This package provides 2 new hooks.
This hook :
- manages mlflow settings at the beginning and the end of the run (run start / end).
- log useful informations for reproducibility as
mlflow tags
(including kedroJournal
information and the commands used to launch the run). - register the pipeline as a valid
mlflow model
if it is aPipelineML
instance
This hook :
- must be used with the
MlflowPipelineHook
- autolog nodes parameters each time the pipeline is run (with
kedro run
or programatically).