Datasets:

Icold
/

kron-corpus

Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
source: string
licence: string
url: string
language: string
domain: string
sha256: string
bytes: int64
docs: int64
path: string
note: string
text: string
title: string
to
{'title': Value('string'), 'text': Value('string'), 'language': Value('string'), 'domain': Value('string'), 'source': Value('string'), 'licence': Value('string'), 'url': Value('string'), 'note': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1835, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 621, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              source: string
              licence: string
              url: string
              language: string
              domain: string
              sha256: string
              bytes: int64
              docs: int64
              path: string
              note: string
              text: string
              title: string
              to
              {'title': Value('string'), 'text': Value('string'), 'language': Value('string'), 'domain': Value('string'), 'source': Value('string'), 'licence': Value('string'), 'url': Value('string'), 'note': Value('string')}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 956, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1702, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1887, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

title string	text string	language string	domain string	source string	licence string	url string	note string
A Rickconvenient Mort/Transcript	Ah, I love that new T-shirt smell. Chicks are gonna be all up in our biz when they see us wearing these Rick and Morty custom tees. I bet we could sell these. Hells, yeah. I fucking love merch! The rain! lt-it burns! Oh, shit. Quick, Morty, protect the shirts! The door! Open the door! Rick! Is it-Morty, will yo...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/A Rickconvenient Mort/Transcript	easter egg: series dialogue
Air Force Wong/Transcript	Wong: Before we move off this topic, can I ask if you're currently a hologram? Rick: Yeah, yeah, h-hold on, one sec, left a bag of funyuns down there last week. Didn't want to waste our time. Okay therapy me Wong: Do you need to take that? Rick: No just the president Wong: Maybe we should turn off the phone Rick: ...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Air Force Wong/Transcript	easter egg: series dialogue
Amortycan Grickfitti/Transcript	Beth: Jerry, can you watch the kids tonight? Someone played Barry White over the racetrack speakers, and now there's a sevenhorse pregnancy. Jerry: Oh, um-- Rick: No, he can't watch the kids, it's guys' night. Beth: How did you two start having guys' nights? Rick: So now, you're complaining that I like your stupid ...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Amortycan Grickfitti/Transcript	easter egg: series dialogue
Analyze Piss/Transcript	(Rick and Morty are on an island firing at small creatures) Morty: I-I can't hold them off, Rick! Th-There's too many! ( Rick gets slapped by a creature's arm) Mr. Nimbus: Richard! Your fortress is a salty waste of coral and clam. Relinquish it at once! (Rick runs up to him and punches him. He breaks a hole in a shi...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Analyze Piss/Transcript	easter egg: series dialogue
Anatomy Park (episode)/Transcript	[Open Ext. Smith residence] (Outside, there is snow on the ground. The Smith house is decorated for Christmas, and Jerry can be heard singing Last King Christmas.) Jerry: singing Last King Christmas last arrived! [Trans Int. Kitchen] (The kitchen is a mess of food in various stages of preparation. Jerry pulls a h...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Anatomy Park (episode)/Transcript	easter egg: series dialogue
Auto Erotic Assimilation/Transcript	This article is a transcript of the Season 2 episode Auto Erotic Assimilation. ‎Open Ext. Rick's Ship in flying through a planet's rings (Morty is in the backseat, Summer is in the starboard seat, Rick is driving) All: (singing) Love. Connection. Experience. (Morty claps) Rick: Yeah Morty All: (singing) Come together....	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Auto Erotic Assimilation/Transcript	easter egg: series dialogue
Bethic Twinstinct/Transcript	Morty: Oh, my God, I cannot fucking believe this! Here we fucking go! Oh! Jeez! Rick: Pardon the interruption. We are once again in the federal clear. Morty: Rick, what the hell?! A-Are you a Turkey Dracula now? Rick: Morty, is that a Pooplickian GamePod? Morty: Actually it's the GamePod XL. They say it's the most ...	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Bethic Twinstinct/Transcript	easter egg: series dialogue
Big Trouble in Little Sanchez/Transcript	"This article is a transcript of the Rick and Morty episode Big Trouble In Little Sanchez. It aired (...TRUNCATED)	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Big Trouble in Little Sanchez/Transcript	easter egg: series dialogue
Childrick of Mort/Transcript	"Summer: What kind of parents drag their kids away for the entire week of spring break? I'm missing (...TRUNCATED)	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Childrick of Mort/Transcript	easter egg: series dialogue
Claw and Hoarder: Special Ricktim's Morty/Transcript	"Rick, we did it!\nWe -- we got it!\n\nHurry up, morty.\nGet in!\n\nChachi!\n[bleep] chachi, morty.\(...TRUNCATED)	en	humour	Fandom wiki (api.php)	CC-BY-SA-3.0	https://rickandmorty.fandom.com/wiki/Claw and Hoarder: Special Ricktim's Morty/Transcript	easter egg: series dialogue

End of preview.

