keras-team/keras · error · NotImplementedError

`sample_weight` property method has not been implemented in

Error message

`sample_weight` property method has not been implemented in {}.

What it means

The base Iterator class exposes sample_weight as an abstract property; no stock legacy iterator implements it, so accessing it always raises NotImplementedError unless a custom subclass overrides it. It is a hook for user-defined iterators that carry per-sample weights.

Source

Thrown at keras/src/legacy/preprocessing/image.py:390

    def filepaths(self):
        """List of absolute paths to image files."""
        raise NotImplementedError(
            "`filepaths` property method has not "
            "been implemented in {}.".format(type(self).__name__)
        )

    @property
    def labels(self):
        """Class labels of every observation."""
        raise NotImplementedError(
            "`labels` property method has not been implemented in {}.".format(
                type(self).__name__
            )
        )

    @property
    def sample_weight(self):
        raise NotImplementedError(
            "`sample_weight` property method has not "
            "been implemented in {}.".format(type(self).__name__)
        )


@keras_export("keras._legacy.preprocessing.image.DirectoryIterator")
class DirectoryIterator(BatchFromFilesMixin, Iterator):
    """Iterator capable of reading images from a directory on disk.

    DEPRECATED.
    """

    allowed_class_modes = {"categorical", "binary", "sparse", "input", None}

    def __init__(
        self,
        directory,
        image_data_generator,

View on GitHub (pinned to 7a34a03db6)

Solutions

In custom iterators, override sample_weight to return the weights array (or None)
Pass sample_weight directly to model.fit(x, y, sample_weight=...) instead of reading it off the iterator
Wrap access in try/except NotImplementedError with a None default

Example fix

# before
class MySeq(keras.utils.Sequence):
    ...
sw = it.sample_weight  # NotImplementedError

# after
class MySeq(keras.utils.Sequence):
    def __init__(self, w):
        self._w = w
    @property
    def sample_weight(self):
        return self._w

Defensive patterns

Strategy: try-catch

Validate before calling

try:
    sw = iterator.sample_weight
except NotImplementedError:
    sw = None

Type guard

def has_sample_weight(it):
    try:
        it.sample_weight
        return True
    except NotImplementedError:
        return False

Try / catch

try:
    sw = iterator.sample_weight
except NotImplementedError:
    sw = None
model.fit(iterator, sample_weight=sw)

Prevention

Pass sample_weight to fit() explicitly instead of reading it from the iterator
Override sample_weight in custom Sequence subclasses

When it happens

Trigger: Reading iterator.sample_weight on any stock iterator (DirectoryIterator, NumpyArrayIterator), or failing to override it in a custom Iterator/Sequence subclass that supplies weights.

Common situations: Passing sample_weight to fit while using a custom Sequence/Iterator without implementing the property; generic introspection code enumerating iterator attributes.

Related errors

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/74c1563b36806096. Report an issue: GitHub.