blazefl.core

Core module of BlazeFL framework.

This module imports and defines the core components of the BlazeFL framework, including client trainers, model selectors, partitioned datasets, and server handlers.

Functions

serialize_model(model[, cpu])	Serialize a PyTorch model's parameters into a flat tensor.
deserialize_model(model, serialized_parameters)	Deserialize a flat tensor back into a PyTorch model's parameters.
process_tensors_in_object(obj, mode[, max_depth])	Recursively traverses an object to process torch.Tensor instances.
reconstruct_from_shared_memory(handle_obj, ...)	Recursively reconstructs an object from a handle-based object and a shared memory buffer object.

Classes

BaseClientTrainer(*args, **kwargs)	Abstract base class for serial client training in federated learning.
FilteredDataset(indices, original_data[, ...])	A dataset wrapper that filters and transforms a subset of the original dataset.
ProcessPoolClientTrainer(*args, **kwargs)	Abstract base class for parallel client training in federated learning.
ThreadPoolClientTrainer(*args, **kwargs)
ModelSelector(*args, **kwargs)	Abstract base class for selecting models in federated learning.
PartitionedDataset(*args, **kwargs)	Abstract base class for partitioned datasets in federated learning.
BaseServerHandler(*args, **kwargs)	Abstract base class for server-side operations in federated learning.
SHMHandle()	A lightweight, serializable handle to a tensor stored in shared memory.