sgl-project/sglang · error · ValueError

{key}.{field} is required

Error message

{key}.{field} is required

What it means

forward() reads fields from per-modality position/psp structures via _psp_field, which requires the field to exist (dict key or object attribute). The error message names both the modality key and the missing field, e.g. 'audio.psp_lengths is required'.

Source

Thrown at python/sglang/multimodal_gen/runtime/models/dits/minimax_h3.py:2121

    @staticmethod
    def _pos_ids(pos_info: Any, key: str) -> torch.Tensor:
        if isinstance(pos_info, dict):
            ids = pos_info.get("position_ids")
        else:
            ids = getattr(pos_info, "position_ids", None)
        if ids is None:
            raise ValueError(f"{key}.position_ids is required")
        return ids.view(-1).to(torch.long)

    @staticmethod
    def _psp_field(psp: Any, key: str, field: str) -> Any:
        if isinstance(psp, dict):
            value = psp.get(field)
        else:
            value = getattr(psp, field, None)
        if value is None:
            raise ValueError(f"{key}.{field} is required")
        return value

    @staticmethod
    def _psp_optional_field(psp: Any, field: str) -> Any:
        if isinstance(psp, dict):
            return psp.get(field)
        return getattr(psp, field, None)

    def refine_prompt_embeds(
        self,
        prompt_embeds: torch.Tensor,
        refiner_cu_seqlens: torch.Tensor,
        *,
        device: torch.device,
    ) -> torch.Tensor:
        """Project and refine request-static text conditioning once."""
        self.materialize_mps_non_layer_weights(
            "condition_proj", "token_refiner.final_norm"

View on GitHub (pinned to 0132848349)

Solutions

1. Inspect the error's {key}.{field} and add that field with the correct tensor to the input structure
2. Cross-check against the model's input dataclass / docs for the full required field list per modality
3. Write a small pre-flight validator that walks required fields before calling forward

Example fix

# before
psp = {"video": {"psp_lengths": lens}}  # missing companion field
# after
psp = {"video": {"psp_lengths": lens, REQUIRED_FIELD: vals}}  # per error message

REQUIRED = ("field_a", "field_b")  # per error message
for key, psp in psp_inputs.items():
    for f in REQUIRED:
        v = psp.get(f) if isinstance(psp, dict) else getattr(psp, f, None)
        assert v is not None, f"{key}.{f}"

def psp_complete(psp: Any, fields: tuple) -> bool:
    for f in fields:
        v = psp.get(f) if isinstance(psp, dict) else getattr(psp, f, None)
        if v is None:
            return False
    return True

When it happens

Trigger: Passing a psp dict like {"audio": {"psp_lengths": x}} where 'audio.field_name' is absent — e.g. missing 'psp_values', 'lengths', or similar required companion field.

Common situations: Hand-built multimodal batches missing companion fields; schema changes where a field was renamed or made required; partial serialization dropping None-valued keys.

Related errors

• Unsupported image type: {type(image)}
• When using multiple prompts with multiple input images, prov
• {key}.position_ids is required
• You have to specify input_ids
• Hunyuan3D requires 'image_path' input.

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/fc6f5f8cad540164. Report an issue: GitHub.