Update README.md

README.md

@@ -14,38 +14,35 @@ datasets:
 - Christoph911/German-legal-SQuAD
 - philschmid/test_german_squad
 ---
-# Llama 2 13b Chat German | GGUF
+# Summary
 
-This repository contains the model [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german) converted in GGUF format.
-The original model was created by [jphme](https://huggingface.co/jphme). It is a fine-tune of Meta's [Llama2 13b Chat](https://huggingface.co/meta-llama/Llama-2-13b-chat), trained on a dataset of instructions in German language. 
+This repository contains the model [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german) in GGUF format, for fast and easy inference with llama.cpp and similar LLM inference tools.
+This model was created and trained by [jphme](https://huggingface.co/jphme). It is a fine-tuned variant of Meta's [Llama2 13b Chat](https://huggingface.co/meta-llama/Llama-2-13b-chat) with a compilation of multiple instruction datasets in German language. 
 
-## Model Profile
-The Model Profile describes relevant properties of the model in a standardized, digestible and easy-to-read way.
+## Model Profile (.aiml)
 
-|Property|Value|
+|Attribute|Details|
 |----------------------------|--------------------------------------------------------------------------------------------------------------|
+| **ID**                     | 1                                                                                                            |
 | **Model**                  | [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german)                        |
-| **Creator**                | [jphme](https://huggingface.co/jphme)                                                                        |
-| **Type**                   | LLM                                                                                                          |
-| **Use Case**               | Text Generation                                                                                              |
-| **Class**                  | Fine-tuned model                                                                                             |
-| **Parameters**             | 70B                                                                                                          |
-| **Finetuning Method**      | Full                                                                                                         |
-| **Finetuning Datasets**    | Prorietary German Conversation Dataset, German SQuAD, German legal SQuAD data, augmented with "wrong" contexts, to improve factual RAG. For details see original model link. |
+| **Source**                 | https://huggingface.co/                                                                                      |
+| **Type**                   | Large Language Model                                                                                         |
+| **Functions**              | Text Generation                                                                                              |
+| **Filetype**               | GGUF                                                                                                         |
+| **Modifications**          | Quantized FP16-accuracy model to 8, 4 (K_M) and 5 (K_M) Bit                                                                                                         |
 | **Architecture**           | Transformers                                                                                                 | 
-| **File Format**            | GGUF                                                                                                         |
+| **Model**                  | [jphme/Llama-2-13b-chat-german](https://huggingface.co/jphme/Llama-2-13b-chat-german)                        |
+| **Creator**                | [jphme](https://huggingface.co/jphme)                                                                        |
+| **Compression**            | GGUF 8 Bit<br/>GGUF 5 Bit (K_M)                                                                              |
+| **Tuning**                 | full_finetune                                                                                                |
+| **Tuning Datasets**        | Prorietary German Conversation Dataset, German SQuAD, German legal SQuAD data, augmented with "wrong" contexts to improve factual RAG. |
 | **Quantization Types**     | 8 Bit <br>5 Bit (K_M)                                                                                        |
 | **Utilized Tools**         | llama.cpp (Commit 9e20231) for quantization to 8, 5 and 4 bit                                                |
 | **Deployment**             | #/bin/sh<br/>chmod +x run.sh && ./run.sh # Script to be published soon                                       |
 
-**Details**
-We are using **AI Markdown Language *(.aiml)*** to standardize the description of our AI models, to make deploying and inferencing as easy and quick as possible.
-
-AIDF is a novel format proposed by Morgendigital that aims to summarize and standardize the information needed to run an AI model. More information on AIDF is coming soon.
-
-**Metadata**
-*Profile Type:* General AI Profile
-*Profile Version:* AIDF v1.0
+**About AI Model Profiles (.aiml Files)**
+We are experimenting with a novel descriptive format called **AI Markdown Language *(.aiml)***. An AIML file contains all the necessary configuration parameters and rules to automatically deploy and serve production-ready AI models.
+Our aim is to unify and standardize the description of AI models and make deployments and inference of AI much easier and quicker. 
 
 ## Replicate
 1. Clone and install llama.cpp *(Commit: 9e20231)*.