How to Install the Python* Version of Intel® Data Analytics Acceleration Library (Intel® DAAL) in Linux*
Published: 05/11/2016
Last Updated: 05/11/2016
Introduction
The Intel® Data Analytics Acceleration Library (Intel® DAAL) 1, 2 is a software solution for data analytics. It provides building blocks for data preprocessing, transformation, modeling, predicting, and so on.
The beta version of Intel DAAL 2017 provides support for the Python* scripting language.
This article shows you a simple way to install Intel Distribution for Python and Intel DAAL in Linux.
Installation
There is more than one way to install the Python version of Intel DAAL in Linux*. Intel DAAL is included with the Intel® Distribution for Python* and in the Intel® Parallel Studio XE 2017 beta. However, for the Intel Parallel Studio 2017 beta 3, users will need to compile the Python source before it can be used.
Thanks to Zhang for showing a simple way to install Intel DAAL through the Anaconda* 4 distribution.
Step-by-Step Instructions
The following steps show you how to install Intel Distribution for Python and Intel DAAL:
-
Install the Anaconda distribution.
- Go to the Anaconda download website 4, and then download either the 2.7 or 3.5 Python version of Anaconda for 64-bit Linux.
-
Execute the following command to install Anaconda:
bash Anaconda2-4.0.0-Linux-x86_64.sh
(for Python 2.7)
or
bash Anaconda3-4.0.0-Linux-x86_64.sh
(for Python 3.5)
-
Add the Intel channel to Anaconda.
-
Before adding the Intel channel into Anaconda, make sure that Anaconda has been correctly installed. One way to do that is to run Python at the command prompt to see if this Python version is from Anaconda:
-
If step 2.1 is correct, use the following command to add the Intel channel to Anaconda using the conda utility included in Anaconda:
conda config --add channels intel
-
Check to ensure that the Intel channel has been added using the following command:
conda config --get channels
If the Intel channel has been added, the result will look something like this:
-
-
Install Python version of Intel® DAAL under Anaconda.
Using the following command to create a new environment for the Intel version Python installation and to install the Intel Distribution for Python and Intel DAAL:
conda create -n intelpython --override-channels --channel intel python=3.5 intelpython scipy pydaal
where:
intelpython (after the switch '-n'): The name of the Intel version Python environment. This can be any name. This name will be used to switch to the Intel version Python environment later on.
Channel intel: the Intel channel.
python=3.5: the Intel version Python installation uses Python version 3.5, not 2.7.
intelpython: the Intel version Python installation name. This cannot be changed.
pydaal: Python version of Intel DAAL.
scipy: pydaal needs this package. -
Check to make sure the Python's environment exists.
Use the following command to ensure the Intel version Python environment “intelpython” has been created:
conda env list
The screen should look like this:
-
Switch to the Intel Python environment.
To use the Python version of Intel DAAL, you need to be in the Intel version Python environment. Use the following command to switch to the Intel version Python environment “intelpython” created above:
Source activate intelpython
If the environment is switched successfully, you will see something like this:
To switch back to the default environment, issue the following command:
Source deactivate
-
Check to make sure the Intel version Python installation is selected.
Before using Intel Distribution for Python, we need to switch to the Intel version Python environment as shown above. Run the following command to ensure the Intel version Python installation is selected, not the version of Python that comes with Anaconda:
Python3.5
The screen should look like this:
Note: The Python version of Intel DAAL only works with Python 3.5 and above. It does not work with Python 2.7.
References
4 https://www.continuum.io/download
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.