This page provides the current Release Notes for the Intel® Distribution for Python*. The notes are categorized by year, from newest to oldest, with individual releases listed within each year.
Click a version to expand it into a summary of new features and changes in that version since the last release, and access the download buttons for the detailed release notes, which include important information, such as pre-requisites, software compatibility, installation instructions, and known issues.
You can copy a link to a specific version's section by clicking the chain icon next to its name.
All files are in PDF format - Adobe Reader* (or compatible) required.
- Updated packages to latest versions with CVE fixes
- dpnp updates
- Implemented the "compute follows data" programming model for the dpnp library
- dpnp package on Windows become available
- performance improvement of element-wise functions in dpnp in case of input data with strides
- numba-dppy updates
- Implemented the "compute follows data" programming model for the kernel API in numba-dppy
- Device private memory support in numba-dppy
- numba-dppy support for any array that supports the `__sycl_usm_array_interface__` protocol for the kernel API
- Provided support for Numba 0.55 and new debugging features in numba-dppy
- Enable DPNP support in numba-dppy on Windows
- Intel® Distribution for Python now supports Python version 3.9
- The dpctl package offers developers increased debugging capabilities with improved error handing and reporting
- Data Parallel Python technology now provides zero copy data exchange performance across packages
New in 2022.1.2 Product Release for Windows* only:
- Fixes Intel® Distribution for Python installation issue
- Numba-dppy works as an extension to off-the-shelf Numba 0.54.0.
- Pandas.MultiIndex support added in SDC.
- Numba-dppy’s @dppy.kernel now support __sycl_usm_array_interface through dpctl’s usm_ndarray.
- Updated all LLVM packages to LLVM 11
- Enabled @vectorize for target dppy in Numba-dppy
- Dpctl now has the ability to get command status and get profiling information from events.
- Dpctl has added queue.submit_barrier method to provide advanced synchronization mechanism to users.
- Expanded the Python C-API for working with dpctl Python objects in native extensions written in C/C++, example added for Pybind11.
- Implemented multi-versioning of DPCTLSyclInterface library on Linux.
- Dpctl can now be built using Open Source LLVM Sycl compiler.
- Providing initial DPC++ compiler conda packages
- SDC with extended Pandas API support and reduced compilation time
- Numba-dppy improvements to documentation, code quality, Python 3.8 support, profiling support
- dpctl improvements to API usability, filter selector support, explicit SYCL context and queue creation, and Level Zero program creation support for Windows
- Support for Level Zero driver to Numba-dppy
- Support SYCL filter selector in queue manager
- Better support for __sycl_usm_array_interface__ in numba-dppy
- Improvements to GDB support in Numba-DPPY generated kernels
- numpy package is updated to v1.20.1
- Improved performance of the following Intel scikit-learn CPU algorithms: Random forest, PCA, SVM
- Introduced bit-to-bit results reproducibility for Scikit-learn patches on CPU
- First public XGBoost version with GPU support
- Bug-fixes, improvements to documentations, user guides and examples
- Machine Learning
- New accelerated Scikit-learn functionality: Random Forest Classification/Regression, kNN Search/Classification/Regression, tSNE, SVC, LASSO, ElasticNet, train_test_split, assert_all_finite, sparse K-means.
- Scikit-learn and daal4py additional optimizations for DBSCAN, SVM, ElasticNet/LASSO, K-Means, train_test_split, Support Vector Classification (SVC), Random Forest, Logistic Regression, F-contiguous inputs.
- Conversion of trained XGBoost and LightGBM models into daal4py Gradient Boosted Trees model for fast prediction.
- XGBoost 1.2 with additional CPU optimizations with ‘hist’-tree method.
- Initial GPU support
- dpnp – GPU-enabled Data Parallel NumPy, a collection of many NumPy algorithms accelerated for GPUs
- dpctl – new Python package for device, queue, and USM data management with initial support in dpnp, scikit-learn, daal4py, and numba
- daal4py optimizations for GPU: KNN Classification, batch and streaming Covariance, DBSCAN, GBT Regression, K-Means, Linear & Logistic Regression, batch and streaming Low Order Moments, PCA, and binary SVM Classification
- GPU support in scikit-learn for DBSCAN, K-Means, Linear Regression and Logistic Regression algorithms
- numba – initial support for automatic GPU offload and GPU kernel semantics
- Numerical computing and image processing
- New mkl_sparse package for Intel® MKL-powered sparse matrix computations in NumPy.
- New mkl_umath package for acceleration of NumPy universal functions.
- Releasing scikit-ipp 1.2.0 for Intel® IPP-accelerated image warping , image filtering, and morphological operations
- Intel® Scalable Dataframe Compiler (Intel® SDC) Beta – Numba extension for accelerating Pandas*
- Releasing scikit-ipp 1.2.0 for image warping, image filtering and morphological operations with scikit-image like API. Support for multi-threading for transform functions and partial multi-threading for filters using OpenMP.
- Releasing mkl_umath Python package for Intel® technologies-powered NumPy universal function.
- Added new features for accelerated KNeighborsClassifier, RandomForestClassifier and RandomForestRegressor estimators, Sparse input support for KMeans and SVC, Probabilities prediction for SVC, Support of ‘normalize’ parameter for Lasso and ElasticNet in scikit-learn.
- Optimizations of train_test_split and Support Vector Classification (SVC) fit and prediction in scikit-learn.
- Conversion of trained XGBoost* and LightGBM* models into daal4py Gradient Boosted Trees model for fast prediction.
- Added new features for Brute Force method for k-Nearest Neighbors classification, new parameters support for k-Nearest Neighbors and Decision Forest in daal4py.
