Is it possible to version files independently?

Hello DVC community!

I have a tricky use case and I’m not sure this is feasible with DVC.
My project features several datasets that I would like to version independently. I would like to keep track of:

  • dataset 1 with its own set of tags
  • datset 2 with its own set of tags
  • dataset 3 with its own set of tags
    so that I can execute a script with any combination of the versions of the three datasets (ex: dataset 1 v1, dataset 2 v2, dataset 3 v3 or dataset 1 v3, dataset 2 v2, dataset 3 v3, etc.).

Is this use case feasible with DVC?

Hello @hugoehlinger !
DVC can’t easily handle independently-versioned datasets within the same project. I guess the best way would be to have a data preparation stage that has the dataset versions as params and pulls the right combination of data from elsewhere (possibly auxiliary DVC repos containing only one dataset each).

Hello @ronan ! Thanks for answering me and giving me this tip