传递给 `fit` 的模型只能将 `training` 和 `call` 中的第一个参数作为位置参数，发现

Question

我正在尝试遵循此代码，但在另一个数据集上：https://www.tensorflow.org/tutorials/text/transformer#encoder_layer 我需要编译和拟合模型。但是，我在运行时收到此错误；我不知道这是什么意思：

 Models passed to `fit` can only have `training` and the first argument in `call` as positional arguments, found: ['tar', 'enc_padding_mask', 'look_ahead_mask', 'dec_padding_mask'].

这是模型：

class Transformer(tf.keras.Model):
  def __init__(self, num_layers, d_model, num_heads, dff, input_vocab_size, 
               target_vocab_size, pe_input, pe_target, rate=0.1,**kwargs,):
    super(Transformer, self).__init__(**kwargs)

    self.encoder = Encoder(num_layers, d_model, num_heads, dff, 
                           input_vocab_size, pe_input, rate)

    self.decoder = Decoder(num_layers, d_model, num_heads, dff, 
                           target_vocab_size, pe_target, rate)

    self.final_layer = tf.keras.layers.Dense(target_vocab_size)
  def get_config(self):

        config = super().get_config().copy()
        config.update({
            'dff':self.dff,
            'input_vocab_size':self.input_vocab_size,
            'target_vocab_size':self.target_vocab_size,
            'pe_input':self.pe_input,
            'pe_target':self.pe_target,
            #'vocab_size': self.vocab_size,
            'num_layers': self.num_layers,
            #'units': self.units,
            'd_model': self.d_model,
            'num_heads': self.num_heads,
            'rate': self.rate,
        })
        return config

  def call(self, inp, tar, training, enc_padding_mask, 
           look_ahead_mask, dec_padding_mask):

    enc_output = self.encoder(inp, training, enc_padding_mask)  # (batch_size, inp_seq_len, d_model)

    # dec_output.shape == (batch_size, tar_seq_len, d_model)
    dec_output, attention_weights = self.decoder(
        tar, enc_output, training, look_ahead_mask, dec_padding_mask)

    final_output = self.final_layer(dec_output)  # (batch_size, tar_seq_len, target_vocab_size)
    #    return final_output, attention_weights


    return tf.keras.Model(inputs=[inputs, dec_inputs], outputs=outputs, name=name)

并创建模型，编译它，并按如下方式拟合：

transformer = Transformer(num_layers, d_model, num_heads, dff,
                          input_vocab_size, target_vocab_size, 
                          pe_input=input_vocab_size, 
                          pe_target=target_vocab_size,
                          rate=dropout_rate)

transformer.compile(optimizer=optimizer, loss=loss_function, metrics=[accuracy])

transformer.fit(dataset, epochs=EPOCHS)

编辑：基于@Geeocode 将转换器类中的 def 函数更新为：

def call(self, inp, tar, enc_padding_mask,look_ahead_mask, dec_padding_mask, training=False,):

    enc_output = self.encoder(inp, training, enc_padding_mask)  # (batch_size, inp_seq_len, d_model)

    # dec_output.shape == (batch_size, tar_seq_len, d_model)
    dec_output, attention_weights = self.decoder(
        tar, enc_output, training, look_ahead_mask, dec_padding_mask)

    final_output = self.final_layer(dec_output)  # (batch_size, tar_seq_len, target_vocab_size)
    return final_output, attention_weights

但是，我仍然遇到相同的错误

Answer 1

你得到错误的原因是因为self.call只需要两个变量 aninput和一个training标志。如果您有多个输入变量，它们将作为元组传递。因此，您可以拥有类似于以下内容的函数定义：

def call(self, input, training):
  inp, tar, enc_padding_mask,look_ahead_mask, dec_padding_mask = input
  ...

Answer 2

这可能是文档中的一个错误，正如我们在 github 上看到的那样，如果我们想要继承，那么除了方法的定义之外，tf.keras.Model我们只能有一个参数，这就是错误消息抱怨的内容：traininginputscall()

如果您对 进行子类化Model，则可以选择training在 中包含一个参数（布尔值）call，您可以使用它来指定训练和推理中的不同行为

  """
  If you subclass `Model`, you can optionally have
  a `training` argument (boolean) in `call`, which you can use to specify
  a different behavior in training and inference:
  ```python
  import tensorflow as tf
  class MyModel(tf.keras.Model):
    def __init__(self):
      super(MyModel, self).__init__()
      self.dense1 = tf.keras.layers.Dense(4, activation=tf.nn.relu)
      self.dense2 = tf.keras.layers.Dense(5, activation=tf.nn.softmax)
      self.dropout = tf.keras.layers.Dropout(0.5)
    def call(self, inputs, training=False):
      x = self.dense1(inputs)
      if training:
        x = self.dropout(x, training=training)
      return self.dense2(x)
  model = MyModel()
  ```
  """    

我建议你在 github 上创建一个问题来引用它。