--- language: - de - en - it - fr - pt - nl - ru - ar - es license: apache-2.0 tags: - spectrum model-index: - name: SauerkrautLM-Nemo-12b-Instruct results: - task: type: text-generation name: Text Generation dataset: name: ENEM Challenge (No Images) type: eduagarcia/enem_challenge split: train args: num_few_shot: 3 metrics: - type: acc value: 70.05 name: accuracy source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: BLUEX (No Images) type: eduagarcia-temp/BLUEX_without_images split: train args: num_few_shot: 3 metrics: - type: acc value: 58.41 name: accuracy source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: OAB Exams type: eduagarcia/oab_exams split: train args: num_few_shot: 3 metrics: - type: acc value: 52.53 name: accuracy source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: Assin2 RTE type: assin2 split: test args: num_few_shot: 15 metrics: - type: f1_macro value: 92.65 name: f1-macro source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: Assin2 STS type: eduagarcia/portuguese_benchmark split: test args: num_few_shot: 15 metrics: - type: pearson value: 75.99 name: pearson source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: FaQuAD NLI type: ruanchaves/faquad-nli split: test args: num_few_shot: 15 metrics: - type: f1_macro value: 83.18 name: f1-macro source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: HateBR Binary type: ruanchaves/hatebr split: test args: num_few_shot: 25 metrics: - type: f1_macro value: 81.98 name: f1-macro source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: PT Hate Speech Binary type: hate_speech_portuguese split: test args: num_few_shot: 25 metrics: - type: f1_macro value: 75.67 name: f1-macro source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: tweetSentBR type: eduagarcia/tweetsentbr_fewshot split: test args: num_few_shot: 25 metrics: - type: f1_macro value: 72.31 name: f1-macro source: url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct name: Open Portuguese LLM Leaderboard --- ![SauerkrautLM-Nemo-12b-Instruct]( https://vago-solutions.ai/wp-content/uploads/2024/07/Sauerkraut-Nemo.png "SauerkrautLM-Nemo-12b-Instruct") ## VAGO solutions SauerkrautLM-Nemo-12b-Instruct **Fine-tuned Model** - *to showcase the potential of resource-efficient Fine-Tuning of Large Language Models using **Spectrum Fine-Tuning*** Introducing **SauerkrautLM-Nemo-12b-Instruct** – our Sauerkraut version of the powerful [mistralai/Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407)! - Fine-tuning on German-English data with [**Spectrum**](https://github.com/cognitivecomputations/spectrum) Fine-Tuning **targeting 25% of the layers.** - Utilized unique German-English Sauerkraut Mix v2 - Implemented bespoke, precision-engineered fine-tuning approach # Table of Contents 1. [Overview of all SauerkrautLM-Nemo-12b-Instruct](#all-SauerkrautLM-Nemo-12b-Instruct) 2. [Model Details](#model-details) - [Training procedure](#training-procedure) 3. [Evaluation](#evaluation) 5. [Disclaimer](#disclaimer) 6. [Contact](#contact) 7. [Collaborations](#collaborations) 8. [Acknowledgement](#acknowledgement) ## All SauerkrautLM-Nemo-12b-Instruct | Model | HF | EXL2 | GGUF | AWQ | |-------|-------|-------|-------|-------| | SauerkrautLM-Nemo-12b-Instruct | [Link](https://huggingface.co/VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct) | coming soon | coming soon | coming soon | ## Model Details **SauerkrautLM-Nemo-12b-Instruct** - **Model Type:** SauerkrautLM-Nemo-12b-Instruct is a fine-tuned Model based on [mistralai/Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407) - **Language(s):** German, English - **License:** Apache 2.0 - **Contact:** [VAGO solutions](https://vago-solutions.ai) ## Training Procedure This model showcases the potential of resource-efficient fine-tuning of large language models using Spectrum Fine-Tuning. Here's a brief on the procedure: **Fine-tuning on German-English Data**: - Utilized Spectrum Fine-Tuning, targeting 25% of the model's layers - Introduced the model to a unique German-English Sauerkraut Mix v2 - Implemented a bespoke, precision-engineered fine-tuning approach **Sauerkraut Mix v2**: - Premium Dataset for Language Models, focusing on German and English - Meticulously selected, high-quality dataset combinations - Cutting-edge synthetic datasets created using proprietary, high-precision generation techniques ## Objective and Results The primary goal of this training was to demonstrate that with Spectrum Fine-Tuning targeting 25% of the layers, a 12 billion parameter model can significantly enhance the capabilities while using a fraction of the resources of the classic fine-tuning approach. The model has substantially improved skills in German and English, as demonstrated by impressive benchmarks on the new Hugging Face leaderboard. At the same time, our fine-tuning improved skills in all other languages that Nemo can speak, showing inter-language effects in LLM performance. **Spectrum Fine-Tuning can efficiently enhance a large language model's capabilities in multiple languages while preserving the majority of its previously acquired knowledge.** ## Evaluation **AGIEVAL** ![SauerkrautLM-Nemo-12b-Instruct-AGIEVAL]( https://vago-solutions.ai/wp-content/uploads/2024/07/agieval2.png "SauerkrautLM-Nemo-12b-Instruct-AGIEVAL") **GPT4ALL** ![SauerkrautLM-Nemo-12b-Instruct-GPT4ALL]( https://vago-solutions.ai/wp-content/uploads/2024/07/gpt4all2.png "SauerkrautLM-Nemo-12b-Instruct-GPT4ALL") **TRUTHFULQA** ![SauerkrautLM-Nemo-12b-Instruct-TRUTHFULQA]( https://vago-solutions.ai/wp-content/uploads/2024/07/tqa2.png "SauerkrautLM-Nemo-12b-Instruct-TRUTHFULQA") **OPENLEADERBOARD 2** ![SauerkrautLM-Nemo-12b-Instruct-OPENLEADERBOARD]( https://vago-solutions.ai/wp-content/uploads/2024/07/hf2.png "SauerkrautLM-Nemo-12b-Instruct-OPENLEADERBOARD") **MMLU 5-Shot** ![SauerkrautLM-Nemo-12b-Instruct-MMLU]( https://vago-solutions.ai/wp-content/uploads/2024/07/mmlu.png "SauerkrautLM-Nemo-12b-Instruct-MMLU") ## Disclaimer We must inform users that despite our best efforts in data cleansing, the possibility of uncensored 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. ## Contact If you are interested in customized LLMs for business applications, please get in contact with us via our website. 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 at [VAGO solutions](https://vago-solutions.ai) ## Acknowledgement Many thanks to [Mistral AI](https://huggingface.co/mistralai) for providing such a valuable model to the Open-Source community. # Open Portuguese LLM Leaderboard Evaluation Results Detailed results can be found [here](https://huggingface.co/datasets/eduagarcia-temp/llm_pt_leaderboard_raw_results/tree/main/VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct) and on the [🚀 Open Portuguese LLM Leaderboard](https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard) | Metric | Value | |--------------------------|---------| |Average |**73.64**| |ENEM Challenge (No Images)| 70.05| |BLUEX (No Images) | 58.41| |OAB Exams | 52.53| |Assin2 RTE | 92.65| |Assin2 STS | 75.99| |FaQuAD NLI | 83.18| |HateBR Binary | 81.98| |PT Hate Speech Binary | 75.67| |tweetSentBR | 72.31|