daal4py - Intel® oneAPI Data Analytics Library

2024.7.0 verified Fri May 01 auth: no python

daal4py provides a convenient Python API to Intel® oneAPI Data Analytics Library (oneDAL) for accelerating machine learning and data analysis. Current stable version is 2024.7.0 on PyPI, with newer releases (2025.11.0) available via Intel's Extension for Scikit-learn. Release cadence is approximately monthly.

pip install daal4py

Common errors

error ModuleNotFoundError: No module named 'daal4py' ↓

cause daal4py not installed or installed in wrong environment.

fix

Run 'pip install daal4py' or 'conda install -c intel daal4py'

error AttributeError: module 'daal4py' has no attribute 'kmeans' ↓

cause Using daal4py version older than 2020 where API changed.

fix

Upgrade daal4py to at least version 2020.0, e.g., 'pip install -U daal4py'

error RuntimeError: Provided context is not sycl_context instance. ↓

cause Passing wrong object to sycl_context or not importing it correctly.

fix

Use 'from daal4py.oneapi import sycl_context' and wrap code in 'with sycl_context("cpu"):'

error TypeError: compute() missing 1 required positional argument: 'data' ↓

cause Calling compute on a method that expects data but none provided.

fix

Ensure algo.compute(data) passes the input array.

Warnings

breaking daal4py 2020.x has a different API compared to newer versions. Upgrade may require code changes. ↓

fix Refer to migration guide: https://intelpython.github.io/daal4py/migration.html

deprecated Internal dpctl tensor handling functionality is deprecated and will be removed in future releases. ↓

fix Use standard numpy or dpctl arrays directly.

gotcha daal4py objects (e.g., from compute) must not be reused after passing to another compute; create fresh instances. ↓

fix Always create a new algorithm object for each call.

gotcha SYCL context must be set before calling any daal4py compute, otherwise it defaults to CPU host. ↓

fix Use 'with sycl_context("gpu"):' to ensure GPU usage.

Install

conda install -c intel daal4py

Imports

daal4py
```
import daal4py
```
No submodule imports; use daal4py.sklearn for patching

Quickstart

Simple K-Means using oneDAL native API with SYCL context.

import numpy as np
import daal4py
from daal4py.oneapi import sycl_context
# Example: K-Means clustering (oneDAL native)
data = np.random.rand(100, 5).astype(np.float64)
with sycl_context('cpu'):
    n_clusters = 3
    algo = daal4py.kmeans(nClusters=n_clusters, maxIterations=100)
    result = algo.compute(data)
print(result.centroids)