TensorFlow 在生成带有 Conv1D 层的简单 GAN 时返回 ValueError

Question

I´m trying to setup simple GAN with TF including Conv1D layer in the discriminator model. To achieve correct output shape, I included Flatten layer.

Unfortunately, while adding generator and discriminator layer together, TF returns error "ValueError: Input tensor must be of rank 3, 4 or 5 but was 2." I tried to do the same with simplest dummy net and compiling of the GAN worked. I expect the trouble is in the input shape of the discriminator layer, but the error description does not give too much lead.

How could I deal with this type of error? Thank you in advance for your help.

def define_discriminator(n_inputs=2):
    model = Sequential()
    model.add(Conv1D(filters = 128, kernel_size = 2, strides=1, input_shape = (n_inputs,1) ))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Flatten())
    model.add(Dense(25, kernel_initializer='he_uniform'))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dense(1, activation='sigmoid'))

    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    model.summary()
    return model

# simple dummy net
"""
model = Sequential()
model.add(Dense(25, activation='relu', kernel_initializer='he_uniform', input_dim=n_inputs))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
model.summary()
return model
"""

def define_generator(latent_dim, n_outputs=2):
    model = Sequential()
    model.add(Dense(15, activation='relu', kernel_initializer='he_uniform', input_dim=latent_dim))
    model.add(Dense(n_outputs, activation='linear'))
    model.summary()
    return model

def define_gan(generator, discriminator):
    discriminator.trainable = False
    model = Sequential()
    model.add(generator)
    model.add(discriminator)
    model.compile(loss='binary_crossentropy', optimizer='adam')
    model.summary()
    return model

Complete Error message here:

Traceback (most recent call last):
  File "C:/pracovni_addr/python_projects/GAN_1D.py", line 181, in <module>
    gan_model = define_gan(m_gen, m_disc)
  File "C:/pracovni_addr/python_projects/GAN_1D.py", line 115, in define_gan
    model.add(discriminator)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\engine\sequential.py", line 182, in add
    output_tensor = layer(self.outputs[0])
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\backend\tensorflow_backend.py", line 75, in symbolic_fn_wrapper
    return func(*args, **kwargs)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\engine\base_layer.py", line 489, in __call__
    output = self.call(inputs, **kwargs)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\engine\network.py", line 583, in call
    output_tensors, _, _ = self.run_internal_graph(inputs, masks)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\engine\network.py", line 740, in run_internal_graph
    layer.call(computed_tensor, **kwargs))
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\layers\convolutional.py", line 163, in call
    dilation_rate=self.dilation_rate[0])
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\keras\backend\tensorflow_backend.py", line 3671, in conv1d
    **kwargs)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 917, in convolution_v2
    name=name)
  File "C:\Users\CLIENT\AppData\Local\Programs\Python\Python37\lib\site-packages\tensorflow_core\python\ops\nn_ops.py", line 969, in convolution_internal
    "Input tensor must be of rank 3, 4 or 5 but was {}.".format(n + 2))
ValueError: Input tensor must be of rank 3, 4 or 5 but was 2.

Answer 1

So the error was in the wrong shape of the generator output (discriminator expects the input shape as (None, 2, 1), but only (None, 2) was given.

Problem solved with:

    model.add(Reshape((n_outputs,1)))

before

    model.sumary() 

in define_generator block