tensorflow/models · error · ValueError

More than one maxpool should be specified in SPP block

Error message

More than one maxpool should be specified in SPP block

What it means

Truthiness guard in SPP.__init__: the sizes argument (max-pooling kernel sizes for the Spatial Pyramid Pooling branches) is empty. An SPP layer with zero pooling branches is degenerate — it produces no multi-scale context — so construction is rejected. The message wording is itself buggy ('More than one' though the check requires at least one), but the remedy is the same: supply at least one pool size.

Solutions

1. Provide a non-empty list of pool sizes, e.g. SPP(sizes=[1, 5, 9, 13])
2. Check config plumbing if sizes arrives empty from YAML (a missing key defaulting to [])
3. Omit the SPP layer entirely if no pooling branches are wanted

When it happens

Trigger: Thrown at official/projects/yolo/modeling/layers/nn_blocks.py:1136 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/2a9725a9ebd54ad1. Report an issue: GitHub.

Appendix: source

Thrown at official/projects/yolo/modeling/layers/nn_blocks.py:1136

  def call(self, inputs, training=None):
    # done this way to prevent confusion in the auto graph
    if self._inverted:
      return self._call_reversed(inputs, training=training)
    else:
      return self._call_regular(inputs, training=training)


class SPP(tf_keras.layers.Layer):
  """Spatial Pyramid Pooling.

  A non-agregated SPP layer that uses Pooling.
  """

  def __init__(self, sizes, **kwargs):
    self._sizes = list(reversed(sizes))
    if not sizes:
      raise ValueError('More than one maxpool should be specified in SPP block')
    super().__init__(**kwargs)

  def build(self, input_shape):
    maxpools = []
    for size in self._sizes:
      maxpools.append(
          tf_keras.layers.MaxPool2D(
              pool_size=(size, size),
              strides=(1, 1),
              padding='same',
              data_format=None))
    self._maxpools = maxpools
    super().build(input_shape)

  def call(self, inputs, training=None):
    outputs = []
    for maxpool in self._maxpools:
      outputs.append(maxpool(inputs))

View on GitHub (pinned to e006f5f0d5)