All Downloads are FREE. Search and download functionalities are using the official Maven repository.

ai.djl.metric.Metrics Maven / Gradle / Ivy

There is a newer version: 0.30.0
Show newest version
/*
 * Copyright 2019 Amazon.com, Inc. or its affiliates. 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. A copy of the License is located at
 *
 * http://aws.amazon.com/apache2.0/
 *
 * or in the "license" file accompanying this file. This file 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.
 */
package ai.djl.metric;

import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.function.BiConsumer;
import java.util.stream.Collectors;

/**
 * A collection of {@link Metric} objects organized by metric name.
 *
 * 

{@code Metric} is a utility class that is used in the {@link ai.djl.training.Trainer} and * {@link ai.djl.inference.Predictor} to capture performance and other metrics during runtime. * *

It is built as a collection of individual {@link Metric} classes. As a container for * individual metrics classes, {@code Metrics} stores them as time series data so that * metric-vs-timeline analysis can be performed. It also provides convenient statistical methods for * getting aggregated information, such as mean and percentile. The metrics is used to store key * performance indicators (KPIs) during inference and training runs. These KPIs include various * latencies, CPU and GPU memory consumption, losses, etc. * *

For more details about using the metrics, see the metrics * tutorial. */ public class Metrics { private Map> metrics; private int limit; private BiConsumer onLimit; /** Constructs an empty {@code Metrics} instance. */ public Metrics() { metrics = new ConcurrentHashMap<>(); } /** * Sets the max size for each metric. * * @param limit the max size for each metric */ public void setLimit(int limit) { this.limit = limit; } /** * Sets the callback function when hit the limit. * * @param onLimit the callback function */ public void setOnLimit(BiConsumer onLimit) { this.onLimit = onLimit; } /** * Adds a {@link Metric} to the collection. * * @param metric the {@link Metric} to be added */ public void addMetric(Metric metric) { List list = metrics.computeIfAbsent( metric.getMetricName(), v -> Collections.synchronizedList(new ArrayList<>())); if (limit > 0 && list.size() >= limit) { if (onLimit != null) { onLimit.accept(this, metric.getMetricName()); } list.clear(); } list.add(metric); } /** * Adds a {@code Metric} given the metric's {@code name} and {@code value}. * * @param name the metric name * @param value the metric value */ public void addMetric(String name, Number value) { addMetric(new Metric(name, value)); } /** * Adds a {@code Metric} given the metric's {@code name}, {@code value}, and {@code unit}. * * @param name the metric name * @param value the metric value * @param unit the metric unit */ public void addMetric(String name, Number value, Unit unit) { addMetric(new Metric(name, value, unit)); } /** * Returns {@code true} if the metrics object has a metric with the given name. * * @param name the name to check for * @return {@code true} if the metrics object has a metric with the given name */ public boolean hasMetric(String name) { return metrics.containsKey(name); } /** * Returns all {@link Metric}s with the specified metric name. * * @param name the name of the metric * @return a list of {@link Metric} with the specified metric name */ public List getMetric(String name) { List list = metrics.get(name); if (list == null) { return Collections.emptyList(); } return list; } /** * Returns a set of {@link String} metric names. * * @return a set of {@link String} metric names */ public Set getMetricNames() { return metrics.keySet(); } /** * Returns the latest {@link Metric} with the specified metric name. * * @param name the name of the metric * @return the {@link Metric} with the specified metric name * @throws IllegalArgumentException if the given name is not found */ public Metric latestMetric(String name) { List list = metrics.get(name); if (list == null || list.isEmpty()) { throw new IllegalArgumentException("Could not find metric: " + name); } return list.get(list.size() - 1); } /** * Returns a percentile {@link Metric} object for the specified metric name. * * @param metricName the name of the metric * @param percentile the percentile * @return the {@link Metric} object at specified {@code percentile} */ public Metric percentile(String metricName, int percentile) { List metric = metrics.get(metricName); if (metric == null || metrics.isEmpty()) { throw new IllegalArgumentException("Metric name not found: " + metricName); } List list = new ArrayList<>(metric); list.sort(Comparator.comparingDouble(Metric::getValue)); int index = metric.size() * percentile / 100; return list.get(index); } /** * Returns the average value of the specified metric. * * @param metricName the name of the metric * @return the average value of the specified metric */ public double mean(String metricName) { List metric = metrics.get(metricName); if (metric == null || metrics.isEmpty()) { throw new IllegalArgumentException("Metric name not found: " + metricName); } return metric.stream().collect(Collectors.averagingDouble(Metric::getValue)); } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy