| 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). |