Running 1.01k 1.01k FineWeb: decanting the web for the finest text data at scale π· Generate high-quality web text data for LLM training