What's New: Intel® Distribution for Python* 2025.2

[StefR_Intel]

The Intel® Distribution for Python, 2025.2 has been updated to include functional and security updates. Users should update to the latest version. For more details on security and known issues, please check out Intel® Distribution for Python* Known Issues.

Deprecation notice

Starting with the 2025.2 release, the standalone installer will no longer be provided. You can check out all the available installation methods here.

What’s new

The mkl_fft package includes the following improvements:

• FFT functions now support the out keyword argument.
• The interface for numpy (mkl_fft.interfaces.numpy_fft) is aligned with NumPy 2.x.x.
• The interface for scipy (mkl_fft.interfaces.scipy_fft) is aligned with SciPy 1.15.x.

Data Parallel Control Library* includes the following improvements:

• Support for the composite devices extension.
• Support for calling kernels with local_accessor, work_group_memory, and raw_kernel_arg from Python.
• Dpctl.tensor conformity to 2024.12 array API specification.
• Enabling the use of compiler flags to build dpctl for specific Intel architectures.

Data Parallel Extension for Python* language includes the following improvements:

• Conformity to 2024.12 array API specification.
• NumPy/CuPy functions gap closure, new support for:
  • Statistics functions: median, percentile, quantile + nan-like
  • Windows functions
  • Elementwise functions: bitwise_count, convolve, divmod, frexp, interp, modf
  • Indexing routine: putmask & special classes
• Compatibility with NumPy 2.2.4.

Other improvements:

• The mkl_umath package is now aligned with numpy-2.x.x.
• The mkl_service package now supports Python 3.13.

[codefarmer]

Really glad to see support for NumPy 2.x with IDP 2025.2 release! Quick notes from my installation attempts:

The download page for IDP provides installation commands for Python 3.11 and 3.12 however with the `pip` installation method, downloads succeed only for Python 3.13 (which is welcome, btw).  Here are the errors with Python 3.12.10 which indicate that the specified NumPy wheels are available:

❯ python -V
Python 3.12.10
❯ python -m pip install -i https://software.repos.intel.com/python/pypi numpy scipy dpnp dpctl tbb4py
Looking in indexes: https://software.repos.intel.com/python/pypi
Collecting numpy
ERROR: HTTP error 404 while getting https://software.repos.intel.com/python/pypi/numpy/numpy-2.2.5-5-cp312-cp312-win_amd64.whl#sha256=513e0cbc15832363985ceecc0f453bfe17de9d99028047b7d87d107d1d218ac4 (from https://software.repos.intel.com/python/pypi/numpy/)
ERROR: Could not install requirement numpy from https://software.repos.intel.com/python/pypi/numpy/numpy-2.2.5-5-cp312-cp312-win_amd64.whl#sha256=513e0cbc15832363985ceecc0f453bfe17de9d99028047b7d87d107d1d218ac4 because of HTTP error 404 Client Error: Not Found for url: https://software.repos.intel.com/python/pypi/numpy/numpy-2.2.5-5-cp312-cp312-win_amd64.whl for URL https://software.repos.intel.com/python/pypi/numpy/numpy-2.2.5-5-cp312-cp312-win_amd64.whl#sha256=513e0cbc15832363985ceecc0f453bfe17de9d99028047b7d87d107d1d218ac4 (from https://software.repos.intel.com/python/pypi/numpy/)

As an aside I wonder if the Python 3.12 and 3.11 choice on the download page is still relevant? Maybe for the conda/mamba installs?

And finally, a request: it would be great if my choice of Platform (Windows) wouldn't revert back to Linux each time I choose any of the other configuration options.

Thanks.

[StefR_Intel]

Hi @codefarmer ,

Thank you for bringing this to our attention. We are looking into it and will follow up accordingly.

Regards,

Stef

[StefR_Intel]

Hi @codefarmer ,

The pip installation should be fixed now. Sorry for the inconvenience.

Regards,

Stef