hydromodpy.results.catalog.dataset_loader#
Catalog-wide ML/DL dataset loader.
Joins scalar tables (simulations, parameters, metrics,
runs_environment) with optional per-simulation Zarr field arrays
into a single xarray.Dataset indexed by sim_id.
Field arrays stay lazy: each per-sim Zarr store is opened through
hydromodpy.results.simulation_group._open_simulation_lazy() which
wraps each registered field in a dask-backed xr.DataArray and
routes the resulting dataset through xr.decode_cf() for CF time
decoding. Per-sim datasets are concatenated along sim_id without
any np.stack / np.concatenate, so the loader scales to N=1000+
runs without OOM.
Classes
|
Build an |