查看keras各种网络结构各层的名字方式

た入场券 2023-02-20 02:09 9阅读 0赞

更多python视频教程请到菜鸟教程https://www.piaodoo.com/

**举例**

> base\_model = ResNet50(weights=‘imagenet', include\_top=True）  
> print(base\_model.summary())

**得到这个结果**

![20200611095048.jpg][]

**补充知识：****使用keras，在load\_model()时，出现NameError: name '\*\*\*' is not defined**

是因为在构造模型是，使用了自定义的层，如Lambda()

# 文本相似度评估方式
    def exponent_neg_manhattan_distance(sent_left, sent_middle, sent_right):
      '''基于曼哈顿空间距离计算两个字符串语义空间表示相似度计算'''
      return ((K.exp(-K.sum(K.abs(sent_left - sent_middle), axis=1, keepdims=True)) - K.exp(-K.sum(K.abs(sent_left - sent_right), axis=1, keepdims=True))) + 1) / 2

def bilstm\_siamese\_model():  
‘’‘搭建孪生网络’’’  
\#可以在这里调参  
embedding\_layer = Embedding(VOCAB\_SIZE + 1,  
EMBEDDING\_DIM,  
weights=\[embedding\_matrix\],  
input\_length=MAX\_LENGTH,  
trainable=True, \#原本为False  
mask\_zero=True)  
\#输入层  
left\_input = Input(shape=(MAX\_LENGTH,), dtype=‘float32’, name=“left\_x”) \#(?, 25)  
middle\_input = Input(shape=(MAX\_LENGTH,), dtype=‘float32’, name=“middle\_x”) \#(?, 25)  
right\_input = Input(shape=(MAX\_LENGTH,), dtype=‘float32’, name=‘right\_x’)  
\#嵌入层  
encoded\_left = embedding\_layer(left\_input) \#(?, 25, 300)  
encoded\_middle = embedding\_layer(middle\_input) \#(?, 25, 300)  
encoded\_right = embedding\_layer(right\_input)

# print(encoded\_left) #

\#孪生网络  
shared\_lstm = create\_base\_network(input\_shape=(MAX\_LENGTH, EMBEDDING\_DIM))  
left\_output = shared\_lstm(encoded\_left)  
middle\_output = shared\_lstm(encoded\_middle)  
right\_output = shared\_lstm(encoded\_right)

# 文本相似度评估方式 #

distance = Lambda(lambda x: exponent\_neg\_manhattan\_distance(x\[0\], x\[1\], x\[2\]),output\_shape=lambda x: (x\[0\]\[0\], 1))(\[left\_output, middle\_output, right\_output\]) \#第二个lambda函数的输入参数不清楚，x\[0\]\[0\]  
print(‘distance’,distance)  
model = Model(\[left\_input, middle\_input, right\_input\], distance)  
model.compile(loss=‘binary\_crossentropy’,  
optimizer=‘nadam’,  
metrics=\[‘accuracy’\])  
model.summary()  
return model

虽然之前已经定义好了exponent\_neg\_manhattan\_distance()，但是在load\_model()依然会报NameError: name 'exponent\_neg\_manhattan\_distance' is not defined

**解决办法：**

![20200611095326.jpg][]

在load\_model的时候，加一个custom\_objects参数就可以了，即

> model = load\_model(model\_path,custom\_objects=\{'exponent\_neg\_manhattan\_distance': exponent\_neg\_manhattan\_distance\}) \#对自定义层一定要说明

**注：**用Google搜bug比用baidu搜质量会高不少。

以上这篇查看keras各种网络结构各层的名字方式就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多

茂名论坛https://www.hnthzk.com/

化州橘红http://www.sfkyty.com/

茂名论坛http://www.nrso.net/

源码搜藏网http://www.cntkd.net/

茂名市高级技工学校(茂名一技)http://www.szsyby.net/

茂名一技https://www.mmbbu.com/

[20200611095048.jpg]: https://img.jbzj.com/file_images/article/202006/20200611095048.jpg
[20200611095326.jpg]: https://img.jbzj.com/file_images/article/202006/20200611095326.jpg