deepfakes/faceswap · error · FaceswapError

You are attempting to load weights from a '{retval[0].name}'

Error message

You are attempting to load weights from a '{retval[0].name}' model into a '{self._name}' model. This is not supported.

What it means

Raised in _get_weights_model when the weights file's first sub-model name does not equal the current model plugin's name (self._name). Faceswap only supports warm-starting within the same architecture, and uses the sub-model name as the identity check, so cross-model weight loading is refused before any layers are touched.

Source

Thrown at plugins/train/model/_base/io.py:522

        Returns
        -------
        List of all models contained within the .keras file

        Raises
        ------
        FaceswapError
            In the event of a failure to load the weights, or the weights belonging to a different
            model
        """
        retval = get_all_sub_models(k_models.load_model(    # pyright:ignore[reportArgumentType]
            self._weights_file,
            compile=False))
        if not retval:
            raise FaceswapError(f"Error loading weights file {self._weights_file}.")

        if retval[0].name != self._name:
            raise FaceswapError(f"You are attempting to load weights from a '{retval[0].name}' "
                                f"model into a '{self._name}' model. This is not supported.")
        return retval

    def _load_layer_weights(self,
                            layer: layers.Layer,
                            sub_weights: layers.Layer,
                            model_name: str) -> T.Literal[-1, 0, 1]:
        """Load the weights for a single layer.

        Parameters
        ----------
        layer
            The layer to set the weights for
        sub_weights
            The list of layers in the weights model to load weights from
        model_name
            The name of the current sub-model that is having it's weights loaded

View on GitHub (pinned to f530cb7508)

Solutions

1. Use weights saved from the same model plugin as the one being trained.
2. If the architecture changed deliberately, start fresh without load-weights.
3. Verify which model plugin the weights came from (the error prints both names) and align your -m model selection with it.

Example fix

# before
faceswap train -m original_model -wm dfl_sae_weights.keras

# after
faceswap train -m original_model -wm original_weights.keras  # same plugin

import keras.models as km
subs = get_all_sub_models(km.load_model(weights_file, compile=False))
assert subs[0].name == my_model_name, f"{subs[0].name} weights cannot load into {my_model_name}"

When it happens

Trigger: Selecting load-weights with a file saved from a different model plugin (e.g. 'original' weights into a 'dfl-sae'/'realface' session) — retval[0].name differs from self._name.

Common situations: Copying a weights command from a tutorial built around another model; renaming the model plugin folder while keeping old weights; experimenting with cross-architecture transfer, which Faceswap does not support.

Related errors

• No weights were successfully loaded from your weights file:
• Error loading weights file {self._weights_file}.
• Load weights selected, but the path '{weights_file}' does no
• Landmark based masks cannot be created for {self._landmark_t
• There is a mismatch between the number of frames found in th

AI-assisted analysis of deepfakes/faceswap@f530cb7508 (2026-08-15). Data as JSON: /api/errors/ca21c14e6684dbfa. Report an issue: GitHub.