GradCAM算法的实现

Question

我有下一个Python代码：

tubby_output=model.output[：，281]

last_conv_layer=model.get_layer（'block5_conv3'））

with tf.GradientTape() as tape:

    grads = tape.gradient(tubby_output, last_conv_layer.output)
    pooled_grads = K.mean(grads, axis=(0,1,2))
    iterate = K.function([model.input],
                         [pooled_grads, last_conv_layer.output[0]])
    pooled_grads_value, conv_layer_output_value = iterate([img_tensor]) 
for i in range(512):
    conv_layer_output_value[:, :, i] *= pooled_grads_value[i]
heatmap = np.mean(conv_layer_output_value, axis=-1)`

我得到以下错误：

    <ipython-input-128-5b66da230a26> in <module>
     13     # Вектор с формой (512,), каждый элемент которого определяет
     14     # интенсивность градиента для заданного канала в карте признаков
---> 15     pooled_grads = K.mean(grads, axis=(0,1,2))
     17     iterate = K.function([model.input],
c:\users\79278\appdata\local\programs\python\python38\lib\site-packages\tensorflow\python\keras\backend.py in mean(x, axis, keepdims)
   2112       A tensor with the mean of elements of `x`.
   2113   """
-> 2114   if x.dtype.base_dtype == d types_module.bool:
   2115     x = math_ops.cast(x, floatx())
   2116   return math_ops.reduce_mean(x, axis, keepdims)
AttributeError: 'None Type' object has no attribute 'dtype'

帮我解决这个问题！谢谢