v0.19.0

Summary

This release introduces a set of new dpnp.ndarray methods and SciPy-compatible functions to improve CuPy compatibility.

It also enhances the performance of existing functions and improves documentation completeness.

Additionally, it extends support for building dpnp from the source for NVIDIA GPUs, with optional architecture selection.

This release is compatible with NumPy 2.3.3.

Details

Added

• Added --target-cuda[=ARCH] option to replace the deprecated --target=cuda, allowing users to build for CUDA devices with optional architecture selection using CodePlay oneAPI plug-in #2478
• Added several new pre-commit rules, including protection against direct commits to master/maintenance branches #2500
• Added implementation of dpnp.ndarray.view method #2520
• Added a new backend routine syrk from oneMKL to perform symmetric rank-k update which is used for a specialized matrix multiplication where the result is a symmetric matrix 2509
• Added timeout-minutes property to GitHub jobs #2526
• Added implementation of dpnp.ndarray.data and dpnp.ndarray.data.ptr attributes #2521
• Added dpnp.ndarray.__contains__ method #2534
• Added implementation of dpnp.linalg.lu_factor (SciPy-compatible) #2557, #2565
• Added implementation of dpnp.piecewise #2550
• Added implementation of dpnp.linalg.lu_solve for 2D inputs (SciPy-compatible) #2575
• Added implementation of dpnp.special.erfc #2588
• Added dpnp.scipy submodule to aggregate new SciPy-compatible functions from linalg and special namespaces #2603

Changed

• Adjusted the pre-commit configuration to run autoupdate weekly #2479
• Improved validation of --target-hip build option to only accept a gfx-prefixed value #2481
• Simplifies backend implementation of dpnp.kaiser by getting rid of unnecessary template #2472
• --onemkl-interfaces and --onemkl-interfaces-dir options for building script are deprecated, instead --onemath and --onemath-dir are introduced to be aligned with oneMath specification #2487
• Clarified description of xp keyword in docstring of dpnp.interp #2506
• Updated existing GitHub workflows to add testing with Python 3.13 #2510
• Aligned the license expression with PEP-639 #2511
• Bumped oneMKL version up to v0.8 #2514
• Removed the use of class template argument deduction for alias template to conform to the C++17 standard #2517
• Changed the order of individual FFTs over axes for dpnp.fft.irfftn to be in forward order #2524
• Replaced the use of numpy.testing.suppress_warnings with appropriate calls from the warnings module #2529
• Improved documentations of dpnp.ndarray class and added a page with description of supported constants #2422
• Updated dpnp.size to accept tuple of ints for axes argument #2536
• Replaced ci section in .pre-commit-config.yaml with a new GitHub workflow with scheduled run to autoupdate the pre-commit configuration #2542
• FFT module is updated to perform in-place FFT in intermediate steps of ND FFT #2543
• Reused dpctl tensor include to enable experimental SYCL namespace for complex types #2546
• Changed Windows-specific logic in dpnp initialization #2553
• Added missing includes to files in ufunc and VM pybind11 extensions #2571
• Refactored backend implementation of dpnp.linalg.solve to use oneMKL LAPACK gesv directly #2558
• Improved performance of dpnp.isclose function by implementing a dedicated kernel for scalar rtol and atol arguments #2540
• Extended dpnp.pad to support pad_width keyword as a dictionary #2535
• Redesigned dpnp.erf function through pybind11 extension of OneMKL call or dedicated kernel in ufunc namespace #2551
• Improved performance of batched implementation of dpnp.linalg.det and dpnp.linalg.slogdet #2572
• Improved documentations of dpnp.tril_indices and dpnp.triu_indices to clarify the returned order of indices #2586
• dpnp uses pybind11 3.0.1 #2594

Deprecated

• --onemkl-interfaces and --onemkl-interfaces-dir options for building script are deprecated, instead --onemath and --onemath-dir are introduced to be aligned with oneMath specification #2487

Removed

• Cleaned up backend code to remove obsolete and unused parts of functionality #2485

