keras-team/keras · error · ValueError

Invalid `curve` argument value "{curve}". Expected one of: {

Error message

Invalid `curve` argument value "{curve}". Expected one of: {list(metrics_utils.AUCCurve)}

What it means

Raised by keras.metrics.AUC's __init__ when curve is an AUCCurve enum instance that is not a supported member (ROC or PR). This fires only when you pass an enum instance; plain strings are handled by a different code path.

Source

Thrown at keras/src/metrics/confusion_metrics.py:1209

        num_thresholds=200,
        curve="ROC",
        summation_method="interpolation",
        name=None,
        dtype=None,
        thresholds=None,
        multi_label=False,
        num_labels=None,
        label_weights=None,
        from_logits=False,
    ):
        # Metric should be maximized during optimization.
        self._direction = "up"

        # Validate configurations.
        if isinstance(curve, metrics_utils.AUCCurve) and curve not in list(
            metrics_utils.AUCCurve
        ):
            raise ValueError(
                f'Invalid `curve` argument value "{curve}". '
                f"Expected one of: {list(metrics_utils.AUCCurve)}"
            )
        if isinstance(
            summation_method, metrics_utils.AUCSummationMethod
        ) and summation_method not in list(metrics_utils.AUCSummationMethod):
            raise ValueError(
                "Invalid `summation_method` "
                f'argument value "{summation_method}". '
                f"Expected one of: {list(metrics_utils.AUCSummationMethod)}"
            )

        # Update properties.
        self._init_from_thresholds = thresholds is not None
        if thresholds is not None:
            # If specified, use the supplied thresholds.
            self.num_thresholds = len(thresholds) + 2
            thresholds = sorted(thresholds)

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Pass a string instead: curve='ROC' or curve='PR'.
2. If enums are needed, import AUCCurve from the same Keras version's metrics_utils.
3. Regenerate pickled/saved configs containing enum objects.

Example fix

# before
from keras.src.metrics import metrics_utils
auc = keras.metrics.AUC(curve=metrics_utils.AUCCurve('SQ'))

# after
auc = keras.metrics.AUC(curve='PR')

assert curve in ('ROC', 'PR'), f'curve must be ROC or PR, got {curve}'

def is_valid_curve(v) -> bool:
    return v in ('ROC', 'PR')

When it happens

Trigger: Passing a stale or custom AUCCurve enum member (e.g. pickled from an older TF/Keras version, or a dynamically constructed enum) to keras.metrics.AUC(curve=...).

Common situations: Code migrated from tensorflow.keras that persisted enum objects across versions; monkeypatched or dynamically created enum members.

Understand the failure class

Background: Invalid enum value errors: "Unknown type", "Invalid scope", "must be one of" — when a string is not on the library's allowed list — this error's family across 23 libraries.

Related errors

• Invalid `summation_method` argument value "{summation_method
• Argument `num_thresholds` must be an integer > 1. Received:
• `num_labels` is needed only when `multi_label` is True.
• `y_pred` must have rank 2 when `multi_label=True`. Found ran
• Invalid AUC curve value: "{key}". Expected values are ["PR",

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/93acb8e0ffecb8f4. Report an issue: GitHub.