[Vecr]

It's probable you can train the same system on the same data multiple times and still get an improvement. You could also train on universal sequence prediction data in between as well.