This package provides 1 new hook.
This hook :
- manages mlflow settings at the beginning and the end of the run (run start / end).
- autolog nodes parameters each time the pipeline is run (with
kedro run
or programatically). - log useful informations for reproducibility as
mlflow tags
(including kedroJournal
information for old kedro versions and the commands used to launch the run). - register the pipeline as a valid
mlflow model
if it is aPipelineML
instance