tensorflow/models · error · ValueError

TransformerLayer expects a three-dimensional input of shape

Error message

TransformerLayer expects a three-dimensional input of shape [batch, sequence, width].

What it means

Error "TransformerLayer expects a three-dimensional input of shape [batch, sequence, width]." thrown in tensorflow/models.

Source

Thrown at official/projects/detr/modeling/transformer.py:667

    self._activity_regularizer = tf_keras.regularizers.get(activity_regularizer)
    self._kernel_constraint = tf_keras.constraints.get(kernel_constraint)
    self._bias_constraint = tf_keras.constraints.get(bias_constraint)
    self._use_bias = use_bias
    self._norm_first = norm_first
    self._norm_epsilon = norm_epsilon
    self._intermediate_dropout = intermediate_dropout
    if attention_initializer:
      self._attention_initializer = tf_keras.initializers.get(
          attention_initializer)
    else:
      self._attention_initializer = tf_utils.clone_initializer(
          self._kernel_initializer)
    self._cross_attention_cls = layers.attention.MultiHeadAttention

  def build(self, input_shape):
    target_tensor_shape = tf.TensorShape(input_shape[0])
    if len(target_tensor_shape.as_list()) != 3:
      raise ValueError("TransformerLayer expects a three-dimensional input of "
                       "shape [batch, sequence, width].")
    hidden_size = target_tensor_shape[2]
    if hidden_size % self.num_attention_heads != 0:
      raise ValueError(
          "The hidden size (%d) is not a multiple of the number of attention "
          "heads (%d)" % (hidden_size, self.num_attention_heads))
    self.attention_head_size = int(hidden_size) // self.num_attention_heads
    common_kwargs = dict(
        bias_initializer=self._bias_initializer,
        kernel_regularizer=self._kernel_regularizer,
        bias_regularizer=self._bias_regularizer,
        activity_regularizer=self._activity_regularizer,
        kernel_constraint=self._kernel_constraint,
        bias_constraint=self._bias_constraint)
    # Self attention.
    self.self_attention = layers.attention.CachedAttention(
        num_heads=self.num_attention_heads,
        key_dim=self.attention_head_size,

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/detr/modeling/transformer.py:667 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/791b918dababd0df. Report an issue: GitHub.