tensorflow/models · error · ValueError

The passed network's output length is %s, which is less than

Error message

The passed network's output length is %s, which is less than the requested num_token_predictions %s.

What it means

Error "The passed network's output length is %s, which is less than the requested num_token_predictions %s." thrown in tensorflow/models.

Source

Thrown at official/nlp/modeling/models/bert_pretrainer.py:90

    network_inputs = network.inputs
    inputs = copy.copy(network_inputs)

    # Because we have a copy of inputs to create this Model object, we can
    # invoke the Network object with its own input tensors to start the Model.
    # Note that, because of how deferred construction happens, we can't use
    # the copy of the list here - by the time the network is invoked, the list
    # object contains the additional input added below.
    sequence_output, cls_output = network(network_inputs)

    # The encoder network may get outputs from all layers.
    if isinstance(sequence_output, list):
      sequence_output = sequence_output[-1]
    if isinstance(cls_output, list):
      cls_output = cls_output[-1]
    sequence_output_length = sequence_output.shape.as_list()[1]
    if sequence_output_length is not None and (sequence_output_length <
                                               num_token_predictions):
      raise ValueError(
          "The passed network's output length is %s, which is less than the "
          'requested num_token_predictions %s.' %
          (sequence_output_length, num_token_predictions))

    masked_lm_positions = tf_keras.layers.Input(
        shape=(num_token_predictions,),
        name='masked_lm_positions',
        dtype=tf.int32)
    inputs.append(masked_lm_positions)

    if embedding_table is None:
      embedding_table = network.get_embedding_table()
    masked_lm = layers.MaskedLM(
        embedding_table=embedding_table,
        activation=activation,
        initializer=tf_utils.clone_initializer(initializer),
        output=output,
        name='cls/predictions')

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/nlp/modeling/models/bert_pretrainer.py:90 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/ef393fae1adc3e90. Report an issue: GitHub.