File size: 746 Bytes
cf78da5 072edee |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
---
dataset_info:
features:
- name: input_ids
sequence: int32
- name: attention_mask
sequence: int8
- name: labels
sequence: int64
splits:
- name: train
num_bytes: 366738831440
num_examples: 13768540
- name: test
num_bytes: 479847540
num_examples: 18015
download_size: 105689403490
dataset_size: 367218678980
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
---
The data to train [fietje-2](https://huggingface.co/BramVanroy/fietje-2). It contains around 28B tokens taken from the Dutch parts of [wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) and [CulturaX](https://huggingface.co/datasets/uonlp/CulturaX). |