Skip to content

Dataset · recursal

Fanatic Fandom cleaned wiki corpus

recursal/Fanatic-Fandom

Large English text corpus of around 7.4B tokens sourced from fandom wikis.

What's inside

Collected by scraping publicly available, primarily English fandom wiki pages and then cleaned. Total volume is about 7.43B tokens (using the llama-2-7b-chat tokenizer).

What you can do with it

  • Pretraining or fine-tuning LLMs (such as LLaMA or RWKV)
  • Training generative models and LoRA adapters on fan-written texts and pop culture
Tasks
Text generation, Fill mask
Languages
English
Size
1M–10M rows
Downloads this month
591
Likes
8

How to load it

from datasets import load_dataset

ds = load_dataset("recursal/Fanatic-Fandom")