keras-team/keras · error · ValueError

Unexpected object from deserialization, expected a layer or…

Error message

Unexpected object from deserialization, expected a layer or operation, got a {type(layer)}

What it means

Guard during Functional model deserialization: after instantiating an entry from the saved layer config, the object is not an Operation/Layer (e.g. the class_name resolved to a loss, metric, or plain function because of a custom_objects mapping mistake). The graph reconstruction only knows how to wire Layer/Operation nodes, so anything else is rejected.

Solutions

1. Check custom_objects / registrations: the layer name in the config must map to an actual Layer (or Operation) subclass
2. Ensure a custom class registered with the same name didn't shadow the layer
3. Verify the serialized model wasn't produced with a mismatched Keras version

When it happens

Trigger: Thrown at keras/src/models/functional.py:573 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/38107ede870e8e0e. Report an issue: GitHub.

Appendix: source

Thrown at keras/src/models/functional.py:573

        Args:
            layer_data: layer config dict.
        """
        layer_name = layer_data["name"]

        # Instantiate layer.
        if "module" not in layer_data:
            # Legacy format deserialization (no "module" key)
            # used for H5 and SavedModel formats
            layer = saving_utils.model_from_config(
                layer_data, custom_objects=custom_objects
            )
        else:
            layer = serialization_lib.deserialize_keras_object(
                layer_data, custom_objects=custom_objects
            )
        if not isinstance(layer, Operation):
            raise ValueError(
                "Unexpected object from deserialization, expected a layer or "
                f"operation, got a {type(layer)}"
            )
        created_layers[layer_name] = layer

        # Gather layer inputs.
        inbound_nodes_data = layer_data["inbound_nodes"]
        for node_data in inbound_nodes_data:
            # We don't process nodes (i.e. make layer calls)
            # on the fly because the inbound node may not yet exist,
            # in case of layer shared at different topological depths
            # (e.g. a model such as A(B(A(B(x)))))
            add_unprocessed_node(layer, node_data)

    # Extract config used to instantiate Functional model from the config. The
    # remaining config will be passed as keyword arguments to the Model
    # constructor.
    functional_config = {}

View on GitHub (pinned to 7a34a03db6)