v2.0.2 (Jul 29, 2026)

Minor Bugfix

Convert all kwargs (such as sigma) of the GaussianFilterOperator to python native (e.g. list of floats for sigma).
E.g., enables passing of torch tensors for sigma.

v2.0.1 (Jul 24, 2026)

Bug fix

• GaussianFilterOperator now works with PyTorch (and array-api-strict) CPU
  arrays under modern scipy. The CPU path previously passed the input array
  straight to scipy.ndimage.gaussian_filter; with array-API-aware scipy
  (SCIPY_ARRAY_API / scipy ≥ 1.16) this made scipy compute natively in the
  input's namespace, where gaussian_filter1d reverses the kernel with a
  negative-step slice (weights[::-1]) that PyTorch does not support. The
  operator now filters on a NumPy view (zero-copy on CPU) and converts back, so
  it is robust regardless of the scipy version or SCIPY_ARRAY_API. This
  fixes downstream use via e.g. DeepInv.

v2.0.0 (Jul 17, 2026)

Major v2.0.0 release including many new features but also a few minor breaking changes.
Please see our changelog for details.

v1.10.2 (Aug 20, 2025)

summary of changes

• add compatibility for latest cupy version (>= 13.5) which require from_dlpack to convert
  from torch tensors
• fix minor issues to be compatible with array-api-strict~=2.0

merged PRs

• update binder postbuild by @gschramm in #110
• remove pdm by @gschramm in #111
• add gradient field projection operator by @gschramm in #112
• WIP example that shows how to implement a gradient descent layer by @gschramm in #113
• restrict cupy version in install instructions by @gschramm in #117
• fix incompatibilities with new cupy version (13.5) by @gschramm in #118
• bump version and update changelog by @gschramm in #119

Full Changelog: v1.10.1...v1.10.2

v1.10.1

1.10.1 (Jan 15, 2025)

• add a check whether sum of tof bins along LOR is non-zero before running
  TOF sinogram back projector
• update installation instructions after conda-forge recipe was updated
• clean up RTD docs build

v1.10.0

v1.10.0

• add support for numpy>=2.0
• add tests with numpy 2.0 on python 3.9 and 3.12
• remove tox.ini

v1.9.1

• BUGFIX: fix missing devices in BlockGeometry LOR descriptor (needed for pytorch + cuda backend)

v1.9.0

• add functionality to create scanners, LOR descriptors and projectors for scanners consiting of equal "block" modules
• BUGFIX: correct behavior of TOF kernel truncation which was wrong in the case that the tof bin width was >> tof resolution

v.1.8.0

v1.8.0 (March 20, 2024)

• add function to count event multiplicity
• add more examples (e.g. DePierro and LM SPDHG)
• re-organize folder structure and pyproject.toml
• force array-api-compat<1.5 (bug in 1.5.0)
• use array-api-strict instead of numpy.array_api

v1.7.3

1.7.3 (January 26, 2024)

• BUGFIX: import importlib.util instead of import importlib
• print banner
• test also on Windows