The following are key question and answers in supporting common steps needed for using the Intel® Xeon Phi™ processor powered Remote Access Cluster for Artificial Intelligence. If you’ve been granted node access to this system to test projects, and build out Machine Learning or Deep Learning models, the FAQs below will serve to help you get started on the system.
- How do I get access to my Intel node?
- You will receive an email from Colfax International that provides a link to some instructions you need to follow in order to setup your system. The instructions are simple to follow and should only take about 10 minutes. It is important that you read each section that pertains to you carefully so that you avoid any confusion or the need to redo the setup process.
- How do I install system packages?
- You can install system packages using the “sudo yum” command.
- Does my Intel node have Intel® Distribution for Python* installed?
- To check if your node has Intel Distribution for Python* installed, look in the “/opt/intel” directory. There should be “intelpython*” directories in “/opt/intel.” The installed version is determined by the remainder of the directory name. For example, intelpython27 means that this directory contains Intel Distribution for Python* 2.7.
- How can I use Intel® Distribution for Python* and Pip instead of the default system Python* and Pip?
You can add “intel-python” and “intel-pip” aliases in the “.bashrc” file in your home directory. The following examples show how to add an alias for Intel Distribution for Python* 2.7 and its corresponding Pip. To use an alias with any other Intel Distribution for Python*, you must substitute the version digits (that is, ’27’) for the installed version you would like to use.
To add an “intel-python” alias:
To add an “intel-pip” alias:
- How do I install Intel® Software Optimization for Theano*?
Each Intel node should already have Intel Software Optimization for Theano* in “/opt/theano”. To use this Intel Software Optimization for Theano* version, use the following commands.
NOTE: If you have not installed these prerequisite packages, please install them using the following command.
pip install nose-parameterized, Pygments, requests, docutils, snowballstemmer, alabaster, Jinja2, imagesize, pytz, babel, Sphinx —user —no-deps # create an installs directory if one does not already exist [[ -d ~/.installs/node_copies ]] || mkdir -p ~/.installs/node_copies # remove the previous installation of theano if it exists rm -rf ~/.installs/node_copies/intel-theano # copy Intel Theano from the opt directory cp -r /opt/theano ~/.installs/node_copies/intel-theano # install Intel Theano using Pip cd ~/.installs/node_copies/intel-theano python setup.py build pip install . —user cp theanorc_icc_mkl ~/.theanorc echo “Test if Intel Theano is installed by running python and importing theano” echo “For more information, see the Theano README.md or Installation files”
- How do I install Keras*?
A version of the Keras deep learning library is in the “/opt/“ directory. To install this version of Keras, make sure that Theano or TensorFlow is installed first, then use the following commands.
# install Keras dependencies. pip install h5py —user —no-deps # create an installs directory if one does not already exist [[ -d ~/.installs/node_copies ]] || mkdir -p ~/.installs/node_copies # remove the previous installation of keras if it exists rm -rf ~/.installs/node_copies/keras # copy from the opt directory cp -r /opt/keras ~/.installs/node_copies/keras # install Keras using Pip cd ~/.installs/node_copies/keras python setup.py build pip install . —user echo “Test if Keras is installed by running python and importing keras” echo “For more information, see the Keras README.md or Installation files”