KronCorpus

The bilingual English/Russian pretraining corpus of KronLLM Q10 — the 10M-parameter model this project ships (9,938,886 parameters, Kronecker-factorised weights). It is published separately because the data is useful on its own: a small, fully documented, science-heavy bilingual corpus with a per-language quota contract and a provenance record for every shard.

What is in it

property	value
tokens	2,783,383,150 train tokens (8K byte-level BPE, 3.60 bytes/token)
text	10.02 GB, 933,522 documents (EN 517,973 / RU 415,549)
languages	EN 59.7 % / RU 40.3 % of tokens — EN 46.5 % / RU 53.5 % of bytes
held-out	46,197,636 tokens, sha256 `a45bbb03…`, byte-identical since v3
system prompts	none inserted (the 46,820 `<
bins	uint16 (`train.bpe.bin`, `val.bpe.bin`), one epoch = 169,884 steps of 16,384 tokens
index	`train_index.npy`: span + token hash per document, for resumable training
provenance	source, licence, URL and SHA-256 for every shard in `manifest.jsonl`

Why bytes and not tokens. The quota is stated in bytes because that is the only quantity independent of the tokenizer, and the token share is measured and published next to it. Russian costs 4.96 bytes/token against 2.92 for English, so a byte quota is not a token quota: at 53.5 % of the bytes, Russian is 40.3 % of the tokens. The 60/40 target is therefore enforced per language (QUOTAS_EN / QUOTAS_RU in kronllm/bilingual.py), not on one shared table.

Composition

Quotas are per language; "realised" is what the mixer actually selected after dedup. ⚠ marks a domain where the supply cannot fill the quota.

English

domain	quota	realised	main sources
mathematics	26 %	26.8 %	OpenWebMath, Wikipedia math
encyclopedic	16 %	16.5 %	Wikipedia EN (3.4 GB of shards)
literature	15 %	15.5 %	Project Gutenberg
code	11 %	11.3 %	permissively licensed code corpora, SFT code slices
CS / software engineering	10 %	10.3 %	Stack Overflow (207,646 answers), IETF RFCs
science	9 %	9.3 %	arXiv, Wikipedia physics/astronomy
machine learning	7 %	7.2 %	arXiv cs.LG/CL, Cosmopedia, framework docs
logic	4 %	2.8 % ⚠	Wikipedia logic, Lean 4 mathlib
humour / identity	2 %	0.2 % ⚠	Fandom, synthetic identity slice

Russian

domain	quota	realised	main sources
encyclopedic	34 %	35.5 %	Wikipedia RU
CS / software engineering	21 %	21.9 %	Habr (49,116 posts), ru Stack Overflow (291,165 answers)
literature	15 %	15.7 %	Wikisource, Gutenberg translations, FineWeb-2
mathematics	13 %	13.6 %	FineWeb-2 math, Wikipedia
science	10 %	10.4 %	FineWeb-2 science, Wikipedia
logic	2 %	2.1 %	Wikipedia logic, FineWeb-2
code	3 %	0.4 % ⚠	Russian code tasks with translated questions
ML / identity	1.5 %	0.3 % ⚠	FineWeb-2, synthetic identity slice

Structural shortfalls, reported rather than padded: en/logic 131 of 192 MB, en/humour 11 of 48, en/identity 1 of 48, ru/code 20 of 168, ru/ml 17 of 56, ru/humour nothing, ru/identity 1 of 28. There is no more Russian code text on the Hub under a usable licence; more would have to be translated.

How documents are selected

Quality ranking inside each domain. Every candidate gets a cheap score on its first 4,000 characters (alphabetic density, digit ratio, vocabulary diversity, word length, length bonus) and the quota takes the best ones, not a random sample. This is the change that separates this corpus from earlier builds, which took documents in SHA-256 order.
Exact dedupe. sha256 of the whole text plus sha256 of its 2 KB prefix. The MinHash/Jaccard pass it replaced cost 55 minutes and 20 GB of RAM.
Two-pass streaming build. Pass 1 streams all 87 shards and keeps only (sha, lang, domain, bytes, path, offset) plus the score — peak RSS ~1.1 GB instead of the 25 GB that killed the in-memory builder. Pass 2 seeks to the selected offsets and writes the mix.
Filters before scoring. Minimum length, symbol ratio, repetition ratio and a script check per language.
Provenance is mandatory. A shard without a stated licence and a URL is rejected by record_shard; manifest.jsonl is the attribution.

