Skip to content

Ongoing work on making a purely out-of-place version of ODL#1723

Draft
Emvlt wants to merge 17 commits intoodlgroup:masterfrom
Emvlt:no_in_place
Draft

Ongoing work on making a purely out-of-place version of ODL#1723
Emvlt wants to merge 17 commits intoodlgroup:masterfrom
Emvlt:no_in_place

Conversation

@Emvlt
Copy link
Member

@Emvlt Emvlt commented Jan 16, 2026

To begin the work, I changed the default support_in_place_ops property and disabled the __setitem__ of the TensorSpaceElement as False and saw where things broke.

I think that providing an out_of_place and an in_place version of each _call and setting the default argument out as None is a good way to make progress on this.

At present, all the core test suite passes, but many improvements are still required.

Emvlt added 17 commits January 5, 2026 15:36
… a bug when calling the python built-in repr (None returned)
1) Fix the numpy version in install_requires to 2.3 (changes in 2.4 will break some of the functions).
2) Fix the torch version in the all to 2.9 (changes in 2.10 might break some of the functions).
…rted by the backend.

1) Changed the  to  which is shorter and does not loose inclarity.
2) Default use_in_place_ops as False in the TensorSpace
3) Began inserting checks on in_place support of operation where a method or function accepts an out argument.
4) Began changing tests to be skipped if the in-place is not supported.
…ation in Numpy 2.4 and modified tests accordingly
… used by the ProductSpaceElement if its space does not support it
It depends on the domain and the range supporting in-place operations
This involves skipping some tests (the __setitem__ for instance) and chaning the way some test elements are created (without using in-place updates).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant