keras-team/keras · error · NotImplementedError

Unknown `interpolation

Error message

Unknown `interpolation` {interpolation}. Expected of one {self._SUPPORTED_INTERPOLATION}.

What it means

RandomElasticTransform validates its interpolation argument in __init__ against the fixed set stored in _SUPPORTED_INTERPOLATION. An unsupported string raises NotImplementedError immediately at construction, so an augmentation that cannot be realized never reaches training. The supported values are those enumerated by the layer (e.g. 'nearest' and 'bilinear').

Solutions

Use one of the exact strings in RandomElasticTransform._SUPPORTED_INTERPOLATION, typically 'nearest' or 'bilinear'
Print the supported set first: print(layers.RandomElasticTransform._SUPPORTED_INTERPOLATION)
If you need 'bicubic', pick a layer that supports it or implement a custom transform outside this layer

Example fix

# before
layers.RandomElasticTransform(interpolation="bicubic")
# after
layers.RandomElasticTransform(interpolation="bilinear")

Defensive patterns

Strategy: validation

Validate before calling

from keras.src.layers.preprocessing.image_preprocessing.random_elastic_transform import RandomElasticTransform
interp = "bilinear"
assert interp in RandomElasticTransform._SUPPORTED_INTERPOLATION, f"use one of {RandomElasticTransform._SUPPORTED_INTERPOLATION}"

Type guard

def is_valid_interpolation(v: str) -> bool:
    return v in {"nearest", "bilinear"}

Try / catch

try:
    layer = RandomElasticTransform(interpolation=interp)
except NotImplementedError as e:
    raise ValueError(f"bad interpolation {interp!r}: {e}") from e

Prevention

Pin interpolation vocabulary to the layer's _SUPPORTED_INTERPOLATION constant
Add config schema validation when values come from YAML/JSON

When it happens

Trigger: Constructing layers.RandomElasticTransform(interpolation=...) with a misspelled or unsupported string, e.g. interpolation='bicubic', interpolation='Bilinear' (case matters), or interpolation='linear'.

Common situations: Copying interpolation names from other frameworks (PyTorch/torchvision or tf.image use different vocabularies such as 'bicubic'); upgrading Keras versions where the accepted set changed; assuming case-insensitive matching.

Understand the failure class

Background: "invalid argument", "unknown mode", "not supported": invalid enum-like argument errors explained — this error's family across 19 libraries.

Related errors

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/b48a2e3849314109. Report an issue: GitHub.

Appendix: source

Thrown at keras/src/layers/preprocessing/image_preprocessing/random_elastic_transform.py:107

        fill_mode="reflect",
        fill_value=0.0,
        value_range=(0, 255),
        seed=None,
        data_format=None,
        **kwargs,
    ):
        super().__init__(data_format=data_format, **kwargs)
        self._set_factor(factor)
        self.scale = self._set_factor_by_name(scale, "scale")
        self.interpolation = interpolation
        self.fill_mode = fill_mode
        self.fill_value = fill_value
        self.value_range = value_range
        self.seed = seed
        self.generator = SeedGenerator(seed)

        if interpolation not in self._SUPPORTED_INTERPOLATION:
            raise NotImplementedError(
                f"Unknown `interpolation` {interpolation}. Expected of one "
                f"{self._SUPPORTED_INTERPOLATION}."
            )

        if fill_mode not in self._SUPPORTED_FILL_MODES:
            raise NotImplementedError(
                f"Unknown `fill_mode` {fill_mode}. Expected of one "
                f"{self._SUPPORTED_FILL_MODES}."
            )

        if self.data_format == "channels_first":
            self.height_axis = -2
            self.width_axis = -1
            self.channel_axis = -3
        else:
            self.height_axis = -3
            self.width_axis = -2
            self.channel_axis = -1

View on GitHub (pinned to 7a34a03db6)