1.4.1

• [#164] fix: njit fallback for broken runtime with torch (@JhonatanFelix)

v1.4

• [#155] feat/perf: use parallelized numba functions if possible (@ilan-gold)
• [#140] chore: SPEC0 update – Python 3.12+ (@flying-sheep)
• [#142] ci: switch to dep groups (@flying-sheep)
• [#161] chore: bump ci to 3.14 (@flying-sheep)

1.3.1

• [#135] fix: support Dask and cupy/scipy sparse matrices in min/max (@quentinblampey, @flying-sheep)

1.3

• [#116] feat: add min and max (@quentinblampey)
• [#132] fix: use accurate lower bound (@flying-sheep)

1.2.5

• [#127] fix: improve variance precision (@flying-sheep)

1.2.4

• [#119] remove numpy 1 support (@flying-sheep)
• [#125] fix: Have sum operate on the passed dtype (@flying-sheep)

v1.2.3

• [#113,#114] Patch dask for sparray support in the same way as the upstream fix was done (@flying-sheep)

v1.2.1

• [#96] Add support for numpy 1.25+ (as opposed to just 2.0+) and correct minimum supported scipy version (1.8→1.11) (@flying-sheep)

Dev changes:

• [#98] Derive test configs from hatch (@flying-sheep)

v1.2

• [#94] Add order parameter to to_dense (@flying-sheep)

v1.1.2

• [#91] Make array_type a session-scoped fixture (@flying-sheep)