Source code for avalanche.evaluation.metrics.topk_acc

################################################################################
# Copyright (c) 2021 ContinualAI.                                              #
# Copyrights licensed under the MIT License.                                   #
# See the accompanying LICENSE file for terms.                                 #
#                                                                              #
# Date: 29-03-2022                                                             #
# Author(s): Rudy Semola                                                       #
# E-mail: contact@continualai.org                                              #
# Website: www.continualai.org                                                 #
################################################################################

from typing import List, Union, Dict

import torch
from torch import Tensor
from torchmetrics.functional import accuracy

from avalanche.evaluation import Metric, PluginMetric, GenericPluginMetric
from avalanche.evaluation.metrics.mean import Mean
from avalanche.evaluation.metric_utils import phase_and_task

from collections import defaultdict


[docs]class TopkAccuracy(Metric[float]): """ The Top-k Accuracy metric. This is a standalone metric. It is defined using torchmetrics.functional accuracy with top_k """
[docs] def __init__(self, top_k): """ Creates an instance of the standalone Top-k Accuracy metric. By default this metric in its initial state will return a value of 0. The metric can be updated by using the `update` method while the running top-k accuracy can be retrieved using the `result` method. :param top_k: integer number to define the value of k. """ self._topk_acc_dict = defaultdict(Mean) self.top_k = top_k
@torch.no_grad() def update( self, predicted_y: Tensor, true_y: Tensor, task_labels: Union[float, Tensor], ) -> None: """ Update the running top-k accuracy given the true and predicted labels. Parameter `task_labels` is used to decide how to update the inner dictionary: if Float, only the dictionary value related to that task is updated. If Tensor, all the dictionary elements belonging to the task labels will be updated. :param predicted_y: The model prediction. Both labels and logit vectors are supported. :param true_y: The ground truth. Both labels and one-hot vectors are supported. :param task_labels: the int task label associated to the current experience or the task labels vector showing the task label for each pattern. :return: None. """ if len(true_y) != len(predicted_y): raise ValueError("Size mismatch for true_y and predicted_y tensors") if isinstance(task_labels, Tensor) and len(task_labels) != len(true_y): raise ValueError("Size mismatch for true_y and task_labels tensors") true_y = torch.as_tensor(true_y) predicted_y = torch.as_tensor(predicted_y) if isinstance(task_labels, int): total_patterns = len(true_y) self._topk_acc_dict[task_labels].update( accuracy(predicted_y, true_y, top_k=self.top_k), total_patterns ) elif isinstance(task_labels, Tensor): for pred, true, t in zip(predicted_y, true_y, task_labels): self._topk_acc_dict[t.item()].update( accuracy(pred, true, top_k=self.top_k), 1 ) else: raise ValueError( f"Task label type: {type(task_labels)}, " f"expected int/float or Tensor" ) def result(self, task_label=None) -> Dict[int, float]: """ Retrieves the running top-k accuracy. Calling this method will not change the internal state of the metric. :param task_label: if None, return the entire dictionary of accuracies for each task. Otherwise return the dictionary `{task_label: topk_accuracy}`. :return: A dict of running accuracies for each task label, where each value is a float value between 0 and 1. """ assert task_label is None or isinstance(task_label, int) if task_label is None: return {k: v.result() for k, v in self._topk_acc_dict.items()} else: return {task_label: self._topk_acc_dict[task_label].result()} def reset(self, task_label=None) -> None: """ Resets the metric. :param task_label: if None, reset the entire dictionary. Otherwise, reset the value associated to `task_label`. :return: None. """ assert task_label is None or isinstance(task_label, int) if task_label is None: self._topk_acc_dict = defaultdict(Mean) else: self._topk_acc_dict[task_label].reset()
class TopkAccuracyPluginMetric(GenericPluginMetric[float]): """ Base class for all top-k accuracies plugin metrics """ def __init__(self, reset_at, emit_at, mode, top_k): self._topk_acc = TopkAccuracy(top_k=top_k) super(TopkAccuracyPluginMetric, self).__init__( self._topk_acc, reset_at=reset_at, emit_at=emit_at, mode=mode ) def reset(self, strategy=None) -> None: if self._reset_at == "stream" or strategy is None: self._metric.reset() else: self._metric.reset(phase_and_task(strategy)[1]) def result(self, strategy=None) -> float: if self._emit_at == "stream" or strategy is None: return self._metric.result() else: return self._metric.result(phase_and_task(strategy)[1]) def update(self, strategy): # task labels defined for each experience task_labels = strategy.experience.task_labels if len(task_labels) > 1: # task labels defined for each pattern task_labels = strategy.mb_task_id else: task_labels = task_labels[0] self._topk_acc.update(strategy.mb_output, strategy.mb_y, task_labels)
[docs]class MinibatchTopkAccuracy(TopkAccuracyPluginMetric): """ The minibatch plugin top-k accuracy metric. This metric only works at training time. This metric computes the average top-k accuracy over patterns from a single minibatch. It reports the result after each iteration. """