Fixed

• Updated pre-commit GitHub workflow to pass no-commit-to-branch check #2501
• Updated the math formulas in summary of dpnp.matvec and dpnp.vecmat to correct a typo #2503
• Avoided negating unsigned integers in ceil division used in dpnp.resize implementation #2508
• Fixed dpnp.unique with 1d input array and axis=0, equal_nan=True keywords passed where the produced result doesn't collapse the NaNs #2530, #2587
• Resolved issue when dpnp.ndarray constructor is called with dpnp.ndarray.data as buffer keyword #2533
• Fixed dpnp.linalg.cond to always return a real dtype #2547
• Resolved the issue in dpnp.random functions to allow any value of size where each element is castable to Py_ssize_t type #2578
• Resolved conda build --test issue in python 3.9 environment #2583
• Fixed tests for the rounding functions to depend on minimum required numpy version #2589
• Fixed tests for the ufuncs to depend on minimum required numpy version #2590
• Added missing permission definition in Autoupdate pre-commit GitHub workflow #2591
• Resolved issue with the cyclic import in linalg submodule #2608

Contributors

• @antonwolfy
• @vlad-perevezentsev
• @vtavana
• @ndgrigorian

Full Changelog: 0.18.1...0.19.0

v0.18.1

Summary

This release achieves dpnp compatibility with Python 3.13 and enables distributing dpnp packages with the latest Python version.

Moreover, the release provides compatibility with NumPy 2.3.0 and includes several bug fixes.

Details

Added

• Enabled support of Python 3.13 #2490

Changed

• Updated the tests scope to exclude several matmul tests in case of numpy 2.3.0 due to known NumPy issue #2495

Fixed

• Fixed a bug for calculating the norm (dpnp.linalg.norm) of empty arrays when keepdims=True is passed #2477
• Updated the tests for hyperbolic and trigonometric elementwise functions to set correct tolerance for float16 dtype #2483

Contributors

• @antonwolfy
• @vtavana

Full Changelog: 0.18.0...0.18.1

v0.18.0

Summary

This release achieves 100% compliance with Python Array API specification (revision 2024.12).

The release provides enhanced compatibility with NumPy 2.2.5. Window and mathematical routines are complemented by a set of new functions.

Moreover, it adds support to build dpnp from the source for AMD GPUs.

Details

Added

• Added implementation of dpnp.hamming #2341, #2357
• Added implementation of dpnp.hanning #2358
• Added implementation of dpnp.blackman #2363
• Added implementation of dpnp.bartlett #2366
• Added implementation of dpnp.convolve #2205
• Added implementation of dpnp.kaiser #2387
• Added implementation of dpnp.bitwise_count #2308
• Added implementation of dpnp.common_type #2391
• Added implementation of dpnp.interp #2417
• Added support to build dpnp for specified AMD GPU architecture using CodePlay oneAPI plug-in #2302

Changed

• Improved performance of dpnp.nansum, dpnp.nanprod, dpnp.nancumsum, and dpnp.nancumprod by reusing dpnp.nan_to_num function in implementation of the functions #2339
• Allowed input array of uint64 dtype in dpnp.bincount #2361
• The vector norms ord={None, 1, 2, inf} and the matrix norms ord={None, 1, 2, inf, "fro", "nuc"} now consistently return zero for empty arrays, which are arrays with at least one axis of size zero. This change affects dpnp.linalg.norm, dpnp.linalg.vector_norm, and dpnp.linalg.matrix_norm. Previously, dpnp would either raise errors or return zero depending on the parameters provided #2371
• Extended dpnp.fft.fftfreq and dpnp.fft.rfftfreq functions to support dtype keyword per Python Array API spec 2024.12 #2384
• Updated dpnp.fix to return output with the same data-type of input #2392
• Updated dpnp.einsum to add support for order=None #2411
• Updated Python Array API specification version supported to 2024.12 #2416
• Removed einsum_call keyword from dpnp.einsum_path signature #2421
• Updated dpnp.vdot to return a 0-D array when one of the inputs is a scalar #2295
• Updated dpnp.outer to return the same dtype as NumPy when multiplying an array with a scalar #2295
• Changed "max dimensions" to None in array API capabilities #2432
• Updated kernel header i0.hpp to expose cyl_bessel_i0 function depending on build target #2440
• Added MKL functions arg, copysign, i0, and inv from VM namespace to be used by implementation of the appropriate element-wise functions #2445
• Clarified details about conda install instructions in Quick start quide and README #2446
• Bumped oneMKL version up to 0.7 #2448
• The parameter axis in dpnp.take_along_axis function has now a default value of -1 #2442
• Updates the list of required python versions documented in Quick Start Guide #2449
• Updated FFT module to ensure an input array is Hermitian before calling complex-to-real FFT #2444
• Aligned black configuration with the list of supported python versions #2457
• Use pyproject.toml instead of setup.py aligning with current packaging best practices #2462
• Added a clarification to dpnp.linalg.cond docstring about its behavior with singular matrices #2460

