tensorflow/models · error · ValueError

Batch size cannot be more than 1.

Error message

Batch size cannot be more than 1.

What it means

Error "Batch size cannot be more than 1." thrown in tensorflow/models.

Source

Thrown at official/vision/serving/semantic_segmentation.py:95

            dtype=tf.float32,
        )
        image_info_spec = tf.TensorSpec(shape=[4, 2], dtype=tf.float32)

        images, image_info = tf.nest.map_structure(
            tf.identity,
            tf.map_fn(
                self._build_inputs,
                elems=images,
                fn_output_signature=(images_spec, image_info_spec),
                parallel_iterations=32))

    outputs = self.inference_step(images)

    # Optionally resize prediction to the input image size.
    if self.params.task.export_config.rescale_output:
      logits = outputs['logits']
      if logits.shape[0] != 1:
        raise ValueError('Batch size cannot be more than 1.')

      image_shape = tf.cast(image_info[0, 0, :], tf.int32)  # pyrefly: ignore[unsupported-operation]
      if self.params.task.train_data.preserve_aspect_ratio:
        rescale_size = tf.cast(
            tf.math.ceil(image_info[0, 1, :] / image_info[0, 2, :]), tf.int32)  # pyrefly: ignore[unsupported-operation]
        offsets = tf.cast(image_info[0, 3, :], tf.int32)  # pyrefly: ignore[unsupported-operation]
        logits = tf.image.resize(logits, rescale_size, method='bilinear')
        outputs['logits'] = tf.image.crop_to_bounding_box(
            logits, offsets[0], offsets[1], image_shape[0], image_shape[1])
      else:
        outputs['logits'] = tf.image.resize(
            logits, [image_shape[0], image_shape[1]], method='bilinear')
    else:
      outputs['logits'] = tf.image.resize(
          outputs['logits'], self._input_image_size, method='bilinear')

    if image_info is not None:
      outputs.update({'image_info': image_info})

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Export with batch size 1 for the semantic segmentation serving module.
2. Reshape the serving input signature to a single image per request.

When it happens

Trigger: Thrown at official/vision/serving/semantic_segmentation.py:95 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/3baa043c79d01b21. Report an issue: GitHub.