vllm_omni.diffusion.offloader.startup ¶

Generic startup handoff between model loaders and offload backends.

OffloadStartupState `dataclass` ¶

Loader-owned state consumed by the offloader startup boundary.

allow_fresh_retry: bool = False

fresh_model_loader: Callable[[], Module] | None = None

host_weight_plan: HostWeightPlan | None = None

close_loader_ownership() -> None

Release a plan that never reached a backend.

attach_offload_startup_state(
    model: Module, state: OffloadStartupState
) -> None

Attach one process-local startup handoff to a loaded pipeline.

take_offload_startup_state(
    model: Module,
) -> OffloadStartupState | None

Take and remove the loader handoff from a pipeline exactly once.