stukenov/sozkz-corpus-tokenized-kk-llama50k-v3

SozKZ Corpus Tokenized v3 (50K)

Pre-tokenized Kazakh corpus. Each sample = 1024 packed tokens.

Source: stukenov/sozkz-corpus-clean-v3

Property	Value
Train blocks	8,787,709
Val blocks	88,300
Block size	1024
Total tokens	~9.09B
Tokenizer	stukenov/sozkz-core-gpt2-50k-kk-base-v1

Source distribution

Source	Documents	Share
culturax	2,707,214	19.8%
hplt_new	2,204,165	16.1%
mc4	1,906,763	13.9%
madlad400	1,794,308	13.1%
kazparc_sync	1,370,243	10.0%
cc100	1,365,739	10.0%
md_leipzig	1,128,122	8.2%
md_kazakhNews	288,247	2.1%
md_oscar	239,807	1.8%
wikipedia	235,915	1.7%
moscar	231,693	1.7%
kazparc	165,754	1.2%
kazsandra	40,236	0.3%
md_kazakhBooks	20,482	0.1%
sib200	648	0.0%
belebele	488	0.0%
wikiann	194	0.0%

Usage

from datasets import load_dataset
ds = load_dataset("stukenov/sozkz-corpus-tokenized-kk-llama50k-v3")
sample = ds["train"][0]
print(len(sample["input_ids"]))  # 1024