Parallelization is up and running :)
TopoPyScale
is improving and being tested by more users. New features, improvements and bug fixes justify a new release. The whole downscaling process gain speed by spreading the load on multicores. This was tested on laptops and servers.
WARNING: This new release requires Python >= 3.8. Otherwise, python runs into a memory leak and crashes.
New features:
- multicore, paralleization of
topo_scale
- tools to explore optimal number of clusters in
topo_sub
- implementation of feature importance for clustering. New field in config file needed
- added precipitation lapse-rate option, commit
Improvements:
- simplification of path management using the library pathlib
- remove all usage of
open_mfdataset()
orto_mfdataset
that are not as efficient as expected - simplification of
tmp
files during downscaling. 85a4149 - added error message in case upper pressure level is not high enough
Bug fixes:
to_netcdf()
default case bug fixed- plotting routines for variable mapping fixed
- fixed bug in time-batch handling of files
- bug in
fetch_WMO()
fixed