Source code for nvflare.app_opt.sklearn.recipes.fedavg

# Copyright (c) 2025, 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 Optional

from nvflare.apis.dxo import DataKind
from nvflare.app_common.abstract.aggregator import Aggregator
from nvflare.app_opt.sklearn.joblib_model_param_persistor import JoblibModelParamPersistor, validate_model_path
from nvflare.client.config import ExchangeFormat, TransferType
from nvflare.fuel.utils.constants import FrameworkType
from nvflare.recipe.fedavg import FedAvgRecipe as UnifiedFedAvgRecipe


[docs] class SklearnFedAvgRecipe(UnifiedFedAvgRecipe): """A recipe for implementing Federated Averaging (FedAvg) with Scikit-learn. Recipe parameters, including ``train_args`` and nested ``per_site_config`` values, must never contain actual secrets. Read secrets from site environment variables or mounted files; references are supported only where documented in :mod:`nvflare.recipe.secrets`. This recipe sets up a complete federated learning workflow with memory-efficient InTime aggregation, specifically designed for scikit-learn models. The recipe configures: - A federated job with initial parameters - FedAvg controller with InTime aggregation for memory efficiency - Optional early stopping and model selection - Script runners for client-side training execution Args: name: Name of the federated learning job. Defaults to "sklearn_fedavg". min_clients: Minimum number of clients required to start a training round. num_rounds: Number of federated training rounds to execute. Defaults to 2. model_params: Model hyperparameters as a dictionary. For SGDClassifier, can include: n_classes, learning_rate, eta0, loss, penalty, fit_intercept, etc. model_path: Optional absolute path to a saved model file (.joblib, .pkl). If provided, the model is loaded from this path at runtime (file must exist). Takes precedence over model_params when loading. train_script: Path to the training script that will be executed on each client. train_args: Command line arguments to pass to the training script. aggregator: Custom aggregator for combining client updates. If None, uses InTimeAccumulateWeightedAggregator with aggregator_data_kind. aggregator_data_kind: Data kind expected from client results. Clients that return differences must label the result with FLModel.params_type=ParamsType.DIFF. Defaults to DataKind.WEIGHTS. launch_external_process: Whether to launch the script in external process. Defaults to False. command: If launch_external_process=True, command to run script (prepended to script). Defaults to "python3 -u". per_site_config: Deprecated constructor form of per-site configuration. New code should call ``set_per_site_config(recipe, config)`` immediately after construction. Nested values become part of the generated job definition and must not contain secrets. key_metric: Metric used to determine if the model is globally best. If validation metrics are a dict, key_metric selects the metric used for global model selection. Higher values must indicate a better model; for lower-is-better metrics such as a loss, report a negated value from the client (e.g., "neg_loss"). Defaults to "accuracy". launch_once: Whether the external process will be launched only once at the beginning or on each task. Only used if `launch_external_process` is True. Defaults to True. shutdown_timeout: If provided, will wait for this number of seconds before shutdown. Only used if `launch_external_process` is True. Defaults to 0.0. Example: Basic usage with same config for all clients: ```python recipe = SklearnFedAvgRecipe( name="sklearn_linear", min_clients=5, num_rounds=50, model_params={ "n_classes": 2, "learning_rate": "constant", "eta0": 1e-4, "loss": "log_loss", "penalty": "l2", "fit_intercept": 1, }, train_script="client.py", train_args="--data_path /tmp/data/HIGGS.csv", ) from nvflare.recipe import SimEnv env = SimEnv(num_clients=5) run = recipe.execute(env) print("Result:", run.get_result()) ``` Per-site configuration: ```python from nvflare.app_opt.sklearn import SklearnFedAvgRecipe from nvflare.recipe import set_per_site_config recipe = SklearnFedAvgRecipe( name="sklearn_linear", min_clients=3, num_rounds=50, model_params={"n_classes": 2, "learning_rate": "constant", "eta0": 1e-4}, train_script="client.py", ) set_per_site_config( recipe, { "site-1": {"train_args": "--data_path /tmp/data/site1.csv"}, "site-2": {"train_args": "--data_path /tmp/data/site2.csv"}, "site-3": {"train_args": "--data_path /tmp/data/site3.csv"}, }, ) ``` Note: By default, this recipe implements the standard FedAvg algorithm where model updates are aggregated using weighted averaging based on the number of training samples provided by each client. If you want to use a custom aggregator, you can pass it in the aggregator parameter. The custom aggregator must be a subclass of the Aggregator class. """ def __init__( self, *, name: str = "sklearn_fedavg", min_clients: int, num_rounds: int = 2, model_params: Optional[dict] = None, model_path: Optional[str] = None, train_script: str, train_args: str = "", aggregator: Optional[Aggregator] = None, aggregator_data_kind: DataKind = DataKind.WEIGHTS, launch_external_process: bool = False, command: str = "python3 -u", per_site_config: Optional[dict[str, dict]] = None, key_metric: str = "accuracy", launch_once: bool = True, shutdown_timeout: float = 0.0, ): validate_model_path(model_path) # Create sklearn-specific persistor persistor = JoblibModelParamPersistor( initial_params=model_params or {}, model_path=model_path, ) # Call the unified FedAvgRecipe with sklearn-specific settings super().__init__( name=name, min_clients=min_clients, num_rounds=num_rounds, train_script=train_script, train_args=train_args, aggregator=aggregator, aggregator_data_kind=aggregator_data_kind, launch_external_process=launch_external_process, command=command, framework=FrameworkType.RAW, # sklearn uses RAW framework server_expected_format=ExchangeFormat.RAW, # sklearn uses RAW exchange format params_transfer_type=TransferType.FULL, model_persistor=persistor, # Pass sklearn-specific persistor per_site_config=per_site_config, key_metric=key_metric, launch_once=launch_once, shutdown_timeout=shutdown_timeout, )