Intel's distribution of optimized python is available on ARGO. Intel has optimized the most commonly used python packages, i.e., Numpy, Scipy, Scikit-learn using Intel MKL (Math Kernel Library) and Intel Data Analytics Acceleration (DAAL) libraries. On Intel architecture these optimized packages are expected to provide a significant performance boost. To make use of Intel's distribution of packages users need to load the Intel module 'intel/python/2018-10-p36' for Python version 3.6 or 'intel/python/2018-10-p27' for Python version 2.7.
In order to check whether the intel optimized version of numpy and other packages has been loaded, the configuration of the packages can be checked as follows,
$ module load intel/python/2018-10-p36 $ pip freeze daal==2019.0 icc-rt==2019.0 impi==2019.0 intel-numpy==1.15.1 intel-openmp==2019.0 intel-scikit-learn==0.19.2 intel-scipy==1.1.0 ...
$ python3 Python 3.6.4 (default, Jun 7 2018, 10:05:32) [GCC 7.3.1 20180303 (Red Hat 7.3.1-5)] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import numpy as np >>> np.show_config() mkl_info: libraries = ['mkl_rt', 'pthread'] library_dirs = ['/opt/anaconda1anaconda2anaconda3/lib'] define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)] include_dirs = ['/opt/anaconda1anaconda2anaconda3/include']
Optimized Tensorflow and Caffe
Intel-optimized Tensorflow and Caffe are also available on ARGO. Users are encouraged to use these optimized versions if the code is written for cpus only.