0.10.2

Fixed

• Move torch import inside s2vec related functions to avoid import errors when torch is not installed

0.10.1

Fixed

• S2Regionalizer failing with geopandas >= 1.0.0 #558

0.10.0

Added

• Open Benchmark for Spatial Representations (OBSR) - added benchmark module with 7 public datasets, implemented by the Kraina AI team (@mskaa3, @Repcak00, @Kubson900, @SzymonLeszkiewicz, @ofurman)
• New optional dependency group: datasets
• Weight-based examples sorting in documentation with mkdocs-nav-weight plugin

Changed

• Reduced line-length in example notebooks from 100 to 88
• Improved typing in the codebase

0.9.9

Added

• S2Vec model as an S2VecEmbedder implemented by @hubkrieb, proposed by Google Research team (Choudhury et al.)
• timm to optional dependencies for S2VecEmbedder

Changed

• Locked gtfs-kit version to 0.6.*

Fixed

• Added __all__ const to spatial split module

0.9.8

Added

• New spatial splitting logic for machine learning model training

Changed

• Bumped minimal overturemaestro version to 0.5.0

Fixed

• Enabled CountEmbedder to parse dataframes with indexes other than string type #537

0.9.7

Added

• Option to pass trainer_kwargs to Gtfs2VecEmbedder's fit and fit_transform functions

Fixed

• Change CRS setting for GeoDataFrame in OSMPbfLoader

0.9.6

Added

• Functions convert_to_regions_gdf and convert_to_features_gdf to transform any existing geo input to an object with the correct index name

0.9.5

Fixed

• Catched OSMnx InsufficientResponseError in OSMOnlineLoader

0.9.4

Fixed

• Saving expected target features in GeoVexEmbedder

0.9.3

Added

• Option to change default aggregation in ContextualCountEmbedder from average to median, sum, min or max