why does it say this: "TypeError: call() missing 1 required positional argument: 'inputs'"

Question

我正在尝试制作一个自定义的类似 vgg 的模型，具有 dropout 和 swish 激活功能。我完全不明白这里的问题是什么。我是初学者，所以也许我完全错了。不过我不这么认为，因为我以教程和指南为例。

这里是代码：

from keras.models import Sequential
from keras.layers import Dense, Conv2D, MaxPooling2D, Dropout
from keras.layers import Flatten, Activation
from keras.applications.vgg16 import VGG16
from keras.models import Model
from keras import backend as K

def swish(x):
    return K.sigmoid(x) * x

custom_vgg = Sequential()
img_width = 224
img_height = 224
vgg_model = VGG16(include_top=False, weights='imagenet',input_shape=(img_width, img_height, 3))
print(vgg_model.get_config())
#vgg_model.save_weights('models/vgg_weights.h5')
for layer in vgg_model.layers:
    if layer.__class__.__name__=='MaxPooling2D':
        layer.trainable = False
        custom_vgg.add(layer(activation = swish))
        custom_vgg.add(Dropout(0.4))
    else :
        custom_vgg.add(layer)
custom_vgg.add(Flatten())
custom_vgg.add(Dense(1024,activation=swish))
custom_vgg.add(Dense(1024,activation=swish))
custom_vgg.add(Dense(128, activation = "softmax"))

custom_vgg.compile(optimizer = "adam", loss = "categorical_crossentropy", metrics = ["accuracy"])

custom_vgg.summary()

控制台这样说：

Traceback (most recent call last):
  File "/Users/shavkatsultanov/Documents/SCHUETZ/tensor_flow/vgg_swish.py", line 20, in <module>
    custom_vgg.add(layer(activation = swish))
TypeError: __call__() missing 1 required positional argument: 'inputs'

我想知道，因为我尝试过类似的事情，遍历图层并添加到另一个模型，并且它奏效了。此外，当它位于第 20 行时，不应期待任何 'inputs' 或其他什么？

如果有人能在这里提供帮助，那就太好了，谢谢！

edit：我想知道关于 if 语句的问题，询问 class_name，是否有问题？真的希望有人能解决这个问题

Answer 1

custom_vgg.add(layer(activation = swish))

您正在调用一个 layer 对象（这不是创建层，它是调用您从现有模型获得的层）。显然 keras 层是可调用的，layer(...) 是一堆内部逻辑的包装器（那是......我认为不是很好 API）。

无论哪种方式，只是...不要调用图层。

why does it say this: "TypeError: call() missing 1 required positional argument: 'inputs'"

why does it say this: "TypeError: call() missing 1 required positional argument: 'inputs'"

python

python-3.x

conv-neural-network

keras

vgg-net

why does it say this: "TypeError: __call__() missing 1 required positional argument: 'inputs'"

why does it say this: "TypeError: __call__() missing 1 required positional argument: 'inputs'"

python

python-3.x

conv-neural-network

keras

vgg-net

why does it say this: "TypeError: call() missing 1 required positional argument: 'inputs'"

why does it say this: "TypeError: call() missing 1 required positional argument: 'inputs'"