This AI Paper Proposes an Effective Paradigm for Large Scale Vision-and-Language Navigation (VLN) Training and Quantitatively Evaluates the Influence of Each Component in the Pipeline