Best practices for a new deep learning project

[asvskartheek]

Is this a duplicate issue to this?