What’s your approach to versioning datasets and models in production?

Discussion

Posted by

h/jessica_m_99 • Mar 30, 2026

I am currently working on an ML project that is moving closer to production, and I am starting to realize how messy things can get without proper versioning.

Right now, we have multiple versions of datasets, different model iterations, and it is getting difficult to track what was used where. It also becomes confusing when we need to debug or roll back.

What’s your approach to versioning datasets and models in production? Do you use specific tools or follow a certain structure?

Would really appreciate practical workflows or setups that have worked for you.

1 COMMENTS

THE LOOP (1)

h/Surya • Apr 1, 2026

I usually version datasets and models together using tools like MLflow or DVC, and keep a clear mapping of data → model → metrics so rollback and debugging are easy.

0 REPLY

Machine Learning

c/machine-learning

A community for machine learning, focused on building, experimenting, and solving real-world problems using data. This community is focused on practical learning, not just theory. Ask questions, share projects, and discuss models, datasets, algorithms, and tools. Talk about model training, evaluation, deployment, and challenges you face while working on ML tasks. Share resources, use cases, and insights. Whether you are just starting or already working in the field, this is a space to learn, apply, and grow in machine learning.

h/tariq_the_techie

24 Aura

h/xiao_chen88

13 Aura

h/captain_uae

12 Aura

Stay on the timeline. Keep discussions relevant to the circle's core theme.
Respect the vibe. Toxicity will result in a drop in Aura and potential ban.