keras-team/keras · error · ValueError

All cells must have a `state_size` attribute. Received cell

Error message

All cells must have a `state_size` attribute. Received cell without a `state_size`: {cell}

What it means

StackedRNNCells requires every cell to expose a `state_size` attribute (checked via `dir(cell)` in `__init__`), because the RNN layer needs to know how many state tensors and their sizes to initialize and pass between timesteps. A cell with `call` but no `state_size` cannot be stepped by the RNN machinery, so construction aborts.

Source

Thrown at keras/src/layers/rnn/stacked_rnn_cells.py:43

    rnn_cells = [keras.layers.LSTMCell(128) for _ in range(2)]
    stacked_lstm = keras.layers.StackedRNNCells(rnn_cells)
    lstm_layer = keras.layers.RNN(stacked_lstm)

    result = lstm_layer(x)
    ```
    """

    def __init__(self, cells, **kwargs):
        super().__init__(**kwargs)
        for cell in cells:
            if "call" not in dir(cell):
                raise ValueError(
                    "All cells must have a `call` method. "
                    f"Received cell without a `call` method: {cell}"
                )
            if "state_size" not in dir(cell):
                raise ValueError(
                    "All cells must have a `state_size` attribute. "
                    f"Received cell without a `state_size`: {cell}"
                )
        self.cells = cells

    @property
    def state_size(self):
        return [c.state_size for c in self.cells]

    @property
    def output_size(self):
        if getattr(self.cells[-1], "output_size", None) is not None:
            return self.cells[-1].output_size
        elif isinstance(self.cells[-1].state_size, (list, tuple)):
            return self.cells[-1].state_size[0]
        else:
            return self.cells[-1].state_size

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Add state_size to the custom cell, e.g. self.state_size = units for a single-state cell (or a tuple/list for multiple states)
2. If the object is not meant to be a cell, replace it with a proper Keras cell (SimpleRNNCell, LSTMCell, GRUCell)
3. For multi-state cells, make sure state_size lengths match what call() returns as new_states

Example fix

# before
class MyCell:
    def __init__(self, units):
        self.units = units
    def call(self, inputs, states):
        ...

# after
class MyCell:
    def __init__(self, units):
        self.units = units
        self.state_size = units
    def call(self, inputs, states):
        ...

assert all(hasattr(c, 'state_size') for c in cells), 'cells missing state_size'
stacked = keras.layers.StackedRNNCells(cells=cells)

def has_state_size(cell) -> bool:
    return hasattr(cell, 'state_size') and cell.state_size is not None

When it happens

Trigger: Passing a custom object that implements call() but not state_size to keras.layers.StackedRNNCells(cells=[...]); porting a PyTorch-style RNN module or generic Layer into a Keras cell list.

Common situations: Writing custom RNN cells and forgetting state_size; migrating from tf.keras v1 or other frameworks whose cell interfaces differ; refactoring a Layer into a cell without adding the RNN cell contract.

Related errors

• Received an invalid value for argument `units`, expected a p
• All cells must have a `call` method. Received cell without a
• Unknown activation function '{activation}' cannot be seriali
• Could not interpret activation function identifier: {identif
• ConvNeXt does not support the `channels_first` image data fo

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/c6a4875afe10e9f3. Report an issue: GitHub.