A tutorial on how to use kedro-mlflow plugin (https://github.com/Galileo-Galilei/kedro-mlflow) to synchronize training and inference and serve kedro pipeline
37
stars
5
forks
source link
catalog and pipeline_ml_factory have inconsistent artifact_path #17
log_model_kwargs
are passed to thepipeline_ml_factory
function:https://github.com/Galileo-Galilei/kedro-mlflow-tutorial/blob/67941e00c6a87dec3d2754291d86087ce5d0813b/src/kedro_mlflow_tutorial/pipeline_registry.py#L30-L38
But they are ignored by kedro-mlflow and we need to use the default values when retrieveing the model.
https://github.com/Galileo-Galilei/kedro-mlflow-tutorial/blob/67941e00c6a87dec3d2754291d86087ce5d0813b/conf/base/catalog.yml#L64
Once the issue is solved in kedro-mlflow, we should update it here.