---
license: mit
---
This is the finewebedu dataset 10 Billion tokens. Tokenized using gpt-2 tokenizer. Each shards are a numpy file and contains 100 million tokens.