krisbailey's picture
Update README.md
da4cb1b verified
metadata
license: odc-by
task_categories:
  - text-generation
language:
  - en
tags:
  - falcon
  - 100M
  - parquet
  - web-refined
  - text-generation
  - clean-web-corpus
  - llm-pretrain
  - domain-agnostic
  - sentence-quality-filtered
  - huggingface-refinedweb
size_categories:
  - 100M<n<1B

falcon-refinedweb-100M

Dataset Description

This is a 100.0 Million token subset of krisbailey/falcon-refinedweb-1B, which is a subset of tiiuae/falcon-refinedweb.

Motivation

100M tokens is a standard size for:

  • CI/CD Pipelines: Fast enough to download and train for unit tests.
  • Debugging: Verifying training loops without waiting for hours.
  • Scaling Laws: The first step in a logarithmic scaling series (100M -> 1B -> 10B).

Dataset Details

  • Total Tokens: 99,999,978
  • Source: krisbailey/falcon-refinedweb-1B
  • Structure: First ~10% of the randomized 1B dataset.
  • Format: Parquet (Snappy compression) - Single File
  • Producer: Kris Bailey (kris@krisbailey.com)

Usage

from datasets import load_dataset

ds = load_dataset("krisbailey/falcon-refinedweb-100M", split="train")
print(ds[0])

Citation

@article{penedo2023refinedweb,
  title={The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only},
  author={Penedo, Guilherme and Malartic, Quentin and Hesslow, Daniel and Cojocaru, Ruxandra and Cappelli, Alessandro and Alobeidli, Hamza and Pannier, Baptiste and Almazrouei, Ebtesam and Launay, Julien},
  journal={arXiv preprint arXiv:2306.01116},
  year={2023}
}