Skip to content

Installaton of MNE Python

Purpose

Demonstrating the steps for installing and running MNE Python on ARC

Prerequisites

  • A SSO (Single Sign-On) account at the University of Oxford
  • VPN access to the University of Oxford network (required for off-site access)
  • An account on the project psyc-neuosc at ARC
  • A project on RFS drive

Steps

Cloning github to ARC $HOME directory

Start a terminal in ARC under nx.arc.ox.ac.uk:

Make a directory under $HOME that for where you want the local repository and move there. 

mkdir $DATA/Python/FLUX
cd $DATA/Python/FLUX

Log into your github.com account and find the repository do clone; e.g.

git clone https://github.come/FLUX-pipeline/Cerca

Clone only needs to be done once. To later pull latest changes from central to local repository 

git pull https://github.come/FLUX-pipeline/Cerca

Changes can now be made to e.g. FILE.PY

To add latest changes from local to central repository 

git add FILE.PY
git commit -m "DESCRIPTION OF CHANGES"
git push origin main

Running MNE Python in Jupyter Notebook on ARC

Before you can install or run the environment you must switch to a compute node:

srun --mem=16G --time=02:00:00 --pty bash

Note: Using srun you submit an interactive job to the compute node. This is equivalent to sbatch, which you use to submit jobs to the compute node that you cannot interact with.

If you are on a gateway node (via remote desktop or VPN) you first need to log in to a login node before you can submit the interactive job above:

ssh htc-login

Installation of MNE Python and MNE BIDS

Note, that you MUST store everything (including environments or cache files) on the data storage ($DATA) and not your home folder. $HOME has only 5GB of storage. In order to make conda store everything in the data folder and not home, we need to make some modifications to the current environment.

First, create and specify the new conda cache directories outside $HOME:

rm -f ~/.condarc # Yes, this one still lives in $HOME

# Create directories (change to your liking)
ENV_DIR="${DATA}/environments"
mkdir -p "${ENV_DIR}"

mkdir -p "${ENV_DIR}/.conda"
mkdir -p "${ENV_DIR}/.conda/pkgs"
mkdir -p "${ENV_DIR}/.conda/envs"

cat >> ~/.condarc <<EOF
pkgs_dirs:
  - $ENV_DIR/.conda/pkgs

envs_dirs:
  - $ENV_DIR/.conda/envs
EOF

Now you are set for createing a new environment.

First time:

module load Anaconda3/2024.02-1 
conda create -p $DATA/mne_env python=3.11 mne jupyter jupyterlab -c conda-forge

After installation activate

source activate mne_env

First time you need to nstall mne-bids:

conda install --channel conda-forge mne-bids

And activate again

source activate mne_env

The final step is link the MNE environtment to Jupyter:

python -m ipykernel install --user --name=mne --display-name "Python (MNE)"

Jupyter

Now start Jupyher and change to the MNE Kernel:

  1. Launch Jupyter Notebook (go to '>' in lower left corner of the NoMachine interface and find Jupyter under applications (use the command /apps/common/arc-app-startup/jupyter/Anaconda3-2022.05).
  2. Open your desired notebook (e.g. the FLUX scripts).
  3. Go to the top menu and select Kernel 4.Change kernel.Choose "Python (MNE)" from the list.

This will allow you to run the FLUX pipeline

The ARC user guide