sgl-project/sglang · error · ValueError

Pipeline '{pipeline_name}' is already registered; pass overw

Error message

Pipeline '{pipeline_name}' is already registered; pass overwrite=True to replace it

What it means

register_pipeline refuses to silently overwrite an existing entry: if pipeline_cls.pipeline_name already exists in _PIPELINE_REGISTRY and overwrite=False (the default), it raises ValueError telling you to pass overwrite=True.

Source

Thrown at python/sglang/multimodal_gen/registry.py:396

    pipeline_cls: Type[ComposedPipelineBase],
    *,
    sampling_param_cls: Any,
    pipeline_config_cls: Type[PipelineConfig],
    hf_model_paths: Optional[List[str]] = None,
    model_detectors: Optional[List[Callable[[str], bool]]] = None,
    overwrite: bool = False,
) -> None:
    """Register an out-of-tree native diffusion pipeline and its configs."""
    _discover_and_register_pipelines()
    if not issubclass(pipeline_cls, ComposedPipelineBase):
        raise TypeError("pipeline_cls must inherit from ComposedPipelineBase")
    if not issubclass(pipeline_config_cls, PipelineConfig):
        raise TypeError("pipeline_config_cls must inherit from PipelineConfig")

    pipeline_name = pipeline_cls.pipeline_name
    existing_pipeline = _PIPELINE_REGISTRY.get(pipeline_name)
    if existing_pipeline is not None and not overwrite:
        raise ValueError(
            f"Pipeline '{pipeline_name}' is already registered; pass overwrite=True to replace it"
        )
    for model_path in hf_model_paths or []:
        if model_path in _MODEL_HF_PATH_TO_NAME and not overwrite:
            raise ValueError(f"Model path '{model_path}' is already registered")
        registered_pipeline = KNOWN_NON_DIFFUSERS_DIFFUSION_MODEL_PATTERNS.get(
            model_path.lower()
        )
        if registered_pipeline is not None and not overwrite:
            raise ValueError(
                f"Model path '{model_path}' is already registered for pipeline "
                f"'{registered_pipeline}'"
            )

    _PIPELINE_REGISTRY[pipeline_name] = pipeline_cls
    _PIPELINE_CONFIG_REGISTRY[pipeline_name] = (
        pipeline_config_cls,
        sampling_param_cls,

View on GitHub (pinned to 0132848349)

Solutions

1. Pass overwrite=True if you intentionally want to replace the existing pipeline
2. Guard registration with a lookup (e.g. check the registry / get_pipeline first) to make it idempotent
3. In notebooks or hot-reload scenarios, restructure so registration runs once (module-level flag or __main__ guard)

Example fix

# before
register_pipeline(MyPipeline, MyConfig)  # second call raises
# after
register_pipeline(MyPipeline, MyConfig, overwrite=True)
# or
if "my_pipeline" not in get_registered_pipeline_names():
    register_pipeline(MyPipeline, MyConfig)

from sglang.multimodal_gen import registry
name = MyPipeline.pipeline_name
already = name in getattr(registry, "_PIPELINE_REGISTRY", {}) or registry.get_pipeline(name) is not None
if not already:
    registry.register_pipeline(MyPipeline, MyConfig)
# else: skip or pass overwrite=True

try:
    register_pipeline(MyPipeline, MyConfig)
except ValueError as e:
    if "already registered" in str(e) and "overwrite=True" in str(e):
        register_pipeline(MyPipeline, MyConfig, overwrite=True)
    else:
        raise

When it happens

Trigger: Calling register_pipeline twice with classes sharing the same pipeline_name attribute — typically re-running registration in notebooks, hot-reload, or a plugin being loaded more than once.

Common situations: Notebook cell re-execution; module reimport triggering registration again; two plugins registering the same pipeline name; duplicate registration in tests without cleanup.

Related errors

• Model path '{model_path}' is already registered
• pipeline_cls must inherit from ComposedPipelineBase
• The quantization method `{quantization}` is already exists.
• LoRA with name {lora_ref.lora_name} already exists. Loaded L
• scalar_type_id {scalar_type_id} doesn't exists.

AI-assisted analysis of sgl-project/sglang@0132848349 (2026-08-28). Data as JSON: /api/errors/dc0f3dffe4094d30. Report an issue: GitHub.