW/finepdfs | spa_Latn | odc-by | null |
finepdfs/swe_Latn | 25,335,103,612 | multilingual | false | false | true | HuggingFaceFW/finepdfs | swe_Latn | odc-by | null |
finepdfs/tha_Thai | 17,395,175,374 | multilingual | false | false | true | HuggingFaceFW/finepdfs | tha_Thai | odc-by | null |
finepdfs/ukr_Cyrl | 25,531,614,138 | multilingual | false | false | true | HuggingFaceFW/finepdfs | ukr_Cyrl | odc-by | null |
finetranslations/multilingual | 1,520,069,046,906 | multilingual | false | false | false | HuggingFaceFW/finetranslations | null | odc-by | 50% of tokens: original multilingual text |
finetranslations/web | 1,520,069,046,906 | web | false | true | false | HuggingFaceFW/finetranslations | null | odc-by | 50% of tokens: machine-translated English |
gpt-oss-rollouts | 3,199,721,925 | specialized | true | false | false | andyrdt/gpt-oss-20b-rollouts | null | mit | null |
hplt_v3 | 612,703,846,013 | web | false | false | false | HPLT/HPLT3.0 | null | cc0-1.0 | null |
institutional_books | 203,634,225,997 | web | false | false | false | institutional/institutional-books-1.0 | null | null | Pages concatenated from text_by_page_gen into full books |
molmo2-cap | 361,102,745 | specialized | false | false | false | allenai/Molmo2-Cap | null | odc-by | null |
nemotron-terminal | 6,079,232,184 | code | true | false | false | nvidia/Nemotron-Terminal-Corpus | null | cc-by-4.0 | null |
nemotron_cc_code_v1/all | 399,405,424,602 | code | false | false | false | nvidia/Nemotron-CC-Code-v1 | null | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_math_v1/3 | 78,901,978,124 | math | false | false | false | nvidia/Nemotron-CC-Math-v1 | 3 | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_math_v1/4plus_mind | 72,200,557,381 | math | false | false | false | nvidia/Nemotron-CC-Math-v1 | 4plus_mind | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/diverse_qa | 676,573,079,794 | web | true | false | false | nvidia/Nemotron-CC-v2 | diverse_qa | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/high_quality | 608,956,318,001 | web | false | false | false | nvidia/Nemotron-CC-v2 | high_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/high_quality_synthetic | 1,223,460,107,713 | web | true | false | false | nvidia/Nemotron-CC-v2 | high_quality_synthetic | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/medium_high_quality | 535,448,458,858 | web | false | false | false | nvidia/Nemotron-CC-v2 | medium_high_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/medium_quality | 2,114,325,066,023 | web | false | false | false | nvidia/Nemotron-CC-v2 | medium_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2/translated_diverse_qa | 592,845,357,611 | multilingual | true | true | false | nvidia/Nemotron-CC-v2 | translated_diverse_qa | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/high_quality | 25,147,272,083 | web | false | false | false | nvidia/Nemotron-CC-v2.1 | high_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/high_quality_dqa | 7,809,566,030 | web | true | false | false | nvidia/Nemotron-CC-v2.1 | high_quality_dqa | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/high_quality_synthetic | 90,857,570,128 | web | true | false | false | nvidia/Nemotron-CC-v2.1 | high_quality_synthetic | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/high_quality_translated | 38,645,703,584 | multilingual | true | true | false | nvidia/Nemotron-CC-v2.1 | high_quality_translated | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/high_quality_translated_synthetic | 153,411,766,753 | multilingual | true | true | false | nvidia/Nemotron-CC-v2.1 | high_quality_translated_synthetic | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/medium_high_quality | 16,354,290,042 | web | false | false | false | nvidia/Nemotron-CC-v2.1 | medium_high_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/medium_high_quality_synthetic | 2,065,376,892,626 | web | true | false | false | nvidia/Nemotron-CC-v2.1 | medium_high_quality_synthetic | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/medium_high_quality_translated | 26,029,329,558 | multilingual | true | true | false | nvidia/Nemotron-CC-v2.1 | medium_high_quality_translated | NVIDIA Data Agreement for Model Training | null |
nemotron_cc_v2_1/medium_quality | 51,666,624,592 | web | false | false | false | nvidia/Nemotron-CC-v2.1 | medium_quality | NVIDIA Data Agreement for Model Training | null |
nemotron_code_v2/synthetic_code_review | 74,236,199,104 | code | true | false | false | nvidia/Nemotron-Pretraining-Code-v2 | synthetic_code_review | NVIDIA Data Agreement for Model Training | null |
nemotron_code_v2/synthetic_question_answering | 233,030,892,341 | code | true | false | false | nvidia/Nemotron-Pretraining-Code-v2 | synthetic_question_answering | NVIDIA Data Agreement for Model Training | null |
