---
license: apache-2.0
task_categories:
- fill-mask
- text-generation
language:
- fr
- en
- de
- rm
- it
---
# Swiss Pretrain Data

This dataset provides a large collection of open-access and license-compliant Swiss data sources for language model training.

The dataset includes the following sources:
| Name | Internal ID | Tokens (B) | Description |
|------|-------------|------------|-------------|
| [Curia Vista](https://www.parlament.ch/en/ratsbetrieb/curia-vista) | `curiavista` | 0.5 | Legal and administrative documents from the Swiss database of parliamentary proceedings. |
| [enscheidsuche](http://entscheidsuche.ch/) | `enscheidsuche_html` | 4.5 | Swiss court decisions, sampled at 50% for balance. |
| [FineWeb-2](https://huggingface.co/datasets/HuggingFaceFW/fineweb-2) | `fineweb-2-swissfilter-quality_10-filterrobots` | 1.8 | A Swiss-only subset of FineWeb-2, filtered for top 10% quality. | 
| [Romansh Data](https://huggingface.co/datasets/swiss-ai/romansh_data) | `roh_data` |0.1 | Various Romansh-language data sources, as detailed in the [dataset documentation](https://huggingface.co/datasets/swiss-ai/romansh_data). |

All sampling, filtering and data-preparation scripts can be found in [our dedicated GitHub repository](https://github.com/swiss-ai/data-pipeline-pretrain).

Feel free to [reach out](mailto:sven.najem-meyer@epfl.ch) for any questions or suggestions 😊