avalanche.evaluation.metrics.class_accuracy_metrics

avalanche.evaluation.metrics.class_accuracy_metrics(*, minibatch=False, epoch=False, epoch_running=False, experience=False, stream=False, classes=None) List[PluginMetric][source]

Helper method that can be used to obtain the desired set of plugin metrics.

Parameters:
  • minibatch – If True, will return a metric able to log the per-class minibatch accuracy at training time.

  • epoch – If True, will return a metric able to log the per-class epoch accuracy at training time.

  • epoch_running – If True, will return a metric able to log the per-class running epoch accuracy at training time.

  • experience – If True, will return a metric able to log the per-class accuracy on each evaluation experience.

  • stream – If True, will return a metric able to log the per-class accuracy averaged over the entire evaluation stream of experiences.

  • classes – The list of classes to track. See the corresponding parameter of ClassAccuracy for a precise explanation.

Returns:

A list of plugin metrics.