keras-team/keras · error · ValueError

ImageNet weights can only be loaded with NASNetLarge or…

Error message

ImageNet weights can only be loaded with NASNetLarge or NASNetMobile

What it means

NASNet weight-loading branch in keras/src/applications/nasnet.py: pretrained ImageNet checkpoints are only published for NASNetLarge (331px) and NASNetMobile (224px) configurations. A custom NASNet variant (different penultimate_filters/default_size) reaches the weight-loading code without a matching checkpoint, so ImageNet weights cannot be provided for it.

Solutions

  1. Use the stock NASNetLarge or NASNetMobile entry points when requesting weights="imagenet"
  2. Use weights=None for custom NASNet architectures and train from scratch
  3. Check that penultimate_filters and default_size match a configuration with published weights
Defensive patterns

Strategy: fallback

When it happens

Trigger: Thrown at keras/src/applications/nasnet.py:304 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/af4b1221f7fbe0f4. Report an issue: GitHub.

Appendix: source

Thrown at keras/src/applications/nasnet.py:304

            model.load_weights(weights_path)
        elif default_size == 331:  # large version
            if include_top:
                weights_path = file_utils.get_file(
                    "nasnet_large.h5",
                    NASNET_LARGE_WEIGHT_PATH,
                    cache_subdir="models",
                    file_hash="11577c9a518f0070763c2b964a382f17",
                )
            else:
                weights_path = file_utils.get_file(
                    "nasnet_large_no_top.h5",
                    NASNET_LARGE_WEIGHT_PATH_NO_TOP,
                    cache_subdir="models",
                    file_hash="d81d89dc07e6e56530c4e77faddd61b5",
                )
            model.load_weights(weights_path)
        else:
            raise ValueError(
                "ImageNet weights can only be loaded with NASNetLarge"
                " or NASNetMobile"
            )
    elif weights is not None:
        model.load_weights(weights)

    return model


@keras_export(
    [
        "keras.applications.nasnet.NASNetMobile",
        "keras.applications.NASNetMobile",
    ]
)
def NASNetMobile(
    input_shape=None,
    include_top=True,

View on GitHub (pinned to 7a34a03db6)