keras-team/keras · error · ValueError

Only input tensors may be passed as positional arguments. Th

Error message

Only input tensors may be passed as positional arguments. The following argument value should be passed as a keyword argument: {arg} (of type {type(arg)})

What it means

When calling a layer, only backend tensors (or None/symbolic tensors) may be passed as positional arguments; everything else must be a keyword argument. This guard catches e.g. passing an int or a Python object positionally where only inputs belong.

Source

Thrown at keras/src/layers/layer.py:908

                backend.set_keras_mask(y, mask)
            return y

        # Used to avoid expensive `tree` operations in the most common case.
        if (
            kwargs
            or len(args) != 1
            or not is_backend_tensor_or_symbolic(args[0], allow_none=False)
            or backend.standardize_dtype(args[0].dtype) != self.input_dtype
        ) and self._convert_input_args:
            args = tree.map_structure(maybe_convert, args)
            kwargs = tree.map_structure(maybe_convert, kwargs)

        ##########################################################
        # 2. Enforce that only tensors can be passed positionally.
        if not self._allow_non_tensor_positional_args:
            for arg in tree.flatten(args):
                if not is_backend_tensor_or_symbolic(arg, allow_none=True):
                    raise ValueError(
                        "Only input tensors may be passed as "
                        "positional arguments. The following argument value "
                        f"should be passed as a keyword argument: {arg} "
                        f"(of type {type(arg)})"
                    )

        # Caches info about `call()` signature, args, kwargs.
        call_spec = CallSpec(
            self._call_signature, self._call_context_args, args, kwargs
        )

        ############################################
        # 3. Check input spec for 1st positional arg.
        # TODO: consider extending this to all args and kwargs.
        self._assert_input_compatibility(call_spec.first_arg)

        ################
        # 4. Call build

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Pass non-tensor arguments as keywords: layer(x, training=True, mask=m)
2. If your custom layer legitimately takes non-tensor positionals, set self._allow_non_tensor_positional_args = True in __init__

Example fix

# before
out = layer(x, mask)
# after
out = layer(x, mask=mask)

assert all(keras.ops.is_tensor(a) for a in tree.flatten(args)), 'pass non-tensors as kwargs'

When it happens

Trigger: layer(x, training) instead of layer(x, training=training); layer(x, mask, True); custom layers called with config objects positionally when _allow_non_tensor_positional_args is False.

Common situations: Converting Keras 2 call signatures where mask was a valid 2nd positional; refactoring call() to accept new non-tensor options; passing Python bools/ints positionally.

Related errors

• `add_loss()` can only be called from inside `build()` or `ca
• `initializer` was passed both positionally and as a keyword
• `dtype` was passed both positionally and as a keyword argume
• `Sequential.layers` attribute is reserved and should not be
• Unknown activation function '{activation}' cannot be seriali

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/cf0b431fae13c4c8. Report an issue: GitHub.