plotter

SAFE-PFL plotter example

What safe-pfl-plotter does

safe-pfl-plotter is a pip package that helps developers / researchers to plot the SAFE-PFL log file.

To submit any issues / pull requests please visit the safe-pfl-plotter repository.

Quick Start

To install safe-pfl-plotter package, run

pip install safe-pfl-plotter --user

Then clone this repository via

git clone https://github.com/safe-pfl/examples

go to the plotter directory anf put the *.log file from the safe-pfl project in it.

Important

• If you ran safe-pfl example on different dates then the log file is different respective to the its running date. The code with find the number of clients with regex so please DO NOT change the log file and you DO NOT need to pass the number of contributing nodes / clients in Federated Learning environment.

Then run the logs_plotter.py in the plotter directory with the log file you want to plot each clients accuracies.

cd plotter # Go to the plotter directory
python ./logs_plotter.py --log_path *.log --plot_result_path ./results/plot # don't add .png extension

the output of the plotter/logs_plotter.py will store in plotter/results/plot.png path.

Note

Other input parameters are --all_test_accuracy_mean which is a boolean, if set to true this wil show the mean of clients / nodes test accuracies. The --all_test_accuracy_std also show the standard deviations of clients / nodes test accuracies id set to true. the --distinct_colors will plot the chart with distinguishable colors if set to true.