tensorflow/models · error · ValueError

Fusion type {} not supported.

Error message

Fusion type {} not supported.

What it means

Error "Fusion type {} not supported." thrown in tensorflow/models.

Source

Thrown at official/vision/modeling/decoders/fpn.py:145

    feats = {str(backbone_max_level): feats_lateral[str(backbone_max_level)]}
    for level in range(backbone_max_level - 1, min_level - 1, -1):
      feat_a = spatial_transform_ops.nearest_upsampling(
          feats[str(level + 1)], 2, use_keras_layer=use_keras_layer)
      feat_b = feats_lateral[str(level)]

      if fusion_type == 'sum':
        if use_keras_layer:
          feats[str(level)] = tf_keras.layers.Add()([feat_a, feat_b])
        else:
          feats[str(level)] = feat_a + feat_b
      elif fusion_type == 'concat':
        if use_keras_layer:
          feats[str(level)] = tf_keras.layers.Concatenate(axis=-1)(
              [feat_a, feat_b])
        else:
          feats[str(level)] = tf.concat([feat_a, feat_b], axis=-1)
      else:
        raise ValueError('Fusion type {} not supported.'.format(fusion_type))

    # TODO(fyangf): experiment with removing bias in conv2d.
    # Build post-hoc 3x3 convolution kernel.
    for level in range(min_level, backbone_max_level + 1):
      feats[str(level)] = conv2d(
          filters=num_filters,
          strides=1,
          kernel_size=3,
          padding='same',
          kernel_initializer=kernel_initializer,
          kernel_regularizer=kernel_regularizer,
          bias_regularizer=bias_regularizer,
          name=f'post_hoc_{level}')(
              feats[str(level)])

    # TODO(fyangf): experiment with removing bias in conv2d.
    # Build coarser FPN levels introduced for RetinaNet.
    for level in range(backbone_max_level + 1, max_level + 1):

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Use fusion_type 'sum' or 'concat' in the FPN decoder config.
2. Check the fusion type value for typos.

When it happens

Trigger: Thrown at official/vision/modeling/decoders/fpn.py:145 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/a175613ee8fbcd0b. Report an issue: GitHub.