**Pygenesis Unity** is a specialized, fine-tuned LLM based on the Qwen 2.5 Coder architecture, optimized for **Unity game development and advanced C# scripting**. This repository contains the model weights in **GGUF format**, making it perfect for efficient local inference.
The model was fine-tuned using a curated dataset of nearly 1,000 high-quality, domain-specific instruction-response pairs sourced from official Unity documentation, advanced C# manuals, and high-tier synthetic data.
## Model Description
-**Developed by:** Pygenesis Association
-**Base Model:** Qwen 2.5 Coder
-**Format:** GGUF (Optimized for local deployment)
-**Specialization:** Unity Engine & C# Language
## Training Details
The training pipeline focused heavily on structure and logic:
* Full coverage of Unity Manual best practices, Monobehaviours, Scriptable Objects, and performance optimization.
* Advanced C# scripting patterns applied to game design.
* Instructional data distillation using frontier models to maximize code accuracy and deep reasoning capabilities.
## Intended Use
Pygenesis Unity is tailored for indie developers and technical leads who want a privacy-first, offline assistant to:
* Generate clean, optimized C# scripts for Unity loops and systems.
* Debug engine-specific code and refactor legacy scripts.
* Implement performance-oriented architecture (such as object pooling, memory management, or basic DOTS structures).
## How to Use
Since this model is provided in GGUF format, you can run it locally using various inference engines.
### Example using Llama.cpp CLI:
```bash
./llama-cli -m pygenesis-unity-qwen2.5-coder.gguf -p "Write a highly optimized C# script for an object pooling system in Unity." -n 512
