metadata
dataset_info:
features:
- name: file_name
dtype: string
- name: id
dtype: string
- name: language
dtype: string
- name: language_score
dtype: float64
- name: source
dtype: string
- name: subsource
dtype: string
- name: text
dtype: string
- name: translated
dtype: string
- name: fineweb-nemotron-edu-score
dtype: float32
- name: fineweb-nemotron-edu-score-label
dtype: string
- name: fineweb-mixtral-edu-score
dtype: float32
- name: fineweb-mixtral-edu-score-label
dtype: string
- name: fasttext-quality-score
dtype: float32
- name: __index_level_0__
dtype: int64
splits:
- name: train
num_bytes: 26246248313
num_examples: 3000000
download_size: 13587653626
dataset_size: 26246248313
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: cc-by-4.0
task_categories:
- text-generation
- text-classification
language:
- uk
tags:
- ukrainian
- nlp
- data
- scoring
Dataset Card for Lapa High Quality Pretraining Dataset
Dataset Description
Dataset Summary
This dataset is a random sample of both https://huggingface.co/datasets/lapa-llm/pretraining-lower-quality and https://huggingface.co/datasets/lapa-llm/pretraining-high-quality to transfer classifiers from English language to Ukrainian.
It was used to transfer the following models from this collection https://huggingface.co/collections/lapa-llm/lapa-v012-pretraining:
- lapa-llm/fineweb-nemotron-edu-score
- lapa-llm/fineweb-mixtral-edu-score
- lapa-llm/fasttext-quality-score
Languages
- Ukrainian (uk)
Dataset Creation
Source Data
- Base dataset: Kobza, FinePDFs, FineWeb, UberText
Considerations for Using the Data
Social Impact
Aims to strengthen the Ukrainian-language LLM ecosystem and improve accessibility of language technology for Ukrainian speakers.
Citation
BibTeX
TBD
License
CC-BY-SA-4.0
This dataset is part of the "Lapa" - Ukrainian LLM initiative to advance natural language processing for the Ukrainian language.