Dataset Viewer
Auto-converted to Parquet Duplicate
total_tokens
int64
shards
list
789,234,784,071
[ { "filename": "en_multi_style/part-00000-f36f5a53-4a77-434b-b9bc-67ed69b93fe2-c000.bin", "total_tokens": 435019551 }, { "filename": "en_multi_style/part-00001-f36f5a53-4a77-434b-b9bc-67ed69b93fe2-c000.bin", "total_tokens": 435597025 }, { "filename": "en_multi_style/part-00002-f36f5a53-4a...

This is the data from the open-source community, derived from ultra-fineweb-l3 after tokenization.

The L3 refined data built on Ultra-FineWeb via Q&A Pair Generation and Multi-style Rewriting.

  • data/en_multi_style: 240B tokens.
  • data/en_qa: 341B tokens.
  • data/zh_multi_style: 85B tokens.
  • data/zh_qa: 121B tokens.

Everyone is free to use it.

Downloads last month
18