Fixed

• Resolved an issue with an incorrect result returned due to missing dependency from the strided kernel on a copy event in dpnp.erf #2378
• Updated conda create commands build and install instructions of Quick start guide to avoid a compilation error #2395
• Added handling of empty string passed to a test env variable defining data type scope as a False value #2415
• Resolved build issues on non-Intel targets in dpnp.i0 and dpnp.kaiser #2439
• Ensure consistency in the dpnp.linalg.LinAlgError exception raised on singular input matrices for both non-batched and batched cases in dpnp.linalg.inv [#2458] (#2458)
• Updated test f/w to correct a check of array interface while converting to numpy.ndarray for comparison [#2467] (#2467)

New Contributors

• @david-cortes-intel made their first contribution in #2446

Contributors

• @AlexanderKalistratov
• @antonwolfy
• @david-cortes-intel
• @ndgrigorian
• @vlad-perevezentsev
• @vtavana

Full Changelog: 0.17.0...0.18.0

v0.17.0

Summary

This release achieves 100% compliance with Python Array API specification (revision 2023.12).

The release provides enhanced compatibility with NumPy 2.2.3. Array manipulation, mathematical, logic, and statistics routines are complemented by a set of new functions.

Furthermore, a number of issues relating to running on NVIDIA GPUs have been resolved.

Details

Added

• Added implementation of dpnp.gcd and dpnp.lcm functions #2091
• Added implementation of dpnp.pad function #2093
• Added implementation of dpnp.linalg.svdvals function #2094
• Added implementation of dpnp.matrix_transpose function and dpnp.ndarray.mT attribute #2095
• Exposed cross, diagonal, matrix_norm, outer, tensordot, trace and vector_norm functions as part of dpnp.linalg namespace #2099
• Added implementation of dpnp.unstack function #2106
• Added implementation of dpnp.ldexp function #2110
• Added implementation of dpnp.vecdot and dpnp.linalg.vecdot functions #2112
• Added implementation of dpnp.i0 function #2118
• Added implementation of dpnp.isfortran function #2122
• Added implementation of dpnp.spacing function #2125
• Added implementation of dpnp.sinc function #2133
• Added implementation of dpnp.corrcoef function #2139
• Added implementation of dpnp.delete function #2142
• Added implementation of dpnp.histogramdd function #2143
• Added implementation of dpnp.bincount function #2145
• Added support of inplace matrix multiplication via the @= operator #2147
• Added implementation of dpnp.insert function #2151
• Added implementation of dpnp.broadcast_shapes function #2153
• Added implementation of dpnp.byte_bounds function #2155
• Added implementation of dpnp.ndindex class #2157
• Added implementation of dpnp.histogram2d function #2262
• Added implementation of dpnp.binary_repr function #2168
• Added implementation of dpnp.apply_along_axis function #2169
• Added implementation of dpnp.cumulative_sum and dpnp.cumulative_prod functions #2171
• Added implementation of dpnp.apply_over_axes function #2174
• Added implementation of dpnp.compress function and dpnp_array.compress method #2177
• Added implementation of dpnp.correlate function #2180, #2203
• Added implementation of dpnp.nanmedian function #2191
• Added implementation of dpnp.ndarray.__iter__ method #2206
• Added implementation of dpnp.iterable function #2208
• Added missing aliases on integer data types #2230
• Enabled validation of dpnp conda/wheel packages with Python 3.13 (limited support) #2249
• Added implementation of dpnp.ndarray.__array_namespace__ method #2252
• Added implementation of dpnp.ndarray.__usm_ndarray__ protocol #2261
• Added implementation of dpnp.isdtype function #2274
• Added implementation of Python Array API Inspection namespace #2275
• Added implementation of dpnp.matvec and dpnp.vecmat functions #2288
• Added implementation of dpnp.unique_all, dpnp.unique_counts, dpnp.unique_inverse and dpnp.unique_values functions #2320

