Tags: alexshtf/torchcurves
Tags
Refactor B-Spline implementation: merge two implementations into one (#… …14) * Refactor B-Spline implementation: renamed BSplineCurve and BSplineEmbeddings to BSpline, updated function calls in tests and modules, and optimized gradient computation in _bspline.py. This unifies the B-Spline layer to handle backpropagation more efficiently. * Fix unit tests to use new class name. * Refactor: Replace `BSpline` with `BSplineCurveBase` across modules, tests, and documentation, removing redundant `_forward_core` implementation. * added pre-commit hook installation to the makefile * Refactor: Rename `BSplineCurveBase` to `BSplineCurve` and `bspline` to `bspline_curves` across modules, tests, benchmarks, and documentation for consistency. * Update docs. * Fix: Replace `torch.addcmul` deprecated call with the new calling convention. * Reran notebooks with the new implementations to verify everything works. * Refactor: Replace `BSplineEmbeddings` with `BSplineCurve` in `README.md` for consistency with recent renaming * Fix: Update type hints in `backward` method to support `None` return annotations * Update: Simplify docstring for `BSplineCurve` module by removing redundant "Base" descriptor * Fix: Update `backward` method's type hints to use `Optional` to support older python versions.