[docs] def __init__(self, top_k): """ Creates an instance of the MinibatchTopkAccuracy metric. """ super(MinibatchTopkAccuracy, self).__init__( reset_at="iteration", emit_at="iteration", mode="train", top_k=top_k ) self.top_k = top_k
def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_MB"
[docs]class EpochTopkAccuracy(TopkAccuracyPluginMetric): """ The average top-k accuracy over a single training epoch. This plugin metric only works at training time. The top-k accuracy will be logged after each training epoch by computing the number of correctly predicted patterns during the epoch divided by the overall number of patterns encountered in that epoch. """
[docs] def __init__(self, top_k): """ Creates an instance of the EpochTopkAccuracy metric. """ super(EpochTopkAccuracy, self).__init__( reset_at="epoch", emit_at="epoch", mode="train", top_k=top_k ) self.top_k = top_k
def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_Epoch"
[docs]class RunningEpochTopkAccuracy(TopkAccuracyPluginMetric): """ The average top-k accuracy across all minibatches up to the current epoch iteration. This plugin metric only works at training time. At each iteration, this metric logs the top-k accuracy averaged over all patterns seen so far in the current epoch. The metric resets its state after each training epoch. """
[docs] def __init__(self, top_k): """ Creates an instance of the RunningEpochTopkAccuracy metric. """ super(RunningEpochTopkAccuracy, self).__init__( reset_at="epoch", emit_at="iteration", mode="train", top_k=top_k ) self.top_k = top_k
def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_Epoch"
[docs]class ExperienceTopkAccuracy(TopkAccuracyPluginMetric): """ At the end of each experience, this plugin metric reports the average top-k accuracy over all patterns seen in that experience. This metric only works at eval time. """
[docs] def __init__(self, top_k): """ Creates an instance of the ExperienceTopkAccuracy metric. """ super(ExperienceTopkAccuracy, self).__init__( reset_at="experience", emit_at="experience", mode="eval", top_k=top_k, ) self.top_k = top_k
def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_Exp"
[docs]class TrainedExperienceTopkAccuracy(TopkAccuracyPluginMetric): """ At the end of each experience, this plugin metric reports the average top-k accuracy for only the experiences that the model has been trained on so far. This metric only works at eval time. """
[docs] def __init__(self, top_k): """ Creates an instance of the TrainedExperienceTopkAccuracy metric. """ super(TrainedExperienceTopkAccuracy, self).__init__( reset_at="stream", emit_at="stream", mode="eval", top_k=top_k ) self._current_experience = 0 self.top_k = top_k
def after_training_exp(self, strategy) -> None: self._current_experience = strategy.experience.current_experience # Reset average after learning from a new experience TopkAccuracyPluginMetric.reset(self, strategy) return TopkAccuracyPluginMetric.after_training_exp(self, strategy) def update(self, strategy): """ Only update the top-k accuracy with results from experiences that have been trained on """ if strategy.experience.current_experience <= self._current_experience: TopkAccuracyPluginMetric.update(self, strategy) def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_On_Trained_Experiences"
[docs]class StreamTopkAccuracy(TopkAccuracyPluginMetric): """ At the end of the entire stream of experiences, this plugin metric reports the average top-k accuracy over all patterns seen in all experiences. This metric only works at eval time. """
[docs] def __init__(self, top_k): """ Creates an instance of StreamTopkAccuracy metric """ super(StreamTopkAccuracy, self).__init__( reset_at="stream", emit_at="stream", mode="eval", top_k=top_k ) self.top_k = top_k
def __str__(self): return "Topk_" + str(self.top_k) + "_Acc_Stream"
[docs]def topk_acc_metrics( *, top_k=3, minibatch=False, epoch=False, epoch_running=False, experience=False, trained_experience=False, stream=False, ) -> List[PluginMetric]: """ Helper method that can be used to obtain the desired set of plugin metrics. :param minibatch: If True, will return a metric able to log the minibatch top-k accuracy at training time. :param epoch: If True, will return a metric able to log the epoch top-k accuracy at training time. :param epoch_running: If True, will return a metric able to log the running epoch top-k accuracy at training time. :param experience: If True, will return a metric able to log the top-k accuracy on each evaluation experience. :param trained_experience: If True, will return a metric able to log the average evaluation top-k accuracy only for experiences that the model has been trained on :param stream: If True, will return a metric able to log the top-k accuracy averaged over the entire evaluation stream of experiences. :return: A list of plugin metrics. """ metrics = [] if minibatch: metrics.append(MinibatchTopkAccuracy(top_k=top_k)) if epoch: metrics.append(EpochTopkAccuracy(top_k=top_k)) if epoch_running: metrics.append(RunningEpochTopkAccuracy(top_k=top_k)) if experience: metrics.append(ExperienceTopkAccuracy(top_k=top_k)) if trained_experience: metrics.append(TrainedExperienceTopkAccuracy(top_k=top_k)) if stream: metrics.append(StreamTopkAccuracy(top_k=top_k)) return metrics
__all__ = [ "TopkAccuracy", "MinibatchTopkAccuracy", "EpochTopkAccuracy", "RunningEpochTopkAccuracy", "ExperienceTopkAccuracy", "StreamTopkAccuracy", "TrainedExperienceTopkAccuracy", "topk_acc_metrics", ] """ UNIT TEST """ if __name__ == "__main__": metric = topk_acc_metrics(trained_experience=True, top_k=5) print(metric)