Spherinator & HiPSter
Spherinator, HiPster and PEST are modules of an higher-level framework Project X for the analysis of
astrophysical data. The modulare design allows to use them independently or in combination.
Apache Parquet is used as internal data format, which allows to store
large amounts of data efficiently.
PEST preprocess simulation data and generate training data for Spherinator and HiPSter, including arbitrary single- and multi-channel images, 3D PPP and PPV cubes, and point clouds.
Spherinator is a Python package providing variational autoencoders (VAE) reduction generic data to a spherical latent space. It is designed to be used with PyTorch Lightning.
HiPSter creates the HiPS tilings and catalogs which can be visualized interactively on the surface of a sphere with Aladin Lite.
- PEST: Preprocessing Engine for Spherinator Training
- Spherinator: Model Training
- HiPSter: Generate HiPS and catalog
- API Reference
AutoencoderConsecutiveConv1DLayerConsecutiveConv2DLayerConsecutiveConvTranspose1DLayerConsecutiveConvTranspose2DLayerConvolutionalDecoder1DConvolutionalDecoder2DConvolutionalEncoder1DConvolutionalEncoder2DEmbeddingReconstructionHuggingFaceResNetEncoderHuggingFaceViTEncoderMLPSequentialUpsamplingDecoder2DVariationalAutoencoderWeightsProviderexport_onnx()yaml2model()
- Contributing
- Workflow Orchestration
- References