keras-team/keras · error · ValueError

Sequential model '{self.name}' has already been configured t

Error message

Sequential model '{self.name}' has already been configured to use input shape {self._layers[0].batch_shape}. You cannot build it with input_shape {input_shape}

What it means

Sequential.build(input_shape) refuses to build when the model already starts with an InputLayer whose batch_shape differs from the requested shape. The InputLayer fixes the input shape at construction time, so a conflicting later build is rejected instead of silently rebuilding.

Source

Thrown at keras/src/models/sequential.py:178

    def _obj_type(self):
        return "Sequential"

    def build(self, input_shape=None):
        try:
            input_shape = standardize_shape(input_shape)
        except:
            # Do not attempt to build if the model does not have a single
            # input tensor.
            return
        if not self._layers:
            raise ValueError(
                f"Sequential model {self.name} cannot be built because it has "
                "no layers. Call `model.add(layer)`."
            )
        if isinstance(self._layers[0], InputLayer):
            if self._layers[0].batch_shape != input_shape:
                raise ValueError(
                    f"Sequential model '{self.name}' has already been "
                    "configured to use input shape "
                    f"{self._layers[0].batch_shape}. You cannot build it "
                    f"with input_shape {input_shape}"
                )
        else:
            dtype = self._layers[0].compute_dtype
            self._layers = [
                InputLayer(batch_shape=input_shape, dtype=dtype)
            ] + self._layers

        # Build functional model
        inputs = self._layers[0].output
        x = inputs
        for layer in self._layers[1:]:
            try:
                x = layer(x)
            except NotImplementedError:

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Pass the exact same shape as the InputLayer's batch_shape, including the batch dimension (e.g. (None, 28, 28))
2. Recreate the model without the InputLayer and rely on build(input_shape) alone
3. Recreate the model with keras.Input(new_shape) as the first layer

Example fix

# before
model = keras.Sequential([keras.Input((32,)), keras.layers.Dense(10)])
model.build((None, 28, 28))  # ValueError

# after
model = keras.Sequential([keras.Input((28, 28)), keras.layers.Dense(10)])
# or: model = keras.Sequential([Dense(10)]); model.build((None, 28, 28))

first = model._layers[0] if model._layers else None
if isinstance(first, keras.layers.InputLayer):
    assert first.batch_shape == input_shape

try:
    model.build(input_shape)
except ValueError as e:
    if 'already been configured' in str(e):
        shape = model._layers[0].batch_shape

When it happens

Trigger: model = keras.Sequential([keras.Input((32,)), ...]); model.build((None, 64,))

Common situations: Loading checkpoints, writing tutorials, or migrating from Keras 2 where build behaved differently

Related errors

• Sequential model '{self.name}' has no defined input shape ye
• To call stateless_call, {self.__class__.__name__} must be bu
• Cannot quantize a layer that isn't yet built. Layer '{self.n
• Layer '{self.name}' was never built and thus it doesn't have
• You tried to call `count_params` on layer '{self.name}', but

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/7dc34e0c83eace6e. Report an issue: GitHub.