Distributed Advanced Indexing, 2D Convolution, Complex Sorting and more

Release notes

• Overview
• Highlights
• Changes
  • Breaking changes
  • Features

Overview

Version 1.9.0 expands Heat’s distributed array framework for large-scale data science on CPU and GPU clusters, bringing native support for distributed advanced indexing and 2D convolution. This release also advances interoperability across the scientific Python ecosystem with improved Python Array API compliance, zero-copy array sharing via DLPack, NumPy-like complex sorting, and support for PyTorch 2.14. In addition, we launch our new landing page and enable direct tutorial execution on Jupyter4NFDI.

Heat is now a NumFOCUS affiliated project!

Highlights

• Distributed indexing by @ClaudiaComito, @JuanPedroGHM, @brownbaerchen, @mtar in PR #938
• 2D convolution by @lolacaro, @JuanPedroGHM in PR #1937
• Improved compliance with array API standard by @mtar, @brownbaerchen in PR #2439, PR #2480, and PR #2539
• argsort and sort NumPy API compliance, with complex support for ht.sort() by @Berkant03 in PR #2386 and PR #2482
• New project landing page by @ClaudiaComito in PR #2361

Changes

Breaking changes

• Implement argsort and make sort NumPy API compliant by @Berkant03 in PR #2386
• Adapt nonzero and where APIs to match NumPy by @ClaudiaComito in PR #2332
• Deprecate DNDarray.lloc by @brownbaerchen in PR #2494

Features

• Add vectorized_sort function by @Berkant03 in PR #2431
• Implementation for complex sorting of dndarrays and take by @Berkant03 in PR #2482
• Add unique_inverse by @mtar in PR #2460
• Add unique_values by @mtar in PR #2452

New Contributors

• @Shabasovich made their first contribution in #2199
• @maliesen made their first contribution in #2376
• @Newer1107 made their first contribution in #2504
• @mrprohack made their first contribution in #2506
• @cnYui made their first contribution in #2528
• @MBK-fr made their first contribution in #2558

Full Changelog: https://github.com/helmholtz-analytics/heat/blob/stable/CHANGELOG.md