An LLM pre-training dataset produced by filtering and deduplicating the raw text collected in the Common Pile v0.1