pytorch-topological follows the semantic versioning.
This changelog contains all notable changes in the project.
- Fixed bug in
make_tensorthat caused the loss of device information. Device handling should now be transparent for clients.
- Fixed various documentation typos
- Fixed bug in
make_tensorcreation for single batches
- A bunch of new test cases
- Alpha complex class (
AlphaComplex) - Dimension selector class (
SelectByDimension) - Discussing additional packages in documentation
- Linting for pull requests
- Improved contribution guidelines
- Improved documentation of summary statistics loss
- Improved overall maintainability
- Improved test cases
- Simplified multi-scale kernel usage (distance calculations with different exponents)
- Test case for cubical complexes
- Usage of seed parameter for shape generation (following
numpy guidelines)
- Batch handler for point cloud complexes
- Sliced Wasserstein distance kernel
- Support for pre-computed distances in Vietoris--Rips complex
- Device compatibility issue (tensor being created on the wrong device)
- Use of
dimflag for alpha complexes - Various documentation issues and coding style problems