Skip to content

vllm_omni.diffusion.models.diffusers_adapter.quantization_utils ¶

Quantization helpers for the Diffusers backend.

logger `module-attribute` ¶

logger = logging.getLogger(__name__)

apply_diffusers_quantization_config ¶

apply_diffusers_quantization_config(
    od_config: Any,
    load_kwargs: dict[str, Any],
    component_names: dict[str, Any],
) -> bool

Inject a courtesy-converted quantization_config into load kwargs.

diffusers_load_kwargs is the canonical Diffusers backend configuration path, so an explicit quantization_config already present in load_kwargs is never replaced. Returns whether a config was injected by this helper.

convert_diffusers_quantization_config ¶

convert_diffusers_quantization_config(
    load_kwargs: dict[str, Any],
) -> None

Build Diffusers-native quantization config objects from dict kwargs.

ensure_supported_diffusers_quantization ¶

ensure_supported_diffusers_quantization(
    quant_config: Any,
) -> None

Validate that a vLLM-Omni quantization config has a Diffusers mapping.