README.md

---
license: apache-2.0
language:
- en
- de
library_name: transformers
pipeline_tag: text-generation
---

![SauerkrautLM](images/hero.png "SauerkrautLM-7b-HerO-multilingual")
## VAGO solutions SauerkrautLM
Introducing SauerkrautLM-v1 - Your German Language Powerhouse! 

We are thrilled to unveil our **very first release**, **SauerkrautLM-v1**. This remarkable creation marks a significant milestone as it is specifically **tailored for the German-speaking community**. In a landscape where German language models are scarce, we are proud to offer a solution that fills this void. 
What sets SauerkrautLM-v1 apart is its versatility. Whether you are an individual looking to harness its capabilities for personal use or a business seeking to integrate it into your projects, our model is designed to accommodate all. It operates under the Apache 2.0 License, providing you with the freedom to explore its potential in both private and commercial applications. 
Performance is at the heart of SauerkrautLM-v1. We put it to the **test using a customized version of MT-Bench for the German language**, and the results speak volumes. It currently stands as the most robust German Language Model on Hugging Face (based on german mt-bench results), showcasing its exceptional capabilities. Rest assured, this model is here to shine and set new standards. And the best thing is it comes in four different sizes (3B, 7B, 13B, 70B) to address your individual needs. 
Our model's journey began with meticulous training using an **augmented dataset within the QLoRA approach**. This is just the beginning of our model series, promising even more innovative and powerful solutions in the future. 

Join us on this exciting adventure as we redefine the possibilities of language modeling for the German-speaking world. 
SauerkrautLM-v1 is here to empower your language-related endeavors like never before. 

## All HerO Models

| Model | HF    | GPTQ  | GGUF  | AWQ  |
|-------|-------|-------|-------|-------|
| SauerkrautLM-7b-HerO   | [Link](https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-HerO) | coming soon | coming soon | coming soon |
| SauerkrautLM-7b-HerO-multilingual   | [Link](https://huggingface.co/VAGOsolutions/SauerkrautLM-7b-HerO-multilingual) | coming soon | coming soon | coming soon |

## Model Details
**SauerkrautLM-7b-HerO**

**Training Dataset:**

SauerkrautLM-7b-HerO-multilingual was trained with mix of German data augmentation and translated data. 
We found, that only a simple translation of training data can lead to unnatural German phrasings. 
Data augmentation techniques were used to grant grammatical, syntactical correctness and a more natural German wording in our training data. 

**Merge Procedure:**

SauerkrautLM-7b-HerO was merged on 1 A100 with [mergekit](https://github.com/cg123/mergekit).
The merged model contains [OpenHermes-2.5-Mistral-7B](https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B) and [Open-Orca/Mistral-7B-OpenOrca](https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca).
We used the gradient SLURP method.


- **Model Type:** SauerkrautLM-7b-HerO is an auto-regressive language model based on the transformer architecture 
- **Language(s):** English, German
- **License:** APACHE 2.0
- **Contact:** [Website](https://vago-solutions.de/#Kontakt) [David Golchinfar](mailto:golchinfar@vago-solutions.de)

**Prompt Template:**
```
<|im_start|>system
Du bist Sauerkraut-HerO, ein großes Sprachmodell, das höflich und kompetent antwortet. Schreibe deine Gedanken Schritt für Schritt auf, um Probleme sinnvoll zu lösen.
<|im_end|>
<|im_start|>user
Wie geht es dir?<|im_end|>
<|im_start|>assistant
Mir geht es gut!<|im_end|>
<|im_start|>user
Bitte erkläre mir, wie die Zusammenführung von Modellen durch bestehende Spitzenmodelle profitieren kann.<|im_end|>
<|im_start|>assistant
```
## Evaluation
**MT-Bench (German)**

 ![First Turn](images/de-1turn.png "First Turn")
 ![Second Turn](images/de-2turn.png "Second Turn")
 ![Average](images/de-avg.png "Average")
 
**MT-Bench (English)**

 ![First Turn](images/eng-1turn.png "First Turn")
 ![Second Turn](images/eng-2turn.png "Second Turn")
 ![Average](images/eng-avg.png "Average")

