README.md

metadata

license: cc-by-4.0

gwBenchmarks Dataset Files

This repository hosts dataset artifacts for the gwBenchmarks codebase:

https://github.com/tousifislam/gwBenchmarks

The GitHub repository contains benchmark code, curation scripts, prompts, and evaluators. This Hugging Face dataset repository stores the binary benchmark artifacts that should not be committed to GitHub.

Directory Layout

Use one top-level directory per benchmark:

waveform/
remnant/
dynamics/
ringdown/
analytic/
validity/
template_bank/

Each benchmark directory should contain:

• the data files used by that benchmark
• a short README.md describing file names, shapes, and provenance
• checksums or generation details when available

Recommended file names match the paths expected by gwBenchmarks:

waveform/waveform_training.h5
waveform/waveform_validation.h5
remnant/remnant_training.h5
remnant/remnant_validation.h5
dynamics/dynamics_training.h5
dynamics/dynamics_validation.h5
ringdown/ringdown_training.h5
ringdown/ringdown_validation.h5
analytic/analytic_training.h5
analytic/analytic_validation.h5
validity/validity_training.h5
validity/validity_validation.h5
template_bank/f_amp.npy
template_bank/Aref_weights.npy
template_bank/bank_wf_params.npy
template_bank/bank_wf_params_test.npy
template_bank/calpha_grid_params.npy

Uploading New Benchmark Data

Clone this dataset repository, add files under the appropriate benchmark directory, and push:

git clone https://huggingface.co/datasets/GWagents/gwBenchmarks
cd gwBenchmarks
mkdir -p waveform
cp /path/to/waveform_training.h5 waveform/
cp /path/to/waveform_validation.h5 waveform/
git add waveform
git commit -m "Add waveform benchmark data"
git push

The repository uses Git LFS for binary artifacts including .h5, .hdf5, .npy, and .npz files.