Skip to content

Dataset · Tree-AI-lab

Multilingual corpus with rare African languages

Tree-AI-lab/smollm-corpus-multilingue

A text corpus of 10 to 100M records for pre-training models in English, French, and West African languages.

What's inside

Contains texts in five languages: English, French, Baoulé, Dioula, and Nushi, with a total volume ranging from 10 to 100 million records. Created by Tree AI Lab as part of the BABA project based on HuggingFaceTB/smollm-corpus. The base data is supplemented with math and deductive texts from finemath and OpenThoughts-114k, as well as large corpora in rare African languages.

What you can do with it

  • Pre-training compact language models with support for African languages
  • Continual pre-training of existing LLMs in Baoulé, Dioula, and Nushi
  • Improving logic and mathematics skills in multilingual models
Tasks
Text generation
Languages
English, French, bci, Dyula, nch
Size
10M–100M rows
Downloads this month
32.4K
Likes
2
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("Tree-AI-lab/smollm-corpus-multilingue")