Sources

source	licence	role
Wikipedia (dump 2023-11-01, EN/RU)	CC BY-SA 3.0/4.0	encyclopedic, domains via keywords
OpenWebMath	ODC-By 1.0	mathematics
arXiv (abstracts + full text)	arXiv non-exclusive	science, ML
Project Gutenberg, Wikisource	public domain / CC BY-SA	literature
FineWeb-2 `rus_Cyrl`	ODC-By 1.0	Russian domains without a primary source
Stack Overflow (EN, CC BY-SA 4.0)	CC BY-SA 4.0	CS / software engineering
Habr, ru Stack Overflow	Habr content / Stack Exchange CC BY-SA 4.0	Russian CS
Lean 4 `mathlib4`	Apache-2.0	formal proofs (logic)
Cosmopedia v2	Apache-2.0	ML/science textbook prose
Code corpora (permissive)	per file, see the dataset pages	code
Fandom wiki API	CC BY-SA 3.0	humour slice only

Decay-pass corpus (Q10A)

A second bin is published next to the epoch one: 809,120,240 tokens (29.1 % of an epoch, 249,868 documents) — the same domain profile selected by quality, plus 5 % chat spread through the pass so the four chat tokens (<|system|>, <|user|>, <|assistant|>, <|eot|>) are not cold when SFT starts. Same held-out bin, so both stages are scored on the same blocks.

Loading

import numpy as np

# Kaggle: the tokenised bins (uint16) live there, GitHub is not a data host
#   kaggle datasets download -d icold21/kron-corpus-2b        # epoch, v10
#   kaggle datasets download -d icold21/kron-q10a-anneal-bin  # decay pass
tokens = np.memmap("train.bpe.bin", dtype=np.uint16, mode="r")
index = np.load("train_index.npy")     # (start, n, hash) per document

# Hugging Face: card, manifest and the identity slice -- not the bins
#   ds = load_dataset("Icold/kron-corpus")

Reproducing the build

# 1. fetch: append-only, sharded, one timeout per source
python scripts/build_bilingual.py fetch-hf --source all
python scripts/build_bilingual.py fetch-gutenberg-ids ...
python scripts/build_bilingual.py fetch-wikibooks ...
python scripts/gen_identity.py --samples-per-language 800

# 2. mix: quality-ranked quotas, per language (two passes, ~1 GB RSS)
python scripts/build_streaming.py --budget-en-mb 4800 --budget-ru-mb 5600

# 3. bins: reuse the tokenizer and the held-out bin so bpb stays comparable
python scripts/prepare_bilingual_bin.py --out-dir bins \
    --reuse-tokenizer data/bilingual/tokenizer.json \
    --reuse-val-dir held_out --phase epoch --no-domain-prompts

The decay pass adds the chat pools (scripts/build_anneal_chat.py in the code repository), the same mix build at 30 % of the budget, and --phase anneal --interleave chat.

Version history

version	what changed
v10 (current)	prose rebalance: EN literature 15 %, EN encyclopedic 16 %, RU technical up, SFT samples, quality-ranked quotas, 2.78B tokens
v1-v9	earlier rebuilds of the same pipeline. The held-out bin stayed byte-identical from v3 onward; versions 4-7 carried inserted domain system prompts, which v8 removed — the trainer must never spend gradient on its own boilerplate

Licence

The corpus is a collection: each shard keeps the licence of its source, as recorded in manifest.jsonl. CC BY-SA text requires attribution and share-alike; Gutenberg texts are public domain in the US; OpenWebMath and FineWeb-2 are ODC-By 1.0; mathlib4 is Apache-2.0. The build scripts are MIT, like KronLLM.

If you redistribute a mix, keep manifest.jsonl next to it — that file is the attribution.

_{The identity slices are synthetic: the model calls itself Kron from a
fixed fact sheet (no numeric facts), 800 plain + 850 pastiche documents per
language, under 0.1 % of the corpus and marked in the manifest.}

Downloads last month: 177

Total file size:

5.12 MB