metadata
configs:
- config_name: default
data_files:
- split: train
path: data/clean-*.parquet
websoup
This dataset is a collection of websites that were scraped based off of their simplicity (size, quality, obfuscation level) and inlined into simple & large html files.
If you want to build a stochastic parrot that spews random web pages (including js and css in a single output) then this dataset is decent (not perfect by any means).
Load with:
from datasets import load_dataset
ds = load_dataset("lowerkey/websoup", split="train")