nemotron_code_v2/synthetic_rewriting | 73,731,123,746 | code | true | false | false | nvidia/Nemotron-Pretraining-Code-v2 | synthetic_rewriting | NVIDIA Data Agreement for Model Training | null |
nemotron_code_v2/synthetic_student_teacher | 25,197,919,681 | code | true | false | false | nvidia/Nemotron-Pretraining-Code-v2 | synthetic_student_teacher | NVIDIA Data Agreement for Model Training | null |
nemotron_code_v2/synthetic_transpilation | 27,781,912,469 | code | true | false | false | nvidia/Nemotron-Pretraining-Code-v2 | synthetic_transpilation | NVIDIA Data Agreement for Model Training | null |
nemotron_sft/sft_code | 56,653,188,193 | code | true | false | false | nvidia/Nemotron-Pretraining-SFT-v1 | sft_code | NVIDIA Data Agreement for Model Training | null |
nemotron_sft/sft_general | 85,200,897,503 | specialized | true | false | false | nvidia/Nemotron-Pretraining-SFT-v1 | sft_general | NVIDIA Data Agreement for Model Training | null |
nemotron_sft/sft_math | 199,939,068,103 | math | true | false | false | nvidia/Nemotron-Pretraining-SFT-v1 | sft_math | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/infinibyte_reasoning | 18,694,414,109 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | infinibyte_reasoning | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/math_textbooks | 25,594,276,808 | math | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | math_textbooks | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/rqa | 135,168,966,675 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | rqa | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/scientific_coding | 1,182,503,855 | code | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | scientific_coding | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/stem_sft | 81,201,425,145 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | stem_sft | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized/wiki_rewrite | 7,257,109,784 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1 | wiki_rewrite | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized_v1_1/code_concepts | 7,034,542,470 | code | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1.1 | code_concepts | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized_v1_1/economics | 72,304,885 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1.1 | economics | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized_v1_1/formal_logic | 125,799,844 | math | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1.1 | formal_logic | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized_v1_1/multiple_choice | 1,557,203,358 | specialized | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1.1 | multiple_choice | NVIDIA Data Agreement for Model Training | null |
nemotron_specialized_v1_1/unconditional_algorithmic | 190,621,205 | code | true | false | false | nvidia/Nemotron-Pretraining-Specialized-v1.1 | unconditional_algorithmic | NVIDIA Data Agreement for Model Training | null |
nsf_awards | 171,336,575 | web | false | false | false | null | null | Public Domain | null |
numinamath | 375,831,015 | math | false | false | false | AI-MO/NuminaMath-1.5 | null | apache-2.0 | null |
sec-edgar | 334,898,505,739 | web | false | false | false | TeraflopAI/SEC-EDGAR | null | apache-2.0 | null |
starcoder2/documentation | 1,401,237,659 | code | false | false | false | bigcode/StarCoder2-Extras | null | null | null |
starcoder2/ir_cpp | 39,008,936,386 | code | false | false | false | bigcode/StarCoder2-Extras | null | null | null |
End of preview. Expand in Data Studio
Marin Token Counts
Token counts for all datasets used in Marin pretraining runs.
Schema
| Column | Type | Description |
|---|---|---|
dataset |
string | Dataset identifier |
marin_tokens |
int | Number of tokens after tokenization |
category |
string | Content domain (web, code, math, academic, books, etc.) |
synthetic |
bool | Whether the data is LLM-generated or LLM-translated |
Categories
- web — Quality-classified Common Crawl text (Nemotron-CC)
- code — Source code and code-related documents
- math — Math-focused extractions and competition problems
- academic — Peer-reviewed papers and abstracts
- reasoning — Cross-domain reasoning and formal logic
- books — Digitized public domain and open access books
- legal — Court decisions, regulations, patents
- government — Parliamentary proceedings and publications
- education — Open educational resources and textbooks
- encyclopedic — Wiki-style reference content
- forum — Q&A sites and chat logs
- documents — PDF-extracted document text
- translation — Parallel translation corpora
- news — CC-licensed news articles
- media — Transcribed audio/video
- supervised — Curated task datasets
- reference — Niche reference sites
- general — General-domain content
Updates
This dataset is updated by running experiments/count_tokens.py from the Marin repo,
which reads tokenized dataset stats from GCS and pushes the results here.
- Downloads last month
- 188