GraphLab, since moving code distribution over to GitHub, appears to use a development cycle of "continuous improvement." By which I mean, the exact version of GraphLab you get depends on the specific day/time you do a "git pull" on the repository. As a developer, I love this. As a user, it terrifies me.
Can we get an official determination from the committee that a GraphLab checkout of a specific commit ID is guaranteed to be able to run all the competition datasets? While we hope that all commits represent useful improvements, there's no guarantee that will be the case... Last minute API changes to GraphLab could result in competition datasets being broken.
Thanks,
Jeff