Linearly stable shear flows first transition to turbulence in the form of localized turbulent patches. At low Reynolds numbers these turbulent patches tend to suddenly decay, following a memory less process. There is no satisfactory explanation as to why turbulence decays in these flows, nor how far in advance decay can be forecasted. Using massive ensembles of simulations of pipe flow and a reduced order model of shear flows we monitor how predictable different initial conditions are to decay events. In this database we include the GPU-codes we use to perform the massive ensembles of direct numerical simulations of pipe and a reduced order model of shear flows. Additionally the database includes time series of the predictability and main variables of many base flow trajectories, and classifications between predictable and unpredictable states. We also include a simple but still accurate model of reduced order model decays.