# Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import List, Union
from nvflare.app_common.app_constant import AppConstants
from nvflare.app_common.response_processors.global_weights_initializer import GlobalWeightsInitializer, WeightMethod
from .broadcast_and_process import BroadcastAndProcess
[docs]class InitializeGlobalWeights(BroadcastAndProcess):
def __init__(
self,
task_name: str = AppConstants.TASK_GET_WEIGHTS,
min_responses_required: int = 0,
wait_time_after_min_received: int = 0,
task_timeout: int = 0,
weights_prop_name=AppConstants.GLOBAL_MODEL,
weight_method: str = WeightMethod.FIRST,
weights_client_name: Union[str, List[str], None] = None,
):
"""A controller for initializing global model weights based on reported weights from clients.
Args:
task_name: name of the task to be sent to clients to collect their model weights
min_responses_required: min number of responses required. 0 means all clients.
wait_time_after_min_received: how long (secs) to wait after min responses are received
task_timeout: max amount of time to wait for the task to end. 0 means never time out.
weights_prop_name: name of the FL Context property to store the global weights
weight_method: method for determining global model weights. Defaults to `WeightMethod.FIRST`.
weights_client_name: name of the client if the method is "client". Defaults to None.
If `None`, the task will be sent to all clients (to be used with `weight_method=WeightMethod.FIRST`).
If list of client names, the task will be only be sent to the listed clients.
"""
if isinstance(weights_client_name, str):
clients = [weights_client_name]
elif isinstance(weights_client_name, list):
clients = weights_client_name
else:
clients = None
BroadcastAndProcess.__init__(
self,
processor=GlobalWeightsInitializer(
weights_prop_name=weights_prop_name, weight_method=weight_method, client_name=weights_client_name
),
task_name=task_name,
min_responses_required=min_responses_required,
wait_time_after_min_received=wait_time_after_min_received,
timeout=task_timeout,
clients=clients,
)