org.jetbrains.kotlinx.dl.api.inference.InferenceModel.kt Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of kotlin-deeplearning-api Show documentation

Deep Learning library for Kotlin.

There is a newer version: 0.6.0-alpha-1

Show newest version

/*
 * Copyright 2020 JetBrains s.r.o. and Kotlin Deep Learning project contributors. All Rights Reserved.
 * Use of this source code is governed by the Apache 2.0 license that can be found in the LICENSE.txt file.
 */

package org.jetbrains.kotlinx.dl.api.inference

import org.jetbrains.kotlinx.dl.api.core.metric.Metrics
import org.jetbrains.kotlinx.dl.dataset.Dataset

/**
 * The basic class for all models which defines the basic interfaces required for inference tasks only.
 */
public abstract class InferenceModel : AutoCloseable {
    /** Input specification for this model. */
    public abstract val inputDimensions: LongArray

    /** Model name. */
    public var name: String? = null

    /**
     * Predicts the class of [inputData].
     *
     * @param [inputData] The single example with unknown label.
     * @return Predicted class index.
     */
    public abstract fun predict(inputData: FloatArray): Int

    /**
     * Predicts vector of probabilities instead of specific class in [predict] method.
     *
     * @param [inputData] The single example with unknown vector of probabilities.
     * @param [predictionTensorName] The name of prediction tensor. It could be changed, if you need to get alternative outputs from model graph.
     * @return Vector that represents the probability distributions of a list of potential outcomes
     */
    public abstract fun predictSoftly(inputData: FloatArray, predictionTensorName: String = ""): FloatArray

    /**
     * Chain-like setter to set up input shape.
     *
     * @param [dims] The input shape.
     */
    public abstract fun reshape(vararg dims: Long)

    /**
     * Creates a copy.
     *
     * @param [copiedModelName] Set up this name to make a copy with a new name.
     * @return A copied inference model.
     */
    public abstract fun copy(
        copiedModelName: String? = null,
        saveOptimizerState: Boolean = false,
        copyWeights: Boolean = true
    ): TensorFlowInferenceModel


    /**
     * Predicts labels for all observation in [dataset].
     *
     * NOTE: Slow method.
     *
     * @param [dataset] Dataset.
     */
    public fun predict(dataset: Dataset): List {
        val predictedLabels: MutableList = mutableListOf()

        for (i in 0 until dataset.xSize()) {
            val predictedLabel = predict(dataset.getX(i))
            predictedLabels.add(predictedLabel)
        }

        return predictedLabels
    }

    /**
     * Evaluates [dataset] via [metric].
     *
     * NOTE: Slow method.
     */
    public fun evaluate(
        dataset: Dataset,
        metric: Metrics
    ): Double {

        return if (metric == Metrics.ACCURACY) {
            var counter = 0
            for (i in 0 until dataset.xSize()) {
                val predictedLabel = predict(dataset.getX(i))
                if (predictedLabel == dataset.getY(i).toInt())
                    counter++
            }

            (counter.toDouble() / dataset.xSize())
        } else {
            Double.NaN
        }
    }
}