- Optimizations of Support Vector Machine training and prediction, Decision Forest classification training in daal4py.
- Latest CVE patches have been applied.
- Implemented Scikit-Learn compatible Gradient Boosted Tree classifier, Decision Tree Classifier and tree-based Adaboost classifier in daal4py.
- Implemented computation of prediction probabilities in Scikit-Learn compatible RandomForest and Gradient Boosted Trees classifiers in daal4py.
- numpy package is updated to v1.18.5
- Scikit-learn package is updated to v 0.23.1
- Support of thunder SVM method for IDP sklearn in daal4py
- Performance optimizations for SVC fit and prediction, Elastic Net and LASSO fit, K-Means fit and prediction, PCA fit and transform, train_test_split
- DBSCAN is accelerated in sklearn.
- Performance improvement for F-contiguous inputs in daal4py.
- Patches updated for compatibility with sklearn 0.22.
- Update to conda 4.7.12.
- Added support for Brownboost, Logistboost as well as Stump regression and Stump classification algorithms to daal4py.
- Added support for Adaboost classification algorithm, including support for method="SAMME" or "SAMMER" for multi-class data in daal4py.
- "Variable Importance" feature has been added to Gradient Boosting Trees in daal4py.
- Ability to compute class prediction probabilities has been added to appropriate classifiers, including logistic regression, tree-based classifiers, etc., in daal4py.
- Single node support for DBSCAN, LASSO, Coordinate Descent (CD) solver algorithms through daal4py package
- Distributed model support for SVD, QR, K-means init++ and parallel++ algorithms through daal4py package
- Additional Scikit-Learn algorithms optimized using Intel® DAAL: Linear, Ridge, Logistic, PCA, KMeans, pairwise_distances and SVC
- New distributed model support for "Moments of low order" and "Covariance" algorithms through daal4py package
- Updated python package versions and their supported platforms
- Extended availability of Intel® DAAL algorithms through daal4py package.
- Daal4py distributed mode support for scale-out to clusters & support for streaming mode for efficient memory handling.
- Updated python packages and their supported platforms.
- Intel® Distribution for Python 2019 Update 2 includes functional and security updates. Users should update to the latest version.
- Scikit-learn optimizations for Logistic Regression, Random Forest Regressor & Classifier.
- New machine learning package (daal4py) with an easy to use API and performance accelerated by Intel® DAAL.
- Introducing Numba* threading layer abstraction that allows to switch between Intel® TBB (default) and OpenMP* threading layer.
- Access to MKL runtime settings available through easy-to-use Python control package (mkl-service)
- Intel® Distribution for Python now integrated into Intel® Parallel Studio XE 2019 installer. Also available as easy command line standalone install.
- Faster Machine learning with Scikit-learn: Support Vector Machine (SVM) and K-means prediction, accelerated with Intel® DAAL.
- XGBoost package included in Intel® Distribution for Python (Linux only).
- Note: The deep learning packages and computer vision packages along with their dependencies will not be included in Intel Distribution for Python, henceforth. However, the packages continue to be available in anaconda/Intel conda channel. Click on the Complete List of Packages for the Intel® Distribution for Python* to learn more.
- SVM classification for Radial Basis Function (RBF) kernel in scikit-learn accelerated with Intel® Data Analytics Acceleration Library for faster training and prediction
- Numba enabled with Short Vector Math Library (SVML) by default, to leverage auto-vectorization and parallelization for performance
- Updated packages include NumPy, SciPy, scikit-learn, Cython, pyDAAL & tbb4py
- Updated 08/06/2018: For users who‘ve already downloaded this product through the windows standalone installers (w_python*_pu3_2018.3.040.exe), due to an installer issue on user group privileges, we have made new packages available with the fix (w_python*_pu3_2018.3.042.exe). Please visit Intel® Registration Center to install and update your existing product.
- Scikit-learn functions: Support Vector Machine (SVM) binary and multiclass algorithms, and K-means prediction, accelerated with Intel® DAAL
- Short Vector Math Library (SVML) optimizations enabled by default in Numba, allowing control of accuracy of SVML functions via fast-math argument.
- XGBoost package included in Intel® Distribution for Python
- mkl_fft and mkl_random have been released as stand-alone packages (originally integrated into Intel's NumPy package)
- Miscellaneous bug fixes and version bumps
- Python 3.6 Support
- Performance accelerations via the latest 2018 Intel® Performance Libraries: Scikit-learn* with Intel® DAAL, FFT in SciPy*, Arithmetic and Transcendental Expressions, and Memory optimizations.
- Updated with Packages for OpenCV, IPP, MLSL, and MKL-DNN
New in this release:
- Includes Intel-optimized Deep Learning frameworks Caffe and Theano, powered by the new Intel® MKL-DNN
- Select Scikit-learn algorithms now accelerated with Intel® Data Analytics Acceleration Library for ~200X speedup
- Arithmetic, transcendental and 1D & multi-dimensional FFT functions significantly faster in NumPy and SciPy
New in this release:
- Announcing the first public release of the performance driven Intel® Distribution for Python* 2017!
- Achieve near-native speedups with performance libraries such as Intel® Math Kernel Library, threading efficiency with TBB, data analytics with pyDAAL.
- Seamless interoperability with conda and Anaconda Cloud. Scale easily with mpi4py and Jupyter notebooks.
Notices and Disclaimers
Intel technologies may require enabled hardware, software or service activation.
No product or component can be absolutely secure.
Your costs and results may vary.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.
No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.
The products described may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request.
Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.