tensorflow/models · error · ValueError

context_level should be specified as odd number.

Error message

context_level should be specified as odd number.

What it means

Error "context_level should be specified as odd number." thrown in tensorflow/models.

Source

Thrown at official/projects/const_cl/modeling/heads/instance_reconstructor.py:117

              4, use_bias=True, activation='relu', name='temporal_mlp_l1'),
          tf_keras.layers.Dense(
              8, use_bias=True, name='temporal_mlp_l2')]

    self._attention_decoder = transformer_decoder.TransformerDecoder(
        num_channels=num_tx_channels,
        num_layers=num_tx_layers,
        num_heads=num_tx_heads,
        use_bias=use_bias,
        activation=activation,
        dropout_rate=dropout_rate,
        layer_norm_epsilon=layer_norm_epsilon)

    self._projection_layer = tf_keras.layers.Dense(num_output_channels)

  def _get_memory_embeddings(self, inputs: tf.Tensor) -> tf.Tensor:
    """Uniformly samples frames to construct memory embeddings."""
    if self._context_level % 2 == 0:
      raise ValueError('context_level should be specified as odd number.')

    num_frames = tf.shape(inputs)[1]
    keyframe_index = num_frames // 2
    stride = num_frames // self._context_level
    start = self._context_level // 2 * -1
    stop = self._context_level // 2 + 1  # exclusive

    memories = []
    for idx in range(start, stop):
      idx = idx * stride + keyframe_index
      memories.append(inputs[:, idx, ...])

    memories = tf.stack(memories, axis=1)
    return memories

  def _add_positional_embedding(self, inputs: tf.Tensor) -> tf.Tensor:
    """Adds positional embeddings to the inputs tensor."""
    # Compute the locations using meshgrid.

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/const_cl/modeling/heads/instance_reconstructor.py:117 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/a6ee189300ff3dcd. Report an issue: GitHub.