get-started-on-windows

Tutorial for getting started with FEDn in Windows

Prerequisites: python environment and install FEDn

1. Go to python's official page https://www.python.org/ftp/python/3.12.6/python-3.12.6-amd64.exe to download the latest release for windows
2. Click on the downloaded executable installer file to start installation
3. Open Windows PowerShell, and test your new python install by running:

python --version

if this does not work or if it shows you a earlier version than <3.12.6, you need to add the python install folder into you PATH environment variable: REPLACE

$env:Path += ';C:\Users\<Username>\AppData\Local\Programs\Python\Python312\'

Try again:

python --version

4. Next, make sure the python package manager "pip" is installed and is pointing to the correct python version (3.12):

pip --version

5. Create a python virtual environment (to avoid changing system wide python dependencies):

python -m venv .venv

A new folder ".venv" should have been created. Activate the new venv:

.venv/Scripts/Activate.ps1

OBS! If you get a permission denied error, it means your user don't have the right permissions to run scripts in the PowerShell environment. Either try to run PowerShell as Admin or in CMD.exe run:

.venv/Scripts/activate.bat

You know that the venv is activated when is says (.venv) in the beginning of the command line. 6. Install Scaleout software FEDn by running:

pip install fedn

7. Test fedn:

fedn --version

During Workshop/hands-on session

1. Sign-up at https://fedn.scaleoutsystems.com
2. In the "Project" for the workshop go to "clients" in the sidebar and click "Connect client". Click the "Download" button. A "client.yaml" file should be located in your downloads folder.
3. Make sure you have the venv activated (see step 5)
4. Make a new folder:

mkdir demo

cd demo

5. Copy or move the downloaded "client.yaml" into the "demo" folder.
6. Start the fedn client:

fedn client start -in client.yaml --secure=True --force-ssl

Tutorial: FEDn package

1. Clone the fedn repository:

git clone https://github.com/scaleoutsystems/fedn.git

2. Move into the mnist-pytorch example:

cd fedn\examples\mnist-pytorch

3. Run the "build" entrypoint of the project:

fedn run build -p client --keep-venv

This command will create and compile the pytoch model needed as the initial model/seed of the federated workflow.

4. Next lets run the startup entrypoint, this entrypoint always runs at client startup:

fedn run startup -p client

This will fetch the training data needed for the client in this example.

5. Finally lets test the train entrypoint:

fedn run train -p -i seed.npz -o updated_model.npz

This command will test the train entrypoint why passing the seed model as input and updated_model.npz as the new updated (trained) model.