Big smape score on arima model

Question

I seem to get weird smape and mse results for my auto_arima model using python\'s pmdarima.

    forecasts = []
    for new_ob in y_test:
        fc = arima.pr