keras-team/keras · error · ValueError

The theta value of a Thresholded ReLU layer should be >=0. R

Error message

The theta value of a Thresholded ReLU layer should be >=0. Received: {theta}

What it means

ThresholdedReLU activates inputs strictly greater than `theta`, and the layer requires theta >= 0 so it behaves as a threshold on positive activations. __init__ raises this ValueError when a negative theta is passed, whether directly or from a loaded config.

Source

Thrown at keras/src/legacy/layers.py:227

            "seed": self.seed,
        }
        base_config = super().get_config()
        return {**base_config, **config}


@keras_export("keras._legacy.layers.ThresholdedReLU")
class ThresholdedReLU(Layer):
    """DEPRECATED."""

    def __init__(self, theta=1.0, **kwargs):
        super().__init__(**kwargs)
        if theta is None:
            raise ValueError(
                "Theta of a Thresholded ReLU layer cannot be None, expecting a "
                f"float. Received: {theta}"
            )
        if theta < 0:
            raise ValueError(
                "The theta value of a Thresholded ReLU layer "
                f"should be >=0. Received: {theta}"
            )
        self.supports_masking = True
        self.theta = tf.convert_to_tensor(theta, dtype=self.compute_dtype)

    def call(self, inputs):
        dtype = self.compute_dtype
        return inputs * tf.cast(tf.greater(inputs, self.theta), dtype)

    def get_config(self):
        config = {"theta": float(self.theta)}
        base_config = super().get_config()
        return {**base_config, **config}

    def compute_output_shape(self, input_shape):
        return input_shape

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Use theta >= 0, e.g. ThresholdedReLU(theta=0.5); theta=0 degenerates toward plain ReLU — prefer keras.layers.ReLU then
2. Clip swept theta values to [0, inf) or constrain the search space
3. Validate deserialized configs before layer construction

Example fix

# before
layer = ThresholdedReLU(theta=-0.5)
# after
layer = ThresholdedReLU(theta=0.5)

theta = cfg.get('theta', 1.0)
assert theta >= 0, f'theta must be >= 0, got {theta}'

def is_valid_theta(t) -> bool:
    return isinstance(t, (int, float)) and t >= 0

When it happens

Trigger: Constructing ThresholdedReLU(theta=-0.5) or loading a serialized model config containing a negative theta value.

Common situations: Tuning theta via a sweep that crosses zero without constraints; hand-editing saved model configs; confusing theta with a bias-like parameter that is allowed to be negative.

Related errors

• Theta of a Thresholded ReLU layer cannot be None, expecting
• Expected `padding` to be a tuple of 3 tuples of 2 integers.
• Expected `padding` to be a tuple of 2 integers. Received: pa
• `factor` argument cannot have an upper bound lesser than the
• `factor` argument must have values larger than -1. Received:

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/b6081c039308d9aa. Report an issue: GitHub.