Quick Start

Getting started

The User Guide covers the full pipeline from data download to training and evaluation. For complete, working examples of how Sage is configured and run in practice, see the scripts under runs/.

Colab tutorials

No local install needed — run Sage on a free Google Colab T4 GPU:

#

Notebook

Launch

1

Signal generation with IMRPhenomD

 Open In Colab

2

Realistic data simulation and whitening

 Open In Colab

3

Training and evaluating a GW detector

 Open In Colab

Repository layout

sage/
├── sage/
│   ├── architecture/       # Frontend, backend, attention, and full networks
│   ├── benchmark/          # Benchmark integrations and comparison utilities
│   ├── core/               # Config, logging, constants, interpolation
│   ├── data/
│   │   ├── noise/          # Real noise samplers and glitch handling
│   │   ├── primer/         # Data download and preparation utilities
│   │   ├── psd/            # PSD generation and loading
│   │   └── waveform/       # Parameter sampling, waveforms, projection, SNR
│   ├── dsp/                # FFT, whitening, PSDs, multirate, multibanding
│   ├── exec/               # Pipeline orchestration
│   ├── factory/            # Training, validation, schedulers, callbacks
│   ├── plotting/           # Diagnostic and publication plotting
│   ├── presets/            # Legacy configs and shared data configs
│   └── utils/              # Checkpointing, timing, Condor utilities
├── runs/                   # Run scripts for specific experiments
├── repro/                  # Reproducibility notebooks and configuration
├── notebooks/              # Exploratory notebooks
├── tests/                  # Lightweight tests and smoke checks
└── docs/                   # Sphinx/ReadTheDocs documentation source