Install City2Graph¶

City2Graph supports Python 3.12–3.14. Install the core package for geospatial graph construction and network analysis, or choose a PyTorch extra when you need PyTorch Geometric tensors and Graph Neural Networks.

Choose an installation¶

Using pip¶

Core installation¶

Install the core package when you want to construct geospatial graphs, convert them to NetworkX or rustworkx, and run spatial network analysis:

pip install city2graph

This keeps the installation smaller by excluding PyTorch and PyTorch Geometric.

With PyTorch (CPU)¶

For PyTorch Geometric conversion and GNN development without a GPU, use:

pip install "city2graph[cpu]"

This installs PyTorch and PyTorch Geometric with CPU support.

With PyTorch and CUDA (GPU)¶

Choose the extra that matches the CUDA wheel required by your environment. For example, install CUDA 13.0 support with:

pip install "city2graph[cu130]"

Supported extras are cu126, cu128, and cu130. The cpu, cu126, and cu130 extras use PyTorch 2.13 or newer. Because PyTorch no longer publishes CUDA 12.8 wheels past 2.11, cu128 uses PyTorch 2.11.

Using conda-forge¶

Core installation¶

Install the core City2Graph package from conda-forge with:

conda install -c conda-forge city2graph

This does not install PyTorch or PyTorch Geometric.

With PyTorch (CPU)¶

To use PyTorch Geometric with the conda-forge package, add the CPU dependencies separately:

# Install City2Graph
conda install -c conda-forge city2graph

# Add PyTorch and PyTorch Geometric
conda install -c conda-forge pytorch pytorch_geometric

With PyTorch and CUDA (GPU)¶

For GPU support, select the appropriate PyTorch version and CUDA build. For example, install PyTorch 2.13.0 with CUDA 13.0 support with:

# Install City2Graph
conda install -c conda-forge city2graph

# Add PyTorch with CUDA support and PyTorch Geometric
conda install -c conda-forge pytorch=2.13.0=*cuda130*
conda install -c conda-forge pytorch_geometric

Browse the conda-forge PyTorch files page to select another supported version and CUDA variant. Ensure that PyTorch, PyTorch Geometric, and the installed CUDA runtime are mutually compatible.

Verify the installation¶

Check that City2Graph imports and report the installed version:

python -c "import city2graph as c2g; print(c2g.__version__)"

For a PyTorch-enabled installation, also verify that PyTorch Geometric is available:

python -c "import torch, torch_geometric; print(torch.__version__, torch_geometric.__version__)"

If these commands fail, confirm that the active Python environment is the same one in which City2Graph was installed.

Core dependencies¶

The core installation includes:

• NetworkX and rustworkx
• GeoPandas, Shapely, and OSMnx
• DuckDB
• libpysal and momepy
• Overture Maps
• SciPy and geopy

PyTorch and PyTorch Geometric are required only for PyG conversion and graph neural network workflows.