Tools/uxlfoundation/scikit-learn-intelex

scikit-learn-intelex

Extension for Scikit-learn is a seamless way to speed up your Scikit-learn application

1.4k ★+1/wkemergingPythonApache License 2.0 →trending

The Lens

By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.

Updated Nov 2026

scikit-learn-intelex makes existing scikit-learn code faster with two extra lines. scikit-learn is the standard Python library for classic machine learning (random forests, logistic regression, clustering), and this extension swaps its internals for Intel's optimized oneDAL math library. Intel built it, the UXL Foundation hosts it, and it is Apache-2.0 and free.

Nothing to host. Install from pip or conda-forge, call the patch_sklearn function before importing scikit-learn, and supported estimators get the accelerated version. The README claims an average 8.5x speedup on training and inference, up to 100x on some workloads, with equivalent accuracy. GPU offload runs through Intel's oneAPI/SYCL stack and a separate scikit-learn-intelex-gpu package.

Verdict: free speed for scikit-learn on x86 servers, worth a benchmark before you pay for bigger instances. Solo: free, try it on your slowest training job. Small teams: free, check your exact estimators are accelerated. Larger teams: free, a cheap win for CPU-heavy pipelines.

The catch: prebuilt wheels are x86-64 only, for Linux and Windows, so Macs and ARM servers are out. The headline numbers are Intel's own benchmarks; your gain depends on algorithm and data size.

Free vs Self-Hosted vs Paid

fully free

Free: The whole extension, from PyPI or conda-forge, under Apache-2.0. The GPU package scikit-learn-intelex-gpu is free too.

Self-hosted: It is a library, so it runs wherever your Python runs, as long as that is x86-64 Linux or Windows (the platforms with prebuilt wheels and conda packages). GPU offload needs Intel's oneAPI system software.

Paid: None. No paid edition or support plan from the project.

Free, and the main financial effect is the opposite of a cost: faster CPU training can mean smaller or fewer instances.

What to do by team size

Solo
free, patch and benchmark your slowest job
Small team
free, verify your estimators are on the accelerated list
Medium team
free, add it to x86 training images
Large team
free, measure on your own data before resizing infrastructure
Self-hosting ops:trivial

Get tools like this every Wednesday

One featured tool, three on the radar. No fluff.

Score
72/100 · B+
Adoption13/30
Maintenance25/25
Community9/20
License15/15
Analysis10/10

A low score is not a verdict on quality. Young and niche tools start low by design. How we calculate scores

Trust Signals

Community discussions enabledOrganization account (23 public repos)

License: Apache License 2.0

Use freely. Patent grant included.

Commercial use: ✓ Yes

About

Owner
Unified Acceleration (UXL) Foundation (Organization)
Stars
1,358
Forks
198

Explore Further

More tools in the directory