Is loading a pickle-trained sklearn model under a different sklearn version a supported compatibility strategy?
Classic MLLevel: IntermediateEnvironment reproduction
Is loading a pickle-trained sklearn model under a different sklearn version a supported compatibility strategy?
Hints
- Serialization captures objects, not a compatibility promise.
- Reproducibility needs the environment.
Answer
No. Reproduce package versions and dependencies from training, and retain data/configuration needed to rebuild. A successful load alone does not establish supported or equivalent behavior.
Source and licence: scikit-learn 1.7.2 — model persistence · BSD-3-Clause · answer: Quantoros editorial