Changed

• Improved performance of dpnp.histogram function by implementing a dedicated kernel #2027
• Improved performance of dpnp.ndarray.fill method by leveraging on dpctl extension exposing fill kernel #2055
• Extended dpnp.ndarray.reshape method and dpnp.reshape function to support shape and newshape keywords #2080
• Extended support of order keyword in like-functions #2088
• Updated dpnp.einsum function to comply with NEP-50 #2120
• Extended dpnp.linalg.pinv and dpnp.linalg.matrix_rank functions to support rtol keyword #2124
• Extended dpnp.array function to support ndmin keyword #2135
• Leveraged dpctl.tensor implementation for dpnp.put_along_axis function #2134
• Corrected dpnp.ndarray.item implemented to return a python scalar instead of zero-dimensional array #2138
• Bumped NumPy, CuPy and Python versions used for building docs #2158
• Extended dpnp.sort and dpnp.argsort functions to support kind="mergesort" and kind="radixsort" values #2159
• Revised and updated information in README.md document #2166
• Permitted "same_kind" casting for elementwise inplace operators #2170
• Bumped oneMKL version up to 0.6 and added new --onemkl-interfaces-dir option to build script #2193
• Updated implementation of dpnp.linalg.solve function to align with numpy >= 2.0 and Python array API #2198
• Improved performance of dpnp.choose function by implementing a dedicated kernel #2201
• Aligned with the functional changes introduced by NumPy 2.2 #2226
• Improved performance of dpnp.nan_to_num function by implementing a dedicated kernel #2228
• Enabled Intel MKL backends when building from the source with --onemkl-interfaces option #2229
• Extended intersphinx_mapping with a link to CuPy documentation to make cupy functions clickable from the rendered pages #2232
• Improved performance of dpnp.nanmedian function when axis is not None passed #2240
• Aligned dpnp.trim_zeros with NumPy 2.2 and added support of a multi-dimensional input array #2241
• Disallowed implicit conversion of dpnp.ndarray to numpy.ndarray #2260
• Extended dpnp.ndarray.to_device method to support stream keyword #2263
• Extended dpnp.sort and dpnp.argsort functions and dpnp.ndarray.sort and dpnp.ndarray.argsort methods to support descending keyword #2269
• Extended dpnp.std and dpnp.var functions and dpnp.ndarray.std and dpnp.ndarray.var methods to support mean keyword #2271
• Aligned qr, eig, eigh, svd and slogdet functions from dpnp.linalg to return namedtuple per Python array API #2276
• Extended dpnp.std and dpnp.var functions and dpnp.ndarray.std and dpnp.ndarray.var methods to support correction keyword #2300
• Extended dpnp.cov function to support all keyword arguments #2303
• Disallowed minlength=None value passed into dpnp.bincount function #2310
• Added build support with oneMath (new name of oneMKL interface) #2313
• Aligned the signature of dpnp.astype function with Python array API #2318

Fixed

• Migrated to experimental extension of DPC++ compiler with group_load/group_store per deprecation build warning #2123
• Fixed DeprecationWarning appearing during running...

v0.16.3

This is a bug fix release which supports building dpnp with DPC++ 2025.0.4 compiler, resolving gh-2243.

Full Changelog: 0.16.2...0.16.3

v0.16.2

This is a bug fix release which supports use of dpnp in virtual environment on Windows, resolving gh-2242.

Full Changelog: 0.16.1...0.16.2

v0.16.1

Summary

This is a maintenance release that fixes bugs and regressions discovered after the 0.16.0 release.

Details

Changed

• Changed to use Miniforge installer in GitHub actions #2057
• Updated README.md to reflect current installation requirements and available options #2166
• Corrected the list of owners and code maintainers #2185
• Bumped the version of oneMKL interface used in dpnp build by default to align it with 2025.0 oneAPI release #2193

Fixed

• Resolved an issue with Compute Follows Data inconsistency in dpnp.extract function #2172
• Resolved an import error when using dpnp in virtual environment on Linux #2199
• Fixed incorrect result produced by dpnp.fft.fft function when input array has negative strides #2202
• Fixed an issue with numpy.ndarray input processing in the dpnp.from_dlpack function and updated the documentation #2209
• Resolved a compilation error when building with DPC++ 2025.1 compiler #2211

Full Changelog: 0.16.0...0.16.1

v0.16.0

Summary

This release reaches an important milestone by making offloading fully asynchronous. Calls to dpnp submit tasks for execution to DPC++ runtime and return without waiting for execution of these tasks to finish. The sequential semantics a user comes to expect from execution of Python script is preserved though.

