genboostgpu

GENBoostGPU.

Two families of entry points live here:

• the local-genetic-variance engine (genboostgpu.lgv, genboostgpu.lsp, genboostgpu.sites), which needs only NumPy and optionally CuPy; and

• the legacy boosting elastic net (boosting_elastic_net, run_windows_with_dask, run_cpgs_*), which needs RAPIDS (cuDF, cuML, dask-cuda). Its final_r2 is an in-sample fit; see the docs.

Legacy names are imported lazily so import genboostgpu works without RAPIDS installed.