Installation

Requirements

These match pyproject.toml (requires-python and runtime dependencies):

Python 3.9+
PyTorch 2.0+
pandas 1.5+
numpy 1.21+ (currently pinned as < 2.0 in the package metadata)
scipy 1.7+
plotly 5.0+
NetworkX 2.6+
scikit-learn 1.0+
statsmodels 0.13+
tqdm 4.60+

Install from PyPI

pip install deepcausalmmm

Install from GitHub

For the latest commit on the default branch:

pip install git+https://github.com/adityapt/deepcausalmmm.git

Optional dependencies

The package declares a test extra (pytest tooling) for contributors:

pip install deepcausalmmm[test]
# or, from a clone:
pip install -e .[test]

Documentation builds use docs/requirements.txt (Sphinx, theme, and doc import dependencies).

Development Installation

For contributors and developers:

# Clone repository
git clone https://github.com/adityapt/deepcausalmmm.git
cd deepcausalmmm

# Install in development mode
pip install -e .

# With test dependencies (pytest, coverage)
pip install -e .[test]

Verify Installation

Test that the package is installed correctly:

from deepcausalmmm import DeepCausalMMM, get_device
from deepcausalmmm.core import get_default_config

print("DeepCausalMMM package imported successfully!")
print(f"Device: {get_device()}")

GPU Support

DeepCausalMMM automatically detects and uses CUDA if available. For GPU acceleration:

Install PyTorch with CUDA support:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Verify GPU detection:

from deepcausalmmm.utils.device import get_device
device = get_device()
print(f"Using device: {device}")

Docker Installation

A Docker image will be available soon for easy deployment:

# Coming soon
docker pull deepcausalmmm/deepcausalmmm:latest

Troubleshooting

Import Errors: Make sure all dependencies are installed. Try reinstalling with pip install --upgrade.
CUDA Issues: Ensure your PyTorch installation matches your CUDA version. Check with torch.cuda.is_available().
Memory Issues: For large datasets, consider reducing batch size or using gradient checkpointing.
Performance Issues: Enable mixed precision training and use appropriate hardware (GPU recommended for large models).