MLflow
Open platform for tracking experiments and managing model lifecycle.
Aperçu
MLflow tracks experiments, packages models and manages their progression through stages. It solves the handover problem between training a model and running it: the artefact needs to be identified, versioned and retrievable, not just present on somebody's disk. It exists because that handover is where most models are quietly lost, and it does it without requiring a hosted service.
Graphe des technologies
Comment MLflow se rattache au reste de la pile.
Appliqué à
- Fine-tuning technique Techniques MLflow used for Fine-tuning High ✓ source
- Pretraining technique Techniques MLflow used for Pretraining High ✓ source
Outils
Logiciels qui implémentent MLflow ou s'y intègrent.