API overview¶

In this page you can find the list of modules/submodules defined in fluke with their classes and functions.

Modules¶

fluke is organized in the following modules:

fluke: contains the core classes and utilities;
fluke.data: contains classes for data handling;
fluke.data.datasets: contains classes for datasets loading;
fluke.client: contains classes for client-side functionalities;
fluke.server: contains classes for server-side functionalities;
fluke.comm: contains classes for communication;
fluke.nets: contains classes for neural networks;
fluke.utils: contains utility classes and functions;
fluke.utils.log: contains classes for logging;
fluke.utils.model: contains classes for model manipulation;
fluke.evaluation: contains classes for evaluation;
fluke.algorithms: contains classes for federated learning algorithms.

`fluke`¶

Classes

`DDict`	A dictionary that can be accessed with dot notation recursively.
`FlukeCache`	A cache class that can store data on disk.
`FlukeCache._ObjectRef`	A reference to an object in the cache.
`FlukeCache._RefCounter`	A reference counter for an object in the cache.
`FlukeENV`	Environment class for the `fluke` framework.
`ObserverSubject`	Subject class for the observer pattern.
`Singleton`	This metaclass is used to create singleton classes.

`fluke.data`¶

Classes

`DataContainer`	Container for train and test data.
`DummyDataContainer`	DataContainer designed for those datasets with a fixed data assignments, e.g., FEMNIST, Shakespeare and FCUBE.
`FastDataLoader`	A DataLoader-like object for a set of tensors that can be much faster than TensorDataset + DataLoader because dataloader grabs individual indices of the dataset and calls cat (slow).
`DataSplitter`	Utility class for splitting the data across clients.

`fluke.data.datasets`¶

Classes

Datasets

Static class for loading datasets.

`fluke.client`¶

Classes

`Client`	Base `Client` class.
`PFLClient`	Personalized Federated Learning client.

`fluke.server`¶

Classes

Server

Basic Server for Federated Learning.

`fluke.comm`¶

Classes

`Message`	This class represents a message that can be exchanged between clients and the server.
`Channel`	A bi-directional communication channel.
`ChannelObserver`	Channel observer interface for the Observer pattern.

`fluke.nets`¶

Classes

`EncoderHeadNet`	Encoder (aka backbone) + Head Network [Base Class] This type of networks are defined as two subnetworks, where one is meant to be the encoder/backbone network that learns a latent representation of the input, and the head network that is the classifier part of the model.
`GlobalLocalNet`	Global-Local Network (Abstract Class).
`HeadGlobalEncoderLocalNet`	This implementation of the Global-Local Network (`GlobalLocalNet`) is meant to be used with the Encoder-Head architecture.
`EncoderGlobalHeadLocalNet`	This implementation of the Global-Local Network (`GlobalLocalNet`) is meant to be used with the Encoder-Head architecture.
`MNIST_2NN`	Multi-layer Perceptron for MNIST.
`MNIST_CNN`	Convolutional Neural Network for MNIST.
`FedBN_CNN`	Convolutional Neural Network with Batch Normalization for CIFAR-10.
`MNIST_LR`	Logistic Regression for MNIST.
`CifarConv2`	Convolutional Neural Network for CIFAR-10.
`ResNet9`	ResNet-9 network for CIFAR-100 classification.
`FEMNIST_CNN`	Convolutional Neural Network for FEMNIST.
`VGG9`	VGG-9 network for FEMNIST classification.
`FedAVGCNN`	Convolutional Neural Network for CIFAR-10.
`ResNet18`	ResNet-18 network as defined in the torchvision library.
`ResNet34`	ResNet-34 network as defined in the torchvision library.
`ResNet50`	ResNet-50 network as defined in the torchvision library.
`ResNet18GN`	ResNet-18 network as defined in the torchvision library but with Group Normalization layers instead of Batch Normalization.
`MoonCNN`	Convolutional Neural Network for CIFAR-10.
`LeNet5`	LeNet-5 for CIFAR.
`Shakespeare_LSTM`	LSTM for Shakespeare.

`fluke.utils`¶

Classes

`OptimizerConfigurator`	This class is used to configure the optimizer and the learning rate scheduler.
`Configuration`	Fluke configuration class.
`ClientObserver`	Client observer interface.
`ServerObserver`	Server observer interface.

Functions

`bytes2human`	Convert bytes to human-readable format.
`cache_obj`	Move the object from the RAM to the disk cache to free up memory.
`clear_cuda_cache`	Clear the CUDA cache.
`get_class_from_str`	Get a class from its name.
`get_class_from_qualified_name`	Get a class from its fully qualified name.
`get_full_classname`	Get the fully qualified name of a class.
`get_loss`	Get a loss function from its name.
`get_model`	Get a model from its name.
`get_optimizer`	Get an optimizer from its name.
`get_scheduler`	Get a learning rate scheduler from its name.
`flatten_dict`	Flatten a nested dictionary.
`import_module_from_str`	Import a module from its name.
`memory_usage`	Get the memory usage of the current process.
`plot_distribution`	Plot the distribution of classes for each client.
`retrieve_obj`	Load an object from the cache (disk).

