问题描述
我正在制作一个深多模自动编码器,该编码器具有两个输入并产生两个输出(它们是重构的输入)。两个输入的形状分别为(1000,50)和(1000,60),模型具有3个隐藏层,目的是将输入1和输入2的两个潜在层连接起来。
这是模型的完整代码:
input_X = Input(shape=(X[0].shape))
dense_X = Dense(40,activation='relu')(input_X)
dense1_X = Dense(20,activation='relu')(dense_X)
latent_X= Dense(2,activation='relu')(dense1_X)
input_X1 = Input(shape=(X1[0].shape))
dense_X1 = Dense(40,activation='relu')(input_X1)
dense1_X1 = Dense(20,activation='relu')(dense_X1)
latent_X1= Dense(2,activation='relu')(dense1_X1)
Concat_X_X1 = concatenate([latent_X,latent_X1])
decoding_X = Dense(20,activation='relu')(Concat_X_X1)
decoding1_X = Dense(40,activation='relu')(decoding_X)
output_X = Dense(X[0].shape[0],activation='sigmoid')(decoding1_X)
decoding_X1 = Dense(20,activation='relu')(Concat_X_X1)
decoding1_X1 = Dense(40,activation='relu')(decoding_X1)
output_X1 = Dense(X1[0].shape[0],activation='sigmoid')(decoding1_X1)
multi_modal_autoencoder = Model([input_X,input_X1],[output_X,output_X1],name='multi_modal_autoencoder')
encoder = Model([input_X,Concat_X_X1)
encoder.save('encoder.h5')
multi_modal_autoencoder.compile(optimizer=keras.optimizers.Adam(lr=0.001),loss='mse')
model = multi_modal_autoencoder.fit([X,X1],[X,epochs=70,batch_size=150)
我想从编码器中返回潜伏表示,该潜伏表示将表现为形状为(1000,4)的numpy数组,然后将其用作另一个模型的输入。希望有人知道这些可以帮助我实现目标。为此,我按照建议尝试了以下操作:
file = h5py.File('encoder.h5','r')
keys = list(file.keys()) #it returns models weights as key
value = file.get('model_weights') #<HDF5 group "/model_weights" (9 members)>
the 9 members are ['concatenate_1','dense_1','dense_2','dense_3','dense_4','dense_5','dense_6','input_1','input_2'].
file['/model_weights/concatenate_1']) returns <HDF5 group "/model_weights/concatenate_1" (0 members)>
value = file['/model_weights/concatenate_1'][:]
AttributeError Traceback (most recent call last)
<ipython-input-18-7bc6cbac9468> in <module>
----> 1 value = file['/model_weights/concatenate_1'][:]
h5py\_objects.pyx in h5py._objects.with_phil.wrapper()
h5py\_objects.pyx in h5py._objects.with_phil.wrapper()
~\Anaconda3\envs\tensorflow\lib\site-packages\h5py\_hl\group.py in __getitem__(self,name)
260 raise ValueError("Invalid HDF5 object reference")
261 else:
--> 262 oid = h5o.open(self.id,self._e(name),lapl=self._lapl)
263
264 otype = h5i.get_type(oid)
~\Anaconda3\envs\tensorflow\lib\site-packages\h5py\_hl\base.py in _e(self,name,lcpl)
135 else:
136 try:
--> 137 name = name.encode('ascii')
138 coding = h5t.CSET_ASCII
139 except UnicodeEncodeError:
AttributeError: 'slice' object has no attribute 'encode'
解决方法
我假设X[0].shape[0]
和X1[0].shape[0]
相等,并且由于它是一个密集层,因此应该为4000。您已经设法进入训练阶段,但是更好的是说Model.fit
是训练过程中造成损失的历史对象。这样,您名为model
的对象实际上不是模型。
要使用此训练有素的模型预测值,您需要致电Model.predict()
,您的情况应类似于:
multi_modal_autoencoder.predict([D1,D2])
Model.predict()
返回预测的numpy数组,在您的情况下为两个数组,并且在检索输入的预测后可能需要reshape方法。然后,您可以将该输出用作下一个网络的输入。
我强烈建议您阅读docs