Keras错误＆＃34;您必须为占位符张量提供值＆quot; bidirectional_1 / keras_learning_phase＆＃39;用dtype bool＆＃34;

Question

我在下面的代码段中收到以下错误：

  

您必须为占位符张量提供值   ＆＃39; bidirectional_1 / keras_learning_phase＆＃39;用dtype bool

如果我添加了dropout图层model.add(Dropout(dropout))，它就可以了。谁知道为什么？后端是Tensorflow，Keras 2.0.1

def prep_model1(embedding_layer1, embedding_layer2, dropout=0.5):

    model0 = Sequential()  
    model0.add(embedding_layer1)
    model0.add(Bidirectional(LSTM(128, return_sequences=False, dropout=dropout)))

    model1 = Sequential() 
    model1.add(embedding_layer2)
    model1.add(Bidirectional(LSTM(128, return_sequences=False, dropout=dropout)))

    model = Sequential()
    model.add(Merge([model0, model1], mode='concat', concat_axis=1))
    #model.add(Dropout(dropout))
    model.add(Dense(1, activation='sigmoid'))

    return model

Answer 1

尝试导入K并在模型之前设置学习阶段。

from keras import backend as K

K.set_learning_phase(1) #set learning phase

来自this issue