我遇到的问题是，当运行下面的keras代码时，我收到此错误：ValueError: Input dimension mis-match.（详细的错误消息和完整的回溯）
只有在使用Theano后端时才会发生这种情况，对于TensorFlow后端，一切正常。
首先是输入：
>>> X_train.shape
(37606, 16)
目标值y_train仅存储在列表中，如下所示为1和0：
[0,1,1,1,0,0, ... ,1,0,1,1]
单词嵌入模型有300暗淡：
>>> w.shape
(702904, 300)
我已经阅读了类似的问题，其中建议使用了两个输出单元（isn＆quot;这只对categorical_crossentropy有用吗？）。我尝试了这一点，同时将目标值重塑为：(37606, 2)。
但这并没有解决问题。
我猜测嵌入层存在一些问题，因为输入被指定为max_review_length = 16，嵌入提供了另外300个dims。但如果是这样，我不知道该改变什么。
如上所述，此代码使用TensorFlow后端工作正常。
（Theano是版本1.0.0和Python 3.4.5上的Keras版本2.1.1）
提前感谢任何帮助或想法！
def create_model(X, Y, summary=False):
batch_size = 512
epochs = 5
sequence_input = Input(shape=(max_review_length,), dtype='int32')
embedding_layer = Embedding(w.shape[0],w.shape[1],
weights=[w],
input_length=max_review_length,
trainable=False)
embedded_sequences = embedding_layer(sequence_input)
attention_probs = Dense(300, activation='softmax', name='attention_vec')(embedded_sequences)
attention_mul = multiply([embedded_sequences, attention_probs], name='attention')
x = Conv1D(256, kernel_size=3, activation='relu', padding='same')(attention_mul)
x = Conv1D(128, kernel_size=3, activation='relu', padding='same')(x)
x = MaxPooling1D(4)(x)
x = Dense(128, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(128, activation='relu')(x)
x = Flatten()(x)
x = Dense(32, activation='relu')(x)
attention_probs = Dense(32, activation='softmax', name='attention2_vec')(x)
attention_mul = multiply([x, attention_probs], name='attention2')
preds = Dense(1, activation='sigmoid')(attention_mul)
model = Model(sequence_input, preds)
model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
if summary: print(model.summary())
model.fit(X, Y, epochs=epochs, batch_size=batch_size, shuffle=True)
return model
seed = 42
np.random.seed(seed)
rn.seed(seed)
model = create_model(X_train, y_train, summary=True)
predictions = model.predict(X_test)
evaluate(predictions, y_test)
完整错误追溯：
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
~/python_env/lib64/python3.4/site-packages/theano/compile/function_module.py in __call__(self, *args, **kwargs)
902 outputs =\
--> 903 self.fn() if output_subset is None else\
904 self.fn(output_subset=output_subset)
ValueError: Input dimension mis-match. (input[0].shape[1] = 16, input[1].shape[1] = 300)
During handling of the above exception, another exception occurred:
ValueError Traceback (most recent call last)
<ipython-input-15-456c0ca552d0> in <module>()
51 #K.set_session(sess)
52
---> 53 model = create_model(X_train, y_train, summary=True)
54 predictions = model.predict(X_test)
55 evaluate(predictions, y_test)
<ipython-input-15-456c0ca552d0> in create_model(X, Y, summary)
39
40 if summary: print(model.summary())
---> 41 model.fit(X, Y, epochs=epochs, batch_size=batch_size, shuffle=True)
42 return model
43 seed = 42
~/python_env/lib64/python3.4/site-packages/Keras-2.1.1-py3.4.egg/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, **kwargs)
1655 initial_epoch=initial_epoch,
1656 steps_per_epoch=steps_per_epoch,
-> 1657 validation_steps=validation_steps)
1658
1659 def evaluate(self, x=None, y=None,
~/python_env/lib64/python3.4/site-packages/Keras-2.1.1-py3.4.egg/keras/engine/training.py in _fit_loop(self, f, ins, out_labels, batch_size, epochs, verbose, callbacks, val_f, val_ins, shuffle, callback_metrics, initial_epoch, steps_per_epoch, validation_steps)
1211 batch_logs['size'] = len(batch_ids)
1212 callbacks.on_batch_begin(batch_index, batch_logs)
-> 1213 outs = f(ins_batch)
1214 if not isinstance(outs, list):
1215 outs = [outs]
~/python_env/lib64/python3.4/site-packages/Keras-2.1.1-py3.4.egg/keras/backend/theano_backend.py in __call__(self, inputs)
1222 def __call__(self, inputs):
1223 assert isinstance(inputs, (list, tuple))
-> 1224 return self.function(*inputs)
1225
1226
~/python_env/lib64/python3.4/site-packages/theano/compile/function_module.py in __call__(self, *args, **kwargs)
915 node=self.fn.nodes[self.fn.position_of_error],
916 thunk=thunk,
--> 917 storage_map=getattr(self.fn, 'storage_map', None))
918 else:
919 # old-style linkers raise their own exceptions
~/python_env/lib64/python3.4/site-packages/theano/gof/link.py in raise_with_op(node, thunk, exc_info, storage_map)
323 # extra long error message in that case.
324 pass
--> 325 reraise(exc_type, exc_value, exc_trace)
326
327
~/python_env/lib64/python3.4/site-packages/six.py in reraise(tp, value, tb)
690 value = tp()
691 if value.__traceback__ is not tb:
--> 692 raise value.with_traceback(tb)
693 raise value
694 finally:
~/python_env/lib64/python3.4/site-packages/theano/compile/function_module.py in __call__(self, *args, **kwargs)
901 try:
902 outputs =\
--> 903 self.fn() if output_subset is None else\
904 self.fn(output_subset=output_subset)
905 except Exception:
ValueError: Input dimension mis-match. (input[0].shape[1] = 16, input[1].shape[1] = 300)
Apply node that caused the error: Elemwise{Add}[(0, 0)](Reshape{3}.0, InplaceDimShuffle{x,0,x}.0)
Toposort index: 105
Inputs types: [TensorType(float32, 3D), TensorType(float32, (True, False, True))]
Inputs shapes: [(512, 16, 300), (1, 300, 1)]
Inputs strides: [(19200, 1200, 4), (1200, 4, 4)]
Inputs values: ['not shown', 'not shown']
Outputs clients: [[Reduce{maximum}{2}(Elemwise{Add}[(0, 0)].0), Elemwise{Composite{exp((i0 - i1))}}(Elemwise{Add}[(0, 0)].0, InplaceDimShuffle{0,1,x}.0), Elemwise{Composite{(i0 + (i1 * EQ(i2, i3) * i4))}}[(0, 0)](Elemwise{Composite{(((i0 / i1) + i2) * i3)}}[(0, 0)].0, TensorConstant{(1, 1, 1) of -1.0}, InplaceDimShuffle{0,1,x}.0, Elemwise{Add}[(0, 0)].0, InplaceDimShuffle{0,1,x}.0)]]
HINT: Re-running with most Theano optimization disabled could give you a back-trace of when this node was created. This can be done with by setting the Theano flag 'optimizer=fast_compile'. If that does not work, Theano optimizations can be disabled with 'optimizer=None'.
HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint and storage map footprint of this apply node.