Torch utilities for training neural networks in gravitational wave physics applications.
Please visit our documentation page to see descriptions and examples of the functions and modules available in ml4gw.
We also have an interactive Jupyter notebook demonstrating much of the core functionality available here.
To run this notebook, download it from the above link and follow the instructions within it to install the required packages.
See also the documentation page for the tutorial to look
through it without running the code.
You can install ml4gw with pip:
pip install ml4gwTo build with a specific version of PyTorch/CUDA, please see the PyTorch installation instructions here to see how to specify the desired torch version and --extra-index-url flag. For example, to install with torch 2.5.1 and CUDA 11.8 support, you would run
pip install ml4gw torch==2.5.1--extra-index-url=https://download.pytorch.org/whl/cu118If you want to develop ml4gw, you can use uv to install the project in editable mode.
For example, after cloning the repository, create a virtualenv using
uv venv --python=3.11Then sync the dependencies from the uv lock file using
uv sync --all-extrasCode changes can be tested using
uv run pytestSee contribution guide for more details.
If you come across errors in the code, have difficulties using this software, or simply find that the current version doesn't cover your use case, please file an issue on our GitHub page, and we'll be happy to offer support. If you want to add feature, please refer to the contribution guide for more details. We also strongly encourage ML users in the GW physics space to try their hand at working on these issues and joining on as collaborators! For more information about how to get involved, feel free to reach out to ml4gw@ligo.mit.edu. By bringing in new users with new use cases, we hope to develop this library into a truly general-purpose tool that makes deep learning more accessible for gravitational wave physicists everywhere.
We are grateful for the support of the U.S. National Science Foundation (NSF) Harnessing the Data Revolution (HDR) Institute for Accelerating AI Algorithms for Data Driven Discovery (A3D3) under Cooperative Agreement No. PHY-2117997.