tensorflow/models · error · ValueError

Looping until exhausted is not supported if `options.use_tf_

Error message

Looping until exhausted is not supported if `options.use_tf_while_loop` is `True`

What it means

Error "Looping until exhausted is not supported if `options.use_tf_while_loop` is `True`" thrown in tensorflow/models.

Source

Thrown at orbit/standard_runner.py:325

    return loop_fn

  def evaluate(self, num_steps: tf.Tensor) -> Optional[runner.Output]:
    """Implements `num_steps` steps of evaluation.

    Args:
      num_steps: The number of evaluation steps to run. When this is -1,
        evaluation proceeds until a call to `eval_step` raises a `StopIteration`
        or `tf.errors.OutOfRangeError`.

    Returns:
      The output of `self.eval_end()`.

    Raises:
      ValueError: If `options.use_tf_while_loop` is `True` and `num_steps` is
        unspecified.
    """
    if self._eval_options.use_tf_while_loop and num_steps == -1:
      raise ValueError("Looping until exhausted is not supported if "
                       "`options.use_tf_while_loop` is `True`")

    outputs = self.eval_begin()  # pylint: disable=assignment-from-no-return

    has_state = outputs is not None
    if self._eval_loop_fn is None:
      self._eval_loop_fn = self.create_eval_loop_fn(has_state)

    # If `recreate_iterator_for_each_eval` is `True`, `self._eval_iter` is
    # always None.
    if self._eval_iter is None:
      eval_iter = tf.nest.map_structure(iter, self.eval_dataset)
      if not self._eval_options.recreate_iterator_for_each_eval:
        self._eval_iter = eval_iter
    else:
      eval_iter = self._eval_iter

    if self._eval_options.use_tf_while_loop and not has_state:

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Pass a finite steps count instead of -1 when use_tf_while_loop is True.
2. Disable use_tf_while_loop to loop until the dataset is exhausted.

When it happens

Trigger: Thrown at orbit/standard_runner.py:325 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/e23cfab1b9ae9fde. Report an issue: GitHub.