连接 keras 层时出错：“连接”层需要具有匹配形状的输入答案

【问题标题】：Error while concatenating keras layers: A `Concatenate` layer requires inputs with matching shapes连接 keras 层时出错：“连接”层需要具有匹配形状的输入
【发布时间】：2019-08-28 20:04:33
【问题描述】：

我有这个模型，叫做Hierarchical Attention Networks：

建议用于文档分类。我对句子单词使用 word2vec 嵌入，我想在 A 点连接另一个句子级嵌入（见图）。

我将它用于包含 3 个句子的文档；模型总结：

word_input = Input(shape=(self.max_senten_len,), dtype='float32')
word_sequences = self.get_embedding_layer()(word_input)
word_lstm = Bidirectional(self.hyperparameters['rnn'](self.hyperparameters['rnn_units'], return_sequences=True, kernel_regularizer=kernel_regularizer))(word_sequences)
word_dense = TimeDistributed(Dense(self.hyperparameters['dense_units'], kernel_regularizer=kernel_regularizer))(word_lstm)
word_att = AttentionWithContext()(word_dense)
wordEncoder = Model(word_input, word_att)
sent_input = Input(shape=(self.max_senten_num, self.max_senten_len), dtype='float32')
sent_encoder = TimeDistributed(wordEncoder)(sent_input)

""" I added these following 2 lines. The dimension of self.training_features is (number of training rows, 3, 512). 512 is the dimension of the sentence-level embedding.  """
USE = Input(shape=(self.training_features.shape[1], self.training_features.shape[2]), name='USE_branch')
merge = concatenate([sent_encoder, USE], axis=1)

sent_lstm = Bidirectional(self.hyperparameters['rnn'](self.hyperparameters['rnn_units'], return_sequences=True, kernel_regularizer=kernel_regularizer))(merge)
sent_dense = TimeDistributed(Dense(self.hyperparameters['dense_units'], kernel_regularizer=kernel_regularizer))(sent_lstm)
sent_att = Dropout(dropout_regularizer)(AttentionWithContext()(sent_dense))
preds = Dense(len(self.labelencoder.classes_))(sent_att)
self.model = Model(sent_input, preds)

当我编译上面的代码时，我得到以下错误：

ValueError: A Concatenate 层需要具有匹配形状的输入除了 concat 轴。得到输入形状：[(None, 3, 128), (None, 3, 514)]

我指定了连接轴 = 1，以连接 (3) 句子的数量，但我不知道为什么我仍然收到错误。

【问题讨论】：

标签： python-3.x keras deep-learning concatenation

【解决方案1】：

这是因为如果您指定该轴，则形状不匹配。如果您这样做，这将起作用：

merge = concatenate([sent_encoder, USE], axis=-1)

现在其余轴上的形状没有冲突

【讨论】：

我收到了这个错误：ValueError: Graph disconnected: cannot obtain value for tensor Tensor("USE_branch:0", shape=(?, 3, 514), dtype=float32) at layer "USE_branch". The following previous layers were accessed without issue: ['input_2', 'time_distributed_2']

【解决方案2】：

错误是由于两行：

merge = concatenate([sent_encoder, USE], axis=1)
# should be:
merge = concatenate([sent_encoder, USE], axis=2) # or -1 as @mlRocks suggested

还有一行：

self.model = Model(sent_input, preds)
# should be:
self.model = Model([sent_input, USE], preds) # to define both inputs

【讨论】：