apache/beam · error · ValueError

Callable create_model_fn must be passedwith…

Error message

Callable create_model_fn must be passedwith ModelType.SAVED_WEIGHTS

What it means

Raised in TFTensorFlowModelHandler (KeyedModelHandler) load_model when model_type is ModelType.SAVED_WEIGHTS but no create_model_fn callable was supplied. Loading raw saved weights requires a callable that instantiates the model architecture into which the weights are loaded.

Solutions

  1. Pass a create_model_fn that returns an instantiated tf.Module/keras model with the expected architecture
  2. Use model_type=ModelType.SAVED_MODEL instead if you have a full SavedModel and don't need weight-only loading

Example fix

// before
handler = TFTensorFlowModelHandlerMRU(
    model_uri='gs://bucket/weights.ckpt', model_type=ModelType.SAVED_WEIGHTS)
// after
handler = TFTensorFlowModelHandlerMRU(
    model_uri='gs://bucket/weights.ckpt',
    model_type=ModelType.SAVED_WEIGHTS,
    create_model_fn=lambda: MyNet())
Defensive patterns

Strategy: validation

Validate before calling

def validate_tf_handler_args(model_type, create_model_fn):
    if model_type == ModelType.SAVED_WEIGHTS and not callable(create_model_fn):
        raise ValueError('create_model_fn must be a callable when model_type is SAVED_WEIGHTS')

Type guard

import collections.abc
create_model_fn_is_valid = lambda fn: isinstance(fn, collections.abc.Callable)

Try / catch

try:
    handler.load_model()
except ValueError as e:
    if 'create_model_fn' in str(e):
        handler._create_model_fn = build_model
        handler.load_model()
    else:
        raise

Prevention

When it happens

Trigger: Constructing the TF handler with model_type=ModelType.SAVED_WEIGHTS and leaving create_model_fn as None, then calling load_model during pipeline setup.

Common situations: Switching a handler from SAVED_MODEL to SAVED_WEIGHTS without adding create_model_fn; assuming weights-only loading can infer the architecture; config-driven handler construction that omits the callable for the weights type.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/f324901bf4c51150. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/ml/inference/tensorflow_inference.py:182

        max_batch_weight=max_batch_weight,
        element_size_fn=element_size_fn,
        batch_length_fn=batch_length_fn,
        batch_bucket_boundaries=batch_bucket_boundaries,
        large_model=large_model,
        model_copies=model_copies,
        **kwargs)
    self._model_uri = model_uri
    self._model_type = model_type
    self._inference_fn = inference_fn
    self._create_model_fn = create_model_fn
    self._load_model_args = {} if not load_model_args else load_model_args
    self._custom_weights = custom_weights

  def load_model(self) -> tf.Module:
    """Loads and initializes a Tensorflow model for processing."""
    if self._model_type == ModelType.SAVED_WEIGHTS:
      if not self._create_model_fn:
        raise ValueError(
            "Callable create_model_fn must be passed"
            "with ModelType.SAVED_WEIGHTS")
      return _load_model_from_weights(self._create_model_fn, self._model_uri)

    return _load_model(
        self._model_uri, self._custom_weights, self._load_model_args)

  def update_model_path(self, model_path: Optional[str] = None):
    self._model_uri = model_path if model_path else self._model_uri

  def run_inference(
      self,
      batch: Sequence[numpy.ndarray],
      model: tf.Module,
      inference_args: Optional[dict[str, Any]] = None
  ) -> Iterable[PredictionResult]:
    """
    Runs inferences on a batch of numpy array and returns an Iterable of

View on GitHub (pinned to 12126d8942)