vllm_omni.diffusion.cache.seacache.backend ¶

CUSTOM_SEACACHE_ENABLERS `module-attribute` ¶

CUSTOM_SEACACHE_ENABLERS = {
    "Cosmos3OmniDiffusersPipeline": enable_cosmos3_seacache,
    "Cosmos3OmniPipeline": enable_cosmos3_seacache,
}

logger = init_logger(__name__)

Backend for spectral-evolution-aware diffusion caching.

enable(pipeline: Any) -> None

refresh(
    pipeline: Any,
    num_inference_steps: int,
    verbose: bool = True,
) -> None

enable_cosmos3_seacache(
    pipeline: Any, config: DiffusionCacheConfig
) -> SeaCacheRootHook