Installation

Sage is currently intended for local editable installs. A CUDA-capable GPU is strongly recommended for on-the-fly waveform generation, training, and large injection studies.

Requirements: Python ≥ 3.9, PyTorch ≥ 2.1, CUDA (optional but recommended).

Option A — conda (recommended for GPU clusters)

The utils/environment.yml file defines the full conda environment (Python 3.11, LALSuite, PyCBC, GWpy, and JupyterLab). utils/create_env.sh automates the three-step setup: conda packages → PyTorch CUDA wheel → editable Sage install.

git clone https://github.com/nnarenraju/sage.git
cd sage/utils
bash create_env.sh
conda activate sage

Option B — pip only

git clone https://github.com/nnarenraju/sage.git
cd sage
python -m pip install -r requirements.txt
python -m pip install -e .

Note

PyTorch installation can depend on your CUDA version. If needed, install the appropriate PyTorch build first using the command from pytorch.org, then install the remaining requirements.

Verifying the installation

Run a quick smoke test to confirm the package is importable:

python -c "import sage; print('Import: OK')"

To run the full test suite:

pytest tests/ -v

For a broad syntax check across the package:

python -m py_compile $(find sage -name '*.py')