apache/beam · error · ValueError

'{type(self).name}' not registered as Specifiable. Decor

Error message

'{type(self).__name__}' not registered as Specifiable. Decorate ({type(self).__name__}) with @specifiable

What it means

apache_beam.ml.anomaly.specifiable.to_spec() can only serialize objects whose class was registered via the @specifiable decorator, which sets a class-level `spec_type`. If the class lacks `spec_type`, the library cannot map the instance back to a constructor spec and raises this ValueError.

Source

Thrown at sdks/python/apache_beam/ml/anomaly/specifiable.py:198

      return subclass

    kwargs = {
        k: _specifiable_from_spec_helper(v, _run_init)
        for k, v in spec.config.items()
    }

    if _run_init:
      kwargs["_run_init"] = True
    return subclass(**kwargs)

  def to_spec(self) -> Spec:
    """Generate a spec from a `Specifiable` subclass object.

    Returns:
      Spec: The specification of the instance.
    """
    if getattr(type(self), 'spec_type', None) is None:
      raise ValueError(
          f"'{type(self).__name__}' not registered as Specifiable. "
          f"Decorate ({type(self).__name__}) with @specifiable")

    args = {
        k: _specifiable_to_spec_helper(v)
        for k, v in self.init_kwargs.items()
    }

    return Spec(type=self.spec_type(), config=args)

  def run_original_init(self) -> None:
    """Invoke the original __init__ method with original keyword arguments"""
    pass

  @classmethod
  def unspecifiable(cls) -> None:
    """Resume the class structure prior to specifiable"""
    pass

View on GitHub (pinned to 12126d8942)

Solutions

Decorate the class with @specifiable (optionally @specifiable(spec_type='my_type')) before instantiating/calling to_spec
If you cannot modify the class, wrap it in a Specifiable-compatible adapter class decorated with @specifiable
Check that the instance's actual (most-derived) class is the decorated one, not an undecorated subclass

Example fix

// before
class MyDetector(AnomalyDetector):
    ...
obj.to_spec()

// after
from apache_beam.ml.anomaly import specifiable

@specifiable.specifiable
class MyDetector(AnomalyDetector):
    ...
obj.to_spec()

Defensive patterns

Strategy: validation

Validate before calling

from apache_beam.ml.anomaly import specifiable
if getattr(type(detector), 'spec_type', None) is None:
    raise TypeError(f"{type(detector).__name__} must be decorated with @specifiable before to_spec()")

Type guard

def is_specifiable(obj) -> bool:
    return getattr(type(obj), 'spec_type', None) is not None

Prevention

Always decorate custom detector/transform classes with @specifiable
Run a unit test that calls to_spec() on every pipeline component
Keep custom class names unique from Beam built-ins

When it happens

Trigger: Calling to_spec() (or specifiable_object_to_spec) on an instance of a class not decorated with @specifiable, e.g. a custom AnomalyDetector subclass or a third-party sklearn model passed into ML Transform without registration.

Common situations: Users write a custom detector subclass and pass it to RunInference/anomaly transforms without decorating it; upgrading Beam adds @specifiable to internal classes that user subclasses now shadow; passing plain sklearn/torch objects inside ensemble detectors.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/9138d681877e8fd4. Report an issue: GitHub.