daal4py

What it is and what it does

daal4py wraps Intel's oneAPI Data Analytics Library (oneDAL) to provide optimized implementations of machine learning algorithms accessible through a Python API. It targets data scientists and ML practitioners who want to accelerate computationally intensive operations like SVM, linear models, and K-means clustering without rewriting their code. The package depends on daal (the underlying C++ library) and numpy, and is distributed as pre-built wheels for Python 3.9–3.12 on Linux (manylinux1) and Windows (x86_64).

Historically, daal4py offered scikit-learn patching to transparently accelerate scikit-learn calls, but that functionality has been deprecated and moved to a separate package (Intel Extension for Scikit-learn). Current users should treat daal4py as a direct API to oneDAL algorithms rather than a drop-in scikit-learn accelerator.

Use it for:

• Accelerate SVM training and inference on large datasets where CPU optimization is a bottleneck.
• Speed up linear regression, logistic regression, and ridge regression on high-dimensional data.
• Optimize K-means clustering performance for exploratory data analysis on large point clouds.
• Integrate oneDAL algorithms directly into custom ML pipelines that don't rely on scikit-learn.
• Benchmark performance of Intel-optimized algorithms against standard implementations.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

daal4py provides a Python API to Intel's oneAPI Data Analytics Library, offering optimized implementations of machine learning and data analytics algorithms.

Yes, if you need accelerated implementations of specific algorithms (SVM, linear models, K-means) and run on Linux or Windows x86_64. Medium install friction and a stable but slowly-evolving codebase are acceptable tradeoffs for the performance gains. Not recommended if you rely on scikit-learn patching—use Intel Extension for Scikit-learn instead. No known vulnerabilities.

Install

daal4py on PyPI

pip install daal4py

uv add daal4py

poetry add daal4py

Installing daal4py

Before you install

Medium install friction due to platform-specific wheels (manylinux1 and Windows x86_64 only). Active maintenance with recent commits; last release 696 days ago suggests a stable but not rapidly evolving codebase.

License in practice

Apache v2.0 permissive license allows commercial and private use with minimal restrictions; suitable for most projects.

Quickstart

pip install daal4py

import daal4py as d4p
from daal4py.sklearn.svm import SVC

# Use optimized SVM classifier
clf = SVC(kernel='rbf')
clf.fit(X_train, y_train)

Requires Linux (manylinux1) or Windows x86_64; macOS and other architectures not supported via PyPI wheels. Requires Python >=3.7.

Verify before relying

• Whether GPU acceleration is available in current PyPI wheels or only through conda channels
• Performance gains over standard scikit-learn for specific algorithm classes
• Current status of scikit-learn patching after deprecation and move to separate package

Package facts

Evidence: daal4py-2024.7.0-py310-none-manylinux1_x86_64.whl; daal4py-2024.7.0-py310-none-win_amd64.whl; daal4py-2024.7.0-py311-none-manylinux1_x86_64.whl; daal4py-2024.7.0-py311-none-win_amd64.whl; daal4py-2024.7.0-py312-none-manylinux1_x86_64.whl; daal4py-2024.7.0-py312-none-win_amd64.whl; daal4py-2024.7.0-py39-none-manylinux1_x86_64.whl; daal4py-2024.7.0-py39-none-win_amd64.whl

Keywords: machine learning, scikit-learn, data science, data analytics

Tags

More Software Development packages