Stage is a collection of algorithms and abstractions (classes) for model-based reinforcement learning research. Implemented are iLQR applied to learned models as well as PETS with various optimizers such as CEM and PI2.