'NoneType' 对象没有属性 '_inbound_nodes' 错误

Question

我必须获取 EfficientNet 最后一个 conv 层的输出，然后计算 H = wT*x+b。我的 w 是 [49,49]。之后，我必须在 H 上应用 softmax，然后进行逐元素乘法 Xì = Hi*Xi。这是我的代码：

common_input = layers.Input(shape=(224, 224, 3))    
x=model0(common_input) #model0 terminate with last conv layer of EfficientNet (7,7,1280)
x = layers.BatchNormalization()(x)

W = tf.Variable(tf.random_normal([49,49], seed=0), name='weight')
b = tf.Variable(tf.random_normal([49], seed=0), name='bias')

x = tf.reshape(x, [-1, 7*7,1280])
H = tf.matmul(W, x,transpose_a=True)
H = tf.nn.softmax(H)
#print(H.shape) (?,49,1280)
#print(x.shape) (?,49,1280)

x=tf.multiply(H, x)

p=layers.Dense(768, activation="relu")(x)
p=layers.Dense(8, activation="softmax", name="fc_out")(p)

model = Model(inputs=common_input, outputs=p)

但我得到这个错误：'NoneType' 对象没有属性 '_inbound_nodes'

<ipython-input-12-6ce3217f045c> in build_model()
     35     p=layers.Dense(8, activation="softmax", name="fc_out")(p)
     36 
---> 37     model = Model(inputs=common_input, outputs=p)
     38 
     39     return model

AttributeError: 'NoneType' object has no attribute '_inbound_nodes'

Answer 1

我在下面的代码中用 Lambda 层替换了操作。请原谅我拙劣的命名。试试这个代码。

W = tf.Variable(tf.random_normal([49,49], seed=0), name='weight')
b = tf.Variable(tf.random_normal([49], seed=0), name='bias')

def all_operations(args):
    x = args[0]
    H = args[1]
    x = tf.reshape(x, [-1, 7*7,1280])
    H = tf.matmul(W, x, transpose_a=True)
    H = tf.nn.softmax(H)
    x = tf.multiply(H, x)
    x = tf.reshape(x, [-1, 49*1280])
    return x

common_input = layers.Input(shape=(224, 224, 3))    
x=model0(common_input) #model0 terminate with last conv layer of EfficientNet (7,7,1280)
x = layers.BatchNormalization()(x)

x = Lambda(all_operations)([x, H])

p=layers.Dense(768, activation="relu")(x)
p=layers.Dense(8, activation="softmax", name="fc_out")(p)

model = Model(inputs=common_input, outputs=p)

'NoneType' 对象没有属性 '_inbound_nodes' 错误

'NoneType' object has no attribute '_inbound_nodes' error

python

conv-neural-network

keras

tensorflow

tf.keras