v0.4.0

Fixed issues

• Fixed FedNH, FedHP and FedProto
• Fixed issues on Dirichlet data distribution when balanced=True
• Removed the required output_size from the networks

New features

• Added get_activation_size in utils
• Added dataset FCUBE from https://arxiv.org/pdf/2102.02079

v0.3.4

Fixed issues

• Fixed problem with the cuda:N setting. Now it works properly and selects the correct device.
• Fixed warning related to the redundancy of wandb.require("core") in the last versions of the wand library

Known issues with torch

If you get the error AttributeError: module 'torch.serialization' has no attribute 'add_safe_globals' please upgrade your torch installation to the latest version.

Important note

Release 0.3.3 has been published and deleted right after due to an error!

v0.3.2

Fixed issues

• Fixed problem with the validation of the eval configuration key
• Fixed minor issue with AllLayerOutputModel class

v0.3.1

Added Kafé as a new federated algorithm.

KAFÉ: Pian Qi, Diletta Chiaro, Fabio Giampaolo, and Francesco Piccialli.
KAFÈ: Kernel Aggregation for FEderated. In ECML-PKDD (2024).
URL: https://link.springer.com/content/pdf/10.1007/978-3-031-70359-1_4.pdf

v0.3.0

[v0.3.0] release (2024-09-09)

Changelog

General

• Federated algorithms can now be saved and resumed through the CLI;
• Clients-only and centralized runs can now be logged into W&B, ClearML and Tensorboard;
• Experiment configuration file now includes a section dedicated to the evaluation configuration;
• FastDataLoader now supports on-the-fly transformations;
• Fixed some issues with the DataSplitter;
• Added some new utility functions in fluke.utils and fluke.utils.model.

Client

• The client logic now is enclosed in the method local_update which comprehend the steps receive_model, fit and send_model. As a consequence, the method fit does not include the communication with the server;
• The Client class is now implementing the Observer pattern, thus it can be observed;
• The client-side evaluation now can happens in three different moments (not mutually exclusive): (i) pre-fit: before the local fit, (ii) post-fit: after the local fit (as it was in the previous release), and (iii) on the server's test set (named 'locals').

Logging

• We added support for Tensorboard and ClearML;
• We improved the logging in general.

Evaluation

• The evaluator is now independent from the client and the server. It must be set in the GlobalSettings and this evaluator is used by both clients and servers.

Algorithms

• We thoroughly tested many of the already included algorithms;
• We added FedALA, FedRod, FedSAM, FedRS, and FedAwS.

v0.1.0

[v0.1.0] First official release (2024-06-03)

Previous PiPy unofficial release [v0.0.5]

Bug Fixes

• Fixed bug related to fluke.data.dataset.Datasets.get
• Fixed issue with fluke.algorithms.fedlc.FedLC.repr
• Fixed minor errors in the documentation
• Fixed erroneous link in the tutorials section of the documentation

Other Changes

• First (almost complete) draft of the API documentation