Interpreter and modules¶
Important
We recommend the use of Intel Python 3 distribution, provided by Intel oneAPI installation, whenever possible. Refer to Discoverer Python Project for more details.
Intel Python 3 interpreter¶
Intel Distribution for Python provides a highly reliable and optimized for productivity Python distribution, including an optimized Python interpreter. To gain access to that Python interpeter, load the module intel.universe
in your Slurm batch script:
module load intel.universe
Afterwards any execution of python
or python3
from within that script, without specifying the full path to the executable, will invoke the Intel Python 3 interpreter.
Note
Intel Python 3 is built without readline support.
Modules¶
Apart from the optimized Python interpreter and its default modules, Intel Distribution for Python installs optimized modules for numerical computations, data analysis, machine learning, and AI. The following modules could be imported within the running Python interpreter process:
- daal4py
- dpctl
- dpnp
- matplotlib
- mkl
- mpi4py
- numba
- numba_dppy
- numexpr
- numpy
- onedal
- pandas
- scipy
- sklearn
- sklearnex
- tbb
Conda and local Conda channel¶
By default Intel Distribution for Python comes with Conda and a local Conda channel. The local Conda channel is located at:
/opt/software/intel/oneapi/conda_channel
The Conda executable can be invoked once the corresponding environment module is successfully loaded:
module load intel.universe
Warning
On Discoverer HPC users should use Conda to create local environment installations ONLY within their Personal scratch and storage folder (/discofs/username). Do not store the Python virtual environments under Home folder (/home/username) since they occupy large ammount of disk space!
To install a virtual environment with Conda on the Discoverer HPC, follow the example:
mkdir -p /discofs/${USER}/jupyter_test_env_numpy
conda install -y -c /opt/software/intel/oneapi/conda_channel -p /discofs/${USER}/jupyter_test_env_numpy jupyterlab seaborn
Other Python versions¶
Users are free to bring their own Python intepreter and modules and install them into their Personal scratch and storage folder (/discofs/username). Note that those installations will not be supported by Discoverer HPC support team.
Getting help¶
See Getting help