scikit-learn-intelex 2021.3.0

Intel(R) Extension for Scikit-learn is a seamless way to speed up your Scikit-learn application.

Intel(R) Extension for Scikit-learn*

Intel(R) Extension for Scikit-learn is a seamless way to speed up your Scikit-learn application. The acceleration is achieved through the use of the Intel(R) oneAPI Data Analytics Library (oneDAL). Patching scikit-learn makes it a well-suited machine learning framework for dealing with real-life problems.

⚠️Intel(R) Extension for Scikit-learn contains scikit-learn patching functionality that was originally available in daal4py package. All future updates for the patches will be available only in Intel(R) Extension for Scikit-learn. We recommend you to use scikit-learn-intelex package instead of daal4py. You can learn more about daal4py in daal4py documentation.

Running the latest scikit-learn test suite with Intel(R) Extension for Scikit-learn:

👀 Follow us on Medium

We publish blogs on Medium, so follow us to learn tips and tricks for more efficient data analysis the help of Intel(R) Extension for Scikit-learn. Here are our latest blogs:

• Intel Gives Scikit-Learn the Performance Boost Data Scientists Need
• From Hours to Minutes: 600x Faster SVM
• Improve the Performance of XGBoost and LightGBM Inference
• Accelerate Kaggle Challenges Using Intel AI Analytics Toolkit
• Accelerate Your scikit-learn Applications
• Accelerate Linear Models for Machine Learning
• Accelerate K-Means Clustering

🔗 Important links

• Documentation
• scikit-learn API and patching
• Benchmark code
• Building from Sources
• About Intel(R) oneAPI Data Analytics Library
• About Intel(R) daal4py

💬 Support

Report issues, ask questions, and provide suggestions using:

• GitHub Issues
• GitHub Discussions
• Forum

You may reach out to project maintainers privately at [e-mail]

🛠 Installation

Intel(R) Extension for Scikit-learn is available at the Python Package Index, on Anaconda Cloud in Conda-Forge channel and in Intel channel.

# PyPi (recommended by default)
pip install scikit-learn-intelex

# Anaconda Cloud from Conda-Forge channel (recommended for conda users by default)
conda install scikit-learn-intelex -c conda-forge

# Anaconda Cloud from Intel channel (recommended for Intel® Distribution for Python users)
conda install scikit-learn-intelex -c intel

[Click to expand] ℹ️ Supported configurations

📦 PyPi channel

OS / Python version	Python 3.6	Python 3.7	Python 3.8	Python 3.9
Linux	[CPU, GPU]	[CPU, GPU]	[CPU, GPU]	❌
Windows	[CPU, GPU]	[CPU, GPU]	[CPU, GPU]	❌
OsX	[CPU]	[CPU]	[CPU]	❌

📦 Anaconda Cloud: Conda-Forge channel

OS / Python version	Python 3.6	Python 3.7	Python 3.8	Python 3.9
Linux	[CPU]	[CPU]	[CPU]	[CPU]
Windows	[CPU]	[CPU]	[CPU]	[CPU]
OsX	[CPU]	[CPU]	[CPU]	[CPU]

📦 Anaconda Cloud: Intel channel

OS / Python version	Python 3.6	Python 3.7	Python 3.8	Python 3.9
Linux	[CPU, GPU]	[CPU, GPU]	[CPU, GPU]	❌
Windows	[CPU, GPU]	[CPU, GPU]	[CPU, GPU]	❌
OsX	[CPU]	[CPU]	[CPU]	❌

You can build the package from sources as well.

