SKA SDP Distributed Self-Cal Prototype
Prototype self-calibration pipeline to distribute processing across HPC cluster nodes, allowing scalability for large datasets and a reduction in overall computation time.
The pipeline can currently perform major and minor cycles to produce a clean image from a calibrated zarr dataset but more functionality is being actively developed.
The project repository can be found here.
This pipeline uses the SKA SDP Processing Function Library (https://gitlab.com/ska-telescope/sdp/ska-sdp-func) for gridding visibilities and the SwiFTly algorithm (https://gitlab.com/ska-telescope/sdp/ska-sdp-exec-swiftly) to perform distributed imaging. Dask (https://github.com/dask/dask) is used as the distributed computing framework to handle the scaling of computations on a HPC cluster.
- API
- deconvolution
- swiftly
- visibility_bin
- visibility_io
- binning
- clean_beam
- gridding
- utils_fits
- pipeline_config
- tasks
bin_data()configure_and_setup_pipeline()configure_pipeline()distributed_hogbom()finish_setup()generate_clean_beam_array()generate_clean_beam_parameters()generate_facets_from_image_data()generate_facets_with_corrections()get_dataset_info()grid_visibilities()hogbom()initialise_dask_client()initialise_gridder()initialise_self_calibration()initialise_swiftly()load_data()predict_residual_visibilities()restore_model()save_image()setup_pipeline()split_image()validate_visibility_bins()
- utils
- constants
- units
Releases