---
license: bigscience-bloom-rail-1.0
datasets:
- OpenAssistant/oasst1
- LEL-A/translated_german_alpaca_validation
- deepset/germandpr
- oscar-corpus/OSCAR-2301
language:
- de
pipeline_tag: conversational
---

# Instruction-fine-tuned German language model (6B parameters; **early alpha version**)

Base model: [malteos/bloom-6b4-clp-german](https://huggingface.co/malteos/bloom-6b4-clp-german) [(Ostendorff and Rehm, 2023)](https://arxiv.org/abs/2301.09626)

Trained on:
- 20B additional German tokens (Wikimedia dumps and OSCAR 2023)
- [OpenAssistant/oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) (German subset)
- [LEL-A/translated_german_alpaca_validation](https://huggingface.co/datasets/LEL-A/translated_german_alpaca_validation)
- [LEL-A's version of deepset/germandpr](https://github.com/LEL-A/EuroInstructProject#instruct-germandpr-dataset-v1-german)

## Chat demo

[![Try out the chat demo](https://huggingface.co/malteos/bloom-6b4-clp-german-oasst-v0.1/resolve/main/chat-demo.png)](https://opengptx.dfki.de/chat/)

[https://opengptx.dfki.de/chat/](https://opengptx.dfki.de/chat/)

Please note that this a research prototype and may not be suitable for extensive use. 


## How to cite

If you are using our code or models, please cite [our paper](https://arxiv.org/abs/2301.09626):

```bibtex
@misc{Ostendorff2023clp,
  doi = {10.48550/ARXIV.2301.09626},
  author = {Ostendorff, Malte and Rehm, Georg},
  title = {Efficient Language Model Training through Cross-Lingual and Progressive Transfer Learning},
  publisher = {arXiv},
  year = {2023}
}

```

## License

[BigScience BLOOM RAIL 1.0](https://bigscience.huggingface.co/blog/the-bigscience-rail-license)


## Acknowledgements

This model was trained during the [Helmholtz GPU Hackathon 2023](https://www.fz-juelich.de/de/ias/jsc/aktuelles/termine/2023/helmholtz-gpu-hackathon-2023). 
We gratefully thank the organizers for hosting this event and the provided computing resources.