Dave Berenbaum
Dave Berenbaum
A useful feature of having a dedicated `log_model` method would be to `dvc add` whatever gets saved.
Closing in favor of #472
Leaving aside what our plans are for DVC checkpoints, the primary need here seems to be auto-integrating with frameworks' existing checkpointing features (at least to start; we might also want...
This feels pretty important now for anyone not using VS Code. Otherwise, it's painful to see the report.
Should we add a test to https://github.com/iterative/dvc/blob/main/tests/integration/test_studio_live_experiments.py to ensure the subdir is sent to studio?
Prioritizing dvc fetch now because it's needed for studio backend to use cloud versioning (https://github.com/iterative/studio/issues/4782). Not prioritizing other items now.
> I think we should wait before implementing #4657 in any case (cc @dberenbaum). Wait for what? Your rationale for `outs_trie` makes sense to me, but I don't follow why...
Okay, makes sense, thanks for the explanation.
Sounds reasonable. @skshetry Can we keep the serialization part as minimal as possible so it doesn't become an entire separate story for now?
Aiming to do push this sprint