tensorflow/models · error · ValueError

decoder_layer should be built.

Error message

decoder_layer should be built.

What it means

Error "decoder_layer should be built." thrown in tensorflow/models.

Source

Thrown at official/projects/nhnet/models.py:78

  # remove <s>
  targets = seq[:, 1:]
  # pad
  pad_ids = tf.ones([batch_size], tf.int32) * pad_token_id
  targets = tf.concat([targets, tf.expand_dims(pad_ids, axis=1)], axis=1)
  tf.assert_equal(tf.shape(targets), (batch_size, seq_len))
  return targets


class Bert2Bert(tf_keras.Model):
  """Bert2Bert encoder decoder model for training."""

  def __init__(self, params, bert_layer, decoder_layer, name=None):
    super(Bert2Bert, self).__init__(name=name)
    self.params = params
    if not bert_layer.built:
      raise ValueError("bert_layer should be built.")
    if not decoder_layer.built:
      raise ValueError("decoder_layer should be built.")
    self.bert_layer = bert_layer
    self.decoder_layer = decoder_layer

  def get_config(self):
    return {"params": self.params.as_dict()}

  def get_decode_logits(self,
                        decoder_inputs,
                        ids,
                        decoder_self_attention_bias,
                        step,
                        cache=None):
    if cache:
      if self.params.get("padded_decode", False):
        bias_shape = decoder_self_attention_bias.shape.as_list()
        self_attention_bias = tf.slice(
            decoder_self_attention_bias, [0, 0, step, 0],
            [bias_shape[0], bias_shape[1], 1, bias_shape[3]])

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/nhnet/models.py:78 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tensorflow/models@e006f5f0d5 (2026-08-24). Data as JSON: /api/errors/caaec78d74683dc9. Report an issue: GitHub.