|
--- |
|
license: apache-2.0 |
|
language: |
|
- en |
|
- de |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
 |
|
|
|
# Occiglot-7B-DE-EN |
|
|
|
> A [polyglot](https://en.wikipedia.org/wiki/Multilingualism#In_individuals) language model for the [Occident](https://en.wikipedia.org/wiki/Occident). |
|
> |
|
|
|
**Occiglot-7B-DE-EN** is a generative language model with 7B parameters for German and English and trained by the [Occiglot Research Collective](https://occiglot.github.io/occiglot/). |
|
It is based on [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) and trained on 114B tokens of additional multilingual and code data with a block size of 8,192 tokens per sample. |
|
Note that the model is a general-purpose base model and was not instruction-fine-tuned nor optimized for chat or other applications. We make an instruction tuned variant available as [occiglot-7b-de-en-instruct](https://huggingface.co/occiglot/occiglot-7b-de-en-instruct) |
|
|
|
This is the first release of an ongoing open research project for multilingual language models. |
|
If you want to train a model for your own language or are working on evaluations, please contact us or join our [Discord server](https://discord.gg/wUpvYs4XvM). **We are open for collaborations!** |
|
|
|
|
|
### Model details |
|
|
|
- **Continued-pretraining from:** [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) |
|
- **Model type:** Causal decoder-only transformer language model |
|
- **Languages:** English, German, and code. |
|
- **License:** [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0.html) |
|
- **Compute resources:** [HessianAI's 42](https://hessian.ai/) |
|
- **Contributors:** Manuel Brack, Patrick Schramowski, Pedro Ortiz, Malte Ostendorff, Fabio Barth, Georg Rehm, Kristian Kersting |
|
- **Research labs:** [Occiglot](https://occiglot.github.io/occiglot/) with support from [SAINT](https://www.dfki.de/en/web/research/research-departments/foundations-of-systems-ai) and [SLT](https://www.dfki.de/en/web/research/research-departments/speech-and-language-technology) |
|
- **Contact:** [Discord](https://discord.gg/wUpvYs4XvM) |
|
|
|
### How to use |
|
|
|
You can use this model directly with a pipeline for text generation. Since the generation relies on some randomness, we |
|
set a seed for reproducibility: |
|
|
|
```python |
|
>>> from transformers import pipeline, set_seed |
|
>>> generator = pipeline('text-generation', model='occiglot/occiglot-7b-de-en') |
|
>>> set_seed(42) |
|
>>> generator("Hallo, Ich bin ein Sprachmodell,", max_length=40, num_return_sequences=1) |
|
[{'generated_text': 'Hallo, Ich bin ein Sprachmodell, das dir bei der Übersetzung von Texten zwischen Deutsch und Englisch helfen kann. Wenn du mir einen Text in Deutsch'}] |
|
``` |
|
|
|
## Dataset |
|
|
|
The training data is the respective subset of the data used for [occiglot-7b-eu5](https://huggingface.co/occiglot/occiglot-7b-eu5), i.e. German plus English and Code. |
|
|
|
The data distribution by language (estimated) is as follows: |
|
- English: ~34% |
|
- Code: ~13% |
|
- German: ~52% |
|
|
|
The training data was prepared using [lm-datasets](https://github.com/malteos/lm-datasets). |
|
The exact data configuration is [here](https://huggingface.co/occiglot/occiglot-7b-eu5/blob/main/lm-datasets-config.yml). |
|
|
|
## Training settings |
|
|
|
- Continual pre-training on 128 x A100-80GB on [HessianAI's 42](https://hessian.ai/). |
|
- Framework: [Determined](https://www.determined.ai/) |
|
- Precision: bf16 |
|
- Optimizer: AdamW (lr: 0.00001, warmup_steps: 420) |
|
- Global batch size: 512 (with 8192 blocksize) split over 128 GPUs |
|
- Cosine Annealing with Warmup |
|
|
|
|
|
## Tokenizer |
|
|
|
Tokenizer is unchanged from [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1). |
|
|
|
## Evaluation |
|
|
|
Preliminary evaluation results can be found below. |
|
Please note that the non-English results are based on partially machine-translated datasets and English prompts ([Belebele](https://huggingface.co/datasets/facebook/belebele) and [Okapi framework](https://github.com/nlp-uoregon/Okapi)) and thus should be interpreted with caution, e.g., biased towards English model performance. |
|
Currently, we are working on more suitable benchmarks for Spanish, French, German, and Italian. |
|
|
|
<details> |
|
<summary>Evaluation results</summary> |
|
|
|
### German |
|
|
|
| | arc_challenge_de | belebele_de | hellaswag_de | mmlu_de | truthfulqa_de | avg | |
|
|:-------------------------------------|-------------------:|--------------:|---------------:|----------:|----------------:|---------:| |
|
| occiglot/occiglot-7b-eu5 | 0.493584 | 0.646667 | 0.666631 | 0.483406 | 0.251269 | 0.508311 | |
|
| occiglot/occiglot-7b-eu5-instruct | 0.529512 | 0.667778 | 0.685205 | 0.488234 | 0.286802 | 0.531506 | |
|
| occiglot/occiglot-7b-de-en | 0.50556 | 0.743333 | 0.67421 | 0.514633 | 0.26269 | 0.540085 | |
|
| occiglot/occiglot-7b-de-en-instruct | 0.54491 | 0.772222 | 0.688407 | 0.515915 | 0.310914 | 0.566474 | |
|
| LeoLM/leo-mistral-hessianai-7b | 0.474765 | 0.691111 | 0.682109 | 0.488309 | 0.252538 | 0.517766 | |
|
| mistralai/Mistral-7B-v0.1 | 0.476476 | 0.738889 | 0.610589 | 0.529567 | 0.284264 | 0.527957 | |
|
| mistralai/Mistral-7B-Instruct-v0.2 | 0.485885 | 0.688889 | 0.622438 | 0.501961 | 0.376904 | 0.535215 | |
|
|
|
|
|
</details> |
|
|
|
## Acknowledgements |
|
|
|
The model training was supported by a compute grant at the [42 supercomputer](https://hessian.ai/) which is a central component in the development of [hessian AI](https://hessian.ai/), the [AI Innovation Lab](https://hessian.ai/infrastructure/ai-innovationlab/) (funded by the [Hessian Ministry of Higher Education, Research and the Art (HMWK)](https://wissenschaft.hessen.de) & the [Hessian Ministry of the Interior, for Security and Homeland Security (HMinD)](https://innen.hessen.de)) and the [AI Service Centers](https://hessian.ai/infrastructure/ai-service-centre/) (funded by the [German Federal Ministry for Economic Affairs and Climate Action (BMWK)](https://www.bmwk.de/Navigation/EN/Home/home.html)). |
|
The curation of the training data is partially funded by the [German Federal Ministry for Economic Affairs and Climate Action (BMWK)](https://www.bmwk.de/Navigation/EN/Home/home.html) |
|
through the project [OpenGPT-X](https://opengpt-x.de/en/) (project no. 68GX21007D). |
|
|
|
|
|
## License |
|
|
|
[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0.html) |
|
|
|
## See also |
|
|
|
- https://huggingface.co/collections/occiglot/occiglot-eu5-7b-v01-65dbed502a6348b052695e01 |
|
|
|
|