⚡️ Get Started

Intel CPU optimizations patching

import numpy as np
from sklearnex import patch_sklearn
patch_sklearn()

from sklearn.cluster import DBSCAN

X = np.array([[1., 2.], [2., 2.], [2., 3.],
              [8., 7.], [8., 8.], [25., 80.]], dtype=np.float32)
clustering = DBSCAN(eps=3, min_samples=2).fit(X)

Intel GPU optimizations patching

import numpy as np
from sklearnex import patch_sklearn
from daal4py.oneapi import sycl_context
patch_sklearn()

from sklearn.cluster import DBSCAN

X = np.array([[1., 2.], [2., 2.], [2., 3.],
              [8., 7.], [8., 8.], [25., 80.]], dtype=np.float32)
with sycl_context("gpu"):
    clustering = DBSCAN(eps=3, min_samples=2).fit(X)

🚀 Scikit-learn patching

Speedups of Intel(R) Extension for Scikit-learn over the original Scikit-learn
Technical details: float type: float64; HW: Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz, 2 sockets, 28 cores per socket; SW: scikit-learn 0.23.1, Intel® oneDAl (2021.1 Beta 10), benchmark code

Intel(R) Extension for Scikit-learn patching affects performance of specific Scikit-learn functionality listed below. In cases when unsupported parameters are used, the package fallbacks into original Scikit-learn. These limitations described below. If the patching does not cover your scenarios, submit an issue on GitHub.

[Click to expand] 🔥 Applying the patching will impact the following existing scikit-learn algorithms:

Task	Functionality	Parameters support	Data support
Classification	SVC	All parameters except kernel = 'poly' and 'sigmoid'.	No limitations.
	RandomForestClassifier	All parameters except warmstart = True and cpp_alpha != 0, criterion != 'gini'.	Multi-output and sparse data is not supported.
	KNeighborsClassifier	All parameters except metric != 'euclidean' or minkowski with p != 2.	Multi-output and sparse data is not supported.
	LogisticRegression / LogisticRegressionCV	All parameters except solver != 'lbfgs' or 'newton-cg', class_weight != None, sample_weight != None.	Only dense data is supported.
Regression	RandomForestRegressor	All parameters except warmstart = True and cpp_alpha != 0, criterion != 'mse'.	Multi-output and sparse data is not supported.
	KNeighborsRegressor	All parameters except metric != 'euclidean' or minkowski with p != 2.	Sparse data is not supported.
	LinearRegression	All parameters except normalize != False and sample_weight != None.	Only dense data is supported, #observations should be >= #features.
	Ridge	All parameters except normalize != False, solver != 'auto' and sample_weight != None.	Only dense data is supported, #observations should be >= #features.
	ElasticNet	All parameters except sample_weight != None.	Multi-output and sparse data is not supported, #observations should be >= #features.
	Lasso	All parameters except sample_weight != None.	Multi-output and sparse data is not supported, #observations should be >= #features.
Clustering	KMeans	All parameters except precompute_distances and sample_weight != None.	No limitations.
	DBSCAN	All parameters except metric != 'euclidean' or minkowski with p != 2, algorithm != brute or auto .	Only dense data is supported.
Dimensionality reduction	PCA	All parameters except svd_solver != 'full'.	No limitations.
	TSNE	All parameters except metric != 'euclidean' or minkowski with p != 2.	Sparse data is not supported.
Unsupervised	NearestNeighbors	All parameters except metric != 'euclidean' or minkowski with p != 2.	Sparse data is not supported.
Other	train_test_split	All parameters are supported.	Only dense data is supported.
	assert_all_finite	All parameters are supported.	Only dense data is supported.
	pairwise_distance	With metric='cosine' and 'correlation'.	Only dense data is supported.
	roc_auc_score	Parameters average, sample_weight, max_fpr and multi_class are not supported.	No limitations.

⚠️ We support optimizations for the last four versions of scikit-learn. The latest release of Intel(R) Extension for Scikit-learn 2021.2.X supports scikit-learn 0.21.X, 0.22.X, 0.23.X and 0.24.X.

📜 Intel(R) Extension for Scikit-learn verbose

To find out which implementation of the algorithm is currently used (Intel(R) Extension for Scikit-learn or original Scikit-learn), set the environment variable:

• On Linux and Mac OS: export SKLEARNEX_VERBOSE=INFO
• On Windows: set SKLEARNEX_VERBOSE=INFO

For example, for DBSCAN you get one of these print statements depending on which implementation is used:

• SKLEARNEX INFO: sklearn.cluster.DBSCAN.fit: running accelerated version on CPU
• SKLEARNEX INFO: sklearn.cluster.DBSCAN.fit: fallback to original Scikit-learn

Read more in the documentation.