Datasets:
Fix: ajouter configs par langue (fr, en, bci, dyu, nch) + corriger description lang column
64a9a59 verified |
Download README.md from Tree-AI-lab/smollm-corpus-multilingue: direct link, hf CLI and curl.
- Browser
- Download file 5.15 kB
-
https://huggingface.co/datasets/Tree-AI-lab/smollm-corpus-multilingue/resolve/main/README.md
- Command line
-
hf download hf://datasets/Tree-AI-lab/smollm-corpus-multilingue/README.md
-
curl -L -H "Authorization: Bearer $HF_TOKEN" -o README.md https://huggingface.co/datasets/Tree-AI-lab/smollm-corpus-multilingue/resolve/main/README.md
5.15 kB
| configs: | |
| - config_name: default | |
| data_files: | |
| - split: train | |
| path: cosmopedia-v2/*.parquet | |
| - config_name: cosmopedia-v2 | |
| data_files: | |
| - split: train | |
| path: cosmopedia-v2/*.parquet | |
| - config_name: fineweb-edu-dedup | |
| data_files: | |
| - split: train | |
| path: fineweb-edu-dedup/*.parquet | |
| - config_name: python-edu | |
| data_files: | |
| - split: train | |
| path: python-edu/*.parquet | |
| - config_name: finemath | |
| data_files: | |
| - split: train | |
| path: finemath/*.parquet | |
| - config_name: open-thoughts | |
| data_files: | |
| - split: train | |
| path: open-thoughts/*.parquet | |
| - config_name: dioula | |
| data_files: | |
| - split: train | |
| path: "*/*_dyu.parquet" | |
| - config_name: nouchi | |
| data_files: | |
| - split: train | |
| path: "*/*_nch.parquet" | |
| - config_name: francais | |
| data_files: | |
| - split: train | |
| path: "*/*_fr*.parquet" | |
| - config_name: anglais | |
| data_files: | |
| - split: train | |
| path: "*/*_en*.parquet" | |
| - config_name: baoule | |
| data_files: | |
| - split: train | |
| path: "*/*_bci*.parquet" | |
| language: | |
| - en | |
| - fr | |
| - bci | |
| - dyu | |
| - nch | |
| license: odc-by | |
| task_categories: | |
| - text-generation | |
| tags: | |
| - tree-ai-lab | |
| - baoulé | |
| - bci | |
| - dioula | |
| - jula | |
| - dyu | |
| - nch | |
| - mandingue | |
| - nouchi | |
| - nch | |
| - argot-ivoirien | |
| - côte-d-ivoire | |
| - burkina-faso | |
| - mali | |
| - french | |
| - multilingual | |
| - pretraining | |
| - causal-lm | |
| - smollm-corpus | |
| - cosmopedia-v2 | |
| - fineweb-edu-dedup | |
| - python-edu | |
| - finemath | |
| - open-thoughts | |
| - chain-of-thought | |
| size_categories: | |
| - 1M<n<10M | |
| # SmolLM-Corpus Multilingue (EN · FR · BCI · DYU · NCH) — Tree AI Lab | |
| Corpus multilingue de pré-entraînement dérivé de **HuggingFaceTB/smollm-corpus** et enrichi avec des piliers de raisonnement mathématique et déductif (**HuggingFaceTB/finemath**, **open-thoughts/OpenThoughts-114k**). | |
| Produit par **Tree AI Lab** dans le cadre du projet **BABA** (IA souveraine africaine). | |
| Enrichi avec des corpus massifs en **Dioula / Jula (`dyu`)**, **Nouchi (`nch`)**, **Baoulé (`bci`)**, **Français (`fr`)** et **Anglais (`en`)**, couvrant plusieurs millions de documents alignés. | |
| ## Langues disponibles | |
| | Code | Langue | Description | | |
| |------|--------|-------------| | |
| | `en` | Anglais | English | | |
| | `fr` | Français | French | | |
| | `bci` | Baoulé | Langue ivoirienne (Akan) | | |
| | `dyu` | Dioula / Jula | Langue mandingue d'Afrique de l'Ouest | | |
| | `nch` | Nouchi | Argot urbain ivoirien | | |
| ## Chargement rapide | |
| ```python | |
| from datasets import load_dataset | |
| # Toutes langues — Cosmopedia-v2 | |
| ds = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'cosmopedia-v2', split='train') | |
| # Par langue uniquement | |
| ds_dyu = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'dioula', split='train') | |
| ds_nch = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'nouchi', split='train') | |
| ds_bci = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'baoule', split='train') | |
| ds_fr = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'francais', split='train') | |
| ds_en = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'anglais', split='train') | |
| # Par sous-corpus source | |
| ds_fineweb = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'fineweb-edu-dedup', split='train') | |
| ds_finemath = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'finemath', split='train') | |
| ds_thoughts = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'open-thoughts', split='train') | |
| ds_python = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'python-edu', split='train') | |
| ``` | |
| ## Architecture des dossiers | |
| | Dossier | Type | Shards / Documents | Format | Langues | | |
| |---|---|:---:|:---:|---| | |
| | `cosmopedia-v2/` | Synthétique éducatif | Shards 001–005 | Parquet Snappy | `en`, `fr`, `bci`, `dyu`, `nch` | | |
| | `fineweb-edu-dedup/` | Web éducatif filtré | Shards 003 & 006 | Parquet Snappy | `en`, `fr`, `bci`, `dyu`, `nch` | | |
| | `python-edu/` | Code informatique | Shards 004 & 007 | Parquet Snappy | `en`, `fr`, `bci`, `dyu` | | |
| | `finemath/` | Raisonnement mathématique | 12 shards | Parquet Snappy | `en`, `fr` | | |
| | `open-thoughts/` | Traces CoT & Backtracking | 2 shards | Parquet Snappy | `en`, `fr` | | |
| ## Schéma des colonnes | |
| | Colonne | Type | Description | | |
| |---|---|---| | |
| | `id` | string | Identifiant unique du document | | |
| | `source` | string | Sous-ensemble d'origine (cosmopedia-v2, fineweb-edu-dedup, python-edu, finemath, open-thoughts) | | |
| | `parquet` | string | Fichier source d'origine | | |
| | `lang` | string | Code ISO de langue : `en`, `fr`, `bci`, `dyu`, `nch` | | |
| | `text` | string | Contenu textuel dans la langue spécifiée | | |
| ## Filtrer par langue en Python | |
| ```python | |
| from datasets import load_dataset | |
| ds = load_dataset('Tree-AI-lab/smollm-corpus-multilingue', 'cosmopedia-v2', split='train') | |
| # Filtrer par langue | |
| ds_bci = ds.filter(lambda x: x['lang'] == 'bci') # Baoulé | |
| ds_dyu = ds.filter(lambda x: x['lang'] == 'dyu') # Dioula | |
| ds_nch = ds.filter(lambda x: x['lang'] == 'nch') # Nouchi | |
| ds_fr = ds.filter(lambda x: x['lang'] == 'fr') # Français | |
| ds_en = ds.filter(lambda x: x['lang'] == 'en') # Anglais | |
| ``` | |
| ## Licence | |
| [ODC-BY](https://opendatacommons.org/licenses/by/1-0/) — Open Data Commons Attribution License. | |
| Données dérivées de HuggingFaceTB/smollm-corpus (Apache 2.0) et enrichies par **Tree AI Lab**. | |