# So apparently the Colab no longer has the option to run RVC from a Drive space and thus training checkpoints are volatile as if you get disconnected, your unsaved (to Drive) work goes "poof" too. How do I prevent that from happening?