Dataset Viewer

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.

FineWeb sample-350BT, GPT-2 tokens

Pre-tokenized FineWeb data for language-model pretraining.

Dataset details

  • Source: HuggingFaceFW/fineweb
  • FineWeb configuration: sample-350BT
  • Tokenizer: tiktoken GPT-2 (gpt2)
  • Token dtype: little-endian uint16
  • Format: NanoGPT binary shards, version 1
  • Total tokens: 361,159,416,052
  • Shards: 1 validation shard and 3,611 training shards

File format

Each .bin file contains a 1 KiB NanoGPT header followed by GPT-2 token IDs stored as uint16. The header fields are:

  1. magic number: 20240520
  2. format version: 1
  3. token count

Usage

These files are directly usable with NanoGPT-compatible shard loaders.

PicoDo expects a single headerless uint16 stream. Strip the headers and concatenate the training shards locally before using it with PicoDo.

Provenance

This is a tokenized derivative of HuggingFaceFW/fineweb, configuration sample-350BT. Please consult the upstream FineWeb dataset card and terms before using or redistributing this dataset.

Downloads last month
-