Skip to content

Releases: ahendriksen/tomosipo

v0.6.0

02 Aug 09:58
Compare
Choose a tag to compare

Added

  • References and links to ts_algorithms were added to the readme

Changed

  • Changed the Tomosipo autograd operators made by the tomosipo.torch_support.to_autograd function. By default they now expect the same input shape (3 dimensions) as a normal Tomosipo operator. New arguments have been added to also support the input shapes expected in Pytorch neural networks.

v0.5.1

15 Jun 14:24
Compare
Choose a tag to compare

Small bugfix to the to_autograd operator. (Pull request 16)

Fixed

  • Fixed to_autograd operator to always output 32 bit results, also with a 64 bi input just like the normal operator. Before the fix it only worked with 32 bit inputs.

v0.5.0

09 Feb 13:14
Compare
Choose a tag to compare

This release is mostly about maintaining existing functionality. The tomosipo paper and Astra toolbox 2 were released, there is a new maintainer, and some bugs were fixed. The conda package is now the default installation option.

Added

  • Added information and Bibtex entry of the tomosipo paper to the readme

Changed

  • Changed repository maintainer from Allard Hendriksen to Dirk Schut
  • The conda package is now the default installation option, instead of cloning from Github with pip. This is changed in the readme
  • Astra toolbox 2 was released so it is now the required version. There were no API changes compared to the recent development version that was required before.
  • Removed cupy and odl packages from the environment presented in the documentation.
    Cupy was removed because of cudatoolkit error messages
    ODL was removed because it changes the global numpy formatting options

Fixed

  • Replaced usage of np.int which was deprecated in numpy 1.20 with int
  • Replaced the usage of collect_ignore in setup.cfg because it was removed from pytest
  • Licence information was updated to show GPLv3

v0.4.1

01 Feb 15:32
Compare
Choose a tag to compare

[0.4.1] - 2021-07-18

Bug fix release for ts.rotate (again)..

Fixed

  • ts.rotate: the rotation matrix had become transposed while fixing #7. This has now been fixed.