不推荐使用 Keras 模型/层正则化属性

Question

我正在使用 Keras 定义一个模型，然后我尝试使用分布式 Tensorflow 对其进行校准，就像 here 所做的那样。

我曾经像链接中那样处理正则化器损失

model = Sequential()
...... #build keras model
loss = tf.reduce_mean(keras.objectives.mean_squared_error(targets, preds))
# apply regularizers if any
if model.regularizers:
    total_loss = loss * 1.  # copy tensor
    for regularizer in model.regularizers:
        total_loss = regularizer(total_loss)
else:
    total_loss = loss

但现在正则化器属性已被贬低，警告建议使用model/layer 的losses 属性，所以我尝试了：

 loss = tf.reduce_mean(keras.objectives.mean_squared_error(targets, preds))
 total_loss = loss * 1.  # copy tensor
 for reg_loss in model.losses:
     tf.assign_add(total_loss, reg_loss)

但这会导致崩溃。请帮忙

Answer 1

正确的做法是：

loss = tf.reduce_mean(keras.objectives.mean_squared_error(targets, preds))
total_loss = loss * 1.  # copy tensor
for reg_loss in model.losses:
total_loss = total8loss + reg_loss