sgl-project/sglang · error · ValueError

Step3-VL image item has num_patches > 0 but no patch_pixel_v

Error message

Step3-VL image item has num_patches > 0 but no patch_pixel_values.

What it means

When an image item declares num_patches > 0 for a given image, Step3-VL expects per-patch features (item_patch_features) to splice in; if none were provided the feature vector cannot be built and it raises ValueError.

Source

Thrown at python/sglang/srt/models/step3_vl.py:885

        patch_offset = 0
        for thumb_count, num_patches_list, patch_count in item_metadata:
            item_thumb_features = all_thumb_features[
                thumb_offset : thumb_offset + thumb_count
            ]
            thumb_offset += thumb_count
            item_patch_features = (
                all_patch_features[patch_offset : patch_offset + patch_count]
                if patch_count > 0
                else None
            )
            patch_offset += patch_count

            cur_patch_idx = 0
            for i, num_patch in enumerate(num_patches_list):
                cur_feature = []
                if num_patch > 0:
                    if item_patch_features is None:
                        raise ValueError(
                            "Step3-VL image item has num_patches > 0 but no patch_pixel_values."
                        )
                    patch_slice = item_patch_features[
                        cur_patch_idx : cur_patch_idx + num_patch
                    ]
                    cur_feature.append(patch_slice.view(-1, patch_slice.shape[-1]))
                cur_feature.append(
                    item_thumb_features[i].view(-1, item_thumb_features.shape[-1])
                )
                cur_patch_idx += num_patch
                merged_image_features.append(
                    torch.cat(cur_feature) if len(cur_feature) > 1 else cur_feature[0]
                )
        return self._flatten_embeddings(merged_image_features)

    def pad_input_ids(self, input_ids: List[int], mm_inputs: MultimodalInputs):
        pattern = MultiModalityDataPaddingPatternMultimodalTokens()
        return pattern.pad_input_tokens(input_ids, mm_inputs)

View on GitHub (pinned to 0132848349)

Solutions

1. Ensure preprocessing produces patch_pixel_values whenever num_patches > 0 and they are attached to the item
2. If patches were intentionally removed, set num_patches entries to 0
3. Align processor and model versions from the same sglang release

Example fix

# before
item.model_specific_data = {"num_patches": [2, 1]}
# after
item.model_specific_data = {"num_patches": [2, 1], "patch_pixel_values": patches}

for n, p in zip(num_patches_list, patches_per_item):
    assert p is not None or n == 0, f"num_patches {n} but no patch features"

When it happens

Trigger: get_image_feature processing an item whose num_patches_list has positive entries while patch_pixel_values / patch features are absent (None) at python/sglang/srt/models/step3_vl.py:885.

Common situations: Dropping the patch branch of preprocessing (only global features kept) while num_patches still says patches exist; version mismatch where the processor emits patch counts but not patch pixels; manually constructed inputs.

Related errors

• Step3-VL image item is missing num_patches.
• Step3-VL image item has num_patches > 0 but no patch_pixel_v
• Expected CHW image tensor, got shape {shape}
• Expected CHW image tensor with 1 or 3 channels, got shape {s
• Unsupported image type: {type}

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/a526cdf471bdb991. Report an issue: GitHub.