Question

我已经训练了这个模型：

from keras.layers import concatenate
from keras.layers import Input, Dense, Masking
from keras.models import Model

x_in = Input(shape=(5,), name='x_in')

s_in = Input(shape=(18,), name='s_in')

s_masked = Masking(0.0)(s_in)

z = concatenate([x_in,s_masked])

dense_1 = Dense(40, kernel_initializer='normal', activation='relu', name='dense_1')(z)
dense_2 = Dense(40, kernel_initializer='normal', activation='relu', name='dense_2')(dense_1)

output = Dense(1, kernel_initializer='normal', name = 'output')(dense_2)

model = Model(inputs=[x_in,s_in], outputs=output)

model.compile(optimizer='adam', loss={'output':'mean_squared_error'})
model.fit({'x_in': x_training,'s_in':s_training},{'output':y_training},batch_size=30, epochs=10, validation_split=0.3, shuffle=True, callbacks=[plot_losses])

我现在要预测。但是由于我有多个输入，因此我不知道如何使用model.predict()。

当我尝试：

predictions = model.predict(x_testing, s_testing)
print predictions

我收到此错误：

该模型需要2个输入数组，但只接收一个数组。发现：形状为（1710，5）的数组

我不明白，因为我给了两个数组x_testing和s_testing

Answer 1

您应该像建立模型（inputs=[x_in,s_in]）一样将输入作为连接数组来输入：

predictions = model.predict([x_testing, s_testing])

预测中的错误-Keras Functional API

1 个答案: