Блог

Я пытаюсь создать однонаправленный RNN для моделирования последовательности и нейронного встраивания. Я создал пользовательский класс для экспериментов с различными архитектурами и параметрами. Класс диспетчера данных в коде — это еще один класс, который в основном считывает текстовые данные, обрабатывает их и преобразует в числовые векторы. tf_train_set — это tensorSliceDataset, содержащий числовые векторы и метки 60% набора данных. остальные 40% находятся в tf_valid_set .
У меня есть следующий код для моего RNN:

 class UniRNN:
    def __init__(self, cell_type= 'gru', embed_size= 128, state_sizes= [128, 64], data_manager= None):
        self.cell_type = cell_type
        self.state_sizes = state_sizes
        self.embed_size = embed_size
        self.data_manager = data_manager
        self.vocab_size = self.data_manager.vocab_size  1 
        
    #return the correspoding memory cell
    @staticmethod
    def get_layer(cell_type= 'gru', state_size= 128, return_sequences= False, activation = 'tanh'):
        if cell_type=='gru':
            return tf.keras.layers.GRU(state_size, return_sequences=return_sequences, activation=activation)
        elif cell_type== 'lstm':
            return tf.keras.layers.LSTM(state_size, return_sequences=return_sequences, activation=activation)
        else:
            return tf.keras.layers.SimpleRNN(state_size, return_sequences=return_sequences, activation=activation)
    
    def build(self):
        x = tf.keras.layers.Input(shape=[None])
        h = tf.keras.layers.Embedding(self.vocab_size, self.embed_size, mask_zero=True, trainable=True)(x)
        num_layers = len(self.state_sizes)
        for i in range(num_layers):
            h = self.get_layer(self.cell_type, self.state_sizes[i], return_sequences=True)(h)
        h = tf.keras.layers.Dense(dm.num_classes, activation='softmax')(h)
        self.model = tf.keras.Model(inputs=x, outputs=h)
    
    def compile_model(self, *args, **kwargs):
        self.model.compile(*args, **kwargs)
    
    def fit(self, *args, **kwargs):
        return self.model.fit(*args, **kwargs)
    
    def evaluate(self, *args, **kwargs):
        self.model.evaluate(*args, **kwargs)
  

Чтобы соответствовать модели, мой код:

 uni_rnn = UniRNN(cell_type='basic_rnn', embed_size=128, state_sizes=[128, 128], data_manager=dm) #Insert your code here
uni_rnn.build()
# uni_rnn.model.summary()
opt = tf.keras.optimizers.RMSprop(learning_rate=0.001)
uni_rnn.compile_model(optimizer=opt, loss='sparse_categorical_crossentropy', metrics=['accuracy'])
uni_rnn.fit(dm.tf_train_set.batch(64), epochs=20, validation_data = dm.tf_valid_set.batch(64))
  

когда я запускаю этот код, я получаю следующую ошибку:

 ---------------------------------------------------------------------------
InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-184-abef9ae0cbcd> in <module>
      3 opt = tf.keras.optimizers.RMSprop(learning_rate=0.001)
      4 uni_rnn.compile_model(optimizer=opt, loss='sparse_categorical_crossentropy', metrics=['accuracy'])
----> 5 uni_rnn.fit(dm.tf_train_set.batch(64), epochs=20, validation_data = dm.tf_valid_set.batch(64))

<ipython-input-170-53f4c12769ab> in fit(self, *args, **kwargs)
     31 
     32     def fit(self, *args, **kwargs):
---> 33         return self.model.fit(*args, **kwargs)
     34 
     35     def evaluate(self, *args, **kwargs):

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythonkerasenginetraining.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, validation_freq, max_queue_size, workers, use_multiprocessing, **kwargs)
    817         max_queue_size=max_queue_size,
    818         workers=workers,
--> 819         use_multiprocessing=use_multiprocessing)
    820 
    821   def evaluate(self,

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythonkerasenginetraining_v2.py in fit(self, model, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_freq, max_queue_size, workers, use_multiprocessing, **kwargs)
    340                 mode=ModeKeys.TRAIN,
    341                 training_context=training_context,
--> 342                 total_epochs=epochs)
    343             cbks.make_logs(model, epoch_logs, training_result, ModeKeys.TRAIN)
    344 

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythonkerasenginetraining_v2.py in run_one_epoch(model, iterator, execution_function, dataset_size, batch_size, strategy, steps_per_epoch, num_samples, mode, training_context, total_epochs)
    126         step=step, mode=mode, size=current_batch_size) as batch_logs:
    127       try:
--> 128         batch_outs = execution_function(iterator)
    129       except (StopIteration, errors.OutOfRangeError):
    130         # TODO(kaftan): File bug about tf function and errors.OutOfRangeError?

