tensorflow/models · error · ValueError

the structure of `anchor_sizes` must be a tuple, list, or di

Error message

the structure of `anchor_sizes` must be a tuple, list, or dict, given {}

What it means

Error "the structure of `anchor_sizes` must be a tuple, list, or dict, given {}" thrown in tensorflow/models.

Source

Thrown at official/vision/ops/anchor_generator.py:179

        self.anchor_generators.append(
            _SingleAnchorGenerator(anchor_size, scale_list, ar_list, stride,
                                   clip_boxes))

  def __call__(self, image_size):
    anchor_generators = tf.nest.flatten(self.anchor_generators)
    results = [anchor_gen(image_size) for anchor_gen in anchor_generators]
    return tf.nest.pack_sequence_as(self.anchor_generators, results)


def maybe_map_structure_for_anchor(params, anchor_sizes):
  """broadcast the params to match anchor_sizes."""
  if all(isinstance(param, (int, float)) for param in params):
    if isinstance(anchor_sizes, (tuple, list)):
      return [params] * len(anchor_sizes)
    elif isinstance(anchor_sizes, dict):
      return tf.nest.map_structure(lambda _: params, anchor_sizes)
    else:
      raise ValueError("the structure of `anchor_sizes` must be a tuple, "
                       "list, or dict, given {}".format(anchor_sizes))
  else:
    return params

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Pass anchor_sizes as a tuple, list, or dict of level -> sizes.
2. Wrap a single size in a list or map it per feature level in a dict.

When it happens

Trigger: Thrown at official/vision/ops/anchor_generator.py:179 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/da9c7f16f6b71d58. Report an issue: GitHub.