Skip to content

vllm_omni.diffusion.sched.sigma_schedule

BASE_SCHEDULE_KEY module-attribute 

BASE_SCHEDULE_KEY = 'base_schedule'

DMD2SigmaSchedule dataclass

Continuous rectified-flow positions pinned by a distilled checkpoint.

A DMD2 student only ever sees the few noise levels it was trained on, so a distilled release ships the exact positions instead of letting the server derive a uniform schedule from num_inference_steps.

This is deliberately distinct from vllm_omni.diffusion.models.dmd2.DMD2Config.denoising_timesteps, which carries integer scheduler timesteps for scheduler-backed pipelines. Here the entries are continuous positions in [0, 1] that still need a per-modality time shift applied, which is what lets one schedule drive several coupled modalities at different shift scales.

base_schedule instance-attribute 

base_schedule: tuple[float, ...]

num_inference_steps property 

num_inference_steps: int

Denoising steps, i.e. one per interval between sigma boundaries.

from_metadata classmethod 

from_metadata(
    metadata: Mapping[str, Any],
    *,
    key: str = BASE_SCHEDULE_KEY,
) -> DMD2SigmaSchedule | None

Read a schedule from checkpoint metadata.

An absent key means the release is not distilled and keeps the legacy uniform schedule. An explicitly empty value is a malformed contract and is rejected rather than silently falling back.

from_positions classmethod 

from_positions(
    base_schedule: Sequence[float],
) -> DMD2SigmaSchedule

shifted_sigmas 

shifted_sigmas(shift_scale: float) -> list[float]

Apply the rectified-flow time shift for one modality.