The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
fineweb16k
A slice of FineWeb (sample-10BT, one parquet file) retokenized with a 16,384-token byte-level BPE, for small-model optimizer research: 300M training tokens in three shards, 4M validation tokens in one, disjoint documents. The tokenizer was trained on 40,000 of the training documents.
Files are in modded-nanogpt's .bin layout: a 256-entry int32 header (magic 20240520, version 1,
token count) followed by uint16 tokens. Every document starts with <|endoftext|> (id 0).
manifest.json lists the sha256 of every file; images that use this data assert against it.
Built by data/build_data.py in the ar-shampoo-bench task repository. FineWeb is released
under the Open Data Commons Attribution License (ODC-By) v1.0, which this derivative inherits.
- Downloads last month
- 35