configuration

def create_configuration(signal: TrainingSignal) -> Configuration:
    """Create a CAREamics configuration from a TrainingSignal.

    Parameters
    ----------
    signal : TrainingSignal
        Signal containing user-set training parameters.

    Returns
    -------
    Configuration
        CAREamics configuration.

    Raises
    ------
    ValueError
        If the algorithm is not supported.
    """
    # experiment name
    if signal.experiment_name == "":
        experiment_name = f"{signal.algorithm}_{signal.axes}"
    else:
        experiment_name = signal.experiment_name

    if signal.is_3d:
        # order of patches is ZYX
        patches: list[int] = [
            signal.patch_size_z,
            signal.patch_size_xy,
            signal.patch_size_xy,
        ]
    else:
        patches = [signal.patch_size_xy, signal.patch_size_xy]

    # model params
    model_params = {
        "depth": signal.depth,
        "num_channels_init": signal.num_conv_filters,
    }

    # augmentations
    augs: list[Union[XYFlipModel, XYRandomRotate90Model]] = []
    if signal.x_flip or signal.y_flip:
        augs.append(XYFlipModel(flip_x=signal.x_flip, flip_y=signal.y_flip))

    if signal.rotations:
        augs.append(XYRandomRotate90Model())

    # create configuration
    if signal.algorithm == SupportedAlgorithm.N2V:
        return create_n2v_configuration(
            experiment_name=experiment_name,
            data_type="tiff" if signal.load_from_disk else "array",
            axes=signal.axes,
            patch_size=patches,
            batch_size=signal.batch_size,
            num_epochs=signal.n_epochs,
            n_channels=signal.n_channels_n2v,
            augmentations=augs,
            independent_channels=signal.independent_channels,
            use_n2v2=signal.use_n2v2,
            logger="tensorboard",
            model_params=model_params,
        )
    elif signal.algorithm == SupportedAlgorithm.N2N:
        return create_n2n_configuration(
            experiment_name=experiment_name,
            data_type="tiff" if signal.load_from_disk else "array",
            axes=signal.axes,
            patch_size=patches,
            batch_size=signal.batch_size,
            num_epochs=signal.n_epochs,
            n_channels_in=signal.n_channels_in_care,
            n_channels_out=signal.n_channels_out_care,
            augmentations=augs,
            independent_channels=signal.independent_channels,
            logger="tensorboard",
            model_params=model_params,
        )
    elif signal.algorithm == SupportedAlgorithm.CARE:
        return create_care_configuration(
            experiment_name=experiment_name,
            data_type="tiff" if signal.load_from_disk else "array",
            axes=signal.axes,
            patch_size=patches,
            batch_size=signal.batch_size,
            num_epochs=signal.n_epochs,
            n_channels_in=signal.n_channels_in_care,
            n_channels_out=signal.n_channels_out_care,
            augmentations=augs,
            independent_channels=signal.independent_channels,
            logger="tensorboard",
            model_params=model_params,
        )
    else:
        raise ValueError(f"Unsupported algorithm: {signal.algorithm}")