keras-team/keras · error · ValueError

Output with path `{path}` is not connected to `inputs`

Error message

Output with path `{path}` is not connected to `inputs`

What it means

When evaluating a Function graph, every declared output must be reachable by traversing from the declared inputs. An output that came from different tensors (constants, another graph) fails this reachability check.

Source

Thrown at keras/src/ops/function.py:221

                    # Use call_fn if provided (e.g., for symbolic execution)
                    op = operation_fn(node.operation)
                    outputs = call_fn(op, *args, **kwargs)
                else:
                    # Use NNX operation mapping
                    operation = self._get_operation_for_node(node)
                    op = operation_fn(operation)
                    outputs = op(*args, **kwargs)

                # Update tensor_dict.
                for x, y in zip(node.outputs, tree.flatten(outputs)):
                    tensor_dict[id(x)] = y

        output_tensors = []
        for i, x in enumerate(self.outputs):
            if id(x) not in tensor_dict:
                path = tree.flatten_with_path(self._outputs_struct)[i][0]
                path = ".".join(str(p) for p in path)
                raise ValueError(
                    f"Output with path `{path}` is not connected to `inputs`"
                )
            output_tensors.append(tensor_dict[id(x)])

        return tree.pack_sequence_as(self._outputs_struct, output_tensors)

    def _assert_input_compatibility(self, inputs):
        try:
            tree.assert_same_structure(inputs, self._inputs_struct)
        except ValueError:
            raise ValueError(
                "Function was called with an invalid input structure. "
                f"Expected input structure: {self._inputs_struct}\n"
                f"Received input structure: {inputs}"
            )
        for x, x_ref in zip(tree.flatten(inputs), self._inputs):
            if len(x.shape) != len(x_ref.shape):
                raise ValueError(

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Compute the output from the input tensors via ops/layers
2. Add the missing source tensor to inputs
3. Remove the stray output from the outputs structure

Example fix

# before
y = some_other_input + 1
fn = keras.ops.Function([x], y)

# after
y = x + 1
fn = keras.ops.Function([x], y)

out_flat = tree.flatten(outputs)
in_ids = {id(t) for t in tree.flatten(inputs)}
assert out_flat, 'outputs empty'

When it happens

Trigger: keras.ops.Function([x], [y]) where y was created independently of x (e.g. y = other_tensor + 1)

Common situations: Functional model construction where a return value comes from a layer applied outside the input graph, or accidental tensor reuse across models

Related errors

• `inputs` argument cannot be empty. Received: inputs={inputs}
• `outputs` argument cannot be empty. Received: inputs={inputs
• Function was called with an invalid input structure. Expecte
• Array inputs to associative_scan must have the same first di
• Invalid reduction: {reduction}. Supported values are: None,

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/6a2733247d031f13. Report an issue: GitHub.