Use numpy.random.RandomGenerator

[cwmeijer]

We could consider using a RandomGenerator object to pass around in our code. See this blog about the motivation. ChatGPT summary: Using a RandomGenerator object avoids the pitfalls of global state and enhances reproducibility, flexibility, and compatibility with modern NumPy features.

I tried this but it will be quite a lot of places where we'd have to change code. Could be worth it. Example:

    def explain(
        self,
        model_or_function: Union[str, callable],
        input_tabular: np.array,
        labels: Optional[Iterable[int]] = None,
        mask_type: Optional[Union[str, callable]] = 'most_frequent',
        batch_size: Optional[int] = 100,
        rng: Optional[np.random.Generator] = None,  # ADD THIS LINE -------------------
    ) -> np.array:
        """Run the RISE explainer.

        Args:
            model_or_function: The function that runs the model to be explained
                                                 or the path to a ONNX model on disk.
            input_tabular: Data to be explained.
            labels: Indices of classes to be explained.
            mask_type: Imputation strategy for masked features
            batch_size: Number of samples to process by the model per batch
            rng: random number generator  # ADD THIS LINE -------------------

        Returns:
            explanation: An Explanation object containing the LIME explanations for each class.
        """
        rng = np.random.default_rng(rng)  # ADD THIS LINE -------------------

        # run the explanation.
        runner = utils.get_function(model_or_function)

        masks = np.stack(
            list(
                generate_tabular_masks(input_tabular.shape, 
                                                      number_of_masks=self.n_masks, p_keep=self.p_keep, 
                                                      rng=rng)))  # PASS RNG ON ------------------
        self.masks = masks if self.keep_masks else None

        masked = mask_data_tabular(input_tabular,