cad recode

Introduction
CAD-Recode is designed to transform point cloud data into 3D CAD models, generating valid Python code. This model builds on the Qwen2-1.5B architecture with an additional linear layer. It utilizes the original tokenizer from Qwen2.

Architecture
The model architecture consists of the Qwen2-1.5B base model with a single linear layer added to enable the transformation of point clouds into CAD code. This design allows it to maintain compatibility with the existing Qwen2 tokenizer, optimizing its functionality for reverse engineering tasks.

Training
Training details for CAD-Recode are not explicitly provided in the document. However, the model is based on fine-tuning the Qwen2-1.5B model, augmented with an additional linear layer for specific tasks related to CAD code generation from point cloud data.

Guide: Running Locally
To run CAD-Recode locally, follow these basic steps:

1. Clone the repository from GitHub: git clone https://github.com/filaPro/cad-recode
2. Install the necessary dependencies as listed in the repository.
3. Load the model and tokenizer from the repository.
4. Prepare your point cloud data and input it into the model to generate CAD code.

For enhanced performance, consider using a cloud GPU service such as AWS, Google Cloud, or Azure, which provides powerful computational resources suited for model inference and training.

License
The CAD-Recode project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (cc-by-nc-4.0), which allows for sharing and adaptation with attribution for non-commercial purposes.