In addition, this release completes implementation of dpnp.fft module and adds several new array manipulation, indexing and elementwise routines. Moreover, it adds support to build dpnp for Nvidia GPUs.

DPNP is now compatible with NumPy 2.0.

Details

Added

• Added implementation of dpnp.gradient function #1859
• Added implementation of dpnp.sort_complex function #1864
• Added implementation of dpnp.fft.fft and dpnp.fft.ifft functions #1879
• Added implementation of dpnp.isneginf and dpnp.isposinf functions #1888
• Added implementation of dpnp.fft.fftfreq and dpnp.fft.rfftfreq functions #1898
• Added implementation of dpnp.fft.fftshift and dpnp.fft.ifftshift functions #1900
• Added implementation of dpnp.isreal, dpnp.isrealobj, dpnp.iscomplex, and dpnp.iscomplexobj functions #1916
• Added support to build dpnp for Nvidia GPU #1926
• Added implementation of dpnp.fft.rfft and dpnp.fft.irfft functions #1928
• Added implementation of dpnp.nextafter function #1938
• Added implementation of dpnp.trim_zero function #1941
• Added implementation of dpnp.fft.hfft and dpnp.fft.ihfft functions #1954
• Added implementation of dpnp.logaddexp2 function #1955
• Added implementation of dpnp.flatnonzero function #1956
• Added implementation of dpnp.float_power function #1957
• Added implementation of dpnp.fft.fft2, dpnp.fft.ifft2, dpnp.fft.fftn, and dpnp.fft.ifftn functions #1961
• Added implementation of dpnp.array_equal and dpnp.array_equiv functions #1965
• Added implementation of dpnp.nan_to_num function #1966
• Added implementation of dpnp.fix function #1971
• Added implementation of dpnp.fft.rfft2, dpnp.fft.irfft2, dpnp.fft.rfftn, and dpnp.fft.irfftn functions #1982
• Added implementation of dpnp.argwhere function #2000
• Added implementation of dpnp.real_if_close function #2002
• Added implementation of dpnp.ndim and dpnp.size functions #2014
• Added implementation of dpnp.append and dpnp.asarray_chkfinite functions #2015
• Added implementation of dpnp.array_split, dpnp.split, dpnp.hsplit, dpnp.vsplit, and dpnp.dsplit functions #2017
• Added runtime dependency on intel-gpu-ocl-icd-system package #2023
• Added implementation of dpnp.ravel_multi_index and dpnp.unravel_index functions #2022
• Added implementation of dpnp.resize and dpnp.rot90 functions #2030
• Added implementation of dpnp.require function #2036