`fluke.utils.log`¶

Classes

`Log`	Basic logger.
`TensorboardLog`	TensorBoard logger.
`WandBLog`	Weights and Biases logger.
`ClearMLLog`	ClearML logger.

Functions

get_logger

Get a logger from its name.

`fluke.utils.model`¶

Classes

`AllLayerOutputModel`	Wrapper class to get the output of all layers in a model.
`ModOpt`	Dataclass to store the model, its associated optimizer and scheduler.
`MMMixin`	Mixin class for model interpolation.
`LinesLinear`	Linear layer with gloabl and local weights.
`LinesConv2d`	Conv2d layer with gloabl and local weights.
`LinesLSTM`	LSTM layer with gloabl and local weights.
`LinesEmbedding`	Embedding layer with gloabl and local weights.
`LinesBN2d`	BatchNorm2d layer with gloabl and local weights.

Functions

`aggregate_models`	Aggregate the models using a weighted average.
`batch_norm_to_group_norm`	Iterates over a whole model (or layer of a model) and replaces every batch norm 2D with a group norm
`check_model_fit_mem`	Check if the models fit in the memory of the device.
`diff_model`	Compute the difference between two model state dictionaries.
`flatten_parameters`	Returns the model parameters as a contiguous tensor.
`get_activation_size`	Get the size of the activations of the model.
`get_global_model_dict`	Get the global model state dictionary.
`get_local_model_dict`	Get the local model state dictionary.
`get_output_shape`	Get the output shape of a model given the shape of the input.
`get_trainable_keys`	Get the keys of the model parameters that are trainable (i.e., require gradients).
`merge_models`	Merge two models using a linear interpolation.
`mix_networks`	Mix two networks using a linear interpolation.
`set_lambda_model`	Set model interpolation constant.
`safe_load_state_dict`	Load a state dictionary into a model.
`state_dict_zero_like`	Create a state dictionary with the same keys as the input state dictionary but with zeros tensors.

`fluke.evaluation`¶

Classes

`Evaluator`	This class is the base class for all evaluators in `fluke`.
`ClassificationEval`	Evaluate a PyTorch model for classification.

`fluke.algorithms`¶

Classes

`CentralizedFL`	Centralized Federated Learning algorithm.
`PersonalizedFL`	Personalized Federated Learning algorithm.

Submodules

`apfl`	Implementation of the APFL [APFL20] algorithm.
`ccvr`	Implementation of the CCVR [CCVR21] algorithm.
`ditto`	Implementation of the DITTO [Ditto21] algorithm.
`dpfedavg`	Implementation of the DPFedAVG: Differential Privacy Federated Averaging [DPFedAVG2017] algorithm.
`fat`	Implementation of the FAT [FAT20] algorithm.
`fedala`	Implementation of the FedALA [FedALA23] algorithm.
`fedamp`	Implementation of the FedAMP [FedAMP21] algorithm.
`fedavg`	Implementation of the Federated Averaging [FedAVG17] algorithm.
`fedavgm`	Implementation of the Federated Averaging with momentum [FedAVGM19] algorithm.
`fedaws`	Implementation of the Federated Averaging with Spreadout [FedAwS20] algorithm.
`fedbabu`	Implementation of the Federated Averaging with Body Aggregation and Body Update [FedBABU22] algorithm.
`fedbn`	Implementation of the FedBN [FedBN21] algorithm.
`feddyn`	Implementation of the FedDyn [FedDyn21] algorithm.
`fedexp`	Implementation of the FedExP [FedExP23] algorithm.
`fedhp`	Implementation of the FedHP: Federated Learning with Hyperspherical Prototypical Regularization [FedHP24] algorithm.
`fedlc`	Implementation of the [FedLC22] algorithm.
`fedld`	Implementation of the FedLD [FedLD24] algorithm.
`fednh`	Implementation of the [FedNH23] algorithm.
`fednova`	Implementation of the [FedNova21] algorithm.
`fedopt`	Implementation of the [FedOpt21] algorithm.
`fedper`	Implementation of the [FedPer19] algorithm.
`fedproto`	Implementation of the FedProto [FedProto22] algorithm.
`fedprox`	Implementation of the [FedProx18] algorithm.
`fedrep`	Implementation of the [FedRep21] algorithm.
`fedrod`	Implementation of the FedROD [FedROD22] algorithm.
`fedrs`	Implementation of the [FedRS21] algorithm.
`fedsam`	Implementation of the [FedSAM22] algorithm.
`fedsgd`	Implementation of the [FedSGD17] algorithm.
`gear`	Implementation of the GEAR [GEAR22] algorithm.
`lg_fedavg`	Implementation of the [LG-FedAVG20] algorithm.
`moon`	Implementation of the [Moon21] algorithm.
`per_fedavg`	Implementation of the [Per-FedAVG20] algorithm.
`pfedme`	Implementation of the [pFedMe20] algorithm.
`scaffold`	Implementation of the [SCAFFOLD20] algorithm.
`superfed`	Implementation of the [SuPerFed22] algorithm.