federer/dist/src/experiments/mnist/coordinator/preprocess.d.ts

Experiments in asynchronous federated learning and decentralized learning

import { Rank } from "@tensorflow/tfjs-node";
import { Dataset, DataSubset } from "../../../common";
import { PreprocessResult } from "../../../coordinator";
import { Environment } from "../../../coordinator/cli";
import { ShardingOptions } from "../coordinator/MnistCoordinator";
import { MnistModelName } from "./model";
import { DatasetName } from "./data/download";
/**
 * Raw MNIST data, as downloaded from the Internet. Images have rank 4 (number
 * of images, width, height, channels), and labels are scalars in the range 0-9,
 * so the list of labels has rank 1.
 */
export declare type RawMnistDataset = Dataset<Rank.R4, Rank.R1>;
/**
 * MNIST data after preprocessing by {@link preprocess}. The images have rank 2
 * if the model is 2NN (list of vectors of flat images), and rank 4 if the model
 * is CNN (number of images, width, height channels). The labels are encoded as
 * one-hot vectors, so have rank 2.
 */
export declare type ProcessedMnistDataset = Dataset<Rank.R2 | Rank.R4, Rank.R2>;
export declare type ProcessedMnistDataSubset = DataSubset<Rank.R4 | Rank.R2, Rank.R2>;
/**
 * Interface specializing {@link PreprocessResult} with a more specific type
 * describing the ranks of the test set.
 */
export interface MnistPreprocessResult extends PreprocessResult {
    testSet: ProcessedMnistDataSubset;
}
export interface MnistPreprocessOptions {
    /** Name of the raw dataset to use. */
    dataset: DatasetName;
    /**
     * Name of the model; preprocessing might vary slightly depending on the model
     */
    modelName: MnistModelName;
    /**
     * Number of label classes to include. Include fewer than 10 to run smaller
     * test experiments.
     */
    numberLabelClasses: number;
    /**
     * Number of clients. Tells us into how many shards to break up the data.
     * @see {@link MnistCoordinatorOptions}
     */
    numberClients: number;
    /** Environment. Dictates where to read data files from. */
    environment: Environment;
    /**
     * Number of digit batches per client. A digit batch is a slice of data,
     * which mostly includes the same digit.
     * @see {@link MnistCoordinatorOptions}
     */
    numberDigitBatchesPerClient: number;
    /**
     * Whether to allow reads from cache (`true`), or to force re-computation of
     * the preprocess pipeline (`false`).
     */
    allowReadFromCache?: boolean;
    shardingOptions: ShardingOptions;
}
/**
 * Preprocess the raw MNIST data into client shards and a test set.
 *
 * @param options Preprocessing options
 * @returns The results of the preprocessing, with all data saved to disk
 */
export declare function preprocess(options: MnistPreprocessOptions): Promise<MnistPreprocessResult>;
//# sourceMappingURL=preprocess.d.ts.map