Changed

• Extended pre-commit pylint check to dpnp.fft module #1860
• Reworked vm vector math backend to reuse dpctl.tensor functions around unary and binary functions #1868
• Extended dpnp.ndarray.astype method to support device keyword argument #1870
• Improved performance of dpnp.linalg.solve by implementing a dedicated kernel for its batch implementation #1877
• Extended dpnp.fabs to support order and out keyword arguments by writing a dedicated kernel for it #1878
• Extended dpnp.linalg module to support usm_ndarray as input #1880
• Reworked dpnp.mod implementation to be an alias for dpnp.remainder #1882
• Removed the legacy implementation of linear algebra functions from the backend #1887
• Removed the legacy implementation of elementwise functions from the backend #1890
• Extended dpnp.all and dpnp.any to support out keyword argument #1893
• Reworked dpnp.repeat to add a explicit type check of input array #1894
• Improved performance of different functions by adopting asynchronous implementation of dpctl #1897
• Extended dpnp.fmax and dpnp.fmin to support order and out keyword arguments by writing dedicated kernels for them #1905
• Removed the legacy implementation of array creation and manipulation functions from the backend #1903
• Extended dpnp.extract implementation to align with NumPy #1906
• Reworked backend implementation to align with non-backward compatible changes in DPC++ 2025.0 #1907
• Removed the legacy implementation of indexing functions from the backend #1908
• Extended dpnp.take implementation to align with NumPy #1909
• Extended dpnp.place implementation to align with NumPy #1912
• Reworked the implementation of indexing functions to avoid unnecessary casting to dpnp_array when input is usm_ndarray #1913
• Reduced code duplication in the implementation of sorting functions #1914
• Removed the obsolete dparray interface #1915
• Improved performance of dpnp.linalg module for BLAS routines by adopting asynchronous implementation of dpctl #1919
• Relocated dpnp.einsum utility functions to a separate file #1920
• Improved performance of dpnp.linalg module for LAPACK routines by adopting asynchronous implementation of dpctl #1922
• Reworked dpnp.matmul to allow larger batch size to be used #1927
• Removed data synchronization where it is not needed #1930
• Leveraged dpctl.tensor implementation for dpnp.where to support scalar as input #1932
• Improved performance of dpnp.linalg.eigh by implementing a dedicated kernel for its batch implementation #1936
• Reworked dpnp.isclose and dpnp.allclose to comply with compute follows data approach #1937
• Extended dpnp.deg2rad and dpnp.radians to support order and out keyword arguments by writing dedicated kernels for them #1943
• dpnp uses pybind11 2.13.1 #1944
• Extended dpnp.degrees and dpnp.rad2deg to support order and out keyword arguments by writing dedicated kernels for them #1949
• Extended dpnp.unwrap to support all keyword arguments provided by NumPy #1950
• Leveraged dpctl.tensor implementation for dpnp.count_nonzero function #1962
• Leveraged dpctl.tensor implementation for dpnp.diff function #1963
• Leveraged dpctl.tensor implementation for dpnp.take_along_axis function #1969
• Reworked dpnp.ediff1d implementation through existing functions instead of a separate kernel #1970
• Reworked dpnp.unique implementation through existing functions when axis is given otherwise through leveraging dpctl.tensor implementation #1972
• Improved performance of dpnp.linalg.svd by implementing a dedicated kernel for its batch implementation #1936
• Leveraged dpctl.tensor implementation for shape.setter ...

v0.15.0

Summary

This release completes implementation of dpnp.linalg module and array creation routine, adds cumulative reductions and histogram functions.

Details

Added

• Implemented dpnp.frombuffer, dpnp.fromfile and dpnp.fromstring functions #1727
• Implemented dpnp.fromfunction, dpnp.fromiter and dpnp.loadtxt functions #1728
• Added implementation of dpnp.linalg.pinv function #1704
• Added implementation of dpnp.linalg.eigvalsh function #1714
• Added implementation of dpnp.linalg.tensorinv function #1752
• Added implementation of dpnp.linalg.tensorsolve function #1753
• Added implementation of dpnp.linalg.lstsq function #1792
• Added implementation of dpnp.einsum and dpnp.einsum_path functions #1779
• Added implementation of dpnp.histogram function #1785
• Added implementation of dpnp.histogram_bin_edges function #1823
• Added implementation of dpnp.digitize function #1847
• Extended pre-commit hooks with pylint configuration #1718
• Extended pre-commit hooks with codespell configuration #1798
• Added a Security policy page #1730
• Implemented nin and nout properties for dpnp elementwise functions #1712
• Implemented outer method for dpnp elementwise functions #1813