**Language Model evaluation Harness**
```
|arc_challenge       |      0|acc     | 0.5555|±  |0.0145|
|                    |       |acc_norm| 0.5956|±  |0.0143|
|arc_easy            |      0|acc     | 0.8388|±  |0.0075|
|                    |       |acc_norm| 0.8262|±  |0.0078|
|boolq               |      1|acc     | 0.8725|±  |0.0058|
|copa                |      0|acc     | 0.9100|±  |0.0288|
|hellaswag           |      0|acc     | 0.6285|±  |0.0048|
|                    |       |acc_norm| 0.8125|±  |0.0039|
|lambada_openai_mt_de|      0|ppl     |45.7314|±  |2.8280|
|                    |       |acc     | 0.4141|±  |0.0069|
|lambada_standard    |      0|ppl     | 3.5467|±  |0.0779|
|                    |       |acc     | 0.6922|±  |0.0064|
|multirc             |      1|acc     | 0.1459|±  |0.0114|
|openbookqa          |      0|acc     | 0.3640|±  |0.0215|
|                    |       |acc_norm| 0.4600|±  |0.0223|
|piqa                |      0|acc     | 0.8123|±  |0.0091|
|                    |       |acc_norm| 0.8281|±  |0.0088|
|race                |      1|acc     | 0.4507|±  |0.0154|
|rte                 |      0|acc     | 0.7040|±  |0.0275|
|truthfulqa_mc       |      1|mc1     | 0.3329|±  |0.0165|
|                    |       |mc2     | 0.4915|±  |0.0150|
|webqs               |      0|acc     | 0.1924|±  |0.0087|
|wic                 |      0|acc     | 0.5752|±  |0.0196|
|winogrande          |      0|acc     | 0.7301|±  |0.0125|
|wsc                 |      0|acc     | 0.6154|±  |0.0479|
|drop                |      1|em      | 0.2140|±  |0.0042|
|                    |       |f1      | 0.4011|±  |0.0041|
|triviaqa            |      3|em      | 0.6259|±  |0.0036|
|wmt16-de-en         |      0|bleu    |39.2043|±  |0.3982|
|                    |       |chrf    | 0.6316|±  |0.0029|
|                    |       |ter     | 0.4816|±  |0.0054|
|wmt16-en-de         |      0|bleu    |25.5745|±  |0.3492|
|                    |       |chrf    | 0.5331|±  |0.0030|
|                    |       |ter     | 0.6463|±  |0.0039|
|xnli_de             |      0|acc     | 0.4547|±  |0.0070|
|xnli_en             |      0|acc     | 0.5595|±  |0.0070|
```
**BBH**
```  
|                      Task                      |Version|       Metric        |Value |   |Stderr|
|------------------------------------------------|------:|---------------------|-----:|---|-----:|
|bigbench_causal_judgement                       |      0|multiple_choice_grade|0.6053|±  |0.0356|
|bigbench_date_understanding                     |      0|multiple_choice_grade|0.6992|±  |0.0239|
|bigbench_disambiguation_qa                      |      0|multiple_choice_grade|0.3721|±  |0.0302|
|bigbench_geometric_shapes                       |      0|multiple_choice_grade|0.1671|±  |0.0197|
|                                                |       |exact_str_match      |0.1003|±  |0.0159|
|bigbench_logical_deduction_five_objects         |      0|multiple_choice_grade|0.2540|±  |0.0195|
|bigbench_logical_deduction_seven_objects        |      0|multiple_choice_grade|0.2043|±  |0.0152|
|bigbench_logical_deduction_three_objects        |      0|multiple_choice_grade|0.4667|±  |0.0289|
|bigbench_movie_recommendation                   |      0|multiple_choice_grade|0.3700|±  |0.0216|
|bigbench_navigate                               |      0|multiple_choice_grade|0.4970|±  |0.0158|
|bigbench_reasoning_about_colored_objects        |      0|multiple_choice_grade|0.6965|±  |0.0103|
|bigbench_ruin_names                             |      0|multiple_choice_grade|0.4152|±  |0.0233|
|bigbench_salient_translation_error_detection    |      0|multiple_choice_grade|0.1443|±  |0.0111|
|bigbench_snarks                                 |      0|multiple_choice_grade|0.6464|±  |0.0356|
|bigbench_sports_understanding                   |      0|multiple_choice_grade|0.6846|±  |0.0148|
|bigbench_temporal_sequences                     |      0|multiple_choice_grade|0.3150|±  |0.0147|
|bigbench_tracking_shuffled_objects_five_objects |      0|multiple_choice_grade|0.2168|±  |0.0117|
|bigbench_tracking_shuffled_objects_seven_objects|      0|multiple_choice_grade|0.1537|±  |0.0086|
|bigbench_tracking_shuffled_objects_three_objects|      0|multiple_choice_grade|0.4667|±  |0.0289|
```  


## Disclaimer
We must inform users that despite our best efforts in data cleansing, the possibility of some such content slipping through cannot be entirely ruled out.
However, we cannot guarantee consistently appropriate behavior. Therefore, if you encounter any issues or come across inappropriate content, we kindly request that you inform us through the contact information provided.
Additionally, it is essential to understand that the licensing of these models does not constitute legal advice. We are not held responsible for the actions of third parties who utilize our models. These models may be employed for commercial purposes, and the Apache 2.0 remains applicable and is included with the model files.
 
## Contact
If you are interested in customized LLMs for business applications, please get in contact with us via our website or contact us at [Dr. Daryoush Vaziri](mailto:vaziri@vago-solutions.de). We are also grateful for your feedback and suggestions.
 
## Collaborations
We are also keenly seeking support and investment for our startup, VAGO solutions, where we continuously advance the development of robust language models designed to address a diverse range of purposes and requirements. If the prospect of collaboratively navigating future challenges excites you, we warmly invite you to reach out to us.

## Acknowledgement
Many thanks to [OpenOrca](https://huggingface.co/Open-Orca) and [teknium](https://huggingface.co/teknium) for providing such valuable models to the Open-Source community.