An attempt to reproduce "CAD Recode"

• arXiv Paper
• GitHub Repository
• Dataset on Hugging Face
• Model on Hugging Face

• Python 3.9 or higher (tested with Python 3.11)

1. Ensure you have Python installed. You can check with:

   python --version  # Should be 3.9 or higher

2. Create a virtual environment:

   python -m venv venv

3. Activate the virtual environment:

   # On Mac/Linux
   source venv/bin/activate
   
   # On Windows
   # venv\Scripts\activate

4. Install required packages:

   pip install -r requirements.txt

Note: Apple Silicon (M1/M2/M3) users may encounter issues with OpenCascade and VTK dependencies. This is a known limitation when working with CadQuery on ARM-based Macs. Consider using an x86 machine or a cloud environment for development.

1. Set up Jupyter kernel:

   python -m ipykernel install --user --name=recode --display-name="ReCode"

2. Login to Hugging Face (recommended for better download rates):

   huggingface-cli login
   # You'll need a token from: https://huggingface.co/settings/tokens

3. Download the dataset:

   # Create datasets directory (it's gitignored)
   mkdir datasets
   
   # Download the dataset
   huggingface-cli download --repo-type dataset filapro/cad-recode --local-dir datasets/cad-recode

4. Open Jupyter notebook and select the "ReCode" kernel

The project uses:

• Python virtual environment (directory: venv/)
• Hugging Face datasets library
• Jupyter notebooks