Also my models beat baseline in evaluations so I'm stoked on that. It took some doing to produce results with time-series data.