在Google Colab构建深度学习模型时遇'Functional'不可下标错误

错误出在for layer in base_model[:]:这一行——MobileNet实例化得到的base_model是Functional类型的模型对象，不支持下标切片访问。要遍历模型的所有层，需使用模型的layers属性，它会返回模型所有层的列表。

修正后的完整代码：

def model_maker():
  base_model = MobileNet(include_top=False,
                         input_shape = (IMG_WIDTH,IMG_HEIGHT,3))
  # 遍历模型所有层并冻结
  for layer in base_model.layers:
    layer.trainable = False
  input = Input(shape=(IMG_WIDTH,IMG_HEIGHT,3))
  custom_model = base_model(input)
  custom_model = GlobalAveragePooling2D()(custom_model)
  custom_model = Dense(64,activation='relu')(custom_model)
  custom_model = Dropout(0.5)(custom_model)
  predictions = Dense(NUM_CLASSES,activation='softmax')(custom_model)
  return Model(inputs=input,outputs=predictions)
# 调用模型
model = model_maker()
model.compile(loss='categorical_crossentropy',
              optimizer=tf.keras.optimizers.Adam(0.001),
              metrics=['acc'])

另外提供一种更简洁的冻结方式：无需循环遍历每一层，直接设置base_model.trainable = False即可冻结整个基础模型的所有层，代码更精简：

def model_maker():
  base_model = MobileNet(include_top=False,
                         input_shape = (IMG_WIDTH,IMG_HEIGHT,3))
  # 直接冻结整个基础模型
  base_model.trainable = False
  input = Input(shape=(IMG_WIDTH,IMG_HEIGHT,3))
  custom_model = base_model(input)
  custom_model = GlobalAveragePooling2D()(custom_model)
  custom_model = Dense(64,activation='relu')(custom_model)
  custom_model = Dropout(0.5)(custom_model)
  predictions = Dense(NUM_CLASSES,activation='softmax')(custom_model)
  return Model(inputs=input,outputs=predictions)

内容的提问来源于stack exchange，提问作者Mainland