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythonkerasenginetraining_v2_utils.py in execution_function(input_fn)
     96     # `numpy` translates Tensors to values in Eager mode.
     97     return nest.map_structure(_non_none_constant_value,
---> 98                               distributed_function(input_fn))
     99 
    100   return execution_function

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerdef_function.py in __call__(self, *args, **kwds)
    566         xla_context.Exit()
    567     else:
--> 568       result = self._call(*args, **kwds)
    569 
    570     if tracing_count == self._get_tracing_count():

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerdef_function.py in _call(self, *args, **kwds)
    630         # Lifting succeeded, so variables are initialized and we can run the
    631         # stateless function.
--> 632         return self._stateless_fn(*args, **kwds)
    633     else:
    634       canon_args, canon_kwds = 

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerfunction.py in __call__(self, *args, **kwargs)
   2361     with self._lock:
   2362       graph_function, args, kwargs = self._maybe_define_function(args, kwargs)
-> 2363     return graph_function._filtered_call(args, kwargs)  # pylint: disable=protected-access
   2364 
   2365   @property

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerfunction.py in _filtered_call(self, args, kwargs)
   1609          if isinstance(t, (ops.Tensor,
   1610                            resource_variable_ops.BaseResourceVariable))),
-> 1611         self.captured_inputs)
   1612 
   1613   def _call_flat(self, args, captured_inputs, cancellation_manager=None):

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerfunction.py in _call_flat(self, args, captured_inputs, cancellation_manager)
   1690       # No tape is watching; skip to running the function.
   1691       return self._build_call_outputs(self._inference_function.call(
-> 1692           ctx, args, cancellation_manager=cancellation_manager))
   1693     forward_backward = self._select_forward_and_backward_functions(
   1694         args,

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerfunction.py in call(self, ctx, args, cancellation_manager)
    543               inputs=args,
    544               attrs=("executor_type", executor_type, "config_proto", config),
--> 545               ctx=ctx)
    546         else:
    547           outputs = execute.execute_with_cancellation(

~anaconda3envsmyenv_tf21_p37libsite-packagestensorflow_corepythoneagerexecute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     65     else:
     66       message = e.message
---> 67     six.raise_from(core._status_to_exception(e.code, message), None)
     68   except TypeError as e:
     69     keras_symbolic_tensors = [

~anaconda3envsmyenv_tf21_p37libsite-packagessix.py in raise_from(value, from_value)

InvalidArgumentError:  assertion failed: [Condition x == y did not hold element-wise:] [x (loss/dense_13_loss/SparseSoftmaxCrossEntropyWithLogits/Shape_1:0) = ] [64 1] [y (loss/dense_13_loss/SparseSoftmaxCrossEntropyWithLogits/strided_slice:0) = ] [64 100]
     [[node loss/dense_13_loss/SparseSoftmaxCrossEntropyWithLogits/assert_equal_1/Assert/Assert (defined at <ipython-input-170-53f4c12769ab>:33) ]] [Op:__inference_distributed_function_111597]

Function call stack:
distributed_function
  

Может кто-нибудь, пожалуйста, объяснить, в чем проблема? и, может быть, как я могу это исправить?

последний RNN в цикле должен иметь return_sequence = False . для этого вы можете просто сделать:

 def build(self):
    x = tf.keras.layers.Input(shape=[None])
    h = tf.keras.layers.Embedding(self.vocab_size, self.embed_size, 
                                  mask_zero=True, trainable=True)(x)
    num_layers = len(self.state_sizes)
    for i in range(num_layers-1):
        h = self.get_layer(self.cell_type, self.state_sizes[i], return_sequences=True)(h)
    h = self.get_layer(self.cell_type, self.state_sizes[i], return_sequences=False)(h)
    h = tf.keras.layers.Dense(dm.num_classes, activation='softmax')(h)
    self.model = tf.keras.Model(inputs=x, outputs=h)