nvflare.app_opt.xgboost.histogram_based_v2.runners.client_runner module

class XGBClientRunner(data_loader_id: str, early_stopping_rounds: int, xgb_params: dict, verbose_eval: bool, use_gpus: bool, model_file_name: str, metrics_writer_id: str | None = None)

Bases: XGBRunner, FLComponent

Constructor.

Parameters:

• early_stopping_rounds – early stopping rounds

• xgb_params – This dict is passed to xgboost.train() as the first argument params. It contains all the Booster parameters. Please refer to XGBoost documentation for details: https://xgboost.readthedocs.io/en/stable/parameter.html

• data_loader_id – the ID points to XGBDataLoader.

• verbose_eval – verbose_eval in xgboost.train

• use_gpus (bool) – A convenient flag to enable gpu training, if gpu device is specified in the xgb_params then this flag can be ignored.

• metrics_writer_id – the ID points to a LogWriter, if provided, a MetricsCallback will be added. Users can then use the receivers from nvflare.app_opt.tracking.

initialize(fl_ctx: FLContext)

Initializes the runner. This happens when the job is about to start.

Parameters:

fl_ctx – FL context

Returns:

None

is_stopped() → Tuple[bool, int]

Checks whether the runner is already stopped.

Returns:

A tuple of (whether the runner is stopped, exit code)

run(ctx: dict)

Runs XGB processing logic.

Parameters:

ctx – the contextual info to help the runner execution

Returns:

None

stop()

Stops the runner.

Returns:

None

xgb_train(params: XGBoostParams, train_data: DMatrix, val_data: DMatrix) → Booster

XGBoost training logic.

Parameters:

• params (XGBoostParams) – xgboost parameters.

• train_data (xgb.core.DMatrix) – training data.

• val_data (xgb.core.DMatrix) – validation data.

Returns:

A xgboost booster.

Note

Users can customize this method for their own training logic.

class XGBoostParams(xgb_params: dict, num_rounds: int = 10, early_stopping_rounds: int = 2, verbose_eval: bool = False)

Bases: object

Container for all XGBoost parameters.

Parameters:

xgb_params – This dict is passed to xgboost.train() as the first argument params. It contains all the Booster parameters. Please refer to XGBoost documentation for details: https://xgboost.readthedocs.io/en/stable/parameter.html