MultilingualUnigramLM/

FineWeb2-tam_Taml-100M

100M-token slice of HuggingFaceFW/fineweb-2 (tam_Taml) for LAS tokenizer training under the MultilingualUnigramLM org.

Source: HuggingFaceFW/fineweb-2 Config: tam_Taml Documents: 263,523 Whitespace-tokens: 100,000,109 Chars: 894,087,067 Format: single data/train-00000-of-00001.parquet, schema (text: string). Mirrors the LangMap-TheStack-*-100M layout.