What does open sourcing have to do with "reproducing"? Last I checked, open sourcing is about allowing others to modify and to distribute the modified version, which you can do with these. Yes, having the full training data and tooling would make it significantly easier, and it is a requirement for GPL, but not for Open Source licenses in general. You may add this as another argument in favor of going back in time and doing more to support Richard Stallman's vision, but this is the world in which we live now.