Client

import com.mindspore.flclient.model.Client

Client defines the execution process object of the end-side federated learning algorithm.

Public Member Functions

initCallbacks

public abstract List<Callback> initCallbacks(RunType runType, DataSet dataSet)

Initialize the callback list.

• Parameters

  • runType: Define run phase.
  • dataSet: DataSet.

• Returns

  The initialized callback list.

initDataSets

public abstract Map<RunType, Integer> initDataSets(Map<RunType, List<String>> files)

Initialize dataset list.

• Parameters

  • files: Data files.

• Returns

  Data counts in different run type.

getEvalAccuracy

public abstract float getEvalAccuracy(List<Callback> evalCallbacks)

Get eval model accuracy.

• Parameters

  • evalCallbacks: Callback used in eval phase.

• Returns

  The accuracy in eval phase.

getInferResult

public abstract List<Object> getInferResult(List<Callback> inferCallbacks)

Get infer phase result.

• Parameters

  • inferCallbacks: Callback used in infer phase.

• Returns

  predict results.

trainModel

public Status trainModel(int epochs)

Execute train model process.

• Parameters

  • epochs: Epoch num used in train process.

• Returns

  Whether the train model is successful.

evalModel

public float evalModel()

Execute eval model process.

• Returns

  The accuracy in eval process.

genUnsupervisedEvalData

public Map<String, float[]> genUnsupervisedEvalData(List<Callback> evalCallbacks)

Generate unsupervised training evaluation data, and the subclass needs to rewrite this function.

• Parameters

  • evalCallbacks: the eval Callback that generates data.

• Returns

  unsupervised training evaluation data

inferModel

public List<Object> inferModel()

Execute model prediction process.

• Returns

  The prediction result.

setLearningRate

public Status setLearningRate(float lr)

Set learning rate.

• Parameters

  • lr: Learning rate.

• Returns

  Whether the set is successful.

setBatchSize

public void setBatchSize(int batchSize)

Set batch size.

• Parameters

  • batchSize: batch size.