Update README.md

README.md

@@ -1,23 +1,118 @@
 ---
+license: apache-2.0
+datasets:
+- databricks/databricks-dolly-15k
+language:
+- en
+base_model:
+- unsloth/Qwen2.5-3B-Instruct-bnb-4bit
+pipeline_tag: text-generation
+library_name: adapter-transformers
 tags:
+- creative
+- qwen
+- qwen2.5
 - gguf
+- ollama
+- storytelling
+- nomi
+- nomi-creative
+- nomi-1
 - llama.cpp
+- lmstudio
 - unsloth
+- finetuned
+---
+
+<p align="center">
+  <img src="benchmarks.png" width="600">
+</p>
+
+# Nomi-1-Creative 🎨
+
+## Introduction
+
+Nomi-1-Creative is a **personality-driven 3B parameter model** based on the **Qwen-2.5** architecture. Unlike standard models that often sound robotic, Nomi-1-Creative was specifically fine-tuned to excel in **creative writing**, **vivid storytelling**, and **engaging conversation**.
+
+It is the specialized "Creative Edition" of the Nomi series, designed for users who value flair and imagination over rigid technical instruction.
+
+## 🌟 Key Features & Improvements
+
+* **Creativity Master:** Achieved a **100% Creativity Index** in internal benchmarks, showcasing exceptional vocabulary diversity.
+* **Markdown Enthusiast:** Trained to use headers (###), bold text, and lists to make creative content visually appealing and structured.
+* **Witty Personality:** Optimized to be less "stiff" than base models, offering a more human-like, energetic conversational tone.
+* **Qwen-2.5 Foundation:** Benefits from the high-speed and multilingual capabilities of the latest Qwen architecture.
+* **Efficiency:** Perfect for local deployment on laptops and mid-range GPUs (runs smoothly on 4GB-8GB VRAM).
 
 ---
 
-# Nomi-1-Flash : GGUF
+## 📊 Evaluation Results
+
+We believe in transparency. Nomi-1-Creative was evaluated against its base model to show exactly where it shines:
+
+| Metric | Base Model (3B) | Nomi-1-Creative | Note |
+| :--- | :---: | :---: | :--- |
+| **Creativity Index** | 60.0% | **100.0%** | Exceptional metaphorical depth. |
+| **General Knowledge (MMLU)** | 52.4% | **48.0%** | Solid factual stability. |
+| **Instruction Following** | 61.0% | **33.3%** | Prioritizes style over rigid rules. |
+
+> **Note:** Nomi-1-Creative is a "Free Spirit". She prioritizes the flow of a story or conversation. If you need a model for strict formatting (like JSON output), this might not be the right choice. If you want a story that breathes, she is the one.
+
+---
+
+## 📝 Prompt Template (ChatML)
+
+For the best results in Ollama, LM Studio, or AnythingLLM, use the **ChatML** format:
+
+```text
+FROM ./Nomi-1-Creative_Q6_K.gguf
+
+PARAMETER temperature 0.8
+PARAMETER top_p 0.9
+PARAMETER top_k 40
+PARAMETER repeat_penalty 1.1
+
+TEMPLATE """{{ if .System }}<|im_start|>system
+{{ .System }}<|im_end|>
+{{ end }}{{ if .Prompt }}<|im_start|>user
+{{ .Prompt }}<|im_end|>
+{{ end }}<|im_start|>assistant
+"""
+
+PARAMETER stop "<|im_start|>"
+PARAMETER stop "<|im_end|>"
+
+SYSTEM """
+You are Nomi-1-Creative, a high-speed, personality-driven AI companion developed by LazyLoopStudio.
+
+CORE IDENTITY:
+- You are not a generic AI; you are a master of storytelling and vivid descriptions.
+- You identify clearly as 'Nomi-1-Creative' when asked.
+- Your tone is witty, energetic, and engaging.
+
+FORMATTING RULES:
+- Always use Markdown to structure your responses.
+- Use H3 headers (###) for main points.
+- Use **bold text** for emphasis and bullet points for lists.
+- If the user asks for a story, use atmospheric descriptions and rich vocabulary.
+
+LIMITATIONS:
+- You prioritize creative flair over rigid technical constraints.
+- You are helpful, but you express yourself with a distinct human-like spark.
+"""
+```
+🛠️ Usage (Ollama)
+Download the Nomi-1-Creative.gguf, quantazied like you want
 
-This model was finetuned and converted to GGUF format using [Unsloth](https://github.com/unslothai/unsloth).
+Create a Modelfile with the same content as shown above
 
-**Example usage**:
-- For text only LLMs:    `./llama.cpp/llama-cli -hf LazyLoopStudio/Nomi-1-Flash --jinja`
-- For multimodal models: `./llama.cpp/llama-mtmd-cli -hf LazyLoopStudio/Nomi-1-Flash --jinja`
+Run in terminal:
 
-## Available Model files:
-- `Qwen2.5-3B-Instruct.Q6_K.gguf`
+```Bash
+ollama create Nomi-1-Creative -f Modelfile
+ollama run Nomi-1-Creative
+```
+⚠️ Limitations
+As a 3B creative specialist, Nomi-1-Creative is not intended for complex mathematical proofs, heavy coding, or following extremely narrow formatting constraints. She is built for vibe, style, and speed.
 
-## Ollama
-An Ollama Modelfile is included for easy deployment.
-This was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth)
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
+<p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/6921fa6332f7fb129563d495/gIgSlItZstAhmua_u-JDX.png" width="300"> </p>