tensorflow/models · error · RuntimeError

{value} has unknown batch.

Error message

{value} has unknown batch.

What it means

Error "{value} has unknown batch." thrown in tensorflow/models.

Source

Thrown at official/modeling/tf_utils.py:284

      be concatenated.
    axis: The axis along which to perform the concatenation as a Python integer
      (not a `Tensor`). E.g., `axis=0` to concatenate along the batch dimension.
    name: A name for the operation (used to create a name scope).

  Returns:
    The result of concatenating `value` along `axis` across replicas.

  Raises:
    RuntimeError: when the batch (0-th) dimension is None.
  """
  with tf.name_scope(name):
    context = tf.distribute.get_replica_context()
    # Typically this could be hit only if the tensor is derived from a
    # dataset with finite epochs and drop_remainder=False, where the last
    # batch could of different batch size and then the dim-0 is of dynamic
    # shape.
    if value.shape.as_list()[0] is None:
      raise RuntimeError(f"{value} has unknown batch.")
    return context.all_gather(value, axis=axis)


def clone_initializer(initializer):
  # Keras initializer is going to be stateless, which mean reusing the same
  # initializer will produce same init value when the shapes are the same.
  if isinstance(initializer, tf_keras.initializers.Initializer):
    return initializer.__class__.from_config(initializer.get_config())
  # When the input is string/dict or other serialized configs, caller will
  # create a new keras Initializer instance based on that, and we don't need to
  # do anything
  return initializer


def serialize_keras_object(obj):
  if hasattr(tf_keras.utils, "legacy"):
    return tf_keras.utils.legacy.serialize_keras_object(obj)
  else:

View on GitHub (pinned to e006f5f0d5)

When it happens

Trigger: Thrown at official/modeling/tf_utils.py:284 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tensorflow/models@e006f5f0d5 (2026-08-24). Data as JSON: /api/errors/2535e75339d6b683. Report an issue: GitHub.