keras-team/keras · error · ValueError

Cannot enable lora on a layer that isn't yet built.

Error message

Cannot enable lora on a layer that isn't yet built.

What it means

enable_lora() must create LoRA A/B variables sized to the built kernel, so it requires the layer to already be built. Calling it on a freshly constructed Dense (no input shape seen yet) raises immediately.

Source

Thrown at keras/src/layers/core/dense.py:265

        output_shape = list(input_shape)
        output_shape[-1] = self.units
        return tuple(output_shape)

    def enable_lora(
        self,
        rank,
        lora_alpha=None,
        a_initializer="he_uniform",
        b_initializer="zeros",
    ):
        if self.kernel_constraint:
            raise ValueError(
                "Lora is incompatible with kernel constraints. "
                "In order to enable lora on this layer, remove the "
                "`kernel_constraint` argument."
            )
        if not self.built:
            raise ValueError(
                "Cannot enable lora on a layer that isn't yet built."
            )
        if self.lora_enabled:
            raise ValueError(
                "lora is already enabled. This can only be done once per layer."
            )
        if self.quantization_mode == "gptq":
            raise NotImplementedError(
                "lora is not currently supported with GPTQ quantization."
            )
        self._tracker.unlock()
        # Determine the correct input dimension for the LoRA A matrix. When
        # the layer has been int4-quantized, `self._kernel` stores a *packed*
        # representation whose first dimension is `ceil(input_dim/2)`. We
        # saved the true, *unpacked* input dimension in `self._orig_input_dim`
        # during quantization. Use it if available; otherwise fall back to the
        # first dimension of `self.kernel`.
        if self.quantization_mode == "int4" and hasattr(

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Build the model first: model.build(input_shape) or model(x) once, then enable LoRA.
2. Guard injection code with if not layer.built: build or skip.
3. For Sequential models, call model.build() with the expected input shape before iterating.

Example fix

# before
dense = keras.layers.Dense(64)
dense.enable_lora(8)  # not built yet

# after
dense = keras.layers.Dense(64)
dense.build((None, 128))
dense.enable_lora(8)

model.build(input_shape)  # or model(x) once
for layer in model.layers:
    if isinstance(layer, keras.layers.Dense) and layer.built:
        layer.enable_lora(rank)

def ready_for_lora(layer) -> bool:
    return layer.built and not getattr(layer, 'lora_enabled', False)

When it happens

Trigger: Calling dense.enable_lora(rank) before layer.build(input_shape) or before the layer has processed a batch — e.g. iterating model.layers immediately after Model() construction.

Common situations: LoRA-injection utilities that run before model.build()/first forward call; functional models where layers construct before input shapes propagate.

Related errors

• You must build the layer before accessing `kernel`.
• Lora is incompatible with kernel constraints. In order to en
• lora is already enabled. This can only be done once per laye
• lora is not currently supported with GPTQ quantization.
• Currently, `_float8_call` doesn't support LoRA

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/d058f52b1d504b6f. Report an issue: GitHub.