---
dataset_info:
  features:
  - name: score
    dtype: float64
  - name: text
    dtype: string
  - name: url
    dtype: string
  - name: dump
    dtype: string
  splits:
  - name: train
    num_bytes: 1784989190
    num_examples: 322895
  download_size: 869613568
  dataset_size: 1784989190
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This is an extract from the first 5M of Arabic FineWeb2 (500K selected then filtered for lang score over 0.9)