Blender L L M

Introduction

BlenderLLM is a large language model designed for computer-aided design (CAD) tasks, utilizing self-improvement techniques for enhanced performance. It is based on the Qwen2.5-Coder-7B-Instruct model and has been fine-tuned using the BlendNet dataset. This project is a collaborative effort from the School of Data Science (SDS) at the Chinese University of Hong Kong, Shenzhen.

Architecture

The architecture of BlenderLLM is built upon Qwen2.5-Coder-7B-Instruct, which serves as the foundational model. This base is enhanced through fine-tuning with the BlendNet dataset, specifically tailored for tasks involving CAD, rendering, and 3D modeling. The model is further optimized using self-improvement methods to ensure high performance.

Training

BlenderLLM's training involves fine-tuning the Qwen2.5-Coder-7B-Instruct model with the BlendNet dataset. Self-improvement techniques are employed to refine the model's capabilities in CAD-related tasks. The training process focuses on improving the model's efficiency and accuracy in generating code and renderings.

Guide: Running Locally

To run BlenderLLM locally, follow these steps:

1. Clone the Repository:
   Clone the BlenderLLM repository from GitHub.

2. Install Dependencies:
   Ensure you have the necessary libraries and dependencies installed. This usually involves setting up a Python environment with packages such as TensorFlow or PyTorch.

3. Download the Model:
   Download the BlenderLLM model files from the repository or Hugging Face model hub.

4. Run the Model:
   Execute the model code to start generating CAD-related outputs.

5. Utilize Cloud GPUs:
   For optimal performance, consider using cloud services like AWS, Google Cloud, or Azure to access powerful GPUs.

License

BlenderLLM is licensed under the Apache-2.0 License. This permits users to freely use, modify, and distribute the software, provided that any copies or substantial portions of the software include the original license.