tensorflow/models · error · ValueError

only support num_test_clips = 1 for action localization task

Error message

only support num_test_clips = 1 for action localization task. 

What it means

Error "only support num_test_clips = 1 for action localization task. " thrown in tensorflow/models.

Source

Thrown at official/projects/videoglue/datasets/action_localization.py:105

      min_resize: Frames are resized so that `min(height, width)` is
        `min_resize`.
      crop_size: Final size of the frame after cropping the resized frames. Both
        height and width are the same.
      zero_centering_image: If `True`, frames are normalized to values in
        [-1, 1]. If `False`, values in [0, 1].
      color_augmentation: Whether to apply color augmentation on video clips.
      augmentation_type: The data augmentation style applied on images.
      augmentation_params: A dictionary of params for data augmentation.
      num_test_clips: Number of test clips (1 by default). If more than 1, this
        will sample multiple linearly spaced clips within each video at test
        time. If 1, then a single clip in the middle of the video is sampled.
        The clips are aggreagated in the batch dimension.
      one_hot_label: Whether to return one-hot label.
      merge_multi_labels: Whether to merge multi_labels.
      import_detected_bboxes: Whether to parse and return detected boxes.
    """
    if num_test_clips != 1:
      raise ValueError('only support num_test_clips = 1 for action '
                       'localization task. ')

    # Parse keyframe index.
    self.parser_builder.parse_feature(
        feature_type=tf.io.FixedLenFeature(1, dtype=tf.int64),
        feature_name=self._KEYFRAME_INDEX_KEY,
        output_name='keyframe_index',
        is_context=True)
    # Add keyframe boxes.
    for dim in ['ymin', 'xmin', 'ymax', 'xmax']:
      utils.add_context_box_dim(
          parser_builder=self.parser_builder,
          sampler_builder=self.sampler_builder,
          preprocessor_builder=self.preprocessor_builder,
          input_box_dim_name=f'{self._GT_PREFIX}/bbox/{dim}',
          output_box_dim_name=f'instances_{dim}',
          num_instances_per_frame=num_instance_per_frame,
          num_frames=num_frames)   # > 1 to duplicate keyframe boxes to all.

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/projects/videoglue/datasets/action_localization.py:105 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/7031a87a23806f1f. Report an issue: GitHub.