ValueError：提供的元素太多。最多需要-761894396，但收到1

Question

当试图在keras中拟合一个相当简单的卷积+密集层模型时，我似乎无法完成fit（）函数。

x训练数据是（批量大小，高度，宽度，通道）格式的形状i_train.shape（10,1070,1230,3）的numpy数组。 y训练数据的形状为m_train.shape（10,1070,1230）（批量，高度，宽度）格式。

nb_filter = 1
nb_row = 3
nb_col = 3

model = Sequential()

model.add(Conv2D(nb_filter, (nb_row, nb_col), padding = 'same', 
activation='relu', input_shape=(h,w,3)))
model.add(MaxPooling2D(pool_size=(2,2), padding = 'valid'))
model.add(Flatten())

model.add(Dense(h*w, activation='relu', input_shape= (329025,)))

model.add(BatchNormalization())

model.add(Reshape((h,w)))
model.compile(loss='mean_squared_error', optimizer='SGD', metrics=['accuracy'])
model.fit(i_train, m_train, batch_size=10, epochs=10, verbose=1)

我得到的错误是

    Traceback (most recent call last):

  File "<ipython-input-9-639284a1176d>", line 1, in <module>
    model.fit(i_train, m_train, batch_size=10, epochs=10, verbose=1)

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\models.py", line 870, in fit
    initial_epoch=initial_epoch)

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\engine\training.py", line 1490, in fit
    self._make_train_function()

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\engine\training.py", line 1014, in _make_train_function
    self.total_loss)

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\optimizers.py", line 162, in get_updates
    moments = [K.zeros(shape) for shape in shapes]

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\optimizers.py", line 162, in <listcomp>
    moments = [K.zeros(shape) for shape in shapes]

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\keras\backend\tensorflow_backend.py", line 601, in zeros
    return variable(tf.constant_initializer(0., dtype=tf_dtype)(shape),

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\tensorflow\python\ops\init_ops.py", line 162, in __call__
    verify_shape=self.verify_shape)

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\tensorflow\python\framework\constant_op.py", line 102, in constant
    tensor_util.make_tensor_proto(value, dtype=dtype, shape=shape, verify_shape=verify_shape))

  File "C:\Users\Jacob.Rose\AppData\Local\Continuum\Anaconda2\envs\TFlow\lib\site-packages\tensorflow\python\framework\tensor_util.py", line 424, in make_tensor_proto
    (shape_size, nparray.size))

ValueError: Too many elements provided. Needed at most -761894396, but received 1

Answer 1

您的模型摘要：

因此，您的密集_1（排除偏见）的参数数量：329025 * 1316100 = 433029802500.一个图层的参数太多了。 您应该添加更多的Conv2D和Pooling图层以减小图像的大小并减小Dense图层的输出大小。