Changed

• Added support of more number of data types and dimensions for input arrays, and all keyword arguments in dpnp.cross function #1715
• Added support of more number of data types and dimensions for input array, and all keyword arguments in dpnp.linalg.matrix_rank function #1717
• Added support of more number of data types and dimensions for input arrays in dpnp.inner function #1726
• Added support of more number of data types and dimensions for input arrays in dpnp.linalg.multi_dot function #1729
• Added support of more number of data types and dimensions for input arrays in dpnp.kron function #1732
• Added support of more number of data types and dimensions for input arrays in dpnp.linalg.matrix_power function #1748
• Added support of more number of data types and dimensions for input array, and all keyword arguments in dpnp.norm function #1746
• Added support of more number of data types and dimensions for input array in dpnp.cond function #1773
• Extended dpnp.matmul function to support axes keyword argument #1705
• Extended dpnp.searchsorted function to support side and sorter keyword arguments #1751
• Extended dpnp.where function to support scalar type by x and y arrays #1760
• Extended dpnp.ndarray.transpose method to support axes keyword as a list #1770
• Extended dpnp.nancumsum function to support axis, dtype and out keyword arguments #1781
• Extended dpnp.nancumprod function to support axis, dtype and out keyword arguments #1812
• Extended dpnp.put function to support more number of data types and dimensions for input arrays #1838
• Extended dpnp.trace function to support axis1, axis2, dtype and out keyword arguments #1842
• Corrected dpnp.ndarray.realand dpnp.ndarray.imag methods to return a view of the array #1719
• Corrected dpnp.nonzero function to raise TypeError exception for input array of unexpected type #1764
• Corrected dpnp.diagonal function to return a view of the array #1817
• Removed dpnp.find_common_type function as it was deprecated since NumPy 1.25.0 #1742
• Removed use of dpctl queue manager API #1735
• Leveraged dpctl.tensor implementation for dpnp.cumsum function #1772
• Leveraged dpctl.tensor implementation for dpnp.cumprod function #1811
• Leveraged dpctl.tensor implementation for dpnp.cumlogsumexp function #1816
• Leveraged dpctl.tensor support of out keyword argument in reduction and dpnp.where functions #1808
• Aligned with dpctl interface changes per Python Array API 2023.12 specification #1774
• Reworked dpnp.linalg.eig and dpnp.linalg.eigvals implementations to fall back on on NumPy calculation due to a lack of required functionality in OneMKL LAPACK #1780
• dpnp uses pybind11 2.12.0 #1783
• Improved dpnp.matmul implementation to use column major gemm layout for F-contiguous input arrays #1793
• Improved performance of dpnp.matmul function by call of dpnp.kron and dpnp.dot for special cases #1815
• Improved performance of dpnp.diag function by use of dpnp.diagonal which returns a view of the array #1822
• Removed limitations from diag_indices, diag_indices_from, fill_diagonal, tril_indices, tril_indices_from, triu_indices, triu_indices_from functions and added implementation of dpnp.mask_indices function #1814

Fixed

• Changed dpnp.linalg.solve to use a pair of getrf and getrs calls from OneMKL library instead of gesv one to mitigate an unexpected RuntimeError exception #1763
• Resolved a hang in batch implementation of dpnp.linalg.solve when computes on CPU device #1778
• Resolved an unexpected TypeError exception raised from dpnp.random.vonmises when used with a scalar kappa argument #1799
• Changed dpnp.flatten to comply with compute follows data approach #1825
• Resolved a hang in batch implementation of dpnp.linalg.eigh when computes on CPU device #1832
• Resolved an unexpected ValueError exception raised from dpnp.linalg.pinv due to a shape issue in dpnp.matmul #1843

New Contributors

• @ZzEeKkAa made their first contribution in #1593

Full Changelog: 0.14.0...0.15.0

v0.14.0

This release will require DPC++ 2024.1.0, which no longer supports Intel Gen9 integrated GPUs found in Intel CPUs of 10th generation and older.

Added

• Added implementation of dpnp.tensordot function #1699
• Added implementation of dpnp.nanmean and dpnp.nanstd functions #1654
• Added implementation of dpnp.angle function #1650
• Added implementation of dpnp.logsumexp and dpnp.reduce_hypot functions #1648
• Added implementation of dpnp.column_stack, dpnp.dstack and dpnp.row_stack functions #1647
• Added implementation of dpnp.nanargmax, dpnp.nanargmin, dpnp.nanmax and dpnp.nanmin functions #1646
• Added implementation of dpnp.clip function, available as well as a method of dpnp array #1645
• Added implementation of dpnp.copysign and dpnp.rsqrt functions #1624
• Added implementation of dpnp.linalg.slogdet function #1607
• Added implementation of dpnp.can_cast function #1600
• Added implementation of dpnp.linalg.solve function #1598
• Added implementation of dpnp.broadcast_arrays function #1594
• Added implementation of dpnp.tile function #1586
• Added implementation of dpnp.iinfo and dpnp.finfo functions #1582
• Added implementation of dpnp.logaddexp function #1561
• Added implementation of dpnp.positive function #1559

