Fed-BioMed software installation

This tutorial gives steps for installing Fed-BioMed components (node, researcher) on a single machine. Deployment documentation explains other available setups.

Hardware requirements

  • 16GB RAM recommended, 8GB RAM minimum for handling PyTorch & ML packages

System requirements

Fed-BioMed is developed and tested under up to date version of :

  • Linux Ubuntu and Fedora, should also work or be easily ported under most Linux distributions
  • MacOS

Check specific guidelines for installation on Windows 11.

Software packages

The following packages are required for Fed-BioMed :

Docker is not required anymore

Recent versions of Fed-BioMed don't require docker anymore for simple installation such as described here. This is now only needed for advanced usage scenarios with additional VPN protection of Fed-BioMed communications.

Install git

Linux Fedora

Install git :

$ sudo dnf install -y git

MacOS

Install git choosing one of the available options for example your favorite third party package manager:

  • macports provides git port
  • homebrew provides git formula

Other

Check platform specific instructions.

Install conda

Linux Fedora

Simply install the package :

$ sudo dnf install conda

Check conda is properly initialized with the following command that should answer the default (base) environment:

$ conda env list

Other

Install conda package manager.

During the installation process, let the conda installer initialize conda (answer "yes" to “Do you wish the installer to initialize Anaconda3 by running conda init ?”)

Check conda is properly initialized with the following command that should answer the default (base) environment:

$ conda env list

Fed-BioMed software

Download Fed-BioMed software by cloning the git repository :

$ git clone -b master https://github.com/fedbiomed/fedbiomed.git
$ cd fedbiomed

In the following tutorials, Fed-BioMed commands use a path relative to the base directory of the clone, noted as ${FEDBIOMED_DIR}. This is not required for Fed-BioMed to work but enables you to run the tutorials more easily.

The way to setup this directory depends on your operating system (Linux, macOSX, Windows), and on your SHELL.

As an example, on Linux/macOSX with bash/zsh it could be done as:

export FEDBIOMED_DIR=$(pwd)

or

export FEDBIOMED_DIR=${HOME}/where/is/fedbiomed

Remember, that this environment variable must be initialized (to the same value) for all running shells (you may want to declare it in your shell initialization file).

Fed-BioMed is provided under Apache 2.0 License.

We don't provide yet a packaged version of Fed-BioMed (conda, pip).

Conda environments

Fed-BioMed uses conda environments for managing package dependencies.

Create or update the conda environments with :

$ ${FEDBIOMED_DIR}/scripts/configure_conda

List the existing conda environments and check the 3 environments fedbiomed-gui fedbiomed-node fedbiomed-researcher were created :

$ conda env list
[...]
fedbiomed-gui            /home/mylogin/.conda/envs/fedbiomed-gui
fedbiomed-node           /home/mylogin/.conda/envs/fedbiomed-node
fedbiomed-researcher     /home/mylogin/.conda/envs/fedbiomed-researcher
[...]

Conda environment for Fed-BioMed Node GUI

Fed-BioMed comes with a user interface that allows data owners (node users) to deploy datasets and manage requested training plans easily. The fedbiomed-gui environment is not going to be used unless the Node GUI is launched. If you don't plan on using the GUI, you can install only the fedbiomed-node and fedbiomed-researcher environments.

${FEDBIOMED_DIR}/scripts/configure_conda node researcher

Please see Node GUI user guide to get more information about launching GUI on your local machine.

The Next Step

After the steps above are completed you will be ready to start Fed-BioMed components. In the following tutorial you will learn how to launch components and add data in Fed-BioMed to prepare an experiment.