Releases: flatironinstitute/nifty-ls
v1.1.0
This release adds nterms > 1 support, implemented in the finufft_chi2 and cufinufft_chi2 backends. Thanks to @YuWei-CH for the feature!
Free-threaded Python (3.13t and 3.14t) is also now supported. More generally, we are now building wheels for Python 3.14 and testing against it.
The finufft version requirement is >= 2.3.
Enhancements
- lombscargle: check that all inputs have the same dtype (#59)
- lombscargle: add support for
nterms > 1(#60) - build: free-threaded python and python 3.14 (#65)
What's Changed [autogenerated, minus bots]
- lombscargle: check that all inputs have the same dtype by @lgarrison in #59
- Nterms by @YuWei-CH in #60
- ci: update test and release pipeline by @lgarrison in #62
- ci: add macos config by @lgarrison in #63
New Contributors
Full Changelog: v1.0.1...v1.1.0
Contributors
Assets 2
v1.0.1
This release updates the upstream finufft requirement to 2.3 and has a few optimizations that make use of it. Thanks to @soichiro-hattori for a few fixes as well!
OpenMP in the C++ helpers on MacOS ARM has also been fixed, which should result in a small performance improvement for users on M1/M2/etc CPUs.
This is the version that matches the submitted research note describing nifty-ls.
The finufft version requirement is >= 2.3.
What's Changed (minus bot updates)
- set default dy=None by @soichiro-hattori in #20
- add version to init.py file by @soichiro-hattori in #19
- ci: switch Docker image to
nvidia/cuda:12.6.0-devel-rockylinux9by @lgarrison in #35 - Bump finufft requirement to 2.3 by @lgarrison in #40
- Wheels for 1.0.1 by @lgarrison in #41
New Contributors
- @dependabot made their first contribution in #15
- @soichiro-hattori made their first contribution in #20
Full Changelog: v1.0.0...v1.0.1
Contributors
Assets 2
v1.0.0
nifty-ls is now production-ready! v1.0.0 has been released on PyPI and includes finufft and cufinufft (CUDA) backends, astropy integration, support for batched periodograms, and more. Performance on the CPU is often 5-10x faster than Astropy LombScargle, and 6 orders of magnitude more accurate (no more twiddling with oversample factors!).
Please file a GitHub issue if you run into problems or need help!