Changed

• Changed exception type from ValueError to NotImplementedError for unsupporting keyword arguments in array creation functions #1695
• Enabled compatibility support against numpy 1.26.4 #1690
• Implemented dpnp.true_divide as an alias on dpnp.divide function #1641
• Added support of more number of data types and dimensions for input array in dpnp.vdot function #1692
• Added support of more number of data types and dimensions for input array in dpnp.linalg.qr function #1673
• Added support of more number of data types and dimensions for input array in dpnp.dot function #1669
• Added support of more number of data types and dimensions for input array in dpnp.linalg.inv function #1665
• Added support of more number of data types for input array in dpnp.sort and dpnp.argsort functions, as well as implementing support of axis keyword #1660
• Added support of more number of data types and dimensions for input array in dpnp.linalg.cholesky function, as well as implementing support of upper keyword #1638
• Added support of more number of data types and dimensions for input array in dpnp.diff, as well as implementing support of prepend and append keywords #1637
• Added support of more number of data types and dimensions for input array in dpnp.matmul function #1616
• Added support of more number of data types and dimensions for input array in dpnp.linalg.det function #1607
• Added support of more number of data types and dimensions for input array in dpnp.linalg.svd function, as well as implementing support of full_matrices, compute_uv and hermitian keywords #1604
• Accepted different data types and dimensions of input arrays in dpnp.put_along_axis and dpnp.take_along_axis functions, as well as available values of axis keyword #1636
• Added keepdims, initial and where keywords to dpnp.amax and dpnp.amin functions #1639
• Extended dpnp.meshgrid function to support sparse and copy keyword arguments #1675
• Extended dpnp.average function to support axis, weights, returned and keepdims keywords and dpnp.nansum function with axis, dtype, keepdims and out keyword arguments #1654
• Extended dpnp.std, dpnp.var and nanvar functions to support axis, dtype, out and keepdims keyword arguments #1635
• Extended dpnp.ogrid and dpnp.mgrid functions with support of device-aware keywords of compute follows data paradigm #1622
• Extended dpnp.indices function to support dtype and sparse keyword arguments, as well as device-aware keywords of compute follows data paradigm #1622
• Extended dpnp.count_nonzero function to support axis and keepdims keyword arguments #1615
• Extended dpnp.put_along_axis and dpnp.take_along_axis functions to support out, dtype and casting keyword arguments #1608
• Extended dpnp.stack and dpnp.concatenate functions to support out, dtype and casting keyword arguments #1608
• Extended dpnp.vstack function to support dtype and casting keyword arguments #1595
• Extended dpnp.diag, dpnp.diagflat, dpnp.ptp and dpnp.vander functions with support of extra keywords to align with compute follows data paradigm #1579
• Extended dpnp.tri and dpnp.identity functions with support of device-aware keywords of compute follows data paradigm #1577
• Added dedicated in-place kernels to dpnp.divide and dpnp.floor_divide functions #1587
• Redesigned dpnp.cbrt and dpnp.exp2 functions through pybind11 extension of OneMKL call where possible or leveraging on dpctl.tensor implementation #1624
• Redesigned dpnp.exp, dpnp.expm1, dpnp.log10, dpnp.log1p and dpnp.log2 functions through pybind11 extension of OneMKL call where possible or leveraging on dpctl.tensor implementation #1576
• Redesigned dpnp.abs function through pybind11 extension of OneMKL call where possible or leveraging on dpctl.tensor implementation #1575
• Redesigned dpnp.hypot function through pybind11 extension of OneMKL call where possible or leveraging on dpctl.tensor implementation #1560
• Leveraged dpctl.tensor implementation for dpnp.reciprocal function #1650
• Leveraged dpctl.tensor implementation for dpnp.mean function #1632
• Leveraged dpctl.tensor implementation for dpnp.repeat function #1614
• Leveraged dpctl.tensor implementation for dpnp.argmax and dpnp.argmin functions #1610
• Leveraged dpctl.tensor implementation for dpnp.geomspace and dpnp.logspace functions #1603
• Leveraged dpctl.tensor implementation for dpnp.max and dpnp.min functions #1602
• Leveraged dpctl.tensor implementation for dpnp.astype function #1597
• Leveraged dpctl.tensor implementation for dpnp.maximum and dpnp.minimum functions #1558

Fixed

• Resolved potential raising of execution placement error from dpnp.take_along_axis and dpnp.put_along_axis functions #1702
• Improved performance of dpnp.matmul and dpnp.dot function when out keyword is passed #1694
• Completed documentation for each array creation functions #1674
• Aligned dpnp.clip where both min and max keywords have None value with NumPy implementation #1670
• Fixed a bug related to out keyword in elementwise functions #1656
• Resolved compilation warnings due to -Wvla-extension option enabled by default #1651
• Replaced deprecated IntelDPCPPConfig.cmake script with vendored IntelSYCLConfig.cmake #1611
• Improved coverage report to include code of pybind11 extensions #1609
• Improved performance of...