The bottom-up prediction of the properties of polymeric materials based on molecular dynamics simulation is a major challenge in soft matter physics. Coarse-grained (CG) models are often employed to access greater spatiotemporal scales required for many applications, but these models normally experience significantly altered thermodynamics and highly accelerated dynamics due to the reduced number of degrees of freedom upon coarse-graining. While CG models can be calibrated to meet certain properties at particular state points, there is unfortunately no temperature transferable and chemically specific coarse-graining method that allows for modeling of polymer dynamics over a wide temperature range. Here, we pragmatically address this problem by "correcting" for deviations in activation free energies that occur upon coarse-graining the dynamics of a model polymeric material (polystyrene). In particular, we propose a new strategy based on concepts drawn from the Adam-Gibbs (AG) theory of glass formation. Namely we renormalize the cohesive interaction strength and effective interaction length-scale parameters to modify the activation free energy. We show that this energy-renormalization method for CG modeling allows accurate prediction of atomistic dynamics over the Arrhenius regime, the non-Arrhenius regime of glass formation, and even the non-equilibrium glassy regime, thus allowing for the predictive modeling of dynamic properties of polymer over the entire range of glass formation. Our work provides a practical scheme for establishing temperature transferable coarse-grained models for predicting and designing the properties of polymeric materials.