Python bundles

Bundle names

• The bundle names reflect the content, like Python packages, and its version, but also which Python version, compilers and libraries that are compatible with it.

“Some well-known bundles by names”

- HPC and big data
    - dask
    - mpi4py
    - numba
- Scientific tools
    - SciPy-bundles: ``numpy``, ``pandas``, ``scipy``
    - xarray
- Biopython
- Interactivity
    - iPython
    - JupyterLab
- Graphics and diagrams
    - Matplotlib
    - Seaborn
- Machine Learning
    - scikit-learn
    - PyTorch
    - TensorFlow
- Bundle of useful packages
    - Python-bundle-PyPI

Package	Bundle module	Also loads	Avail at *
numpy	SciPy-bundle	Python-bundle-PyPI	P, K, C
pandas	SciPy-bundle	Python-bundle-PyPI	P, K, C
scipy	SciPy-bundle	Python-bundle-PyPI	P, K, C
matplotlib	matplotlib	SciPy-bundle	P, K, C
seaborn	Seaborn	Matplotlib	P, K, C
biopython	BioPython	SciPy-bundle	P, K, C
dask	dask	Matplotlib	P, K, C
ipython	IPython	Python-bundle-PyPI	P, K, C
Jupyterlab	JupyterLab	IPython	P, K, C
xarray	xarray	SciPy-bundle	P, , C
numba	numba	SciPy-bundle	P, K, C
mpi4py	mpi4py	OpenMPI	P, K, C
scikit-learn	scikit-learn	SciPy-bundle	P, K, C
torch	PyTorch	OpenMPI	P, K, C

• Dardel (D), Tetralith (T), Alvis (A), Pelle (P), Kebnekaise (K), Cosmos (C)

Bundle versions

• The module endings may contain GCCcore-X.Y.Z and/or [YEAR-a/b]. Example SciPy-bundle/2024.05-gfbf-2024a or Python/3.12.3-GCCcore-13.3.0

  • GCCcore reflects the GCC compiler version that is compatible when using C/C++ “back end” code.

  • The year reflects an EasyBuild toolchain, see FOSS toolchains.

“Some FOSS tool chains and Python version using them”

FOSS	Python version	GCC version	Bundle version
2024a	3.12.3	13.3.0	2024.06/06
2025b	3.13.5	14.3.0	2025.07

• foss is the full level toolchain.

• gfbf means that the libraries FlexiBLAS (incl. LAPACK) + FFTW are included.

• gompi means that the MPI library OpenMPI is included.

• See Toolchain diagram

Danger

• Make sure to use bundles that are compatible with each-other and with needed Python version.

• Otherwise it is better to create isolated environments with Conda or virtual environments, see Isolated environments in Python.

Example Matplotlib

• contourpy

• Cycler

• fonttools

• kiwisolver

• matplotlib

Example Matplotlib 3.10.5-gfbf-2025b

Packages:

• contourpy-1.3.3

• cycler-0.12.1

• fonttools-4.58.5

• kiwisolver-1.4.8

• matplotlib-3.10.5

Dependencies:

• Python/3.13.5-GCCcore-14.3.0

• Python-bundle-PyPI/2025.07-GCCcore-14.3.0

• SciPy-bundle/2025.07-gfbf-2025b

What is Python-bundle-PyPI?

Bundle of Python packages from PyPI

• Homepage: https://python.org/

• Type ml help Python-bundle-PyPI/[version] on Pelle to see an output.

  Among others:

  • chardet

  • future

  • Jinja

  • pkginfo

  • psutil

  • Pygments

  • pydevtool

  • pytest

  • pytz

  • regex

  • Sphinx

  • threadpoolctl

  • toml

  • urllib

Also loads the package module

- virtualenv

Principles with example: matplotlib

• Decide what you need!

  1. Start new project with newest toolchain

  2. Go for version you have used before (reproduce)

  3. Exact versions of many packages may need an isolated environment.

• Load one or several bundles, python is loaded on the fly!

• Check versions

ml spider matplotlib

This is a very good way to find packages that are not in a bundle with the same name. For instance, pandas and numpy are parts of the SciPy-bundle

or

ml avail matplotlib

• Load prerequisites, if needed, and then

ml matplotlib/<version>

or

ml matplotlib/<version>

• Start Python session in a console with

python

• Load a needed library, like

import matplotlib