Customize Evaluator

The Evaluator class is used to evaluate the performance of a model using different evaluation metrics. The class is initialized to accept a list of metrics, where each metric can be a string representing a built-in metric function (e.g., accuracy, precision, recall, f1, auc), or a custom Callable function. If no metrics are provided, accuracy, precision, recall, and f1 will be used by default, and if you want to insert your own custom metrics function into Evaluator, you can do so by creating a Callable function that follows the signature of metric_func(outputs: Tensor, y: Tensor) -> float. Tensor) -> float signature, which is then passed to the Evaluator constructor as part of the metrics list.