Installation

Requirements

• Python 3.8 or higher

• PyTorch 2.0+ (provided by cineca-ai module)

• CUDA-capable GPU (optional, for GPU acceleration)

Installation on Leonardo (CINECA)

1. Load required modules:

This step might need to be ignored in the case that conflict of the environment modules arises. In that case, you can install the dependencies uv via pip (see next section).

module load profile/deeplrn
module load cineca-ai

2. Clone the repository:

   git clone https://github.com/kardaneh/rtnn.git
   cd rtnn
   

3. Create virtual environment:

   uv venv --python 3.8 # python version must be compatible with PyTorch
   source .venv/bin/activate
   

4. Install missing dependencies:

   uv pip install xarray mpltex
   

5. Install the package:

   uv pip install -e .
   

6. Verify installation:

   python -c "import rtnn; print(rtnn.__version__)"
   rtnn --version
   

Installation on other systems

If you’re not on Leonardo, you can install all dependencies via pip (remember to uncomment the dependencies in pyproject.toml):

uv pip install -e .
# This will install all dependencies including PyTorch

Development Installation

For development, install with dev dependencies:

uv pip install -e ".[dev]"
pre-commit install

Dependencies

The following packages are provided by the cineca-ai module: - PyTorch 2.0.0a0 with CUDA 12.1 - NumPy, SciPy, Pandas, Matplotlib - scikit-learn, seaborn, rich, tqdm, tensorboard

Only these need to be installed: - xarray - mpltex