LingweiGu/NanoKnow-Fineweb-Edu-Index

NanoKnow FineWeb-Edu Lucene Index

A pre-built Lucene BM25 index over karpathy/fineweb-edu-100b-shuffle—the exact pre-training corpus used by the nanochat family of language models. Built with Anserini.

This index is part of the NanoKnow project: github.com/castorini/NanoKnow

Index Details

Property	Value
Corpus	karpathy/fineweb-edu-100b-shuffle
Documents	97,230,848
Index Size	~326 GB
Index Type	Lucene (BM25)
Built With	Anserini / Pyserini
Files	1,639 Lucene segment files

Document ID Format

Each document has a unique ID: shard_XXXXX_YYYYY

• XXXXX: zero-padded shard number (0-1822)
• YYYYY: row offset within the parquet shard

For example, shard_00151_20323 refers to row 20,323 in shard 151 of the FineWeb-Edu parquet files.

Usage

Download

huggingface-cli download LingweiGu/NanoKnow-Fineweb-Edu-Index --repo-type dataset --local-dir ./fineweb-edu-index

Search with Pyserini

from pyserini.search.lucene import LuceneSearcher

searcher = LuceneSearcher("./fineweb-edu-index")
print(f"Index contains {searcher.num_docs:,} documents")

hits = searcher.search("What is the capital of France?", k=10)
for hit in hits:
    print(f"{hit.docid}: {hit.score:.4f}")

Retrieve Document Text

import json

doc = searcher.doc("shard_00151_20323")
text = json.loads(doc.raw())["contents"]
print(text[:500])

Part of NanoKnow

This index powers the NanoKnow pipeline, which projects QA benchmarks onto pre-training corpora to study parametric knowledge vs. RAG. See the NanoKnow repository for:

• Pre-built qrels mapping SQuAD and NQ questions to FineWeb-Edu documents
• Scripts to project new benchmarks onto this index
• Scripts to evaluate nanochat checkpoints on in-corpus vs. out-of-corpus questions

Citation

@inproceedings{gu2026nanoknow,
  title={Projecting QA Datasets to FineWeb},
  author={Gu, Lingwei and Jedidi, Nour and Lin, Jimmy},
  booktitle={Proceedings of SIGIR},
  year={2026}
}

License

Apache 2.0