Skip to content

vllm_omni.diffusion.output_metadata

Compatibility imports for diffusion output metadata.

The shared output payload and metadata schema lives in vllm_omni.outputs.output_metadata. Keep this module as a stable import path for existing diffusion callers.

DiffusionMetadata module-attribute 

DiffusionMetadata: TypeAlias = dict[
    str, DiffusionMetadataValue
]

DiffusionMetadataMapping module-attribute 

DiffusionMetadataMapping: TypeAlias = Mapping[
    str, DiffusionMetadataValue
]

DiffusionMetadataValue module-attribute 

DiffusionMetadataValue: TypeAlias = (
    DiffusionScalar
    | DiffusionShape
    | DiffusionTensorMap
    | torch.Tensor
    | list["DiffusionMetadataValue"]
    | tuple["DiffusionMetadataValue", ...]
    | Mapping[str, "DiffusionMetadataValue"]
)

DiffusionMultimodalOutput module-attribute 

DiffusionMultimodalOutput: TypeAlias = dict[
    str, DiffusionPayloadValue | DiffusionMetadata
]

DiffusionPayload module-attribute 

DiffusionPayload: TypeAlias = dict[
    str, DiffusionPayloadValue
]

DiffusionPayloadMapping module-attribute 

DiffusionPayloadMapping: TypeAlias = Mapping[
    str, DiffusionPayloadValue
]

DiffusionPayloadValue module-attribute 

DiffusionPayloadValue: TypeAlias = (
    DiffusionScalar
    | bytes
    | torch.Tensor
    | np.ndarray
    | PIL.Image.Image
    | DiffusionTrajectoryPayload
    | list["DiffusionPayloadValue"]
    | tuple["DiffusionPayloadValue", ...]
    | Mapping[str, "DiffusionPayloadValue"]
)

DiffusionPostprocessRawOutput module-attribute 

DiffusionPostprocessRawOutput: TypeAlias = (
    DiffusionPayloadValue
    | DiffusionPayload
    | DiffusionOutputEnvelope
)

DiffusionScalar module-attribute 

DiffusionScalar: TypeAlias = str | int | float | bool | None

DiffusionShape module-attribute 

DiffusionShape: TypeAlias = list[int] | tuple[int, ...]

DiffusionTensorMap module-attribute 

DiffusionTensorMap: TypeAlias = Mapping[str, torch.Tensor]

FinalOutputModalityType module-attribute 

FinalOutputModalityType: TypeAlias = Literal[
    "text", "image", "audio", "video"
]

DiffusionOutputEnvelope

Bases: TypedDict

metadata instance-attribute 

metadata: DiffusionMetadata

payload instance-attribute 

payload: DiffusionPayload

DiffusionTrajectoryPayload

Bases: TypedDict

decoded instance-attribute 

decoded: list[Image] | None

latents instance-attribute 

latents: Tensor | dict[str, Any] | None

log_probs instance-attribute 

log_probs: Tensor | dict[str, Any] | None

timesteps instance-attribute 

timesteps: Tensor | dict[str, Any] | None

strip_internal_metadata 

strip_internal_metadata(
    metadata: DiffusionMetadataMapping,
) -> DiffusionMetadata

Return public metadata, dropping model-private postprocess state.

validate_diffusion_metadata 

validate_diffusion_metadata(
    metadata: DiffusionMetadataMapping,
) -> None

Validate canonical diffusion output metadata fields.

Unknown model-specific groups and keys are intentionally tolerated. This validator only guards fields that are part of the shared output protocol or consumed by serving code.

validate_public_diffusion_metadata 

validate_public_diffusion_metadata(
    metadata: DiffusionMetadataMapping,
) -> None

Validate metadata that is safe to expose outside model postprocess.