src.core.worker module

src.core.worker.worker(w_id: int, server: ParameterServer, model: Callable[[int], Module], input_dim: int, dataset_builder: Callable[[int, int, int], Tuple[DataLoader, int]], param: ConfigParameters, start_evt) → None

Worker function for Stale Synchronous Parallel training.

Parameters:

• w_id (int) – Worker ID.

• server (ParameterServer) – Parameter server

• model (Callable[[int], nn.Module]) – Model class to be trained.

• input_dim (int) – Input dimension of the model.

• dataset_builder (Callable[[int,int,int], Tuple[torch.utils.data.DataLoader,int]]+) – Function used to build the dataset.

• param (ConfigParameters) – SSP Configuration parameters.

Returns:

None