Clone the reposi)tory
git clone https://github.com/FarrOutLab/GWInferno.gitRecommended to use conda to set up your environment with python>=3.9:
For CPU usage only create an environment and install requirements and GWInferno with:
cd gwinferno
conda create -n gwinferno python=3.10
conda activate gwinferno
conda install -c conda-forge numpyro h5py
pip install --upgrade pip
pip install -r pip_requirements.txt
python -m pip install .To enable JAX access to CUDA enabled GPUs we need to specify specific versions to install (See JAX installation instructions for more details). For a GPU enabled environment use:
cd gwinferno
conda create -n gwinferno_gpu python=3.10
conda activate gwinferno_gpu
pip install numpyro[cuda] -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
conda install h5py cudnn
pip install -r pip_requirements.txt
python -m pip install .Install popsummary dependency:
In a folder outside the gwinferno folder, run (make sure gwinferno conda environemnt is active):
git clone https://git.ligo.org/zoheyr-doctor/popsummary.git
cd popsummary
pip install .Given a catalog of GW Posterior samples in standardized PESummary format, defined by catalog.json file to run population inference with GWInferno one must write a yaml file defining the desired model, hyperparameters, and other auxiliary configuration arguments.
For example a simple config.yml file that defines a Truncated Powerlaw population model over primary masses (i.e. mass_1) we have:
# Run Label
label: Truncated_Powerlaw_mass_1
# Population Parameters, Models, HyperParameters, and Priors
models:
mass_1:
model: gwinferno.numpyro_distributions.Powerlaw
hyper_params:
alpha:
prior: numpyro.distributions.Normal
prior_params:
loc: 0.0
scale: 3.0
minimum:
prior: numpyro.distributions.Uniform
prior_params:
low: 2.0
high: 10.0
maximum:
prior: numpyro.distributions.Uniform
prior_params:
low: 50.0
high: 100.0
# Sampler Configuration Args
sampler:
kernel: NUTS
kernel_kwargs:
dense_mass: true
mcmc_kwargs:
num_warmup: 500
num_samples: 1500
num_chains: 1
# Data Configuration Args
data:
catalog_summary_json: /path/to/catalog/summary/file/catalog.jsonwith this file written and ready to go run to perform inference!
gwinferno_run_from_config.py config.ymlReleased under MIT by @FarrOutLab.
If GWInferno is used please cite this paper
We make use of and build upon other open source software projects, please cite them if using GWInferno.
- Bruce Edelman -- [email protected]
- Jaxen Godfrey -- [email protected]
- Gino Carrillo -- [email protected]
- Ben Farr -- [email protected]