我尝试通过以下代码吸引人们的注意，我试图推断出我创建的模型，它具有我想要的形状和值，并且可以与model.summary（）完美配合。但是，它不会通过CTC损失计算。有人知道我的自定义图层有什么问题吗？ def compute_atten（data，atten_w）： input_vector = data.numpy（） atten_vectors = atten_w.numpy（） all_batch = [] 对于索引，枚举（输入向量）中的one_batch： tmp_w = atten_vectors [index] all_vector = [] 对于j，enumerate（one_batch）中的向量： tmp = np.zeros（input_vector.shape [2]） 对于tmp_w [j]中的w： tmp + =向量* w all_vector.append（tmp） all_batch.append（all_vector） all_batch = np.asarray（all_batch，dtype = np.float32） 返回all_batch
class AttentionLayer(tf.keras.layers.Layer):
def __init__(self,**kwargs):
super(AttentionLayer, self).__init__(**kwargs)
def build(self,input_shape):
super().build(input_shape)
self.built = True
def call(self,inputs):
output_shape = self.compute_output_shape(tf.shape(inputs))
transpose_input = tf.transpose(inputs,perm=[0,2,1])
atten_w = K.backend.batch_dot(inputs,transpose_input)
atten_w = tf.linalg.set_diag(atten_w,tf.zeros(tf.shape(atten_w)[0:-1],dtype=tf.float32))
atten_w = tf.nn.softmax(atten_w,axis=1)
atten_v = tf.py_function(calculate_atten,inp=[inputs,atten_w],Tout=[tf.float32])
atten_v = K.backend.reshape(tf.convert_to_tensor(atten_v),output_shape)
return atten_v
def compute_output_shape(self, input_shape):
return input_shape
2019-06-20 10:42:14.753458: W tensorflow/core/framework/op_kernel.cc:1431] OP_REQUIRES failed at transpose_op.cc:157 : Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 2019-06-20 10:42:14.753536: W
tensorflow/core/common_runtime/base_collective_executor.cc:214] BaseCollectiveExecutor::StartAbort Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] [[replica_1/StatefulPartitionedCall/ExpandDims/_240]] 2019-06-20 10:42:14.753980: E
tensorflow/core/common_runtime/process_function_library_runtime.cc:764] Component function execution failed: Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] [[replica_1/StatefulPartitionedCall/ExpandDims/_240]] 2019-06-20 10:42:14.754115: W
tensorflow/core/common_runtime/base_collective_executor.cc:214] BaseCollectiveExecutor::StartAbort Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] 2019-06-20 10:42:14.754457: E
tensorflow/core/common_runtime/process_function_library_runtime.cc:764] Component function execution failed: Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] 2019-06-20 10:42:14.765733: W
tensorflow/core/common_runtime/base_collective_executor.cc:214] BaseCollectiveExecutor::StartAbort Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] [[replica_1/StatefulPartitionedCall/CTCLoss/_237]] 2019-06-20 10:42:14.765808: E
tensorflow/core/common_runtime/process_function_library_runtime.cc:764] Component function execution failed: Invalid argument: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] [[replica_1/StatefulPartitionedCall/CTCLoss/_237]] Traceback (most recent call last): File "ctc_model4.py", line 1060, in <module> train_ops = ditributred_train(train_iterator) File
"/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 438, in __call__ return self._stateless_fn(*args, **kwds) File "/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 1288, in __call__ return graph_function._filtered_call(args, kwargs) # pylint: disable=protected-access File
"/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 574, in _filtered_call (t for t in nest.flatten((args, kwargs)) File
"/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 627, in _call_flat outputs = self._inference_function.call(ctx, args) File
"/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 415, in call ctx=ctx) File
"/misc/home/usr16/cheesiang_leow/.virtualenvs/tensorflow-2.0/lib/python3.6/site-packages/tensorflow/python/eager/execute.py", line 66, in quick_execute six.raise_from(core._status_to_exception(e.code, message), None) File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: transpose expects a vector of size 0. But input(1) is a vector of size 3 [[{{node transpose}}]] [[replica_1/StatefulPartitionedCall/ExpandDims/_240]] [Op:__inference_ditributred_train_24926]