how to forecast with InnovationMLR model?

[adalisan]

I was successfully able to fit InnovationMLR models to SARS-COV2 seq. count data and get growth advantage models. How would I be able to run forecasting with such a model? There is this function that works for MultinomialLogisticRegression models,

def forecast_frequencies(post_samples, mlr, forecast_L):
    """Use posterior beta to forecast posterior frequencies.

        Args:
            samples (Dict[str,Array]): samples attribute of posterior object PosteriorHandler from evofr
            mlr (ModelSpec): model object from evofr
            forecast_L (int): forecasting length in days

        Returns:
            Array: freq. estimates for posterior
        """

    # Making feature matrix for forecasting
    last_T = post_samples["freq"].shape[1]

    X = mlr.make_ols_feature(start=last_T, stop=last_T + forecast_L)
    # Posterior beta
    beta = jnp.array(post_samples["beta"])

    # Matrix multiplication by sample
    dot_by_sample = vmap(jnp.dot, in_axes=(None, 0), out_axes=0)
    logits = dot_by_sample(X, beta)  # Logit frequencies by variant
    return softmax(logits, axis=-1)

we have the predictor values in the "predictors" array
predictors = np.empty((N_variants, n_features))
#load predictor values into numpy array 
how can I adapt the forecast_frequencies functions to forecast with the model forecast_L days in to the future?