2 ms·

We should. However in this case, isn't it a bit of a stretch to assume they didn't put just about everything in the training data?