---
dataset_info:
  features:
  - name: text
    dtype: string
  - name: timestamp
    dtype: string
  - name: url
    dtype: string
  - name: source
    dtype: string
  splits:
  - name: train
    num_bytes: 3901684924
    num_examples: 741465
  download_size: 1933038896
  dataset_size: 3901684924
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This dataset was created in the following way:

The CulturaX (https://huggingface.co/datasets/uonlp/CulturaX) dataset was loaded, specifically the Greek subset.
Then a subset of the Greek subset was taken, specifically to contain about 0.5 B tokens.
This is to be used as the general domain corpus for the Greek language for continual-pretraining.