sgl-project/sglang · error · ValueError

Image placeholder count does not match image_data

Error message

Image placeholder count does not match image_data

What it means

Raised by _merge_video_media when a rendered image marker is found in the prompt but the iterator over natively supplied image_data is exhausted. The number of image placeholders produced by video rendering must match the count of provided images.

Source

Thrown at python/sglang/srt/multimodal/processors/dots_note_omni.py:335

        pattern = re.compile(
            f"({native_pattern.pattern}|{_EXPANDED_VIDEO_MEDIA_RE.pattern})"
        )
        rendered = []
        last = 0

        for match in pattern.finditer(input_text):
            rendered.append(input_text[last : match.start()])
            marker = match.group(0)
            expanded_media = video_media.get(marker)
            if expanded_media is not None:
                modality, value = expanded_media
            else:
                modality = self.mm_tokens.get_modality_of_token(marker)
                if modality == Modality.IMAGE:
                    try:
                        value = next(native_images)
                    except StopIteration as exc:
                        raise ValueError(
                            "Image placeholder count does not match image_data"
                        ) from exc
                elif modality == Modality.AUDIO:
                    try:
                        value = next(native_audios)
                    except StopIteration as exc:
                        raise ValueError(
                            "Audio placeholder count does not match audio_data"
                        ) from exc
                else:
                    raise ValueError(f"Unsupported dots omni media marker: {marker}")

            if modality == Modality.IMAGE:
                ordered_images.append(value)
                rendered.append(
                    self.image_start_token + self.image_token + self.image_end_token
                )
            else:

View on GitHub (pinned to 0132848349)

Solutions

1. Align the number of image placeholders in the prompt/video content with len(image_data)
2. Pass keyframe images explicitly in image_data when your template references them
3. Update sglang — renderer/marker logic is covered by tests and may already be fixed for your template

Example fix

// before
image_data = []  # template has 2 image markers
// after
image_data = [frame1, frame2]  # one per marker

img_markers = prompt.count(IMAGE_TOKEN) + rendered_video_image_markers
assert img_markers == len(image_data), f'{img_markers} markers vs {len(image_data)} images'

try:
    await processor.process_mm_data_async(...)
except ValueError as e:
    if 'placeholder count does not match' in str(e):
        # recount markers vs media and rebuild the request
        ...

When it happens

Trigger: A video's preprocessed content emits more image markers than there are entries in the request's image_data list — e.g. the prompt/template adds image placeholders while image_data is empty or shorter.

Common situations: Chat template inserts extra image tokens for the video keyframes, but the client sent no image_data; or counts drift after editing the prompt. The bundled tests exercise this exact mismatch path.

Related errors

• Audio placeholder count does not match audio_data
• Video placeholder count does not match video_data: {len(vide
• An exception occurred while loading multimodal data: {e}
• prompt has {num_placeholders} image placeholder token(s) but
• processor image placeholder count mismatch: processor={proce

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/2806fce54aeb168e. Report an issue: GitHub.