High Performance Computing on the VIB Data Core Compute.

Set up environment

Log in to the compute cluster:

ssh -p 2022 user.name@compute.vib.be

Ask for an interactive session, e.g.:

salloc --partition=debug_28C_56T_750GB --ntasks=8 --mem=16G --time=02:00:00

In the default environment, install support for loading environment kernels (no modules should be loaded at this time, nor conda environments):

ml purge
ml Python
while [ ! -z $CONDA_PREFIX ]; do conda deactivate; done

pip install environment_kernels

Load the Mamba module to create a conda env.

ml load Mamba

Next use environment_vib_compute.yml to build a conda environment:

mamba env create -f environment_vib_compute.yml

Activate the environment:

conda activate harpy

Install Harpy:

pip install 'harpy-analysis @ git+https://github.com/saeyslab/harpy.git'

Run the Harpy notebook as an interactive session.

Make an ipython kernel to use in a JupyterLab notebook. The displayname is what you will select in JupyterLab.

ipython kernel install --user --name harpy --display-name "harpy"

Now on https://compute.vib.be, start a JupyterLab on GPU (select Python 3.10); check the conda environment box; select Mamba as the system wide Conda Module; fill in harpy as the name of the Custom Conda Environment.

You should now be able to run the notebook Harpy_feature_calculation.ipynb in an interactive session on the VIB compute cluster.