千层面/茶卷串联激活补丁

Question

我正在尝试连接来自不同conv层的激活补丁。然而，我不断得到错误。我认为这是一个合并的问题。我把我试过的放在下面

下面是一个小例子：

input_var = T.tensor4('inputs')

# Building the network
l_in = InputLayer(shape=(None, 3, 32, 32), input_var=input_var)
l1 = ConvLayer(l_in, num_filters=32, filter_size=(3,3), stride=(1,1), pad='same')
l2 = ConvLayer(l1, num_filters=64, filter_size=(3,3), stride=(1,1), pad='same')

hidden = []
# Sample for hypercolumn
hidden += [lasagne.layers.get_output(l1)[:,:,0:2,0:2]]
hidden += [lasagne.layers.get_output(l2)[:,:,0:2,0:2]]
hypercolumn = lasagne.layers.concat(hidden) # Error on this line

theano.function([input_var], [hypercolumn])

错误：

AttributeError: 'TensorVariable' object has no attribute 'output_shape'