Question

我知道我们可以使用get_layer（）建立一个新模型，但是我的问题有点不同：

我简化了模型：

import keras
import numpy as np
a = np.array([[1,2,3,4,5,6],[1,2,3,4,5,6],[1,2,3,4,5,6]])
x = keras.layers.Input(shape=(6,))
y = keras.layers.Lambda(lambda x: x * 1)(x)
z = keras.layers.Lambda(lambda x: x * 1)(y[:,4])
model = keras.Model(x,[y,z])
model.compile(optimizer='sgd',loss='mean_squared_error')
model.predict(a)

如果我删除“ z”行，则该模型运行正常。否则会遇到以下错误：

AttributeError                            Traceback (most recent call last)
<ipython-input-74-7fc86ac55867> in <module>()
----> 1 model = keras.Model(x,[y,z])
    2 model.compile(optimizer='sgd',loss='mean_squared_error')

~/.conda/envs/21/lib/python3.6/site-packages/keras/legacy/interfaces.py in wrapper(*args, **kwargs)
    89                 warnings.warn('Update your `' + object_name +
    90                               '` call to the Keras 2 API: ' + signature, stacklevel=2)
---> 91             return func(*args, **kwargs)
    92         wrapper._original_function = func
    93         return wrapper

~/.conda/envs/21/lib/python3.6/site-packages/keras/engine/network.py in __init__(self, *args, **kwargs)
    89                 'inputs' in kwargs and 'outputs' in kwargs):
    90             # Graph network
---> 91             self._init_graph_network(*args, **kwargs)
    92         else:
    93             # Subclassed network

~/.conda/envs/21/lib/python3.6/site-packages/keras/engine/network.py in _init_graph_network(self, inputs, outputs, name)
    233         # Keep track of the network's nodes and layers.
    234         nodes, nodes_by_depth, layers, layers_by_depth = _map_graph_network(
--> 235             self.inputs, self.outputs)
    236         self._network_nodes = nodes
    237         self._nodes_by_depth = nodes_by_depth

~/.conda/envs/21/lib/python3.6/site-packages/keras/engine/network.py in _map_graph_network(inputs, outputs)
1410                   layer=layer,
1411                   node_index=node_index,
-> 1412                   tensor_index=tensor_index)
1413 
1414     for node in reversed(nodes_in_decreasing_depth):

~/.conda/envs/21/lib/python3.6/site-packages/keras/engine/network.py in build_map(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)
1397             tensor_index = node.tensor_indices[i]
1398             build_map(x, finished_nodes, nodes_in_progress, layer,
-> 1399                       node_index, tensor_index)
1400 
1401         finished_nodes.add(node)

~/.conda/envs/21/lib/python3.6/site-packages/keras/engine/network.py in build_map(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)
1369             ValueError: if a cycle is detected.
1370         """
-> 1371         node = layer._inbound_nodes[node_index]
1372 
1373         # Prevent cycles.

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

有人可以告诉我为什么会发生这种情况以及如何解决吗？谢谢！

Answer 1

只需在Lambda层中切张量：

a = np.array([[1, 2, 3, 4, 5, 6],
             [1, 2, 3, 4, 5, 6],
             [1, 2, 3, 4, 5, 6]])
x = keras.layers.Input(shape=(6,))
y = keras.layers.Lambda(lambda x: x * 1)(x)
z = keras.layers.Lambda(lambda x: x[:, 4] * 1)(y)
model = keras.Model(x, [y, z])
model.compile(optimizer='sgd', loss='mean_squared_error')
model.predict(a)

如何使用中间层的结果切片（数组）

1 个答案: