nvflare.app_common.abstract.model_processor module

class ModelProcessor

Bases: ABC

abstract apply_model(network, multi_processes: bool, model_params: dict, fl_ctx: FLContext, options=None)

Call to apply the model parameters to the training network.

Parameters:

• network – training network

• multi_processes – boolean to indicates if it’s a multi-processes

• model_params – model parameters to apply

• fl_ctx – FLContext

• options – optional information that can be used for this process

abstract extract_model(network, multi_processes: bool, model_vars: dict, fl_ctx: FLContext) → dict

Call to extract the current model from the training network.

Parameters:

• network – training network

• multi_processes – boolean to indicates if it’s a multi-processes

• model_vars – global model dict

• fl_ctx – FLContext

Returns:

a dictionary representing the model