1bf64db523
- Update to 0.18.0: * Update sparse_finch notebook and upgrade finch-tensor * Allow no-copy construction from SciPy COO arrays. * MAINT: Re-enable COO tests after fix * Implementing repeat function * Adding tile function * Setting boolean indexing flag to False * Adding unstack function * Fixing check_zero_fill_value * Adding diff function * fix: Fix conversions from non-canonical scipy.sparse arrays * fix: Fix for check_zero_fill_value and equivalent * MAINT: Upgrade MLIR version * ENH: Implement reshape function * ENH: CSC and CSF formats for MLIR backend * MNT: Move constructors to a direct style * MNT: Refactor MLIR constructors * ENH: Simple COO format * MNT: Add CI mode to examples * BUG: Skip test_where test * Fix 1D return dtype in argmax/argmin * scipy.sparse fill-value fix. * Add support for NumPy 2.0. * API: Rename PyData backend to Numba * API: Enable finch backend for scipy.sparse * BUG: Fix CSR/CSC matmul * API: Update asarray function * Support Python 3.12 and drop 3.9. * Remove backend kwarg from asarray. * API: Initial finch_backend
Steve Kowalik2026-03-06 01:10:44 +00:00
af10a0653d
- update to 0.15.5: * Make sure [r]matmul__ returns NotImplemented on invalid types. - Limit numpy to < 2. - skip python 3.6 build * Fix TypingError on sparse.dot with complex dtypes. (Issue #403, PR #404) * Improve the performance of sparse.dot. (Issue #331, PR #389, * Added multi-axis 1-D indexing support. (PR #343, thanks * Fix outer for arrays that weren’t one-dimensional. (Issue #346, * Fix for COO constructor accepting invalid inputs. (Issue #385, * Fixed a bug where converting an empty DOK array to COO leads to * Remove broadcasting restriction between sparse arrays and dense * Implement deterministic dask tokenization. (Issue #300, PR * Improve testing around densification (PR #321, thanks * Replace linear_loc with ravel_multi_index, which is 3x faster. * Add error msg to tensordot operation when ndim==0 (Issue #332, * Maintainence fixes for Sphinx 3.0 and Numba 0.49, and dropping * Fixed a bug where indexing with an empty list could lead to * Add the diagonal and diagonalize functions. (Issue #288, PR * Add stack and concatenate for GCXS. (Issue #301, PR #303, * Fix issue where functions dispatching to an attribute access * Add partial support for constructing and mirroring COO objects
Dirk Mueller2025-02-25 20:19:11 +00:00
63922934de
Accepting request 1226481 from devel:languages:python:numeric
Ana Guerrero2024-11-27 21:10:53 +00:00
f9cb499044
Accepting request 1226473 from home:Guillaume_G:branches:devel:languages:python:numeric
Daniel Garcia2024-11-26 11:04:23 +00:00
d53a11ada9
Accepting request 1225339 from devel:languages:python:numeric
Ana Guerrero2024-11-21 14:14:02 +00:00
89826dfebf
Accepting request 1225338 from home:bnavigator:branches:devel:languages:python:numeric
Benjamin Greiner2024-11-20 16:26:21 +00:00
05da4e494c
Accepting request 1202832 from devel:languages:python:numeric
Ana Guerrero2024-09-24 15:33:59 +00:00
a0e60eb714
Accepting request 1202756 from home:bnavigator:branches:devel:languages:python:numeric
Markéta Machová2024-09-24 08:47:37 +00:00