tensorflow - 类别 CategoricalCrossentropy 与函数 categorical_crossentropy

Question

在 Tensorflow2 中，我可以使用类 tf.keras.losses.CategoricalCrossentropy ( defined here ) 或函数 categorical_crossentropy ( defined here ) 计算标签和预测之间的交叉熵损失:

第一个代码是:

     loss_object = tf.keras.losses.SparseCategoricalCrossentropy()
     # ...
     @tf.function
     def train_step(samples, labels):
        with tf.GradientTape() as tape:
          predictions = model(samples)
          loss = loss_object(labels, predictions)
        #...

第二个更直接:

   @tf.function
    def forward(features, training=False):
        predictions = model.call(...)
        loss = tf.losses.categorical_crossentropy(
                y_true=features['label'],
                y_pred=predictions)
        return loss, predictions

数值结果是一样的。但是，我想知道是否有更有效的方法？或者更一般地说，根据某些特定情况应该使用哪一个？

请注意，对于任何类/函数 defined by the API 的问题可能都相同

Answer 1

来自 Keras documentation :

Note that all losses are available both via a class handle and via a function handle. The class handles enable you to pass configuration arguments to the constructor (e.g. loss_fn = CategoricalCrossentropy(from_logits=True)), and they perform reduction by default when used in a standalone way (see details below).