I often get stuck by the calculation time when I work on a model locally. Even sampling quite heavily sometimes still let the running times very long. Ideally, these times must be of the order of the second. Otherwise it makes experimenting a bit tedious.
SageMaker provides with quite a handy way to fire up a familiar model development environment by mean of its notebook instances. To switch from your local to the online env, it recommends using github or its own git platform. I guess the idea is just to pull your branch where you left it locally before you decided to switch.
Problem is, how do you do when your git platform does not integrate with AWS (company constraints) ? Or is there another smooth way to do that ? (scp ?)
Just to be clear, when I talk about developing a model, I do not just talk about training and scoring a model. Most of the time I need to work and analyse in-depth the impact of a pre processing step.
