tensorflow/models · error · ValueError

steps_per_loop should be positive integer.

Error message

steps_per_loop should be positive integer.

What it means

Error "steps_per_loop should be positive integer." thrown in tensorflow/models.

Source

Thrown at official/legacy/bert/model_training_utils.py:83

  return iterator


def _float_metric_value(metric):
  """Gets the value of a float-value keras metric."""
  return metric.result().numpy().astype(float)


def clip_by_global_norm_callback(grads_and_vars):
  """Performs gradient clipping."""
  grads, variables = zip(*grads_and_vars)
  (clipped_grads, _) = tf.clip_by_global_norm(grads, clip_norm=1.0)
  return zip(clipped_grads, variables)


def steps_to_run(current_step, steps_per_epoch, steps_per_loop):
  """Calculates steps to run on device."""
  if steps_per_loop <= 0:
    raise ValueError('steps_per_loop should be positive integer.')
  if steps_per_loop == 1:
    return steps_per_loop
  remainder_in_epoch = current_step % steps_per_epoch
  if remainder_in_epoch != 0:
    return min(steps_per_epoch - remainder_in_epoch, steps_per_loop)
  else:
    return steps_per_loop


def write_txt_summary(training_summary, summary_dir):
  """Writes a summary text file to record stats."""
  if not tf.io.gfile.exists(summary_dir):
    tf.io.gfile.mkdir(summary_dir)
  summary_path = os.path.join(summary_dir, _SUMMARY_TXT)
  with tf.io.gfile.GFile(summary_path, 'wb') as f:
    logging.info('Training Summary: \n%s', str(training_summary))
    f.write(json.dumps(training_summary, indent=4))

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/legacy/bert/model_training_utils.py:83 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/a0f67b975c6cc1e1. Report an issue: GitHub.