keras-team/keras · error · ValueError

Argument `precision` must be in the range [0, 1]. Received:

Error message

Argument `precision` must be in the range [0, 1]. Received: precision={precision}

What it means

Raised by keras.metrics.RecallAtPrecision's __init__ when the precision argument is outside [0, 1]. Precision is a probability-like score; Keras validates the range at construction time.

Source

Thrown at keras/src/metrics/confusion_metrics.py:1032

    ```python
    model.compile(
        optimizer='sgd',
        loss='binary_crossentropy',
        metrics=[keras.metrics.RecallAtPrecision(precision=0.8)])
    ```
    """

    def __init__(
        self,
        precision,
        num_thresholds=200,
        class_id=None,
        name=None,
        dtype=None,
    ):
        if precision < 0 or precision > 1:
            raise ValueError(
                "Argument `precision` must be in the range [0, 1]. "
                f"Received: precision={precision}"
            )
        self.precision = precision
        self.num_thresholds = num_thresholds
        super().__init__(
            value=precision,
            num_thresholds=num_thresholds,
            class_id=class_id,
            name=name,
            dtype=dtype,
        )

    def result(self):
        recalls = ops.divide_no_nan(
            self.true_positives,
            ops.add(self.true_positives, self.false_negatives),
        )

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Pass a fraction in [0, 1], e.g. precision=0.99.
2. Clamp computed values: min(max(v, 0.0), 1.0).
3. Validate config entries before building the model.

Example fix

# before
m = keras.metrics.RecallAtPrecision(precision=99)

# after
m = keras.metrics.RecallAtPrecision(precision=0.99)

if not (0.0 <= precision <= 1.0):
    raise ValueError(f'precision must be in [0,1], got {precision}')

def is_probability(v) -> bool:
    return isinstance(v, (int, float)) and 0.0 <= v <= 1.0

When it happens

Trigger: keras.metrics.RecallAtPrecision(precision=1.05); passing 99 (percent); a computed tp/(tp+fp) that is unclamped above 1.

Common situations: Percent/fraction confusion; unclamped ratios; config typos.

Related errors

• Argument `specificity` must be in the range [0, 1]. Received
• Argument `sensitivity` must be in the range [0, 1]. Received
• Argument `recall` must be in the range [0, 1]. Received: rec
• Invalid `beta` argument value. It should be > 0. Received: b
• Invalid `threshold` argument value. It should verify 0 < thr

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/4a03bd894c37426c. Report an issue: GitHub.