Needed packages for the exercises

[bclaremar]

Please inform here what is needed and, if possible, the different demands for different systems. @bclaremar will gather the packages for the users to install as virtual environments on their cluster

[bbrydsoe]

I am using: Python/3.11.8 on UPPMAX Python/3.11.3 on HPC2N Python/3.11.5 on LUNARC Python/3.10.5 on NSC

I have found working packages with these Python versions for: pandas, matplotlib, numpy, scipy (many are in SciPy-bundle). I also have working tkinter for this. These goes together on NSC: buildtool-easybuild/4.8.0-hpce082752a2 GCC/11.3.0 OpenMPI/4.1.4 Python/3.10.4 SciPy-bundle/2022.05 matplotlib/3.5.2 Tkinter/3.10.4

I need: numba on NSC, so maybe a venv could be created for the above Python version?

[bbrydsoe]

These are the options on NSC:

Python 3.11.5 has pandas, numpy, scipy, jupyterlab To load : ml buildtool-easybuild/4.8.0-hpce082752a2 GCC/13.2.0 Python/3.11.5 SciPy-bundle/2023.11 JupyterLab/4.2.0

or

Python 3.10.4 has pandas, numpy, scipy, matplotlib To load : ml buildtool-easybuild/4.8.0-hpce082752a2 GCC/11.3.0 Python/3.10.4 OpenMPI/4.1.4 SciPy-bundle/2022.05 matplotlib/3.5.2 Tkinter/3.10.4

None has scikit-learn, pytorch, tensorflow, seaborn, numba

[bclaremar]

UPPMAX based on module 3.11.8 and ML we need these in the venv

• spyder
• scikit-build-core cmake
• lightgbm
• mpi4py
• Julia
• xarray

[bclaremar]

UPPMAX requirements file is done continuing with NSC now

[jayant-yadav]

I am using : UPPMAX : Python/3.11.8 HPC2N : Python/3.11.3 LUNARC : Python/3.11.5 NSC : Python/3.11.5

For all 4 centres we need 2 separate envs for tf and torch each. python3 -m pip install torch torchvision python3 -m pip install 'tensorflow[and-cuda]' This is because they contain conflicting cuda toolkits. PS: they download their own numpy. I have checked sklearn yet.

[bclaremar]

let spyder out for ML/dask