Define a Dockerfile

[abcsds]

Idea as follows:

• Choose a a small linux docker image (chose python3.9)
• [.] Install mne via official installation medium (installed directly from pip instead)
• Install missing requirements for mn-bids-pipeline, if any (installed)
• Mount the bids_root and work folder as volumes. (mounting is done on running, but volumes are declared)
• Load config.py. Maybe through loading a workspace volume? (missing)
• Iun command: python run.py --config=./work/config.py(missing)

The black box model would suppose the config.py and the data volume as entry points. The output is writen to the bids_root mounted volume.

Here's an initial sugestion for a dockerfile:

FROM python:3.9

USER root

RUN apt-get -y update && apt-get install python3-pip -y
RUN groupadd -r user && useradd -r -g user mne-user

RUN mkdir -p /home/mne-user/work
RUN mkdir -p /home/mne-user/data
RUN chown mne-user /home/mne-user
RUN chown mne-user /home/mne-user/work
RUN chown mne-user /home/mne-user/data
VOLUME ["/home/mne-user/work", "/home/mne-user/data"]

USER mne-user
WORKDIR ~/work
RUN pip install mne
RUN pip install -r https://raw.githubusercontent.com/mne-tools/mne-bids-pipeline/main/requirements.txt
ENTRYPOINT python -c "import mne; mne.sys_info()" && /bin/bash

This is not the insallation suggested in the documentation, but I couldn't get conda working its environments while building docker images. This installation should be cleaner anyway. I chose python's container, which is 916 MB. After this installation the MNE container is 2.59GB. MNE's install is large anyway.

[welcome]

Hello! 👋 Thanks for opening your first issue here! ❤️ We will try to get back to you soon. 🚴🏽‍♂️

[hoechenberger]

Hello @abcsds, this is a good start, thanks!

• I'd suggest to use the official MNE installers instead, on a Debian or Ububtu base image.
• You forgot to actually fetch the pipeline repository :)

[abcsds]

Hi @hoechenberger,
Thanks for picking it up. I sent the PR.

[larsoner]

Closing for mne-tools/mne-docker#24